Sign in with
Sign up | Sign in
Your question

THE INTEL-AMD PROCESSOR RACE REVS UP

Last response: in CPUs
Share
March 22, 2006 5:44:55 PM

Quote:
There will always be a need for mainstream CPUs that meet the challenge of “damn the kilowatts, full speed ahead!” AMD’s Opteron, Athlon FX, Turion, and Athlon X2 CPUs own the high-performance mantle; Intel will not recover it, and it knows that. Intel is playing on the fact that the coming round of high-performance server and desktop AMD64 processors, due midyear and dubbed Revision F, look like monsters compared with Intel’s tiny Core. AMD64 Revision F uses a bigger die and therefore a bigger socket. Rev F CPUs will consume more power.


I've said a while ago that AMD is NOT worried about Intel's CORE architecture. 8)

http://www.infoworld.com/article/06/03/22/76583_13OPcurve_1.html
March 22, 2006 5:56:10 PM

Quote:
There will always be a need for mainstream CPUs that meet the challenge of “damn the kilowatts, full speed ahead!” AMD’s Opteron, Athlon FX, Turion, and Athlon X2 CPUs own the high-performance mantle; Intel will not recover it, and it knows that. Intel is playing on the fact that the coming round of high-performance server and desktop AMD64 processors, due midyear and dubbed Revision F, look like monsters compared with Intel’s tiny Core. AMD64 Revision F uses a bigger die and therefore a bigger socket. Rev F CPUs will consume more power.


I've said a while ago that AMD is NOT worried about Intel's CORE architecture. 8)

http://www.infoworld.com/article/06/03/22/76583_13OPcurve_1.html

I gotta say something that hopefully people read....WHO CARES ABOUT TDP IN A DESKTOP!!!! Jesus, if you're going to buy these high-end AM2 Chips, or Socket F chips, you already have either 2 PSU's or 1 High-End PSU that's 600w+, so what does a TDP of 85 or 125 vs. 65 matter if you're already pullin 600w+ @ full load? I'll tell you...NOTHING! When it comes to mobile, great, give me 20w TDP, yes! But on a desktop, every website and poster is saying "Conroe and Core have less TDP, so it pwns!" well, if AMD can make me an FX-70 @ 3.6GHz w/ a TDP of 200w, I'll bite it, because I know any Conroe can't match it.

@Topic: That revision will be insane for performance, as well as the ability for power changes. Turion 64 currently is as low as 25w TDP, and Merom is 35w for standard and 5w for Ultra Low Voltage, and no way does a new revision raise TDP enough to matter. It's going to be interesting to see Revision F performance/power, should be nice.

~~Mad Mod Mike, pimpin' the world 1 rig at a time
March 22, 2006 6:59:36 PM

Using Conroe as an example, I think raw performance advantage combined with lower power equals a cooler fast chip that may indeed have alot of untapped overclocking headroom that stirred the excitement, not just the lower TDP itself.
Related resources
March 22, 2006 7:02:06 PM

Quote:
Using Conroe as an example, I think raw performance advantage combined with lower power equals a cooler fast chip that may indeed have alot of untapped overclocking headroom that stirred the excitement, not just the lower TDP itself.


True True, but your overclocking headroom won't be temperature related as much as it will be keeping it cool. Obviously the internal temps will sky-rocket, but what's the point in overclocking, say, a 4GHz Socket AM2? Even a Conroe can't touch that beast. You'll always have those extreme people out there, and I of course love OCing to the max, but I'd rather have a 4GHz at max than having to OC a 2.6 to 4GHz w/o paying $$$ for the type of cooling + energy bill.

~~Mad Mod Mike, pimpin' the world 1 rig at a time
March 22, 2006 7:06:28 PM

Quote:
gotta say something that hopefully people read....WHO CARES ABOUT TDP IN A DESKTOP!!!! Jesus, if you're going to buy these high-end AM2 Chips, or Socket F chips, you already have either 2 PSU's or 1 High-End PSU that's 600w+, so what does a TDP of 85 or 125 vs. 65 matter if you're already pullin 600w+ @ full load? I'll tell you...NOTHING! When it comes to mobile, great, give me 20w TDP, yes! But on a desktop, every website and poster is saying "Conroe and Core have less TDP, so it pwns!" well, if AMD can make me an FX-70 @ 3.6GHz w/ a TDP of 200w, I'll bite it, because I know any Conroe can't match it.

@Topic: That revision will be insane for performance, as well as the ability for power changes. Turion 64 currently is as low as 25w TDP, and Merom is 35w for standard and 5w for Ultra Low Voltage, and no way does a new revision raise TDP enough to matter. It's going to be interesting to see Revision F performance/power, should be nice.


You got that right MadMike. :wink:
I still have the confidence that Rev F chips will run cooler thanks to FD-SSOI.
March 22, 2006 7:38:19 PM

Isnt TDP the reason why you praise and use AMD?

Funny how its totally no big deal now that AMD wants to try "performance at the cost of power consumption".
But AMD will miserably fail at that just how they failed to beat the struggling NetBurst and make people think 64-bit makes your family-planner bigger.

Just on a side note:
AMD will meet the requirements to beat Woodcrest with 65nm or even 45nm production and thats 1.5-2.5 years away. I dont know how those laser-brains at that website make up their stories.
March 22, 2006 7:44:44 PM

Quote:
Isnt TDP the reason why you praise and use AMD?

Funny how its totally no big deal now that AMD wants to try "performance at the cost of power consumption".
But AMD will miserably fail at that just how they failed to beat the struggling NetBurst and make people think 64-bit makes your family-planner bigger.

Just on a side note:
AMD will meet the requirements to beat Woodcrest with 65nm or even 45nm production and thats 1.5-2.5 years away. I dont know how those laser-brains at that website make up their stories.


Can't beat Netburst? Are you a blind moron? Wow...even CompGeek isn't THAT stupid, sheesh.

64-bit is the future, and Intel gets whooped like a government mule in that area. I did say the Heat on AMD is a good reason why AMD is better, but since they also OUTPERFORM, I just as much praise that as the former.

~~Mad Mod Mike, pimpin' the world 1 rig at a time
March 23, 2006 10:48:36 PM

Yeah, I do find it rather funny how the tables have turned here. For the longest time, Intel was flamed for their high power usage. Now, it's no big deal since AMD will sacrifice power for performance.

It does nothing for credibility on the AMD side of things and has really turned into nothing but a soap opera.
March 23, 2006 10:53:43 PM

They are worried that you haven't gotten a new keyboard.
March 23, 2006 11:23:05 PM

My next case will have 2x 121mm fat exaust pipes coming out the top
a c 102 à CPUs
March 24, 2006 12:09:28 AM

You do have a point, but I think it is more a point of the extent that TDP matters or not and why it matters. Desktop users have large power supplies that can just get bigger if they need to- up to about 1500-2000W, which is when you start to need a dedicated circuit in your house to run the machine. That is the upper ceiling for power consumption and we're nowhere near it.

Why it matters somewhat is that power == heat and heat takes noisy fans to get rid of. We all gripe about a CPU that idles at 55 C and brag about how another idles at room temperature plus a degree or two. The lower TDP chips should run cooler and thus be more overclockable and require less for noisy fans to keep quiet. If they developed a CPU that drew 200W at idle but idled at 20 C, I bet few of us would care about TDP. I do hear a few VIA EPIA owners gripe about how their 3-7W CPUs run blisteringly hot.
March 24, 2006 12:17:37 AM

Quote:
You do have a point, but I think it is more a point of the extent that TDP matters or not and why it matters. Desktop users have large power supplies that can just get bigger if they need to- up to about 1500-2000W, which is when you start to need a dedicated circuit in your house to run the machine. That is the upper ceiling for power consumption and we're nowhere near it.

Why it matters somewhat is that power == heat and heat takes noisy fans to get rid of. We all gripe about a CPU that idles at 55 C and brag about how another idles at room temperature plus a degree or two. The lower TDP chips should run cooler and thus be more overclockable and require less for noisy fans to keep quiet. If they developed a CPU that drew 200W at idle but idled at 20 C, I bet few of us would care about TDP. I do hear a few VIA EPIA owners gripe about how their 3-7W CPUs run blisteringly hot.


Yea, I enjoyed hearing a broadcast over the Internet a while ago of an interview with some Systems Managers who maintained the Servers for companies. I remember them talking in the kW's like it was nothing, lol. I found it funny that people ask if a 550w PSU is enough and you enter the server world and you're talking kW's, heh.

Tru64, more heat needs more loud fans to dissipate, but that's why we made Fan Controllers and EIST/CNQ ;) . I can attest to a noise issue, having 4 Vantec Tornado 92mm 55dBa running @ max is 3 Sonne+ (I believe that's the word, maybe spelled wrong).

~~Mad Mod Mike, pimpin' the world 1 rig at a time
March 24, 2006 2:02:04 AM

Quote:
They are worried that you haven't gotten a new keyboard.


ok whats with the keyboard thing???

On subject: If AMD raises the TDP, wouldn't that offset some heat by going to a smaller nm (65) or (45)for the chip design?
Or is it that with the added (metal) their building into the design, is to handle more power through the chip to remain stable at a higher freq.

I'm lost here people, please don't flame, just trying to understand what the possible design could do in socket " F "
a c 102 à CPUs
March 24, 2006 2:05:39 AM

MMM was complaining about his current keyboard sucking. He said he'd like a keyboard that you could glide your hands over with a short key travel (like a laptop's.) I suggested he mod a laptop keyboard into a desktop-capable one and he said he'd get a laptop-like desktop board from Newegg.

It all started with an impossible pimping challenge- look back in this forum to see it.
March 24, 2006 2:18:35 AM

Quote:
MMM was complaining about his current keyboard sucking. He said he'd like a keyboard that you could glide your hands over with a short key travel (like a laptop's.) I suggested he mod a laptop keyboard into a desktop-capable one and he said he'd get a laptop-like desktop board from Newegg.

It all started with an impossible pimping challenge- look back in this forum to see it.


That's one, but I think Action_Man is referring to 9-Inch who constantly puts his subjects in CAPS.

EDIT: That isn't impossible, just need a MASSIVE OC to accomplish...if I felt like ruining a Dual AMD Opteron 64 board, I'd give it a try ;) .

~~Mad Mod Mike, pimpin' the world 1 rig at a time
March 24, 2006 2:32:48 AM

ok got it

look at my prior post please as I did an edit!
March 24, 2006 6:27:32 AM

http://www.infoworld.com/article/06/03/22/76583_13OPcur...

There are a lot of philosophical speculations without any kind of usefull information(like a specifications or something that can prove a little the child stories) that can be found on internet.
There is a content missing in the article, and the good(empty) words about AMD are meaning nothing! Its like a reading childish NewYears wishes.

@jumpingjack
Transition from 90nm SOI to 65nm SOI is benefiting like the transition from 90nm nonSOI to 65nm nonSOI. It will reduce TDP so more cores could be added as more effective than gaining the clock-speed to boost performance.

@MadModMike(with AMD in your heart)
Even a FX-whatever clocked at 4GHz will not beat Conroe EE(comming around September) 3.33GHz FSB1333(DAMN i would like to see DDR2 clocked 1333 and runing 1:1 to CPU). But 4GHz K8 are SF for the 90nm and for the future 65nm also.
The K8L needs a reconstruction in its architecture, optimization for high-latency memory, more processing units and better instructions set in order to keep up with allready revealed and introduced Core architecture.
March 24, 2006 8:01:35 AM

According to available info, the highest clocked conroe on release will be 2.66ghz. Are you trying to tell us that Intel will put out an EE that is more than 25% faster than it's next fastest chip?
Quick calc, yup, you dont have a clue.
March 24, 2006 8:35:24 AM

This is one and half months old article:
http://www.tgdaily.com/2006/02/11/intel_conroe_to_launc...

"Intel will be offering six versions of the Conroe processor at launch:
E6700: 2.66 GHz / FSB 1066/ 4 MB shared L2 cache
E6600: 2.40 GHz / FSB 1066/ 4 MB shared L2 cache
E6400: 2.13 GHz / FSB 1066/ 2 MB shared L2 cache
E6300: 1.86 GHz / FSB 1066/ 2 MB shared L2 cache
E4200: 1.60 GHz / FSB 800/ 2 MB shared L2 cache
Conroe Extreme Edition (XE): Specifications unknown"

The unknown specifications are known:
http://www.xbitlabs.com/news/cpu/display/20060212234350...

this might help also http://www.vr-zone.com/?i=3215
March 24, 2006 9:04:43 AM

Quote:

Tru64, more heat needs more loud fans to dissipate, but that's why we made Fan Controllers and EIST/CNQ ;) . I can attest to a noise issue, having 4 Vantec Tornado 92mm 55dBa running @ max is 3 Sonne+ (I believe that's the word, maybe spelled wrong).


4 of them would make that one loud system, at 55 dB it is already above safe for 1 hour usage OH&S guidelines.

If that is 55 dBa each, then it becomes 64 dB, and that is very loud for a PC, louder than some rackmount servers. (Which normally you sit nowhere near).

If I was LANing near someone with a PC that loud, I'd take noise measurements then advice of potential legal action (OH&S) if they didn't shut it down, or make it quiet (and if that means they need to underclock, that is their problem for building such a frankenstien PC). Liability (even at large LANs, where management become liable at some levels) these days is everything and similar cases have already been won.

2007 is going to be year of the quiet PC, as OH&S is pushing for far quieter office machines. (When you sit near 32+ PCs the noise adds up).

Quote:
According to available info, the highest clocked conroe on release will be 2.66ghz. Are you trying to tell us that Intel will put out an EE that is more than 25% faster than it's next fastest chip?
Quick calc, yup, you dont have a clue.


Intel can clock the Extreme version at 3.4x GHz if they want, as the current TDP of Conroe is so low as is they have +50% TDP headroom, which translates into (SqRt 1.538) about +24% clockspeed headroom, the power usage will rise (still within acceptable levels though), and clock speed could be 25-33% higher easily. So long as the surface area to contact ratio is good, or better than it is now, they'll overclock very well up to 100w TDP. A +33% increase in clock speed will only 'cost' +77% more power (on a given die size) which is sitting pretty at 116 watts (from 65 watts if I recalled correctly).

We all know Intel like to clock things either +25% or +33% higher for the 'hardcore' varients.

Surface area to contact ratio of the processor / heatsink is expected to be good, considering it is going to have 2 x 4-issue cores on it, and a heap of cache.

It is like Intel removed 'Coke Original' from the market with the 2nd half of the Pentium 4s, and now they are re-releasing what people crave as 'Coke Classic'. I suspect they planned this all along just to see if their marketing efforts alone would cause them to lose market share, and while they did, it wasn't that much of a loss to them, if they kept doing it for another 2 years though.... :roll:

================================================

PS: 200 watts TDP would cause a CPU to burst / reach flashpoint. Sure it would raise performance (SqRt 2.00 = 200%) by +41% or so over 100 watts TDP models (3.92 GHz or so), but it would raise 'absolute' temperatures to extreme levels. (eg: Is 50 C +25% hotter than 40 C, what if you convert them both to Fahrenheit from Celsius ?)

Your lightbulb is only 1% efficient at giving off light, the rest of the energy it heat, and it gets so hot it glows. You do not want a 200 watt TDP processor. (Unless there is some way to control, and transfer all of that heat at very high speeds to something it won't destory).

We already know the Conroe EE/XE is going to outperform 3.92 GHz clocked dual-core K8_ CPUs anyway.
March 24, 2006 11:40:35 AM

Nice article 9-inch, thanks for posting it.


Revision F > Conroe

-or-

Conroe > Revision F


Is anyone else excited?
March 24, 2006 1:11:33 PM

...off topic...
Quote:
Is anyone else excited?

I'm so excited and I just cant hide it....come on, come on...
March 24, 2006 3:31:21 PM

Quote:

I'm lost here people, please don't flame, just trying to understand what the possible design could do in socket " F "


I've got news on Socket F indicating it will be using FB-DIMMs, but the interface chip (similar to Registered DRAM) on FB-DIMMs can be used to connect to DDR2, XDR RAMBUS, DDR3 (possibly), and other changed memory technologies.

Socket F could support 4 cores per processor (at 65nm), and possibly 6-8 cores per processor (at 45nm) if they choose to stick with it.

The problem they want to fix is only evident on the newer desktop machines now:


[Other fast dedicated links for general system I/O and video]
|
CPU <=[Dedicated Memory Ctrlr]============> RAM
[ ----------------- Same Protocol ---------------------]

The problem with that design is each time the memory protocols change, they need to change the most expensive part in most systems.

So they are looking at moving towards something like this:

[Other fast dedicated links for general system I/O and video]
|
CPU <=[Dedicated FB-DIMM Interface Ctrlr]====> FB-DIMM interface per DIMM <===> Actual DRAMs
[ ---------------------- Same Protocol -------------------------] (it not wrapped above)


The advantage of the above is, if the memory type changes, the protocol used from CPU to the interface on the DIMM doesn't need to change, but they still get high speed connections, with far more future upgrade options available.

Downside is having, what some may call part of the North-Bridge, on the DIMMs themselves will raise the cost of 'Fully Buffered-[Insert RAM chips of choice here]'. However decent Registered DIMMs (with 128x4 internal layout, Advanced ECC) cost twice per MB when compared to consumer memory anyway, so it isn't really that major an issue. Power consuption of memory will rise, at least initially, but that normally isn't an issue for 'overall / total system power usage' anyway.

It also means they'll be able to scale to 64 GB in budget servers, and 128-192 GB in higher end servers within 5 years. Enough memory for 10 applications to leak memory for a month each and delay reboots. (Like the F-22 Raptor can delay stalls - Is it really a good thing to do it ?, let alone mention it :lol:  ). :p 
March 24, 2006 4:12:16 PM

Socket F will not be the answer to Conroe. Socket F is more or less the 754 of the 64bit world... its the first inital step into the world of DDRII and what looks like future memory types also the addition of a L3 cache to make up for the poor latencies of DDRII.

AMDs answer will come when they intro there new core at 65nm. AMD doesn't have the manufacturing capabilities of Intel, but they have better designers. So they've partnered with IBM to help shrink to 65nm, 45nm, and 32nm, along with SSOI and some other processes. Not to mention there going to implement a co-processor on the core die... what that co-processor will do, i don't know, but i'd like to see it effectively replace the Northbridge on the motherboard.

AMD will have an answer to Conroe. The battle for next-gen CPU's is just getting started. Intel has won round 1, but not the fight.
March 24, 2006 4:20:08 PM

lol,

amd in the LEAD

and intel seems to run faster though, amd has still 6 months to accelerate too

who will win the race ???

depending on where is the finish line
March 24, 2006 4:52:44 PM

I think you mean Socket S1 (638 pins) is basically the 754 of the next platforms. (Useful in HTPCs and the like I'd imagine).

Socket AM2 (formerly M2) is the Socket 939 (UnBuffered) of the next platforms.

Socket F (1207) is the 'new' Socket 940 (Reg) for the new Opterons.

All 3 platforms will have various AMD processors become available at 65nm, with Dual-Channel DDR2 support, and possibly 45nm (and other memory technologies for Socket F, as above) within their lifespan.
March 24, 2006 5:00:54 PM

yea my bad, i was thinking about socket 754 as i was writing and meant to say Revision F... thank you for pointing this out.

And what i meant by Revision F being the 754 of the 64bit world, is that 754 was the AMDs first attempt at a 64bit processor designed for mainstream use. Just like Revision F will be AMDs first attempt at DDRII memory. Just a metaphor didn't mean to make it come out like its literally a breakthough, cause its obviously not.

once again, my bad
March 24, 2006 5:29:57 PM

Oh yeah, I get what you're saying now. (Replaces word 'Socket' with 'Revision' when re-reading).

Hopefully they'll do 100 internal revisions to get the max benefit they can when released.

Prefetching data / instructions from DDR2 (higher latency not cancelled out by higher clock speeds quite yet) requires a more advanced scheduler, so they need to do alot more work than just the memory controller.

It makes sense (now).

As for the co-processor, it is likely to have native support for x87 microcode, although software may need to be especially compiled to it, Hopefully it'll be like SIMD but able to pass to 96 element www.ClearSpeed.com designs. eg: SIMD or MIMD and able to process arrays far, far larger than 4 values, say 256 - 512 value arrays, or 16 different arrays of 16.

Physics may benefit from PPU trickery, and even offloading some parts to the GPU, or combinations thereof, but adding MIMD (vs SIMD) feature sets, and being able to process datasets 64 times the size within one or two processor generations would give physics in games a 'huge' jump. (Even larger than the PPU on the market now, using parts from ClearSpeed that have been available for quite awhile now on PCI-X 64 bit long slots).
March 25, 2006 2:18:52 AM

Jeez you people are mountain of info!

I think I got enough homework for the weekend for chip design..etc... 8O

Thank you, I have learned alot and value your inputs

@jumpingjack: thanks to you as well sir....should have wrote a book and had it published.. :wink:

My hobby is anything with computers...etc....technonut if you will. I still love aircraft though, as my job is aircraft design and engineering safety.
One of the best moments I had with my job was when I was working with engineers on the F-15, The test pilot gave me a nice unforgetable ride in one....yes I hurled my guts out but loved it... :twisted: The one thing I learned during the flight is that the G-suit needs to be a nice tight fit in that man hood area...if not the twins get squeezed... 8O 8O 8O

anyway thanks for the info
!