Pentium-M put up against P4EE and A64FX

Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

The Pentium-M seems to be unusually superb, even going up against the
ultimate gaming processors.

GamePC - Dothan Revisted : Focusing On Gaming Performance
http://www.gamepc.com/labs/view_content.asp?id=dothangaming&page=1

Yousuf Khan
50 answers Last reply
More about pentium p4ee a64fx
  1. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    yjkhan@gmail.com (ykhan) wrote:

    >The Pentium-M seems to be unusually superb, even going up against the
    >ultimate gaming processors.
    >
    >GamePC - Dothan Revisted : Focusing On Gaming Performance
    >http://www.gamepc.com/labs/view_content.asp?id=dothangaming&page=1

    Did I read that right, that the Pentium-M did so well, even though it
    was only running a 64-bit-wide, DDR333 memory interface?
  2. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    yjkhan@gmail.com (ykhan) wrote :

    > The Pentium-M seems to be unusually superb, even going up against the
    > ultimate gaming processors.
    >
    > GamePC - Dothan Revisted : Focusing On Gaming Performance
    > http://www.gamepc.com/labs/view_content.asp?id=dothangaming&page=1

    and if you look at the banners you will see that gamePC sells gamer
    systems based on P-M .. what a surprise


    Pozdrawiam.
    --
    RusH //
    http://randki.o2.pl/profil.php?id_r=352019
    Like ninjas, true hackers are shrouded in secrecy and mystery.
    You may never know -- UNTIL IT'S TOO LATE.
  3. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    "ykhan" <yjkhan@gmail.com> wrote in message
    news:bd84ed0c.0411231258.49ef02be@posting.google.com...

    " The Pentium-M seems to be unusually superb, even going up against the
    ultimate gaming processors. "


    I priced up the three test systems they built for a comparison. Newegg
    didn't have any Corsair 2-2-2 PC2700 DDR, so I put the equivalent PC3200
    with it. They also didn't have any Corsair 4-4-4 PC4200 DDR2, so I put the
    equivalent PC5400 with it.


    - Intel Pentium M 755 2.0GHz http://snipurl.com/aumx $439.00
    - Corsair TWINX1024-3200XLPT http://snipurl.com/a5er $251.00
    - Aopen i855GMEm-LFS http://snipurl.com/aumu $269.00
    - Gigabyte Radeon X800 XT AGP http://snipurl.com/auoq $489.00
    - Western Digital Raptor 74GB SATA http://snipurl.com/auox $151.00
    TOTAL: $1599.00


    - AMD Athlon64 FX-55 2.6GHz http://snipurl.com/aun8 $849.00
    - Corsair TWINX1024-3200XLPT http://snipurl.com/a5er $251.00
    - Asus A8V Deluxe http://snipurl.com/aumz $129.00
    - Gigabyte Radeon X800 XT AGP http://snipurl.com/auoq $489.00
    - Western Digital Raptor 74GB SATA http://snipurl.com/auox $151.00
    TOTAL: $1869.00


    - Intel Pentium 4 3.4GHz EE http://snipurl.com/aune $1019.00
    - Corsair TWIN2X1024-5400C4 http://snipurl.com/auoe $355.00
    - ASUS P5AD2 Premium http://snipurl.com/aun1 $261.00
    - Sapphire Radeon X800 XT PCI-E http://snipurl.com/auos $599.00
    - Western Digital Raptor 74GB SATA http://snipurl.com/auox $151.00
    TOTAL: $2385.00
  4. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    "Cuzman" <cuzNOSPAM@supanet.com> wrote :

    > - AMD Athlon64 FX-55 2.6GHz http://snipurl.com/aun8 $849.00
    > - Intel Pentium 4 3.4GHz EE http://snipurl.com/aune $1019.00

    look at the tests
    "Intel Pentium 4 540 (3.2 GHz)" and "AMD Athlon64 3200+ (2.0 GHz)" are
    just little behind the topscorers, and are significantly cheaper. Only
    "I have loads of money and I'm an ignorant ass" peaple buy FX and EE
    processors for gaming systems.


    Pozdrawiam.
    --
    RusH //
    http://randki.o2.pl/profil.php?id_r=352019
    Like ninjas, true hackers are shrouded in secrecy and mystery.
    You may never know -- UNTIL IT'S TOO LATE.
  5. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    On Wed, 24 Nov 2004 01:28:11 +0000 (UTC), RusH <logistyka1@pf.pl>
    wrote:

    >"Cuzman" <cuzNOSPAM@supanet.com> wrote :
    >
    >> - AMD Athlon64 FX-55 2.6GHz http://snipurl.com/aun8 $849.00
    >> - Intel Pentium 4 3.4GHz EE http://snipurl.com/aune $1019.00
    >
    >look at the tests
    >"Intel Pentium 4 540 (3.2 GHz)" and "AMD Athlon64 3200+ (2.0 GHz)" are
    >just little behind the topscorers, and are significantly cheaper. Only
    >"I have loads of money and I'm an ignorant ass" peaple buy FX and EE
    >processors for gaming systems.
    >
    >
    >Pozdrawiam.

    Opteron 150 (essentially all the same as FX-55, but socket 940 instead
    of 939) - $599 http://www.pricewatch.com/
    Doesn't this make AMD system more price-competitive?
  6. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    "nobody@nowhere.net" <mygarbage2000@hotmail.com> wrote :

    > Opteron 150 (essentially all the same as FX-55, but socket 940
    > instead of 939) - $599 http://www.pricewatch.com/
    > Doesn't this make AMD system more price-competitive?

    + you can O/C AMD 3200 to 2450Hz easily (done that), dont know how O/C
    friendly FXes are


    Pozdrawiam.
    --
    RusH //
    http://randki.o2.pl/profil.php?id_r=352019
    Like ninjas, true hackers are shrouded in secrecy and mystery.
    You may never know -- UNTIL IT'S TOO LATE.
  7. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    RusH wrote:
    > and if you look at the banners you will see that gamePC sells gamer
    > systems based on P-M .. what a surprise

    ? just as they sell gamer systems based on any AMD or Intel CPU on the
    market. gaming is just a common benchmark for non-server rigs.
  8. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    On Wed, 24 Nov 2004 01:28:11 +0000 (UTC), RusH <logistyka1@pf.pl>
    wrote:

    >"Cuzman" <cuzNOSPAM@supanet.com> wrote :
    >
    >> - AMD Athlon64 FX-55 2.6GHz http://snipurl.com/aun8 $849.00
    >> - Intel Pentium 4 3.4GHz EE http://snipurl.com/aune $1019.00
    >
    >look at the tests
    >"Intel Pentium 4 540 (3.2 GHz)" and "AMD Athlon64 3200+ (2.0 GHz)" are
    >just little behind the topscorers, and are significantly cheaper. Only
    >"I have loads of money and I'm an ignorant ass" peaple buy FX and EE
    >processors for gaming systems.
    >

    I would go for the 90nm version
    of the Athlon64 3200+... $189ea (pricewatch)

    As for a case & power supply.. start out with the quiet Antec Sonata..
    http://www.antec-inc.com/pro_details_enclosure.php?ProdID=15138

    You'll save a lot money initially,
    have plenty of future upgradeability,
    reduce your re-occuring costs (power consumption),
    and end up with a much quieter system.
  9. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    nobody@nowhere.net wrote:
    > On Wed, 24 Nov 2004 01:28:11 +0000 (UTC), RusH <logistyka1@pf.pl>
    > wrote:
    >
    >
    >>"Cuzman" <cuzNOSPAM@supanet.com> wrote :
    >>
    >>
    >>>- AMD Athlon64 FX-55 2.6GHz http://snipurl.com/aun8 $849.00
    >>>- Intel Pentium 4 3.4GHz EE http://snipurl.com/aune $1019.00
    >>
    >>look at the tests
    >>"Intel Pentium 4 540 (3.2 GHz)" and "AMD Athlon64 3200+ (2.0 GHz)" are
    >>just little behind the topscorers, and are significantly cheaper. Only
    >>"I have loads of money and I'm an ignorant ass" peaple buy FX and EE
    >>processors for gaming systems.
    >>
    >>
    >>Pozdrawiam.
    >
    >
    > Opteron 150 (essentially all the same as FX-55,

    Not quite. It is the same speed as the FX-53
    and hence 200 MHz slower than the FX-55.

    > but socket 940 instead
    > of 939) - $599 http://www.pricewatch.com/
    > Doesn't this make AMD system more price-competitive?
  10. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    Rob Stow <rob.stow.nospam@shaw.ca> wrote in message news:<pd2pd.313315$Pl.122190@pd7tw1no>...
    > nobody@nowhere.net wrote:
    > > Opteron 150 (essentially all the same as FX-55,
    >
    > Not quite. It is the same speed as the FX-53
    > and hence 200 MHz slower than the FX-55.

    And probably doesn't have access to the 1000Mhz Hypertransport bus of
    the Athlon 64 or FX either. Don't know how much difference that would
    actually make though, since the memory bus is completely separate from
    the Hypertransport bus in any case.

    Yousuf Khan
  11. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    On Wed, 24 Nov 2004 09:21:46 +0000 (UTC), RusH <logistyka1@pf.pl>
    wrote:

    >"nobody@nowhere.net" <mygarbage2000@hotmail.com> wrote :
    >
    >> Opteron 150 (essentially all the same as FX-55, but socket 940
    >> instead of 939) - $599 http://www.pricewatch.com/
    >> Doesn't this make AMD system more price-competitive?
    >
    >+ you can O/C AMD 3200 to 2450Hz easily (done that), dont know how O/C
    >friendly FXes are
    >
    >
    >Pozdrawiam.

    All I know for sure is that Opteron242 in dual config easily takes the
    10% soft O/C provided by the MCI software bundled with Master2FAR
    board. The only problem with it is that every time I reboot the
    system I need to open that funky control panel and set the O/C again
    because it reverts to default speed. I kinda grew tired of it and
    stopped doing it. Besides, I am still waiting for the task that would
    load both CPUs up to 100%
    ;-)
  12. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    "nobody@nowhere.net" <mygarbage2000@hotmail.com> wrote in message
    news:d6t7q0pusfmo0hhdfam6pf5p9hsra9rbt9@4ax.com...
    > On Wed, 24 Nov 2004 01:28:11 +0000 (UTC), RusH <logistyka1@pf.pl>
    > wrote:
    >
    >>"Cuzman" <cuzNOSPAM@supanet.com> wrote :
    >>
    >>> - AMD Athlon64 FX-55 2.6GHz http://snipurl.com/aun8 $849.00
    >>> - Intel Pentium 4 3.4GHz EE http://snipurl.com/aune $1019.00
    >>
    >>look at the tests
    >>"Intel Pentium 4 540 (3.2 GHz)" and "AMD Athlon64 3200+ (2.0 GHz)" are
    >>just little behind the topscorers, and are significantly cheaper. Only
    >>"I have loads of money and I'm an ignorant ass" people buy FX and EE
    >>processors for gaming systems.
    >>
    >>
    >>Pozdrawiam.
    >
    > Opteron 150 (essentially all the same as FX-55, but socket 940 instead
    > of 939) - $599 http://www.pricewatch.com/
    > Doesn't this make AMD system more price-competitive?

    Not so much as you might think, though. While you would save $350 on the
    CPU, Socket 940 needs REGISTERED DIMMs, thus making the memory more
    expensive. For 1GB, it would be $271, and that is 2-2-3-6 timings, little
    more relaxed than the $251 that was given above for non-registered DIMMs.
    http://www.newegg.com/app/ViewProductDesc.asp?description=20-146-940&depa=1
    is what I am referring to specifically. Not much of a difference, but it
    would be more like $70 different if the base system had 2-2-3-6 timings
    instead of 2-2-2-5. Also, as from Newegg's site, $599 is for the OEM
    version. Whereas the price poster quoted $849 for the FX-55, that is a
    Retail item. The actual price for a Retail Opteron 150 is $649, thus
    changing the $350 to $200, to be fair. Further, you'd need a seperate
    heatsink for the Opteron, as it isn't included in OEM, let's say a decent
    one is $25. Further, the given Socket 939 board is $129, whereas the
    cheapest Socket 940 board is $172. So, now the $350 difference has become
    more like $240 less. Unless you go with a Retail CPU, then it is more like
    $140 difference. Since the given price difference between the Pentium-M and
    FX-55 was $250, this won't change the price leader, but it does make AMD
    more competitive.

    Also, I'd like to point out at this point, that an Opteron 150 is the same
    speed as a FX-53, which is actually $54 cheaper than an Opteron 150 (OEM v
    OEM), or $150 cheaper (Retail v Retail). Point being, a FX-53 is $545 OEM ;
    $599 Retail, whereas the Opteron 150 is $599 OEM; $649 Retail. They are
    both Socket 940 parts, so, the adjustments in RAM and Motherboard would
    remain with a FX-53, but, the CPU is cheaper.

    Further, an Athlon64 4000+ is the same as a FX-53, but in Socket 939 form.
    The CPU is, based off the Opteron 150 setup, $165 more expensive based on
    something which changes the price from being $240 less to being just $75
    less, BUT, you have to use the cheaper motherboard, and cheaper RAM, thus
    making that $75 less actually $165 less (the numbers just work out that
    way). That changes the $1849 figure to $1684, a grand total of $85
    difference between the Athlon64 4000+ setup and the Pentium-M setup.

    BTW, a Socket 479 board can only be upgraded to a Pentium-M 2.0GHz,
    officially. From where it stands now, the Socket 939/940 boards will be
    around for a while longer, I don't know how long, but I would imagine
    atleast until the CPUs hit a 3.0GHz speed....Guess they would nickname that
    either a FX-59 or Athlon64 4600+ (assuming 1MB L2).

    Either way, the P4 seems out of it, eh?
  13. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    "ykhan" <yjkhan@gmail.com> wrote in message
    news:bd84ed0c.0411231258.49ef02be@posting.google.com...
    > The Pentium-M seems to be unusually superb, even going up against the
    > ultimate gaming processors.
    >
    > GamePC - Dothan Revisted : Focusing On Gaming Performance
    > http://www.gamepc.com/labs/view_content.asp?id=dothangaming&page=1
    >
    > Yousuf Khan

    Well I certainly hope that Intel does push this as thier new desktop chip of
    choice. I would deffonetly like to see AMD and Intel really start competing
    again pushing the preformance envolope! Wait? What am I saying? That just
    means my new AMD64 3000+ will be religated to being "low end" again even
    more quickly! And I was just getting used to having a mid-range systetm!

    Carlo Razzeto
  14. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    "chrisv" <chrisv@nospam.invalid> wrote in message
    news:g1b7q01k5q1co4b35l91boktukqriqf1e9@4ax.com...
    > yjkhan@gmail.com (ykhan) wrote:
    >
    > >The Pentium-M seems to be unusually superb, even going up against the
    > >ultimate gaming processors.
    > >
    > >GamePC - Dothan Revisted : Focusing On Gaming Performance
    > >http://www.gamepc.com/labs/view_content.asp?id=dothangaming&page=1
    >
    > Did I read that right, that the Pentium-M did so well, even though it
    > was only running a 64-bit-wide, DDR333 memory interface?
    >

    No. You read that wrong. At nominal core frequencies, the
    performance of Dothan family generally aligns with performance
    trends of Athlon-FX family. But Dothan systematically lags in
    core frequency, and therefore in performance, look more carefully
    at the charts. The impression of Dothan performing exceptionally
    well is because they included the 15%-overclocked data point.

    Keep in mind that core frequency disparity (on a given process
    generation) is a consequence of differences in chip core design,
    therefore the disparity bethween Dothan and A-FX will continue
    going forward. So the performance.

    - aap
  15. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    ykhan wrote:
    > Rob Stow <rob.stow.nospam@shaw.ca> wrote in message news:<pd2pd.313315$Pl.122190@pd7tw1no>...
    >
    >>nobody@nowhere.net wrote:
    >>
    >>>Opteron 150 (essentially all the same as FX-55,
    >>
    >>Not quite. It is the same speed as the FX-53
    >>and hence 200 MHz slower than the FX-55.
    >
    >
    > And probably doesn't have access to the 1000Mhz Hypertransport bus of
    > the Athlon 64 or FX either. Don't know how much difference that would
    > actually make though, since the memory bus is completely separate from
    > the Hypertransport bus in any case.

    But... doesn't every Opteron has a Hypertransport bus, and the SMP
    models have two? I'm looking at a spec sheet which seems to indicate that.

    --
    bill davidsen (davidsen@darkstar.prodigy.com)
    SBC/Prodigy Yorktown Heights NY data center
    Project Leader, USENET news
    http://newsgroups.news.prodigy.com
  16. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    "Bill Davidsen" <davidsen@darkstar.prodigy.com> wrote in message
    news:bQRqd.1700$2N1.318@newssvr31.news.prodigy.com...
    > But... doesn't every Opteron has a Hypertransport bus, and the SMP models
    > have two? I'm looking at a spec sheet which seems to indicate that.
    >
    > --
    > bill davidsen (davidsen@darkstar.prodigy.com)
    > SBC/Prodigy Yorktown Heights NY data center
    > Project Leader, USENET news
    > http://newsgroups.news.prodigy.com

    You are right, every A64 based chip uses HT as it's internal interconnect.

    Carlo
  17. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    On Tue, 30 Nov 2004 03:31:19 +0000, Bill Davidsen wrote:

    > ykhan wrote:
    >> Rob Stow <rob.stow.nospam@shaw.ca> wrote in message news:<pd2pd.313315$Pl.122190@pd7tw1no>...
    >>
    >>>nobody@nowhere.net wrote:
    >>>
    >>>>Opteron 150 (essentially all the same as FX-55,
    >>>
    >>>Not quite. It is the same speed as the FX-53
    >>>and hence 200 MHz slower than the FX-55.
    >>
    >>
    >> And probably doesn't have access to the 1000Mhz Hypertransport bus of
    >> the Athlon 64 or FX either. Don't know how much difference that would
    >> actually make though, since the memory bus is completely separate from
    >> the Hypertransport bus in any case.
    >
    > But... doesn't every Opteron has a Hypertransport bus, and the SMP
    > models have two? I'm looking at a spec sheet which seems to indicate that.

    Opterons have three HT links. SMP versions have three "coherent" HT
    busses. Non-SMP versions have three non-coherent HT busses. It's not
    clear what the difference is. It seems it's a product differentiation
    issue.

    IIRC, Athlon64s (socket 754s) have only one HT link.

    --
    Keith
  18. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    On Tue, 30 Nov 2004 03:31:19 GMT, Bill Davidsen
    <davidsen@darkstar.prodigy.com> wrote:

    >ykhan wrote:
    >> Rob Stow <rob.stow.nospam@shaw.ca> wrote in message news:<pd2pd.313315$Pl.122190@pd7tw1no>...
    >>
    >>>nobody@nowhere.net wrote:
    >>>
    >>>>Opteron 150 (essentially all the same as FX-55,
    >>>
    >>>Not quite. It is the same speed as the FX-53
    >>>and hence 200 MHz slower than the FX-55.
    >>
    >>
    >> And probably doesn't have access to the 1000Mhz Hypertransport bus of
    >> the Athlon 64 or FX either. Don't know how much difference that would
    >> actually make though, since the memory bus is completely separate from
    >> the Hypertransport bus in any case.
    >
    >But... doesn't every Opteron has a Hypertransport bus, and the SMP
    >models have two? I'm looking at a spec sheet which seems to indicate that.

    Opterons have hypertransport links that operate at 800MHz (1600MT/s),
    the latest models of Athlon64 have HT links running at 1000MHz.

    It's unclear if there is any performance advantage to this for
    single-processor systems, though perhaps somewhat ironically,
    multiprocessor Opterons should benefit from it.

    -------------
    Tony Hill
    hilla <underscore> 20 <at> yahoo <dot> ca
  19. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    keith wrote:
    > On Tue, 30 Nov 2004 03:31:19 +0000, Bill Davidsen wrote:
    >
    >
    >>ykhan wrote:
    >>
    >>>Rob Stow <rob.stow.nospam@shaw.ca> wrote in message news:<pd2pd.313315$Pl.122190@pd7tw1no>...
    >>>
    >>>
    >>>>nobody@nowhere.net wrote:
    >>>>
    >>>>
    >>>>>Opteron 150 (essentially all the same as FX-55,
    >>>>
    >>>>Not quite. It is the same speed as the FX-53
    >>>>and hence 200 MHz slower than the FX-55.
    >>>
    >>>
    >>>And probably doesn't have access to the 1000Mhz Hypertransport bus of
    >>>the Athlon 64 or FX either. Don't know how much difference that would
    >>>actually make though, since the memory bus is completely separate from
    >>>the Hypertransport bus in any case.
    >>
    >>But... doesn't every Opteron has a Hypertransport bus, and the SMP
    >>models have two? I'm looking at a spec sheet which seems to indicate that.
    >
    >
    > Opterons have three HT links. SMP versions have three "coherent" HT
    > busses. Non-SMP versions have three non-coherent HT busses. It's not
    > clear what the difference is. It seems it's a product differentiation
    > issue.
    >

    The 1xx have three non-coherent.
    The 2xx have one coherent, two non-coherent.
    The 8xx have three coherent.
  20. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    alexi wrote:
    > "chrisv" <chrisv@nospam.invalid> wrote in message
    > news:g1b7q01k5q1co4b35l91boktukqriqf1e9@4ax.com...
    >
    >>yjkhan@gmail.com (ykhan) wrote:
    >>
    >>
    >>>The Pentium-M seems to be unusually superb, even going up against the
    >>>ultimate gaming processors.
    >>>
    >>>GamePC - Dothan Revisted : Focusing On Gaming Performance
    >>>http://www.gamepc.com/labs/view_content.asp?id=dothangaming&page=1
    >>
    >>Did I read that right, that the Pentium-M did so well, even though it
    >>was only running a 64-bit-wide, DDR333 memory interface?
    >>
    >
    >
    > No. You read that wrong. At nominal core frequencies, the
    > performance of Dothan family generally aligns with performance
    > trends of Athlon-FX family. But Dothan systematically lags in
    > core frequency, and therefore in performance, look more carefully
    > at the charts. The impression of Dothan performing exceptionally
    > well is because they included the 15%-overclocked data point.

    What? That's not a projected point, it's measured performance. It looks
    as if it performs well because it does.
    >
    > Keep in mind that core frequency disparity (on a given process
    > generation) is a consequence of differences in chip core design,
    > therefore the disparity bethween Dothan and A-FX will continue
    > going forward. So the performance.
    >
    > - aap
    >
    >


    --
    bill davidsen (davidsen@darkstar.prodigy.com)
    SBC/Prodigy Yorktown Heights NY data center
    Project Leader, USENET news
    http://newsgroups.news.prodigy.com
  21. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    Bill Davidsen wrote:
    > alexi wrote:
    >> The impression of Dothan performing exceptionally
    >> well is because they included the 15%-overclocked data point.
    >
    > What? That's not a projected point, it's measured performance. It looks
    > as if it performs well because it does.

    A lot of people don't consider it fair to compare an overclocked
    processor in a match against commercially availible chips. I can think
    of two good reasons. 1) Not everybody is willing to overclock, so this
    performance is not availible "out of the box". 2) If you get to
    overclock that processor, you need to compare it to overclocked versions
    of the competitors. That is usually never done, leaving an unfair
    playing field.

    I am on the side of Dothan being a great chip. Even the un-overclocked
    version performed very well. And it was still a notebook chip. When
    intel optimizes for desktop use, it will likely stomp the P4-EE and the
    FX. But only time will tell, since that's still two generations down
    the road according to internet sources *ugh*.

    Alex
    --
    My words are my own. They represent no other; they belong to no other.
    Don't read anything into them or you may be required to compensate me
    for violation of copyright. (I do not speak for my employer.)
  22. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    On Tue, 30 Nov 2004 08:18:10 -0500, Alex Johnson <compuwiz@jhu.edu> wrote:

    >Bill Davidsen wrote:
    >> alexi wrote:
    >>> The impression of Dothan performing exceptionally
    >>> well is because they included the 15%-overclocked data point.
    >>
    >> What? That's not a projected point, it's measured performance. It looks
    >> as if it performs well because it does.
    >
    >A lot of people don't consider it fair to compare an overclocked
    >processor in a match against commercially availible chips. I can think
    >of two good reasons. 1) Not everybody is willing to overclock, so this
    >performance is not availible "out of the box". 2) If you get to
    >overclock that processor, you need to compare it to overclocked versions
    >of the competitors. That is usually never done, leaving an unfair
    >playing field.

    So you *do* agree with this opinion... that one should overclock neither or
    both?... *AND* that the results are interesting only to people who like to
    overclock.

    >I am on the side of Dothan being a great chip. Even the un-overclocked
    >version performed very well. And it was still a notebook chip. When
    >intel optimizes for desktop use, it will likely stomp the P4-EE and the
    >FX. But only time will tell, since that's still two generations down
    >the road according to internet sources *ugh*.

    I think you'd better hold judgement on stomping Athlon64 until the EM64T(?)
    version of Pentium-M is released - AMD is gaining "experience" with this
    ISA/mode while Intel apparently twiddles its thumbs and dishes out market
    segmentation gobbledygook... with all the "weight" of the goons at IDC,
    Forrester, et.al. and of course, RCK, behind it. Personally, I think we
    *may* get back to a situation where Intel & AMD play leapfrog on clock
    frequencies, even though neither will apparently be using that in its model
    numbering.:-)

    Rgds, George Macdonald

    "Just because they're paranoid doesn't mean you're not psychotic" - Who, me??
  23. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    On Tue, 30 Nov 2004 05:12:06 +0000, Rob Stow wrote:

    > keith wrote:
    >> On Tue, 30 Nov 2004 03:31:19 +0000, Bill Davidsen wrote:
    >>
    >>
    >>>ykhan wrote:
    >>>
    >>>>Rob Stow <rob.stow.nospam@shaw.ca> wrote in message news:<pd2pd.313315$Pl.122190@pd7tw1no>...
    >>>>
    >>>>
    >>>>>nobody@nowhere.net wrote:
    >>>>>
    >>>>>
    >>>>>>Opteron 150 (essentially all the same as FX-55,
    >>>>>
    >>>>>Not quite. It is the same speed as the FX-53
    >>>>>and hence 200 MHz slower than the FX-55.
    >>>>
    >>>>
    >>>>And probably doesn't have access to the 1000Mhz Hypertransport bus of
    >>>>the Athlon 64 or FX either. Don't know how much difference that would
    >>>>actually make though, since the memory bus is completely separate from
    >>>>the Hypertransport bus in any case.
    >>>
    >>>But... doesn't every Opteron has a Hypertransport bus, and the SMP
    >>>models have two? I'm looking at a spec sheet which seems to indicate that.
    >>
    >>
    >> Opterons have three HT links. SMP versions have three "coherent" HT
    >> busses. Non-SMP versions have three non-coherent HT busses. It's not
    >> clear what the difference is. It seems it's a product differentiation
    >> issue.
    >>
    >
    > The 1xx have three non-coherent.
    > The 2xx have one coherent, two non-coherent.
    > The 8xx have three coherent.

    Thank you. T'was late. The above is obvious when you note the
    segmentation. The real question remaining is what's the difference? Is
    it a paper fifference, intentional crippling, or perhaps testing, or (oh,
    my!) a real difference between the links?

    --
    Keith

    --
    Keith
  24. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    Alex Johnson wrote:
    > Bill Davidsen wrote:
    >
    >> alexi wrote:
    >>
    >>> The impression of Dothan performing exceptionally
    >>> well is because they included the 15%-overclocked data point.
    >>
    >>
    >> What? That's not a projected point, it's measured performance. It
    >> looks as if it performs well because it does.
    >
    >
    > A lot of people don't consider it fair to compare an overclocked
    > processor in a match against commercially availible chips.

    I assume you mean "operating at rated clock speed" but I take your meaning.

    > I can think
    > of two good reasons. 1) Not everybody is willing to overclock, so this
    > performance is not availible "out of the box". 2) If you get to
    > overclock that processor, you need to compare it to overclocked versions
    > of the competitors. That is usually never done, leaving an unfair
    > playing field.

    The "exceptionally well" and "in spec" are not mutually exclusive. I'm
    not disagreeing with you in any way, but the performance is real, and
    who knows if a higher clocked chip will appear on the market. That's a
    marketing rather than technical issue.
    >
    > I am on the side of Dothan being a great chip. Even the un-overclocked
    > version performed very well. And it was still a notebook chip. When
    > intel optimizes for desktop use, it will likely stomp the P4-EE and the
    > FX. But only time will tell, since that's still two generations down
    > the road according to internet sources *ugh*.

    I suspect that Intel could drop the price of the EE if they chose to do
    so, making them more cost effective. I'm not against the Dothan or any
    other chip if it fits a need.

    --
    bill davidsen (davidsen@darkstar.prodigy.com)
    SBC/Prodigy Yorktown Heights NY data center
    Project Leader, USENET news
    http://newsgroups.news.prodigy.com
  25. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    keith wrote:
    > On Tue, 30 Nov 2004 05:12:06 +0000, Rob Stow wrote:
    >
    >
    >>keith wrote:
    >>
    >>>On Tue, 30 Nov 2004 03:31:19 +0000, Bill Davidsen wrote:
    >>>
    >>>
    >>>
    >>>>ykhan wrote:
    >>>>
    >>>>
    >>>>>Rob Stow <rob.stow.nospam@shaw.ca> wrote in message news:<pd2pd.313315$Pl.122190@pd7tw1no>...
    >>>>>
    >>>>>
    >>>>>
    >>>>>>nobody@nowhere.net wrote:
    >>>>>>
    >>>>>>
    >>>>>>
    >>>>>>>Opteron 150 (essentially all the same as FX-55,
    >>>>>>
    >>>>>>Not quite. It is the same speed as the FX-53
    >>>>>>and hence 200 MHz slower than the FX-55.
    >>>>>
    >>>>>
    >>>>>And probably doesn't have access to the 1000Mhz Hypertransport bus of
    >>>>>the Athlon 64 or FX either. Don't know how much difference that would
    >>>>>actually make though, since the memory bus is completely separate from
    >>>>>the Hypertransport bus in any case.
    >>>>
    >>>>But... doesn't every Opteron has a Hypertransport bus, and the SMP
    >>>>models have two? I'm looking at a spec sheet which seems to indicate that.
    >>>
    >>>
    >>>Opterons have three HT links. SMP versions have three "coherent" HT
    >>>busses. Non-SMP versions have three non-coherent HT busses. It's not
    >>>clear what the difference is. It seems it's a product differentiation
    >>>issue.
    >>>
    >>
    >>The 1xx have three non-coherent.
    >>The 2xx have one coherent, two non-coherent.
    >>The 8xx have three coherent.
    >
    >
    > Thank you. T'was late. The above is obvious when you note the
    > segmentation. The real question remaining is what's the difference?

    The coherent can be used for anything - including inter-cpu
    links. The non-c can by used for almost anything /except/
    inter-cpu links.

    I heard a rather vague explanation from an AMD rep at an
    Opteron demo a couple of months before the official Opty
    release. He used the words synchronous and asynchronous
    almost interchangebly with coherent and non-coherent - while
    a colleague of his kept interrupting with "that's too much
    of an oversimplification".

    He also said there is some binning involved. Chips that
    test OK are sold as 8xx, others as 2xx and the bottom
    of the bin as 1xx and Athlon FX.

    > Is
    > it a paper fifference, intentional crippling, or perhaps testing, or (oh,
    > my!) a real difference between the links?

    See above comment about binning. And I'm sure that if
    they have a surplus that test OK as 8xx but need more
    2xx, then they will cripple as necessary.


    Aside:

    I asked the AMD rep what else other than inter-cpu links
    would need a coherent HT link and the only example he could
    come up with was that off-chip memory controllers would
    theoretically be much better with a coherent HT link.
  26. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    On Wed, 01 Dec 2004 05:16:59 +0000, Rob Stow wrote:

    > keith wrote:
    >> On Tue, 30 Nov 2004 05:12:06 +0000, Rob Stow wrote:
    >>
    >>
    >>>keith wrote:
    >>>
    >>>>On Tue, 30 Nov 2004 03:31:19 +0000, Bill Davidsen wrote:
    >>>>
    >>>>
    >>>>
    >>>>>ykhan wrote:
    >>>>>
    >>>>>
    >>>>>>Rob Stow <rob.stow.nospam@shaw.ca> wrote in message news:<pd2pd.313315$Pl.122190@pd7tw1no>...
    >>>>>>
    >>>>>>
    >>>>>>
    >>>>>>>nobody@nowhere.net wrote:
    >>>>>>>
    >>>>>>>
    >>>>>>>
    >>>>>>>>Opteron 150 (essentially all the same as FX-55,
    >>>>>>>
    >>>>>>>Not quite. It is the same speed as the FX-53
    >>>>>>>and hence 200 MHz slower than the FX-55.
    >>>>>>
    >>>>>>
    >>>>>>And probably doesn't have access to the 1000Mhz Hypertransport bus of
    >>>>>>the Athlon 64 or FX either. Don't know how much difference that would
    >>>>>>actually make though, since the memory bus is completely separate from
    >>>>>>the Hypertransport bus in any case.
    >>>>>
    >>>>>But... doesn't every Opteron has a Hypertransport bus, and the SMP
    >>>>>models have two? I'm looking at a spec sheet which seems to indicate that.
    >>>>
    >>>>
    >>>>Opterons have three HT links. SMP versions have three "coherent" HT
    >>>>busses. Non-SMP versions have three non-coherent HT busses. It's not
    >>>>clear what the difference is. It seems it's a product differentiation
    >>>>issue.
    >>>>
    >>>
    >>>The 1xx have three non-coherent.
    >>>The 2xx have one coherent, two non-coherent.
    >>>The 8xx have three coherent.
    >>
    >>
    >> Thank you. T'was late. The above is obvious when you note the
    >> segmentation. The real question remaining is what's the difference?
    >
    > The coherent can be used for anything - including inter-cpu
    > links. The non-c can by used for almost anything /except/
    > inter-cpu links.

    Obviously, according to the press.

    > I heard a rather vague explanation from an AMD rep at an
    > Opteron demo a couple of months before the official Opty
    > release. He used the words synchronous and asynchronous
    > almost interchangebly with coherent and non-coherent - while
    > a colleague of his kept interrupting with "that's too much
    > of an oversimplification".

    I want to get the "vague" part out of the discussion. I'm rather
    interested in the details, though somehow I think my curiosity won't be
    satisfied. ...much the same as I can't/don't say anything about what
    we're doing.

    > He also said there is some binning involved. Chips that test OK are
    > sold as 8xx, others as 2xx and the bottom of the bin as 1xx and Athlon
    > FX.

    That's natural, but I don't see a huge test fallout on the coherent logic
    (whatever it is). More likely it's simply not tested to save a few dimes
    at final test. BTDT.

    > > Is
    >> it a paper fifference, intentional crippling, or perhaps testing, or
    >> (oh, my!) a real difference between the links?
    >
    > See above comment about binning. And I'm sure that if they have a
    > surplus that test OK as 8xx but need more 2xx, then they will cripple as
    > necessary.

    I'm more interested in the *difference* between the links, rather than
    their business plan. Sure, anything pre-destined to be a 1xx won't have
    thecoherency tested (whatever that means). Making my 144 into an 844
    really wasn't my interest here though. I guess what I'm asking for is
    a 40k' view of "coherent HT".

    --
    Keith


    >
    > Aside:
    >
    > I asked the AMD rep what else other than inter-cpu links would need a
    > coherent HT link and the only example he could come up with was that
    > off-chip memory controllers would theoretically be much better with a
    > coherent HT link.
  27. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    On Wed, 01 Dec 2004 03:07:07 GMT, Bill Davidsen
    <davidsen@darkstar.prodigy.com> wrote:

    >Alex Johnson wrote:
    ....snip...
    >
    >I suspect that Intel could drop the price of the EE if they chose to do
    >so, making them more cost effective. I'm not against the Dothan or any
    >other chip if it fits a need.

    Not on this one - it's nothing more (or less) than a repackaged Xeon
    MP with 2MB L3 cache, and these Xeons cost an arm and a leg.
    Understandably so, because all the extra cache eats up shitloads of
    die space. Intel can afford to sell a few below the cost (still above
    and beyond the budget of most normal folks) for PR/benchmarketing
    purposes, but not the many needed to move the P4 Extremely Expensive
    into mainstream.
  28. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    In comp.sys.intel keith <krw@att.bizzzz> wrote:
    > (whatever it is). More likely it's simply not tested to save a few dimes
    > at final test. BTDT.

    The links are all the same. They can all carry coherent traffic,
    even the 1xx. (except ofcourse there's no such traffic on 1xx
    processors)

    The difference is in an internal register that recognizes other
    processor (ID)'s that a given processor has to be coherent to.

    The 1xx won't talk to anyone else, so there's no coherency
    traffic. The 2xx will talk to one other processor, and the
    8xx will talk to 7 others.

    The HT links are generic, it's the processor that will or
    will not generate the traffic.

    The distinction makes it possible to design a 2P Opteron box with
    two 2xx processors that have 2 HT links hooked up to each other.
    Both HT links can carry coherent traffic, but limited to the same guy.

    --
    davewang202(at)yahoo(dot)com
  29. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    nobody@nowhere.net wrote:
    > On Wed, 01 Dec 2004 03:07:07 GMT, Bill Davidsen
    > <davidsen@darkstar.prodigy.com> wrote:
    >
    >
    >>Alex Johnson wrote:
    >
    > ...snip...
    >
    >>I suspect that Intel could drop the price of the EE if they chose to do
    >>so, making them more cost effective. I'm not against the Dothan or any
    >>other chip if it fits a need.
    >
    >
    > Not on this one - it's nothing more (or less) than a repackaged Xeon
    > MP with 2MB L3 cache, and these Xeons cost an arm and a leg.

    No, they SELL for an arm and a leg, it's less documented that the actual
    cost of production. It depends on your belief that it actually cost 3x
    more than the P4 at the same clock speed. NOTE: I don't say you're
    wrong, other than assuming that the selling price reflects the cost of
    production.

    > Understandably so, because all the extra cache eats up shitloads of
    > die space. Intel can afford to sell a few below the cost (still above
    > and beyond the budget of most normal folks) for PR/benchmarketing
    > purposes, but not the many needed to move the P4 Extremely Expensive
    > into mainstream.
    >
    Pass due to lack of data.

    --
    bill davidsen (davidsen@darkstar.prodigy.com)
    SBC/Prodigy Yorktown Heights NY data center
    Project Leader, USENET news
    http://newsgroups.news.prodigy.com
  30. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    keith wrote:
    > On Wed, 01 Dec 2004 05:16:59 +0000, Rob Stow wrote:
    >
    >
    >>keith wrote:
    >>
    >>>On Tue, 30 Nov 2004 05:12:06 +0000, Rob Stow wrote:
    >>>
    >>>
    >>>
    >>>>keith wrote:
    >>>>
    >>>>
    >>>>>On Tue, 30 Nov 2004 03:31:19 +0000, Bill Davidsen wrote:
    >>>>>
    >>>>>
    >>>>>
    >>>>>
    >>>>>>ykhan wrote:
    >>>>>>
    >>>>>>
    >>>>>>
    >>>>>>>Rob Stow <rob.stow.nospam@shaw.ca> wrote in message news:<pd2pd.313315$Pl.122190@pd7tw1no>...
    >>>>>>>
    >>>>>>>
    >>>>>>>
    >>>>>>>
    >>>>>>>>nobody@nowhere.net wrote:
    >>>>>>>>
    >>>>>>>>
    >>>>>>>>
    >>>>>>>>
    >>>>>>>>>Opteron 150 (essentially all the same as FX-55,
    >>>>>>>>
    >>>>>>>>Not quite. It is the same speed as the FX-53
    >>>>>>>>and hence 200 MHz slower than the FX-55.
    >>>>>>>
    >>>>>>>
    >>>>>>>And probably doesn't have access to the 1000Mhz Hypertransport bus of
    >>>>>>>the Athlon 64 or FX either. Don't know how much difference that would
    >>>>>>>actually make though, since the memory bus is completely separate from
    >>>>>>>the Hypertransport bus in any case.
    >>>>>>
    >>>>>>But... doesn't every Opteron has a Hypertransport bus, and the SMP
    >>>>>>models have two? I'm looking at a spec sheet which seems to indicate that.
    >>>>>
    >>>>>
    >>>>>Opterons have three HT links. SMP versions have three "coherent" HT
    >>>>>busses. Non-SMP versions have three non-coherent HT busses. It's not
    >>>>>clear what the difference is. It seems it's a product differentiation
    >>>>>issue.
    >>>>>
    >>>>
    >>>>The 1xx have three non-coherent.
    >>>>The 2xx have one coherent, two non-coherent.
    >>>>The 8xx have three coherent.
    >>>
    >>>
    >>>Thank you. T'was late. The above is obvious when you note the
    >>>segmentation. The real question remaining is what's the difference?
    >>
    >>The coherent can be used for anything - including inter-cpu
    >>links. The non-c can by used for almost anything /except/
    >>inter-cpu links.
    >
    >
    > Obviously, according to the press.
    >
    >
    >>I heard a rather vague explanation from an AMD rep at an
    >>Opteron demo a couple of months before the official Opty
    >>release. He used the words synchronous and asynchronous
    >>almost interchangebly with coherent and non-coherent - while
    >>a colleague of his kept interrupting with "that's too much
    >>of an oversimplification".
    >
    >
    > I want to get the "vague" part out of the discussion. I'm rather
    > interested in the details, though somehow I think my curiosity won't be
    > satisfied. ...much the same as I can't/don't say anything about what
    > we're doing.
    >

    My curiousity isn't satisfied either on this issue
    and there isn't much I can do about the "vague part".
    I got the impression I was talking to a couple of
    junior engineers who were trying to regurgitate an
    explanation that they themselves didn't understand.

    Maybe this will help you out:
    http://www.hypertransport.org/docs/spec/HTC20031217-0036-0005.pdf

    >
    >>He also said there is some binning involved. Chips that test OK are
    >>sold as 8xx, others as 2xx and the bottom of the bin as 1xx and Athlon
    >>FX.
    >
    >
    > That's natural, but I don't see a huge test fallout on the coherent logic
    > (whatever it is). More likely it's simply not tested to save a few dimes
    > at final test. BTDT.
    >
    >
    >> > Is
    >>
    >>>it a paper fifference, intentional crippling, or perhaps testing, or
    >>>(oh, my!) a real difference between the links?
    >>
    >>See above comment about binning. And I'm sure that if they have a
    >>surplus that test OK as 8xx but need more 2xx, then they will cripple as
    >>necessary.
    >
    >
    > I'm more interested in the *difference* between the links, rather than
    > their business plan. Sure, anything pre-destined to be a 1xx won't have
    > thecoherency tested (whatever that means). Making my 144 into an 844
    > really wasn't my interest here though. I guess what I'm asking for is
    > a 40k' view of "coherent HT".
    >
  31. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    On Thu, 02 Dec 2004 00:47:28 GMT, "nobody@nowhere.net"
    <mygarbage2000@hotmail.com> wrote:

    >On Wed, 01 Dec 2004 03:07:07 GMT, Bill Davidsen
    ><davidsen@darkstar.prodigy.com> wrote:
    >
    >>Alex Johnson wrote:
    >...snip...
    >>
    >>I suspect that Intel could drop the price of the EE if they chose to do
    >>so, making them more cost effective. I'm not against the Dothan or any
    >>other chip if it fits a need.
    >
    >Not on this one - it's nothing more (or less) than a repackaged Xeon
    >MP with 2MB L3 cache, and these Xeons cost an arm and a leg.

    The price that the chips SELL for has almost nothing to do with the
    amount that they COST to produce.

    >Understandably so, because all the extra cache eats up shitloads of
    >die space.

    The die size of the 2MB L3 Xeon/P4EE build on a 130nm process is
    237mm^2. Large but not astronomical. This is only about 20% larger
    than AMD's Opteron build on a 130nm process, which they list for as
    little as $163 (Opteron 140) and sell for even less.

    I'd be VERY surprised if it cost Intel much more than $75 to produce a
    Xeon processor. Last estimates I heard had their P4 "Prescott" chips
    costing roughly $25 to $35 a chip. Even with a 200% markup in price,
    Intel could certainly sell these chips for $200 to $250, and Intel's
    standard markup on chips is only 60%.

    > Intel can afford to sell a few below the cost (still above
    >and beyond the budget of most normal folks) for PR/benchmarketing
    >purposes, but not the many needed to move the P4 Extremely Expensive
    >into mainstream.

    The reason why the P4 EE is so expensive is because Intel is selling
    it for that much, not because of any cost of production.

    -------------
    Tony Hill
    hilla <underscore> 20 <at> yahoo <dot> ca
  32. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    nobody@nowhere.net wrote:
    > On Wed, 01 Dec 2004 03:07:07 GMT, Bill Davidsen
    > <davidsen@darkstar.prodigy.com> wrote:
    >
    >
    >>Alex Johnson wrote:
    >
    > ...snip...
    >
    >>I suspect that Intel could drop the price of the EE if they chose to do
    >>so, making them more cost effective. I'm not against the Dothan or any
    >>other chip if it fits a need.
    >
    >
    > Not on this one - it's nothing more (or less) than a repackaged Xeon
    > MP with 2MB L3 cache, and these Xeons cost an arm and a leg.
    > Understandably so, because all the extra cache eats up shitloads of
    > die space. Intel can afford to sell a few below the cost (still above
    > and beyond the budget of most normal folks) for PR/benchmarketing
    > purposes, but not the many needed to move the P4 Extremely Expensive
    > into mainstream.
    >

    Please get your attributions right! I did not say this. I would never
    say this.

    Alex
    --
    My words are my own. They represent no other; they belong to no other.
    Don't read anything into them or you may be required to compensate me
    for violation of copyright. (I do not speak for my employer.)
  33. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    On Thu, 02 Dec 2004 08:24:07 -0500, Tony Hill <hilla_nospam_20@yahoo.ca>
    wrote:

    >On Thu, 02 Dec 2004 00:47:28 GMT, "nobody@nowhere.net"
    ><mygarbage2000@hotmail.com> wrote:
    >
    >>On Wed, 01 Dec 2004 03:07:07 GMT, Bill Davidsen
    >><davidsen@darkstar.prodigy.com> wrote:
    >>
    >>>Alex Johnson wrote:
    >>...snip...
    >>>
    >>>I suspect that Intel could drop the price of the EE if they chose to do
    >>>so, making them more cost effective. I'm not against the Dothan or any
    >>>other chip if it fits a need.
    >>
    >>Not on this one - it's nothing more (or less) than a repackaged Xeon
    >>MP with 2MB L3 cache, and these Xeons cost an arm and a leg.
    >
    >The price that the chips SELL for has almost nothing to do with the
    >amount that they COST to produce.
    >
    >>Understandably so, because all the extra cache eats up shitloads of
    >>die space.
    >
    >The die size of the 2MB L3 Xeon/P4EE build on a 130nm process is
    >237mm^2. Large but not astronomical. This is only about 20% larger
    >than AMD's Opteron build on a 130nm process, which they list for as
    >little as $163 (Opteron 140) and sell for even less.
    >
    >I'd be VERY surprised if it cost Intel much more than $75 to produce a
    >Xeon processor. Last estimates I heard had their P4 "Prescott" chips
    >costing roughly $25 to $35 a chip. Even with a 200% markup in price,
    >Intel could certainly sell these chips for $200 to $250, and Intel's
    >standard markup on chips is only 60%.

    I don't see the economics here as that simple even in terms of just the
    marginal cost of production... without even considering fixed costs. E.g.
    if Intel wanted to produce nothing but Xeons with 2MB of L3 cache, could
    they?... and at what real marginal cost? Is there any estimate available
    of current yields on Xeons/P4EEs and what binning/reassignment arrangements
    are in place?

    Put another way, are they max'd out on production of Xeons/P4EEs? If not,
    why not?:-)

    Rgds, George Macdonald

    "Just because they're paranoid doesn't mean you're not psychotic" - Who, me??
  34. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    On Thu, 02 Dec 2004 08:53:49 -0500, Alex Johnson <compuwiz@jhu.edu> wrote:

    >nobody@nowhere.net wrote:
    >> On Wed, 01 Dec 2004 03:07:07 GMT, Bill Davidsen
    >> <davidsen@darkstar.prodigy.com> wrote:
    >>
    >>
    >>>Alex Johnson wrote:
    >>
    >> ...snip...
    >>
    >>>I suspect that Intel could drop the price of the EE if they chose to do
    >>>so, making them more cost effective. I'm not against the Dothan or any
    >>>other chip if it fits a need.
    >>
    >>
    >> Not on this one - it's nothing more (or less) than a repackaged Xeon
    >> MP with 2MB L3 cache, and these Xeons cost an arm and a leg.
    >> Understandably so, because all the extra cache eats up shitloads of
    >> die space. Intel can afford to sell a few below the cost (still above
    >> and beyond the budget of most normal folks) for PR/benchmarketing
    >> purposes, but not the many needed to move the P4 Extremely Expensive
    >> into mainstream.
    >>
    >
    >Please get your attributions right! I did not say this. I would never
    >say this.

    The way I read his post he was cleary answering and quoting a post from
    Bill Davidsen, which was posted in reply to a post of yours... which was
    snipped. I don't see your problem here.

    Rgds, George Macdonald

    "Just because they're paranoid doesn't mean you're not psychotic" - Who, me??
  35. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    On Thu, 02 Dec 2004 14:21:21 -0500, George Macdonald wrote:

    > On Thu, 02 Dec 2004 08:24:07 -0500, Tony Hill <hilla_nospam_20@yahoo.ca>
    > wrote:
    >
    >>On Thu, 02 Dec 2004 00:47:28 GMT, "nobody@nowhere.net"
    >><mygarbage2000@hotmail.com> wrote:
    >>
    >>>On Wed, 01 Dec 2004 03:07:07 GMT, Bill Davidsen
    >>><davidsen@darkstar.prodigy.com> wrote:
    >>>
    >>>>Alex Johnson wrote:
    >>>...snip...
    >>>>
    >>>>I suspect that Intel could drop the price of the EE if they chose to do
    >>>>so, making them more cost effective. I'm not against the Dothan or any
    >>>>other chip if it fits a need.
    >>>
    >>>Not on this one - it's nothing more (or less) than a repackaged Xeon
    >>>MP with 2MB L3 cache, and these Xeons cost an arm and a leg.
    >>
    >>The price that the chips SELL for has almost nothing to do with the
    >>amount that they COST to produce.
    >>
    >>>Understandably so, because all the extra cache eats up shitloads of
    >>>die space.
    >>
    >>The die size of the 2MB L3 Xeon/P4EE build on a 130nm process is
    >>237mm^2. Large but not astronomical. This is only about 20% larger
    >>than AMD's Opteron build on a 130nm process, which they list for as
    >>little as $163 (Opteron 140) and sell for even less.
    >>
    >>I'd be VERY surprised if it cost Intel much more than $75 to produce a
    >>Xeon processor. Last estimates I heard had their P4 "Prescott" chips
    >>costing roughly $25 to $35 a chip. Even with a 200% markup in price,
    >>Intel could certainly sell these chips for $200 to $250, and Intel's
    >>standard markup on chips is only 60%.
    >
    > I don't see the economics here as that simple even in terms of just the
    > marginal cost of production... without even considering fixed costs.

    Of course not. As Tony has indicated, price has *nothing* to do with
    cost.

    > E.g.
    > if Intel wanted to produce nothing but Xeons with 2MB of L3 cache, could
    > they?...

    Sure, and why not?

    > and at what real marginal cost?

    Irrelevant. Price is greazter than cost and that's what matters here.

    > Is there any estimate
    > available of current yields on Xeons/P4EEs and what binning/reassignment
    > arrangements are in place?
    >
    > Put another way, are they max'd out on production of Xeons/P4EEs? If
    > not, why not?:-)

    Look at it realistically. If they sold *only* Xeon/P4EEs, how many would
    they have to sell at $100 to maintain market share. How many $1000 Xeons
    would they *not* sell because the Xeons cost $100.

    THink not cost, but market differentiation. That's exactly where Intel is
    between a rock and a hard place now. AMD has figured out how to
    differentiate effectively, while kicking Intel's ass in all segments.
    Intel's iceberg lookout is still asleep.

    --
    Keith
  36. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    On Thu, 02 Dec 2004 04:19:57 +0000, David Wang wrote:

    > In comp.sys.intel keith <krw@att.bizzzz> wrote:
    >> (whatever it is). More likely it's simply not tested to save a few dimes
    >> at final test. BTDT.
    >
    > The links are all the same. They can all carry coherent traffic,
    > even the 1xx. (except ofcourse there's no such traffic on 1xx
    > processors)

    Obvioulsly. "Why", or perhaps "why not" was the question (I've read ahead ;-).

    > The difference is in an internal register that recognizes other
    > processor (ID)'s that a given processor has to be coherent to.

    Thank you! That ws the information I was missing. Indeed it's an
    ingenious way of making things (not) work.

    > The 1xx won't talk to anyone else, so there's no coherency traffic. The
    > 2xx will talk to one other processor, and the 8xx will talk to 7 others.

    I see that now.

    > The HT links are generic, it's the processor that will or
    > will not generate the traffic.
    >
    > The distinction makes it possible to design a 2P Opteron box with
    > two 2xx processors that have 2 HT links hooked up to each other.
    > Both HT links can carry coherent traffic, but limited to the same guy.

    Those AMD guys are pretty sharp. ;-)

    Thanks again for making it this clear. The bottom line is that the 1xx is
    "crippled".

    --
    Keith
  37. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    George Macdonald wrote:
    >>>On Wed, 01 Dec 2004 03:07:07 GMT, Bill Davidsen
    >>><davidsen@darkstar.prodigy.com> wrote:
    >>>
    >>>
    >>>
    >>>>Alex Johnson wrote:
    >>>
    >>>...snip...
    >>>
    >>>
    >>>>I suspect that Intel could drop the price of the EE if they chose to do
    >>>>so, making them more cost effective. I'm not against the Dothan or any
    >>>>other chip if it fits a need.

    > The way I read his post he was cleary answering and quoting a post from
    > Bill Davidsen, which was posted in reply to a post of yours... which was
    > snipped. I don't see your problem here.

    The snippage and the inclusion of my name directly above the remaining
    text implies directly that I said the quoted material. If every word I
    said was cut out, then my name should not be attached to any remaining
    quoted text as it was clearly not me the quote is from.

    Now the snip-quote has been copied in half a dozen replies and people
    are reading my name and someone else's quote. Perhaps soon people will
    come into this thread and say "This Alex character is a moron. Look at
    the rediculous thing he said here." If I'm to be considered a moron, I
    demand that it be because of the merits of moronic things I've
    /actually/ said! :)

    Alex
    --
    My words are my own. They represent no other; they belong to no other.
    Don't read anything into them or you may be required to compensate me
    for violation of copyright. (I do not speak for my employer.)
  38. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    On Thu, 02 Dec 2004 14:21:21 -0500, George Macdonald
    <fammacd=!SPAM^nothanks@tellurian.com> wrote:

    >I don't see the economics here as that simple even in terms of just the
    >marginal cost of production... without even considering fixed costs. E.g.
    >if Intel wanted to produce nothing but Xeons with 2MB of L3 cache, could
    >they?... and at what real marginal cost?

    Intel has multiple generations of production technology running at any
    given time, so potential max production and costs would depend on the
    wafer starts available for a given chip's line. On a given technology
    line (say, 90 nm), the cost per die of a Celeron vs a Xeon is mostly
    dependent on die size (which also affects yield from both defects and
    edge exclusion). Older technologies are more heavily amortized and
    optimized, with more emphasis on cost reduction, but can't produce the
    latest and greatest.

    There's some dependence on number of layers, litho steps, and such as
    well, but the capital costs and manpower are roughly the same either
    way, though consumables costs will vary.

    The problem for any manufacturer is forecasting demand while meeting
    contractual commitments, since there's always a lag between starting
    the raw materials and getting the finished product out (maybe 3 months
    for CPUs - the holiday production is already being packaged and
    shipped). Even though the margin on Celerons may be lower, customers
    want them, so they have to be produced. So yes, they could produce
    all Xeon, but it would be a marketing disaster.

    >Is there any estimate available
    >of current yields on Xeons/P4EEs and what binning/reassignment arrangements
    >are in place?

    Not publicly available ones... It's tracked very closely internally,
    of course, and a lot of resources go into maximizing yield and
    binsplits, even when demand is down.

    >Put another way, are they max'd out on production of Xeons/P4EEs? If not,
    >why not?:-)

    Probably because the customer demand isn't high enough to divert
    resources from more mainstream chips to these. If the high-margin
    product demand outstrips supply, internal resources are reallocated to
    balance that with the contractual commitments for other chips, but
    alienating the high-volume customers to serve the high-margin
    customers would probably start a negative spiral. It's a bit of a
    tightrope act when demand is high.


    --
    Neil Maxwell - I don't speak for my employer
  39. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    On Thu, 02 Dec 2004 14:21:21 -0500, George Macdonald
    <fammacd=!SPAM^nothanks@tellurian.com> wrote:

    >On Thu, 02 Dec 2004 08:24:07 -0500, Tony Hill <hilla_nospam_20@yahoo.ca>
    >wrote:
    >
    >>I'd be VERY surprised if it cost Intel much more than $75 to produce a
    >>Xeon processor. Last estimates I heard had their P4 "Prescott" chips
    >>costing roughly $25 to $35 a chip. Even with a 200% markup in price,
    >>Intel could certainly sell these chips for $200 to $250, and Intel's
    >>standard markup on chips is only 60%.
    >
    >I don't see the economics here as that simple even in terms of just the
    >marginal cost of production... without even considering fixed costs. E.g.
    >if Intel wanted to produce nothing but Xeons with 2MB of L3 cache, could
    >they?...

    They probably could, but no they couldn't sell them. While a $200 or
    $250 P4EE might sound like a pretty good deal to most of us in this
    newsgroup, the bulk of computer buyers aren't going to pay that when
    they could get a Celeron for $75.

    > and at what real marginal cost? Is there any estimate available
    >of current yields on Xeons/P4EEs and what binning/reassignment arrangements
    >are in place?

    Intel is notoriously secretive about such things, so any "estimates"
    would be VERY rough ones to say the least. That being said, if
    there's one thing that Intel knows it is how to make a microprocessor,
    even ones with a HUGE amount of cache (case-in-point, the Itaniums
    with up to 9MB of cache). I doubt that there having any serious
    issues with building Xeons or P4EE chips on their, now very mature,
    130nm process.

    Their new 90nm process might be a different story, but if you have a
    little look here:

    http://processorfinder.intel.com/scripts/details.asp?sSpec=SL7RR&ProcFam=483&PkgType=ALL&SysBusSpd=ALL&CorSpd=ALL


    You'll see that Intel already has a 2MB L2 cache P4 built on their
    90nm process. This seems like it was a bit of a stealth-launch, since
    I haven't heard any announcement or seen any reviews (let alone
    product availability), but they've got it listed.

    >Put another way, are they max'd out on production of Xeons/P4EEs? If not,
    >why not?:-)

    Marketing considerations primarily. If they wanted to sell LOTS of
    P4EE chips they would need to lower the price. If they lower the
    price they would have to lower the price of their mid-grade chips
    down. If they lower the mid-grade, then they need to drop out the
    price on low-end chips where the bulk of sales occur. No matter how
    good of a chip Intel sells at $250, still the majority of people are
    likely to buy whatever chip is being sold at $75, regardless of
    performance differences. If Intel tried to ignore that, AMD would eat
    their lunch.

    -------------
    Tony Hill
    hilla <underscore> 20 <at> yahoo <dot> ca
  40. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    On Thu, 02 Dec 2004 21:45:22 -0500, keith <krw@att.bizzzz> wrote:

    >On Thu, 02 Dec 2004 14:21:21 -0500, George Macdonald wrote:
    >
    >> On Thu, 02 Dec 2004 08:24:07 -0500, Tony Hill <hilla_nospam_20@yahoo.ca>
    >> wrote:

    <snip>>

    >>>I'd be VERY surprised if it cost Intel much more than $75 to produce a
    >>>Xeon processor. Last estimates I heard had their P4 "Prescott" chips
    >>>costing roughly $25 to $35 a chip. Even with a 200% markup in price,
    >>>Intel could certainly sell these chips for $200 to $250, and Intel's
    >>>standard markup on chips is only 60%.
    >>
    >> I don't see the economics here as that simple even in terms of just the
    >> marginal cost of production... without even considering fixed costs.
    >
    >Of course not. As Tony has indicated, price has *nothing* to do with
    >cost.

    I'm not even looking at price. According to Tony's numbers Intel should be
    selling high-end(?) Prescotts for $35.+60%. Is that really what Dell pays?

    >> E.g.
    >> if Intel wanted to produce nothing but Xeons with 2MB of L3 cache, could
    >> they?...
    >
    >Sure, and why not?
    >
    >> and at what real marginal cost?
    >
    >Irrelevant. Price is greazter than cost and that's what matters here.

    Price?? We seem to be talking about different things here. I've no idea
    what their manufacturing arrangements are for Xeons but it would be
    interesting to know what the ratio of duds (not mark-downs) to good ones is
    compared to P4s??

    Of course we are only talking about stamping the widgets out after the
    cooky-cutter has been made.

    >> Is there any estimate
    >> available of current yields on Xeons/P4EEs and what binning/reassignment
    >> arrangements are in place?
    >>
    >> Put another way, are they max'd out on production of Xeons/P4EEs? If
    >> not, why not?:-)
    >
    >Look at it realistically. If they sold *only* Xeon/P4EEs, how many would
    >they have to sell at $100 to maintain market share. How many $1000 Xeons
    >would they *not* sell because the Xeons cost $100.

    Price?... cost?<shrug>

    Rgds, George Macdonald

    "Just because they're paranoid doesn't mean you're not psychotic" - Who, me??
  41. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    On Thu, 2 Dec 2004 04:19:57 +0000 (UTC), David Wang <foo@bar.invalid>
    wrote:

    >In comp.sys.intel keith <krw@att.bizzzz> wrote:
    >> (whatever it is). More likely it's simply not tested to save a few dimes
    >> at final test. BTDT.
    >
    >The links are all the same. They can all carry coherent traffic,
    >even the 1xx. (except ofcourse there's no such traffic on 1xx
    >processors)
    >
    >The difference is in an internal register that recognizes other
    >processor (ID)'s that a given processor has to be coherent to.
    >
    >The 1xx won't talk to anyone else, so there's no coherency
    >traffic. The 2xx will talk to one other processor, and the
    >8xx will talk to 7 others.
    >
    >The HT links are generic, it's the processor that will or
    >will not generate the traffic.
    >
    >The distinction makes it possible to design a 2P Opteron box with
    >two 2xx processors that have 2 HT links hooked up to each other.
    >Both HT links can carry coherent traffic, but limited to the same guy.

    Any idea when this register is programmed... between wafer and packaged
    chip?

    Rgds, George Macdonald

    "Just because they're paranoid doesn't mean you're not psychotic" - Who, me??
  42. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    On Fri, 03 Dec 2004 10:18:37 -0500, Tony Hill <hilla_nospam_20@yahoo.ca>
    wrote:

    >On Thu, 02 Dec 2004 14:21:21 -0500, George Macdonald
    ><fammacd=!SPAM^nothanks@tellurian.com> wrote:
    >
    >>On Thu, 02 Dec 2004 08:24:07 -0500, Tony Hill <hilla_nospam_20@yahoo.ca>
    >>wrote:
    >>
    >>>I'd be VERY surprised if it cost Intel much more than $75 to produce a
    >>>Xeon processor. Last estimates I heard had their P4 "Prescott" chips
    >>>costing roughly $25 to $35 a chip. Even with a 200% markup in price,
    >>>Intel could certainly sell these chips for $200 to $250, and Intel's
    >>>standard markup on chips is only 60%.
    >>
    >>I don't see the economics here as that simple even in terms of just the
    >>marginal cost of production... without even considering fixed costs. E.g.
    >>if Intel wanted to produce nothing but Xeons with 2MB of L3 cache, could
    >>they?...
    >
    >They probably could, but no they couldn't sell them.

    I wasn't thinking of all perfect top-speed Xeons. When you can't make many
    of a particular "perfect" item, you charge more for it... in the ideal
    model of course... and in any business endeavor. I'm just trying to get a
    grasp on how far that model is strayed from here.

    For the P4EEs, the high clock rates must limit yield but failure to qualify
    there only involves binning to a lower speed grade. Even for the Xeon, the
    P4 FSB is a hairy piece of work; I can see where guaranteeing MP will work
    with QDR data, dynamic bus inversions etc. and DDR address lines on a
    multi-drop bus *could* limit volume and there's less scope for marking down
    here. Of couse, like you I'm speculating - that's the easy bit.:-)

    I'm not saying you're necessarily wrong here but there *is* room for
    "discussion" given how closely guarded the truth is.

    Rgds, George Macdonald

    "Just because they're paranoid doesn't mean you're not psychotic" - Who, me??
  43. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    On Fri, 03 Dec 2004 08:02:12 -0500, Alex Johnson <compuwiz@jhu.edu> wrote:

    >George Macdonald wrote:
    >>>>On Wed, 01 Dec 2004 03:07:07 GMT, Bill Davidsen
    >>>><davidsen@darkstar.prodigy.com> wrote:
    >>>>
    >>>>
    >>>>
    >>>>>Alex Johnson wrote:
    >>>>
    >>>>...snip...
    >>>>
    >>>>
    >>>>>I suspect that Intel could drop the price of the EE if they chose to do
    >>>>>so, making them more cost effective. I'm not against the Dothan or any
    >>>>>other chip if it fits a need.
    >
    >> The way I read his post he was cleary answering and quoting a post from
    >> Bill Davidsen, which was posted in reply to a post of yours... which was
    >> snipped. I don't see your problem here.
    >
    >The snippage and the inclusion of my name directly above the remaining
    >text implies directly that I said the quoted material. If every word I
    >said was cut out, then my name should not be attached to any remaining
    >quoted text as it was clearly not me the quote is from.

    On a first level quote, I always assume that the name at the top is the one
    associated with the quoted text.

    >Now the snip-quote has been copied in half a dozen replies and people
    >are reading my name and someone else's quote. Perhaps soon people will
    >come into this thread and say "This Alex character is a moron. Look at
    >the rediculous thing he said here." If I'm to be considered a moron, I
    >demand that it be because of the merits of moronic things I've
    >/actually/ said! :)

    Sorry but I didn't see it that way at all. I'm sure my own snippage is
    often a bit err, sloppy - much easier to have a reader which has a limit on
    level of quotes of course... and I'd think there would be very few who
    would consider you a moron.:-)

    Rgds, George Macdonald

    "Just because they're paranoid doesn't mean you're not psychotic" - Who, me??
  44. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    On Fri, 03 Dec 2004 15:13:14 -0500, George Macdonald
    <fammacd=!SPAM^nothanks@tellurian.com> wrote:

    >On Thu, 02 Dec 2004 21:45:22 -0500, keith <krw@att.bizzzz> wrote:
    >
    >>On Thu, 02 Dec 2004 14:21:21 -0500, George Macdonald wrote:
    >>
    >>> On Thu, 02 Dec 2004 08:24:07 -0500, Tony Hill <hilla_nospam_20@yahoo.ca>
    >>> wrote:
    >
    ><snip>>
    >
    >>>>I'd be VERY surprised if it cost Intel much more than $75 to produce a
    >>>>Xeon processor. Last estimates I heard had their P4 "Prescott" chips
    >>>>costing roughly $25 to $35 a chip. Even with a 200% markup in price,
    >>>>Intel could certainly sell these chips for $200 to $250, and Intel's
    >>>>standard markup on chips is only 60%.
    >>>
    >>> I don't see the economics here as that simple even in terms of just the
    >>> marginal cost of production... without even considering fixed costs.
    >>
    >>Of course not. As Tony has indicated, price has *nothing* to do with
    >>cost.
    >
    >I'm not even looking at price. According to Tony's numbers Intel should be
    >selling high-end(?) Prescotts for $35.+60%. Is that really what Dell pays?

    They probably don't pay much more than that for the 2.8A and 2.93GHz
    Prescott chips that make up a lot of their low-end systems being sold
    these days. Consider that they sell complete systems using said
    processors for only $500, including a monitor! Factor in the cost
    that Dell pays for the OS, the hard disk, motherboard, memory, etc.
    etc., plus all of the associated marketing and support costs and they
    obviously are not paying the $168 that Newegg sells this processor
    for.

    Note though that the numbers above are a rough guesstimate on price of
    production. Intel also has some other costs that they factor in to
    things along the way and are generally fairly secretive about such
    things, so it's tough to get a firm grip on real costs.

    -------------
    Tony Hill
    hilla <underscore> 20 <at> yahoo <dot> ca
  45. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    On Thu, 02 Dec 2004 04:29:49 GMT, Bill Davidsen
    <davidsen@darkstar.prodigy.com> wrote:

    >nobody@nowhere.net wrote:
    >> On Wed, 01 Dec 2004 03:07:07 GMT, Bill Davidsen
    >> <davidsen@darkstar.prodigy.com> wrote:
    >>
    >>
    >>>Alex Johnson wrote:
    >>
    >> ...snip...
    >>
    >>>I suspect that Intel could drop the price of the EE if they chose to do
    >>>so, making them more cost effective. I'm not against the Dothan or any
    >>>other chip if it fits a need.
    >>
    >>
    >> Not on this one - it's nothing more (or less) than a repackaged Xeon
    >> MP with 2MB L3 cache, and these Xeons cost an arm and a leg.
    >
    >No, they SELL for an arm and a leg, it's less documented that the actual
    >cost of production. It depends on your belief that it actually cost 3x
    >more than the P4 at the same clock speed. NOTE: I don't say you're
    >wrong, other than assuming that the selling price reflects the cost of
    >production.
    >
    >> Understandably so, because all the extra cache eats up shitloads of
    >> die space. Intel can afford to sell a few below the cost (still above
    >> and beyond the budget of most normal folks) for PR/benchmarketing
    >> purposes, but not the many needed to move the P4 Extremely Expensive
    >> into mainstream.
    >>
    >Pass due to lack of data.
    Well, let's see...
    Fact: Intel is losing market share to AMD, including the server
    market - probably the most lucrative and thus important.

    Fact: Intel's top brass is fuming mad because of it.

    And now question: if Xeon MP is not too expensive in manufacturing,
    then why Intel would not flood the market with 2MB L3 cache parts,
    undercutting Opterons 1xx and 2xx pricewise while providing similar
    performance?

    The possible answers, aside from legal/marketing/other non-technical
    reasons (no hard data, just speculations):
    1. The yields on 2 MB cache part are not so great, therefore
    availability is limited. While the quantities are sufficient for the
    very limited 4P and above market, it's not nearly enough to supply
    2P/1P market.
    2. Mainly because of (1), but also due to some other reasons
    (costlier testing etc) the cost is too high to drop the price much
    lower than it is.
    3. Heat dissipation issues (I have no data handy, but suspect all
    those extra transistors just MUST produce heat) prevent 2 MB parts
    from being used in 1U (pizza box) servers - the main 2P/1P server
    application.
    4. All of the above, and then some ;-)
  46. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    On Sun, 05 Dec 2004 01:56:16 +0000, nobody@nowhere.net wrote:

    > On Thu, 02 Dec 2004 04:29:49 GMT, Bill Davidsen
    > <davidsen@darkstar.prodigy.com> wrote:
    >
    >>nobody@nowhere.net wrote:
    >>> On Wed, 01 Dec 2004 03:07:07 GMT, Bill Davidsen
    >>> <davidsen@darkstar.prodigy.com> wrote:
    >>>
    >>>
    >>>>Alex Johnson wrote:
    >>>
    >>> ...snip...
    >>>
    >>>>I suspect that Intel could drop the price of the EE if they chose to do
    >>>>so, making them more cost effective. I'm not against the Dothan or any
    >>>>other chip if it fits a need.
    >>>
    >>>
    >>> Not on this one - it's nothing more (or less) than a repackaged Xeon
    >>> MP with 2MB L3 cache, and these Xeons cost an arm and a leg.
    >>
    >>No, they SELL for an arm and a leg, it's less documented that the actual
    >>cost of production. It depends on your belief that it actually cost 3x
    >>more than the P4 at the same clock speed. NOTE: I don't say you're
    >>wrong, other than assuming that the selling price reflects the cost of
    >>production.
    >>
    >>> Understandably so, because all the extra cache eats up shitloads of
    >>> die space. Intel can afford to sell a few below the cost (still above
    >>> and beyond the budget of most normal folks) for PR/benchmarketing
    >>> purposes, but not the many needed to move the P4 Extremely Expensive
    >>> into mainstream.
    >>>
    >>Pass due to lack of data.
    > Well, let's see...
    > Fact: Intel is losing market share to AMD, including the server
    > market - probably the most lucrative and thus important.
    >
    > Fact: Intel's top brass is fuming mad because of it.
    >
    > And now question: if Xeon MP is not too expensive in manufacturing,
    > then why Intel would not flood the market with 2MB L3 cache parts,
    > undercutting Opterons 1xx and 2xx pricewise while providing similar
    > performance?
    >
    > The possible answers, aside from legal/marketing/other non-technical
    > reasons (no hard data, just speculations):
    > 1. The yields on 2 MB cache part are not so great, therefore
    > availability is limited. While the quantities are sufficient for the
    > very limited 4P and above market, it's not nearly enough to supply
    > 2P/1P market.
    > 2. Mainly because of (1), but also due to some other reasons
    > (costlier testing etc) the cost is too high to drop the price much
    > lower than it is.
    > 3. Heat dissipation issues (I have no data handy, but suspect all
    > those extra transistors just MUST produce heat) prevent 2 MB parts
    > from being used in 1U (pizza box) servers - the main 2P/1P server
    > application.
    > 4. All of the above, and then some ;-)

    5. Because flooding the desktop market with "server" chips would crash
    their price structure (a.k.a. market segmentation) causing INTC stock to
    go into a tailspin. One thing INTC isn't is *stoopid!

    --
    Keith
  47. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    On Sun, 05 Dec 2004 01:56:16 GMT, "nobody@nowhere.net"
    <mygarbage2000@hotmail.com> wrote:

    >Well, let's see...
    >Fact: Intel is losing market share to AMD, including the server
    >market - probably the most lucrative and thus important.
    >
    >Fact: Intel's top brass is fuming mad because of it.

    Keep in mind that Intel still has roughly 95% of the server market
    share. AMD might have eaten into it a little bit, but they're still a
    pretty small player.

    >And now question: if Xeon MP is not too expensive in manufacturing,
    >then why Intel would not flood the market with 2MB L3 cache parts,
    >undercutting Opterons 1xx and 2xx pricewise while providing similar
    >performance?
    >
    >The possible answers, aside from legal/marketing/other non-technical
    >reasons (no hard data, just speculations):

    There may be no hard data, but those marketing reasons are exactly the
    ones that you need to be looking at here. It's not a technical issue.

    >1. The yields on 2 MB cache part are not so great, therefore
    >availability is limited. While the quantities are sufficient for the
    >very limited 4P and above market, it's not nearly enough to supply
    >2P/1P market.

    Uhh.. are you ignoring the fact that Intel DOES sell a 2MB cache part
    specifically for the 1P/2P market?

    >2. Mainly because of (1), but also due to some other reasons
    >(costlier testing etc) the cost is too high to drop the price much
    >lower than it is.

    The extra cache doesn't add any significant cost to testing.

    >3. Heat dissipation issues (I have no data handy, but suspect all
    >those extra transistors just MUST produce heat) prevent 2 MB parts
    >from being used in 1U (pizza box) servers - the main 2P/1P server
    >application.

    The 3.4GHz P4 built on a 130nm process had a TDP of 89W while the
    3.4GHz P4EE built on the same 130nm process has a TDP of 109.6W. So,
    if we are to assume no other effects, that cache is consuming an extra
    20W of power in these chips. Drop down to the 3.2GHz/533MT/s bus
    speed Xeon with 2MB of L2 cache and you've got a TDP of 93W.

    Is this a problem? Obviously not since Dell will sell you a 1U, 2
    processor server using a pair of 3.2GHz Xeons with 2MB of L2 cache.
    This is one of the configuration options for their PowerEdge 1750.

    >4. All of the above, and then some ;-)

    As mentioned above, you can't ignore the marketing side of things
    because that's where the real answer lies.

    -------------
    Tony Hill
    hilla <underscore> 20 <at> yahoo <dot> ca
  48. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    George Macdonald wrote:
    > Sorry but I didn't see it that way at all. I'm sure my own snippage is
    > often a bit err, sloppy - much easier to have a reader which has a limit on
    > level of quotes of course... and I'd think there would be very few who
    > would consider you a moron.:-)

    "Level of quotes". Made me think. I went back and took a look at what
    was displayed in the viewer and what the source text looked like and
    what do you know. Mozilla's news reader displays different depth
    information. While the quote and my name are at different levels of >,
    Mozilla 1.4.2 (damn, I was on 1.7 before the last reboot...time to hunt
    down duplicate copies) shows them both after the same shaded sidebar |.

    Oh, good :-) I've been trying to stay out of the flamewars and just
    answer questions that have simple, non-opinion answers lately. My
    opinion in the past was not popular around here. At least some people
    noticed. :-)

    Alex
    --
    My words are my own. They represent no other; they belong to no other.
    Don't read anything into them or you may be required to compensate me
    for violation of copyright. (I do not speak for my employer.)
  49. Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

    On Mon, 06 Dec 2004 08:22:31 -0500, Alex Johnson <compuwiz@jhu.edu> wrote:

    >George Macdonald wrote:
    >> Sorry but I didn't see it that way at all. I'm sure my own snippage is
    >> often a bit err, sloppy - much easier to have a reader which has a limit on
    >> level of quotes of course... and I'd think there would be very few who
    >> would consider you a moron.:-)
    >
    >"Level of quotes". Made me think. I went back and took a look at what
    >was displayed in the viewer and what the source text looked like and
    >what do you know. Mozilla's news reader displays different depth
    >information. While the quote and my name are at different levels of >,
    >Mozilla 1.4.2 (damn, I was on 1.7 before the last reboot...time to hunt
    >down duplicate copies) shows them both after the same shaded sidebar |.
    >
    >Oh, good :-) I've been trying to stay out of the flamewars and just
    >answer questions that have simple, non-opinion answers lately. My
    >opinion in the past was not popular around here. At least some people
    >noticed. :-)

    I always figure it's not a "popularity contest".:-) If you have a
    different opinion from me, I still want to hear it and hopefully maybe be
    able to discuss it civilly... which IIRC has always been the case. There
    are always the people who stalk with flame throwers at the ready but we
    mustn't let them rule.

    Rgds, George Macdonald

    "Just because they're paranoid doesn't mean you're not psychotic" - Who, me??
Ask a new question

Read More

CPUs Pentium Gaming