Sign in with
Sign up | Sign in
Your question

THG Prescott Review is online! ~1% slower than NW!

Tags:
Last response: in CPUs
Share
February 1, 2004 4:34:01 PM

<A HREF="http://www.tomshardware.com/cpu/20040201/index.html" target="_new">Link! It's real!</A>

<i>Edit: this was about the german version, which was out some hours before the english one. The above link is now in english.</i>

It would seem that Prescott is roughly on par with northwood... actually, it is a northwood match.

But no better, no worse. Intel has a good card in their hands, if they can get SSE3 to good use and increase clock and enable 64 bit technology, and a bad one if not. Prescott is nothing stellar at this point, it's not above northwood... <b>Northwood beats prescott in 1/3 of benchmarks, and prescott beats northwood in 2/3 of benchmarks,</b> and there is no completely clear winner... When Northwood is faster than prescott, it's by a somewhat bigger margin, but it doesn't happen as often.

:evil:  <font color=red><b>M</b></font color=red>ephistopheles
February 1, 2004 4:47:24 PM

Performance is well bellow expectations... Alas I hope all it requires is a recompileing of the software.

-taitertot

If this post has attitude, seems to be overly aggressive, rude, distasteful to 99% of the users here, and shows a zealous defense of Intel... It’s probably Spud.
February 1, 2004 4:51:42 PM

Well, it's still new technology - SSE3, I mean.

It's below expecations, but at least it's not below Northwood.

I hope it turns out to be better when it matures.

<b>31 stage-pipeline in prescott is true, BTW.</b>

:evil:  <font color=red><b>M</b></font color=red>ephistopheles
Related resources
February 1, 2004 5:02:54 PM

SSE3 isnt much in the end just helps things move along in the weaker areas of the core such as pur x87 and SSE-SSE2 streams, also some integer conversion instructions, and some more pack and stack instructions. Oh and I guess 1 instruction to synchronize threading on the Hyper Threading chips.

Perhaps new compiling will help since most code is designed around the idea of smaller trace l1 data and instruction, and l2 data caches.

-taitertot

If this post has attitude, seems to be overly aggressive, rude, distasteful to 99% of the users here, and shows a zealous defense of Intel... It’s probably Spud.
February 1, 2004 5:09:40 PM

Well, here's the comparison: 3.2C vs 3.2E:

<font color=red>Quake 3 - Northwood still 3% faster.</font color=red>
<font color=green>SPECviewperf 7.1.1, drv-09 - Prescott 13% faster.
SPECviewperf 7.1.1, dx-08 - Prescott 5% faster.
SPECviewperf 7.1.1, light-06 - Prescott 1.5% faster.
SPECviewperf 7.1.1, proe-02 - Prescott 4% faster.
SPECviewperf 7.1.1, ugs-03 - Prescott 7% faster.</font color=green>
<font color=red>Wolfenstein - Enemy Territory - Northwood still 2% faster.</font color=red>
<font color=red>Comanche 4 - Northwood still 16% faster!</font color=red>
<font color=red>UT2003 - Northwood still 2.5% faster.</font color=red>
Splinter Cell - Both with differences inferior to 1%.
3dMark03 - Both with differences inferior to 1%.
X2 Rolling Demo - Both with differences inferior to 1%.
<font color=red>AquaMark 3 - Northwood still 2% faster.</font color=red>
Mainconcept MPEG Encoder 1.4.1 - Exact match. No difference.
<font color=green>Pinnacle Studio 9 - Prescott 7% faster.
XMPEG 5.0.3 / DivX 5.1.1 Pro - Prescott 6% faster.</font color=green>
Windows Media Encoder 9 - Within 1% of each other.
<font color=red>Windows Movie Maker 2.0 - Northwood still 3% faster.</font color=red>
<font color=green>Steinberg Nuendo - Prescott 5% faster.</font color=green>
<font color=red>Lame 3.95.1 - Northwood still 12% faster!</font color=red>
<font color=green>BAPCo SYSmark 2004 - Prescott 2% faster.
WinRAR 3.20 - Prescott 7% faster.</font color=green>
<font color=red>Newtek Lightwave 7.5c - Northwood still 3% faster.
Cinema 4D XL 8.503 - Northwood still 13.6% faster!
Discreet 3DStudio MAX 6.0 - Northwood still 4% faster.
Wolfram Research Mathematica 5.0 - Northwood still 7% faster.</font color=red>

This is the overall picture... So prescott is not earthshattering, but it's nothing to really, really laugh about. It has weaknesses... If it can, however, be brought to clock rates of 4Ghz and will eventually have activated 64-bit extensions, then it will make a difference for Intel alright. Plus, add another prescott @ 3.46Ghz and 1066Mhz FSB, and nobody will be laughing. So, while it is nothing revolutionary at all, it has its own merits.

:evil:  <font color=red><b>M</b></font color=red>ephistopheles
February 1, 2004 5:11:54 PM

Averageing that shows the Northwood is still faster on the old code. Which is fine and dandy I want to see recompiled code running on the chip before I make a sound verdict on this stepping.

-taitertot

If this post has attitude, seems to be overly aggressive, rude, distasteful to 99% of the users here, and shows a zealous defense of Intel... It’s probably Spud.
February 1, 2004 5:18:08 PM

From these results, you could estimate that Northwood is still 0.2% (I did the math) faster than Prescott. Which is, of course, completely negligible...

So on the average, it can be said that Prescott is equal to Northwood... More or less. If you use lots of apps, on average, you'll get the same performance.

In any case, I think Prescott was built with circuitry in it that will prove it to be more than it's now. It is 99% likely that it has 64-bit circuitry in it...

:evil:  <font color=red><b>M</b></font color=red>ephistopheles
February 1, 2004 5:21:43 PM

True.

-taitertot

If this post has attitude, seems to be overly aggressive, rude, distasteful to 99% of the users here, and shows a zealous defense of Intel... It’s probably Spud.
February 1, 2004 6:26:40 PM

My German's a little rusty, does the review talk about overclockability at all?
February 1, 2004 6:33:19 PM

No, it does not. At all.

:evil:  <font color=red><b>M</b></font color=red>ephistopheles
February 1, 2004 7:10:13 PM

So basically if your a gamer stay away from prescott. Comanche + unreal = shocking performance.

My system spec: Fast PC<A HREF="http://service.futuremark.com/compare?2k1=7000747" target="_new"> 3D-2001 </A>
"It's not the spoon that bends, it's only yourself."
February 1, 2004 7:26:57 PM

I must say I'm disappointed. All the waiting and anxiety and... splat. It's not any better. Sure it will reach big Mhz, but I don't like the whole performance through clockspeed policy.

Really good deal if it really does OC to 4+ Ghz on air, once 400+ memory becomes cheaper.

<b>wooooow <font color=red> Killer Klowns </font color=red> ... from <font color=blue>outer space</font color=blue>... HOLY SH¡T!</b>
February 1, 2004 7:37:38 PM

I thought Toms review said that there is no overclocking included because it wont due to heat issues(?)

I'll sum it up, the review basically shows that Prescott sucks.
And it is a major disappointment.
No reason for a Northwood owner to upgrade to it IMO.
It will be a good processor for people who dont know anything, it fits well in a Dell... and from the review it looks like a fitting "upgrade" for Mike Dell and company. :lol: 
And that leaves AMD as the only real choice for an enthusiast as usual.

I'm glad all that hoop-la over this POS is finally over and we can forget about intel again until the next fanboy rally.

----
Support the NV/AMD/IBM axis of evil.
Who cares about HL2/D3 when we have Call of Duty today!!
February 1, 2004 7:44:25 PM

Prescott is no big deal right now, but it doesn't suck. It's no groundbreaker at start. There is no reason to upgrade to it, and that is right, but this doesn't mean that prescott won't be capable of anything useful and that Intel is doomed.
Quote:
And that leaves AMD as the only real choice for an enthusiast as usual.

Funny, I thought the OCed 2.4C or 2.6C was an excellent enthusiast's choice.

:evil:  <font color=red><b>M</b></font color=red>ephistopheles
February 1, 2004 7:54:23 PM

<b><i>Enthusiasts</i> used the <i>2500+</i></b>
:smile:

I'm just giving you a hard time.
But really, you can find an 'ok' enthusiast chip from intel if you dig hard enough.
I meant for those in the know, I dont see any reason to choose Prescott over an A64.
Apologies on my using the word enthusiast, it might not have been the best choice.
I dont always equate enthusiast=overclocker, though I understand many do.

Intels not doomed... its just silly that it is their response to the A64.
Intel fanboys had <b>better</b> hope that Intel <b>really</b> is "holding back" on AMD... and no one really knows, but in either case (if they are/arent), this is pathetic.

So Prescott doesnt suck. Its just pathetic. :lol: 

----
Support the NV/AMD/IBM axis of evil.
Who cares about HL2/D3 when we have Call of Duty today!!
February 1, 2004 7:57:29 PM

BTW, <A HREF="http://www.anandtech.com/cpu/showdoc.html?i=1956&p=13" target="_new">check out</A> prescott's overclocking potential. This is a very strong indication that Intel can ramp up clock speeds with this processor. AMD cannot do that with its current line so easily at this point.

:evil:  <font color=red><b>M</b></font color=red>ephistopheles
February 1, 2004 8:17:31 PM

AMD doesnt need too as much. Its about IPC, not ramping up clockspeeds with 50 stage pipelines.
Especially when your not wanting a $300+ processor and maybe wanting something mid to low end...
processors with higher IPC (Athlons) just seem much 'snappier'.
You can tell the difference even in windows IMO.

I prefer AMDs chosen path, to pioneer AMD64 for the consumer market and design an overall better rounded processor than intels increased pipelines with marchetechure in mind.

----
Support the NV/AMD/IBM axis of evil.
Who cares about HL2/D3 when we have Call of Duty today!!
February 1, 2004 8:27:26 PM

It's not so much the performance loss, but the <b>heat</b> of the processor! Seriously, it's pushing up towards 103 watts for the 3.2ghz. Just imagine how *bad* it will get at 4ghz. Intel must have to package some serious coolers with their procs sooner than later, or they'll be pushing towards the 70 celsius built in temp limiter.

Wait for Tejas :lol: 

:cool: I run my AthlonXfx at 7.65 Exahertz :cool:
February 1, 2004 8:38:01 PM

It's the same old Intel thing ... or at least, the same old Pentium 4 thing. Make the pipeline longer so you can get more MHz out of it to make up for the IPC deficiencies. I suppose it doesn't matter how you get performance, either through MHz or IPC, but I just don't like Intel's approach. I mean, how long can they make the pipeline to keep increasing the speed? I guess time will tell who's approach is better, but I'm sticking with AMD for now.

<i>Money talks. Mine always likes to say "goodbye." :smile: </i>
February 1, 2004 8:42:24 PM

Quote:
Its about IPC, not ramping up clockspeeds with 50 stage pipelines.

Nope, it never was about IPC. IPC is useless.

It is about PERFORMANCE. Sorry, but this cannot be debated at all. IPC and clock alone don't mean squat.
Quote:
Especially when your not wanting a $300+ processor and maybe wanting something mid to low end...
processors with higher IPC (Athlons) just seem much 'snappier'.

Prescotts are rather agressively priced. And as for what higher IPC seems, (snappier or whatever) I don't care what it seems. I care about how it performs.

This "feeling" that IPC is always the smart choice over clock rates is completely bogus (note: I said feeling, 'cause that's what it is). You'll feel right at home with a high IPC processor? Suit yourself, the pentium M has the highest IPC to date on consumer-level processors. Doesn't mean it is the fastest at all. You don't gave to feel at home with anything, you have to reason your way through this. This isn't a love affair, this is an objective field - hardware, that is.

Personally, I reason that a processor that performs better is snappier, be it from AMD or Intel.

In any case, I think that Prescott has yet to show its potential. Maybe within two months, when Alderwood and Grantsdale debut with DDR2-533 and a possible 1066Mhz, we'll get to know this architecture's capabilities.

Bear in mind that Prescott is still a great overclocker.


:evil:  <font color=red><b>M</b></font color=red>ephistopheles
February 1, 2004 8:48:43 PM

Quote:
I mean, how long can they make the pipeline to keep increasing the speed? I guess time will tell who's approach is better, but I'm sticking with AMD for now.

While I understand your point, I just wanted to point a small thing out here: you're saying that you don't like Intel's approach because it is without a future, right? But, when you buy a processor, beware, you're not buying its future, you're buying that specific CPU. This is exactly why you won't go out and buy a Scotty right now - because it doesn't look good <i>just now</i>. This is exactly because, some 6 months ago, you'd have chosen a processor from the P4C line - because it excelled in performance and was faster than equivalent offerings from AMD. So sticking with AMD has little to do with saying that Intel's approach will turn out to be less effective in the future... CPUs have to be bought on an each-case basis, regardless of the company's policies for the future. So it wouldn't really be in your best interest to stick to a company solely because you believe its goals for the market for the future are more to your liking.

I hope I managed to express myself clearly here... :eek: 

:evil:  <font color=red><b>M</b></font color=red>ephistopheles
February 1, 2004 9:07:34 PM

Quote:
So it wouldn't really be in your best interest to stick to a company solely because you believe its goals for the market for the future are more to your liking.

No, I'm saying I think that AMD took the wiser path than Intel for today AND tomorrow.
That is true now and no longer debatable.

As far as my IPC "feeling", its not a feeling its very noticable. I could benchmark it if you would like, well I wouldnt do that really but I am 100% sure the "feeling" is relating to better PERFORMANCE.

I dont know why you would take intels side though as you are setting yourself up to battle against the hordes.
I dont see much reason to go Intel.

Prescott may not look good right now.. but its a big deal because this is Intels response to the A64, and its a lackluster response.

It appears little AMD is the superior semiconductor manufacturer.

----
Support the NV/AMD/IBM axis of evil.
Who cares about HL2/D3 when we have Call of Duty today!!
February 1, 2004 9:08:09 PM

I don't know what you guys were expecting but don't think Prescott "sucks".

It was never meant to be faster then Northwood clock for clock - it was only supposed to allow faster clock speeds and that's what it does. If intel wanted it to be different they would have called it something different (ie: not a P4 2.8E). It was not meant to be an upgrade (ie: you wouldn't buy a 3.2E if you had a 3.0C). Reviews that up now are getting regular 3.6~3.7 overclocks on air, with some cores apparently even hitting well over 4GHz on air. That's not bad at all for the first public stepping using a brand new manufacturing technique.

<A HREF="http://www.anandtech.com/cpu/showdoc.html?i=1956&p=24" target="_new">Anand</A> points out that
Quote:
Prescott 's enhancements actually give it a steeper increase in performance per increase in clock. Not only can Prescott be clocked higher than Northwood, but as its clock speed is increased, it will start to outperform similarly clocked Northwood CPUs.

The first stepping generally doesn't overclock well. What if the next stepping gives us 4.5 overclocks on air? You never know.

*Dual PIII-800 @900 i440BX and Tualeron 1.2 @1.74 i815*
February 1, 2004 9:18:42 PM

Quote:
I don't know what you guys were expecting but don't think Prescott "sucks


I was expecting more based on all of the "pre-game" hype that I had to listen to from these intel-whores.
Now its game time and Intel didnt bring any game, but they brought their usual peformance. :lol: 

----
Support the NV/AMD/IBM axis of evil.
Who cares about HL2/D3 when we have Call of Duty today!!
February 1, 2004 9:25:31 PM

WOW! just thought of something. The 2.8E should have a 21x multiplier. Get some standard pc3200 DDR. 200*21=4200. I doubt you can get it that high, but OCing to 3.6~4.0 should be no problem judging by the reviews. That would be a really easy, really cheap OC. Interesting!

<b>wooooow <font color=red> Killer Klowns </font color=red> ... from <font color=blue>outer space</font color=blue>... HOLY SH¡T!</b>
February 1, 2004 9:35:16 PM

Ouch... Unfortunately, it would seem that I didn't make my point clear enough at all. You missed it entirely.
Quote:
No, I'm saying I think that AMD took the wiser path than Intel for today AND tomorrow.
That is true now and no longer debatable.

What is no longer debatable is that for now, AMD has the performance lead over Intel.
Quote:
As far as my IPC "feeling", its not a feeling its very noticable. I could benchmark it if you would like, well I wouldnt do that really but I am 100% sure the "feeling" is relating to better PERFORMANCE.

I dont know why you would take intels side though as you are setting yourself up to battle against the hordes.
I dont see much reason to go Intel.

I'm not going against the hordes. It's you who sees everything as either pro-AMD or pro-Intel. I won't even try to explain; I'll just quote myself.
Quote:
This is exactly why you won't go out and buy a Scotty right now - because it doesn't look good just now.

Did you actually read that?

I agree with you!!! Or didn't you notice? Yes, it's obvious that AMD has the upper hand now!!!
Quote:
Prescott may not look good right now.. but its a big deal because this is Intels response to the A64, and its a lackluster response.

It appears little AMD is the superior semiconductor manufacturer.

Indeed! This is not a product good enough, as of now, to compete with A64! AMD DOES HAVE THE UPPER HAND NOW!

What I said is that you won't go and buy a 2500+ for the price of a 2.8Ghz Northwood C just because AMD has the upper hand now in the top-performance segment. Nor do you choose your CPU based on the company's policies! You choose a CPU for what it is!

I hope I made myself a bit more clear.

:evil:  <font color=red><b>M</b></font color=red>ephistopheles
February 1, 2004 9:41:27 PM

But from what ive seen @ stock the prescott runs upto and over 60c on air????

Athlon 64 3200
MSI K8T
9700pro @ 378/338
Antec 1000AMG
February 1, 2004 9:44:50 PM

Quote:
I don't know what you guys were expecting but don't think Prescott "sucks".

It doesn't suck at all. It will be a great processor when it's clocked a bit higher.

The problem is that initial reactions are completely clouded by emotions and expectations. That's what this is.

Can you say that current Athlon cores at 2.2Ghz can be clocked at <b>+25%</b> their clock (meaning, 2.8Ghz)?? Of course you cannot. Yet several sites state that clocking Prescott to 3.8Ghz is easy, and some are even speaking about clocking it at 4.0Ghz.

So going out now and buying a prescott over a northwood might not be a great idea for non-OCers, but don't throw the Scotty concept in the trash just yet, guys. This pipeline lengthening was a carefully measured event...

Anyone ever thought of just how impressive it is to increase the pipeline by 55% and still not lose performance?

:evil:  <font color=red><b>M</b></font color=red>ephistopheles
Anonymous
a b à CPUs
February 1, 2004 10:14:19 PM

Am I really the only one that is *not* surprised ? I told you Intel released the 3.4 Northwood to at least stay close with AMD in the benchmarking parade (especially gaming, where prescott s*ucks), because Prescott would *not* be better. Everything we knew showed this, wether it was leaked benchmarks, Intels extremely low profile attitude on Prescott, or industry rumours.. seriously, prescott outperforming Northood would have been a surprise.

Anyway, now you know.. you could say its roughly on par with northwood, but frankly, where it matters most IMHO, gaming, its even almost a speedgrade slower than Northwood. Have a look at Anands test.

Now all of this wouldnt have to be all that bad, if clock ramps up quickly, but even though Prescott seems to be a good overclocker, I fear the heat output is problematic for anything but enthousiast grade HSF's. most of us don't mind spending $25 on a good cooler, and/or don't mind the noise too much, we often have a huge cases full of case fans with plenty of airflow.. but for OEM kit, which is the bulk of the market, those 100+W power figures are a problem. And will only get worse as clock speed ramp. Some people here may start thinking twice about the benefits of these "all new hyperpipelined netburst" designs, I definately don't think it is the future. Bring on those fast clocked Dothans in desktops intel, skewing the process for clockspeed rather than ultra low leakage and voltage should allow 2.2+ Dothan, and that is the intel chip I would want

= The views stated herein are my personal views, and not necessarily the views of my wife. =
February 1, 2004 10:15:07 PM

Quote:
But from what ive seen @ stock the prescott runs upto and over 60c on air????

And my 486DX-33 ran about 25C without a heatsink... but it doesn't mean it's a better processor.

If the chip doesn't die, and the socket doesn't melt, I don't think it really matters. In five years we'll probably look back and say "wow, ONLY ~100W"

The new intel socket is supposed to bring bigger heatsinks.

*Dual PIII-800 @900 i440BX and Tualeron 1.2 @1.74 i815*
February 1, 2004 10:25:45 PM

Quote:
If the chip doesn't die, and the socket doesn't melt, I don't think it really matters. In five years we'll probably look back and say "wow, ONLY ~100W"

True, absolutely. For all I care, the processor could be churning out an absurd 800W if it was providing excellent performance stably and without the noise of a Boeing jet engine.

:evil:  <font color=red><b>M</b></font color=red>ephistopheles
Anonymous
a b à CPUs
February 1, 2004 10:28:54 PM

>Anyone ever thought of just how impressive it is to
>increase the pipeline by 55% and still not lose
>performance?

Anyone thought about how utterly unimpressive it is to have a new core with 150% (!!!) more transistors, 100% more cache and 50% (!) higher thermal density, and not even outperform the previous core ? I don't care the pipeline was incrased tenfold, it still wouldnt come close to impress me.

So, prescott will probably allow higher clockspeeds.. probably 4 GHz versus 3.4 for northwood.. <b>a whopping 17%</b>. >3% of that increase is already annihilated by the decreased IPC (depending on what benchmarks you find most important), leaving you with a ~13% extra performance headroom.. assuming linear scaling. Anything but impressive in my book. 13% used to be a single speedgrade in the old days, 2 speedgrades at most. A 2.2 GHz Athlon64 +13% doesnt even get you to 2.5 GHz. Looks like AMD can screw up its 90 nm transition and still be able to claim the performance crown for the remaining of this year. (especially with S939, faster HT, and unregistered ram support soon to be released)

But hey, everyone say with me "WAIT FOR TEJAS !"

= The views stated herein are my personal views, and not necessarily the views of my wife. =
Anonymous
a b à CPUs
February 1, 2004 10:32:41 PM

>If the chip doesn't die, and the socket doesn't melt, I
>don't think it really matters

It definately does. do the math on what it costs you over 5 years if you run your computer 24/7.

Or do a guestimate what it costs to make heatsinks, power supplies and motherboards than can cope with 100+W and 100+Ampere. May not matter too much in a $3000 computer, but its definately a big factor in the mainstream oem market where prescott is eventuall headed. It may not be a problem for you, but its definately a problem for intel.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
February 1, 2004 10:36:25 PM

Quote:
(especially with S939, faster HT, and unregistered ram support soon to be released)

BBaeyens with his usual speech...

S939 won't bring any performance benefits to AMD's processors.
Unbuffered RAM will not bring a performance benefit more than a very, very few percentage points.
Increased HT bus is irrelevant. Current Bus is already overkill.

Why do you use your pessimistic, conservative evaluation only on Intel? And I suppose Alderwood, with its 5% boost in performance, won't matter? DDR2? PCI-E? Increase in FSB to match synch with DDR2-533? I think the platform advances that are about to happen to prescott will certainly benefit its IPC more than those three factors you mentioned for AMD.

Besides, current 130nm technology from AMD right now probably has difficulties in reaching 2.4Ghz with good yields, let alone 2.5 or 2.6Ghz. They'll definitely need 90nm by the time they're trying to push for 2.5Ghz and beyond.

As for costs, if the processor put out a good performance at good cost, it wouldn't matter if the damned thing was a hellish 1000W! That's our point.

:evil:  <font color=red><b>M</b></font color=red>ephistopheles<P ID="edit"><FONT SIZE=-1><EM>Edited by Mephistopheles on 02/02/04 10:05 AM.</EM></FONT></P>
February 1, 2004 10:38:18 PM

Quote:
I agree with you!!! Or didn't you notice? Yes, it's obvious that AMD has the upper hand now!!!

I know you are agreeing. :smile:
I just wanted to rub it in a bit.

----
Support the NV/AMD/IBM axis of evil.
Who cares about HL2/D3 when we have Call of Duty today!!
February 1, 2004 10:41:55 PM

Post deleted by Mephistopheles
February 1, 2004 10:42:51 PM

Rub it in? That didn't sound too good... :eek: 

:evil:  <font color=red><b>M</b></font color=red>ephistopheles
February 1, 2004 10:49:55 PM

Quote:
But hey, everyone say with me "WAIT FOR TEJAS !"

I thought AMD was supposed to be shaking in their boots waiting for Pretzel?

Quote:
Anyone thought about how utterly unimpressive it is to have a new core with 150% (!!!) more transistors, 100% more cache and 50% (!) higher thermal density, and not even outperform the previous core ?

Wow, I never thought of all that.. what a impotent bunch we've been forced to listen to for some time now.

You know alot about this stuff BB. Are you an engineer?
You'd probably be a good one.

----
Support the NV/AMD/IBM axis of evil.
Who cares about HL2/D3 when we have Call of Duty today!!
February 1, 2004 11:03:33 PM

Be careful, kinney. If you read anandtech's review of the prescott, you won't come to the conclusion that they're an incompetent bunch.

It's all a matter of perspective, really.

But we must still bear in mind that it's all the same reality we're in.

:evil:  <font color=red><b>M</b></font color=red>ephistopheles
Anonymous
a b à CPUs
February 1, 2004 11:03:35 PM

>S939 won't bring any performance benefits to AMD's
>processors

Wanna bet ?

>Unbuffered RAM will not bring a performance benefit more
>than a very, very few percentage points

Not huge no, but a few percent nevertheless. keep in mind going from 3.2 GHz to 3.4 GHz is also only 6% in clock, and even less in performance, so a (pulling number out of my arse) 3-5% speed difference is not all that irrelevant.

>Increased HT bus is irrelevant. Current Bus is already
>overkill.

Really ? Then how come nForce3 performes so noticably worse with it 600 MHz HT bus, if 800 MHz is that much overkill ?

>And I suppose Alderwood, with its 5% boost in performance,
>won't matter?

I don't know a thing about Alderwood, but if it does bring 5%, than that does matter.

> DDR2?

No. I see it decreasing performance, rather than increasing it for quite some time.

> PCI-E?

No, since this applies to both intel and AMD. And no, I don't expect big speed bumps either (though its still nice and flexible technology)

>Increase in FSB to match synch with DDR2-533?

No, for reasons I pointed out above. Faster FSB might help more significantly when matched with DDR1-533 though, but also here, faster memory will help AMD as well as Intel.

>Besides, current 130nm technology from AMD right now
>probably has difficulties in reaching 2.4Ghz with good
>yields, let alone 2.5 or 2.6Ghz.

Got anything to back that up ? Cause I don't.. all i see is AMD roadmaps promising us 3700+ chips next quarter, so still on 130 nm, and most likely 2.4 GHz

>They'll definitely need 90nm by the time they're trying to
>push for 2.5Ghz and beyond.

Maybe, maybe not. I've heard this "EOL thing on process X" argument so often for AMD chips, yet each time AMD manages so squeeze out a few extra speedgrades, especially when their next process or chip is being delayed. Who would have guessed Barton went up to 2.33 GHz ? Tbred to 2.2 or whatever it was ? Hell, even K6 to beyond 500 MHz. Athlon has been pronounced EOL at 1 GHz, 1.4 GHz, 1.6 GHz, 2 GHz.. it never materialized. Athlon 64 now was being assumed almost maxed out already at 2 GHz, 2.2, now you assume 2.4. I'll believe when I see it because if 90 nm would be late, and AMD would be forced to fine tune their 130nm process even further, I firmly expect 2.6 GHz to be doable. Remember SOI is still young, AMD is still learning.

BTW.. ->
>Bob Baeyens with his usual speech...
Yeah, thing is, however much many people hate it, I am useally right.. I've been predicting poor Prescott performance for nearly a year now, when everyone still assumed it would be at least 20% faster per clock. Bookmark my prediction thread, as check back by the end of the year. While you may think otherwise, I'm not an AMD droid, but its AMD that is calling the shots now.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
Anonymous
a b à CPUs
February 1, 2004 11:10:22 PM

>Just switch one of your lamps to a power-saving variety and
>you'll get about 50-70W free power for your dear electrical
>bill...

You forget a few things, like PSU efficiency. A 550W PSU will likely not reach more than 40% (? ) effiency, you are are burning up closer to 200W. Granted, only if you are running SETI or something, but even iddeling its not that glamerous as PSU efficiency gets worse if the delta between actual load and rated load gets bigger (IOW, the PSU get even less effecient if your cpu iddles). This will get even worse as core voltage drop further, and we start seeing 120 or 150A requirements.

>Come on, Bob. CPUs don't get chosen because of their
>electrical bill implications. That was a far-fetched one.

Maybe not, but maybe they *should* take it into consideration. People compare prices down to the dollar, yet never even consider their computer may cost $100 per year in electricity.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
February 1, 2004 11:14:58 PM

Quote:
>S939 won't bring any performance benefits to AMD's
>processors

Wanna bet ?

Why? FX already has everything S939 has.
Quote:
Really ? Then how come nForce3 performes so noticably worse with it 600 MHz HT bus, if 800 MHz is that much overkill ?

Come on, you participated <A HREF="http://forumz.tomshardware.com/hardware/modules.php?nam..." target="_new">in this thread</A> yourself and you said that you were not surprised that it didn't really matter if the HT bus was set to 400, 600 or 800Mhz. Remember?
Quote:
so a (pulling number out of my arse) 3-5% speed difference is not all that irrelevant.

Don't pull numbers out of your arse, bob. It is probably not good for your health. And I won't comment on anything you pull out of your arse either.
Quote:
I don't know a thing about Alderwood, but if it does bring 5%, than that does matter.

How come you are so sure about AMD being superior to Intel this whole year but don't know that the platform transition to LGA775 will bring performance increases? Knowinh the overall scenario before stating such bold predictions helps...
Quote:
I'll believe when I see it because if 90 nm would be late, and AMD would be forced to fine tune their 130nm process even further, I firmly expect 2.6 GHz to be doable. Remember SOI is still young, AMD is still learning.

SOI is still young? And I suppose 90nm and Prescott is old? And SOI being young and AMD learning means they'll get higher clocks comparatively to intel? Isn't intel learning too? About the 90nm process? And about extending P4 to 64-bit?...

:evil:  <font color=red><b>M</b></font color=red>ephistopheles
Anonymous
a b à CPUs
February 1, 2004 11:16:59 PM

>You know alot about this stuff BB.

I know squat, but in my language we have an expression, I'm not sure it exists in English:
In the land of blind, one eye is king. And I'm not even sure I have one eye, I probably just have good ears.

>Are you an engineer?
FWIW, not a technical engineer, but a hmmm "trade/sales engineer", combination of technical and economical studies. But that was a loooooooong time ago.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
February 1, 2004 11:30:06 PM

Quote:
do the math on what it costs you over 5 years if you run your computer 24/7.

I do agree with you, less power consumption would be better from many viewpoints however an extra 20W isn't going to kill most people on their power bill.

In Canada we pay on average 10 cents per kW/hr

so: 80W x 24 hrs = 1.92 kW/hr per day x 10 cents per kW/hr = 19.2 cents per day
19.2 cents per day x 365 days = $70.08 x 5 years = $350.40

and: 100W x 24hrs = 2.4 kW/hr per day x 10 cents per kW/hr = 24 cents per day
24 cents per day x 365 days = $87.60 x 5 years = $438.00

So about an extra CDN$17.50 per year per 20W. Add tax and I'll give you an extra CDN$1 per extra Watt per year.

Sure, you need a few more motherboard traces to supply power to the CPU, but then intel brings in PCI-E which dramatically cuts down on the number of PCI traces needed on a board. It all balances out.

Despite the skyrocketing power consumption of processors in the last five years, OEM computers are cheaper then they've ever been previously. Look at Dell - for $580 you can get:
- P4 2.4G
- 256MB
- 80GB HDD
- CD + CD-R/W
- MS Works
- Keyboard/mouse
- 17" screen
- Sound/speakers
- 6 months internet
- Modem/NIC

20 Months ago that processor alone would have cost you $562 (if you bought a tray of 1000).

*Dual PIII-800 @900 i440BX and Tualeron 1.2 @1.74 i815*
Anonymous
a b à CPUs
February 1, 2004 11:38:36 PM

>Why? FX already has everything S939 has

You mean like unregistered, dual channel ram ? Or 1 GB HT ?

>Come on now bob, you participated in this thread yourself
>and you said that you were not surprised that it didn't
>really matter if the HT bus was set to 400, 600 or 800Mhz.
>Remember?

Okay, you got me there :)  And I forgot about that thread. If indeed this guys benchmarking is representative, then scratch the 1 GHz HT advantage.. But still, I find it odd, since everyone and their dog points to the 600 MHz HT links in the nForce3 to explain its mediocre performance. Oh, well, we'll find out soon enough.

>And I won't comment on anything you pull out of your arse
>either.

:)  well, then google around; I'm sure there are plenty of benchmarks that will give you more precise numbers. I've seen them somewhere, and seemed to remember 3-5%. If they are wrong, feel free to post the correct numbers.

>How come you are so sure about AMD being superior to Intel
>this whole year but don't know that the platform transition
>to LGA775 will bring performance increases?

Well, i guess I don't expect miracles from platform changes. I'm sure intel will release better/faster chipsets, but then I'm also sure there is some potential left in AMD64 chipsets. P4 chipsets have been around for a while, while AMD64 chipsets are about as young at KT133 non A was in the Athlon days.. Overall, this may help close the gap, but we're talking percentages at best. I think the real gap is going to get substantially wider than that over the next 6-12 months. everyone seems to have forgotten how K8 scales so much better with clock speed thanks to its integrated memory controller.

>SOI is still young? And I suppose 90nm and Prescott is old?

No, but Intel rarely does better (or for that matter, a lot worse) than its roadmaps. Those roadmaps show 4 GHz by the end of this year, or the beginning of next year depending which roadmap you use. Intel's main problem is also not clockscaling as such, but heat and leakage. You assume AMD has a clock scaling issue, i don't claim to know that. I do know they are not having major thermal problems though.

> And SOI being young and AMD learning means they'll get
>higher clocks comparatively to intel?

Oh no, their roadmaps show that. Intel is going from 3.4 to ~4 GHz (+17%) over the next 12 months, and that is including a shrink and a new core. AMD still has a shrink to take advantage off, and I just don't believe they won't be able to scale Athlon64 from 2.2 to 2.5 GHz in a year, and with a shrink to 90nm. 2.5 ought to be doable on 130nm if they had to, and a cake walk if they get their 90nm running properly (which seems to be the case so far). Who knows, I could be wrong, but I doubt. And don't forget the scaling issue ! AMD may gain as much as 20% more than intel for the same clock increase as intel.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
Anonymous
a b à CPUs
February 1, 2004 11:51:12 PM

Great, you actually did the math.. I was too lazy, so thanks for that.

2 small comments though, as I posted higher up, 20W extra for the cpu actually translates into >40-50 consumed by the PSU due to its inherent limited efficiency. It may even be worse with high power PSU's and every low voltage (high current) cpu's, but I don't have hard numbers. I think my numbers are pretty conservative, and suspect actual efficiency to be closer to 30%. I'll see if I can dig up some numbers.. BTW, a hot cpu also requires more fans, or faster fans, or both, eating into the power budget again. Nothing major, but it all adds up.

Secondly, here electricity costs roughly twice as much apparently as in Canada (you guys got waterfalls and things I guess, we don't). So, doubling the actual power consumption, double the price/watt, and I'm closer to $70/year extra. Still not huge no, but not entirely negligable either if you run seti 24/7. And things get much worse when you start overclocking and overvolting (and using high performance fans to keep things cool).

= The views stated herein are my personal views, and not necessarily the views of my wife. =
February 2, 2004 12:07:13 AM

I am no switching powersupply expert...but i do know a nicely designed switcher can easily reach into 90%+ efficiency ranges. Of course computer psus are not designed for max efficiency but i would still say 75% efficient at max load is a safe bet. Therefore i think that 50% effiecient is a better estimate...unless of course you can show some graph that relates switchers efficiency to load which i am too lazy to dig up.


If it isn't a P6 then it isn't a procesor
110% BX fanboy
February 2, 2004 12:07:46 AM

Post deleted by Mephistopheles
!