1st Prescott P4 2.8ghz benchies revealed!!

[Kanavit]

Prescott 2.8 Benches (http://www.ocheaven.com/article/0310/readgoodarticle.asp?id=26)

very promising results, considering it's only in ES sample. 1mb L2, PNI(SSE3) ,90nm, and advanced hyperthreading are product highlights.

i can't wait, i think i'm going to get Asus P4C800 just because of prescott.:toast: :toast:

[Soulburner]

Link doesn't work unless you copy/paste for some reason.

And I would not by an 865/875 board now expecting future compatibility, becuase you are not guaranteed anything. The best thing to do would be to buy an actual chipset (Grantsdale) that is made for the processor. This means waiting to see whats available at the time.

The Prescotts won't even be here until Jan-Feb of next year anyway. That article is dated about a month and a half ago.

I like this one though:

http://www.oc.com.tw/article/0309/imgs/p4p-listcpubench2003.gif

[Kanavit]

ok, thx SoulBurner

i fixed the link.

yes, that particular benchmark is very revealing. i see much potential in this chip.

i like this result , look at how well it does to the 3.2
http://www.oc.com.tw/article/0309/imgs/p4p-list3dmark2001se.gif

[Soulburner]

Yeah but its obviously not running right on the current boards/BIOS's setups. I would expect this thing to kick ass when its released.

[drunkenmaster]

not really, remember, Intel themselves have said the release prescott, 3.4Ghz i think supposed to be, WILL be slower than the 3.2EE. Also remember how well p4's did in sisoft with sse2 tests, yet real world performance was nothing to talk about. AFAIK SSE3 provides a tiny amount of the increase in speed(when apps finally became optimised, some apps ) that SSE2 gave.

[drunkenmaster]

also remember, if the EE is supposed to be 10% faster due to 2meg cache onboard than the 3.2 none EE. THen well cache up to 1meg, 5% increase? doesn't quite work like that, but 3dmark 2k1 gives way less than that, 3% increase from the extra cache compared to the 2.8 P4. That doesn't look like almost any improvement clock for clock from the actual core, 3% increase is really expected from a growing cache, as we've seen proven many times.

[Kanavit]

remember, Level 2 cache is more important then Level 3 cache. The impact is greater performance wise.

Prescott has 1mb L2

P4 EE has 512kb L2

so, it remains to be seen how much performance increase this makes.

[climbski]

THIS BLOWS HARD.... I just realized... they arn't going to make a 12 or 13 multi Prescot are they??? Dammit!!!!!:stick:

[Soulburner]

Originally posted by drunkenmaster
not really, remember, Intel themselves have said the release prescott, 3.4Ghz i think supposed to be, WILL be slower than the 3.2EE. Also remember how well p4's did in sisoft with sse2 tests, yet real world performance was nothing to talk about. AFAIK SSE3 provides a tiny amount of the increase in speed(when apps finally became optimised, some apps ) that SSE2 gave.
No, there was no Extreme Edition in any of those tests above.

[Poki]

Seems to be darned hot at only 2.8GHz too. Even with an AC unit blowing on it!

[drunkenmaster]

i wasn't refering to p4ee results in teh test as there obviously ain't any. Just that intel themselves say it won't be faster, but infact slower than the 3.2EE , and the tests show in real world app(ok 3dmark is not real world, but better than the latest beta cpu benchy showing better scores for sse3 which won't show in ANY app that will be out for some time) that its barely faster than an equiv P4, even though intel are spouting on about new core, optimisations, better , sse3. All that was shown was a weak 3% increase which is almost certainly down to the cache and not anything to do with core. Or if the core IS better then its use of cache is crap, either way not good.

L2 cache in theory is much better than L3, but depends entirely on chipset, optimising, how it works. A huge L3 obviously helps a lot, might be less effective than L2(debatable) but it may help the memory as a whole work better, and let the L2 work better two, as when it makes mistakes the needed prefetched stuff might be in L3 rather than way slower ram. I reckon to a point it should even out.

[Soulburner]

Also remember their comments on it not running properly on the current boards. To see what this processor can really do we will have to wait until its released in LGA775 on Grantsdale.

[drunkenmaster]

Hmm, must show that the core is extremely badly designed then, there must be something actually wrong with it. It doesn't do more work than a P4 3.2, yet it looks like its gonna use more power and be hotter, that literally makes no sense as generally, these are teh main production reasons for moving to newer cores as the old core is nearing its end. But they've made it less efficient rather than more.

Dunno, but sure seems odd, sure they'll fix it, maybe someone should send them an E-mail with a AMD letterhead bit and say maybe it needs an extra layer in the core, and they could call the new revisions the prescott-bred B. :p

[drunkenmaster]

Hmm, its pinout may change, but thats more board design issue, the inner core should be the same, no reason it will change, and tbh no real reason the chipset should matter. I mean, of course it does, and we should see an increase in performance, but thats a general chipset thing, P4 worked better on canterwood than 845 and nf2 better than kt333. Grantsdale ain't here, we'll have to see when its finally released and if the chipset is release then( i dunno what their plan is tbh, both versions straight off considering the delay?). when was the new socket supposed to come out assuming the prescott had come out around late oct/early nov? Either way, the prescott will work on teh canterwood, which means it should work at 100%, so any increas grantsdale brings should surely be jsut chipset improvement increases?

I haven't read about it yet at all, will do soon.

[Soulburner]

Why are you saying its poorly designed? Do you have the qualifications and knowledge to work on your own processor design, let alone on the level Intel has reached?

I really don't think you are inclined to say such things. If you read what is posted in that link it says it had problems on the current boards, that were NOT made for the Prescott processor.

[drunkenmaster]

well, i think we've all heard it has problems with being too hot, its screwed intels roadmap and Should, by design of intel, ahve been ready for release and obviously hasn't been as its too hot. If its too hot, there is a problem.

It doesn't matter really that Intel can design chips far better than me, and even that, the top intel design engineers can 99% of people that work for them can't. Because from an outside perspective, something is Clearly wrong with it.

EDIT:- Also, the article only Guesses, that it has problems with all 875 boards, i would assume of the range of boards it tested it used the board that gave the highest results. if so who is to say that it isn't compatible, they didn't say it as its impossible. The chip was always going to be released to work at first with the 857 chipset. If that was teh plan, then even if that created problems, the chipset itself must be able to support it. The problems appear to be the chip uses to much power, to use and assuming thats the case, Intel this is a major problem. This could Well account for the problems of 10-30% diff in scores. This could be power problems not being able to keep cpu at 100%. We already now that of teh "tens" of boards they tested, very few are RATED for prescott. If so then assuming the ones that were gave consistant, and highest results, there is no reason to assume these boards had any problem at all.


EDIT number 2:- this is general reasonably assumed as from teh results themselves, and from INTEL saying it WON't be as fast as a 3.2EE< so it will be less than 10% faster than the 3.2P4 none EE and at 3.4Ghz, INtel said this. SO where do you imagine grantsdale will provide a large boost in performance and currently none working boards. Aquamark and 3dmark show its about 2-3% faster than teh 2.8P4, this is exactly what INtel said it would do. IMHO canterwood, the right one, works with it completely fine. For you to say otherwise would contradict Intel.

[Kanavit]

remember how the P4 1.4ghz was slower than the PIII 1.3 when it first came out. Give Intel some time to develop the Processor and it will become faster and cooler. i see a lot of potential with Prescott 90nm design and SSE3 instructions.

[NoOne]

CPU Mark 99 - Precott 12% Slower
3DMark2001SE - Prescott 3% Faster
3DMark2003 - Prescott 2.5% Faster
PCMark 2002 Pro - Prescott 16% Slower
SuperPI 1M - Prescott 6% Faster
SuperPI 2M - Prescott 10% Faster
Content Creation Winstone 2002 1.0.1 - Prescott 5% Faster
AQUAMark 3 - Prescott 1% Faster
CPUBench 2003 - Prescott 53% Faster
Sandra Memory Bandwidth - Prescott 5.5% Faster

[Kanavit]

it looks good NoOne.

[drunkenmaster]

lol, wait for the BETA CPU bench 2003 to either include INtel in its list of developers in the final product, or for that score to be reduced DRAMTICALLY, or for it to say it only tests for sse2 and 3, and if it has sse3 it gets a 53% boost.

I'm sorry, but thats clearly an error in a buggy program. The rest of the increase are as expected.
Also, Kanavit, the P4 is still slower clock for clock than the P3. Its designed that way to clock higher, that did it easily, the prescott simply isn't clocking, hence the delays, hence the very high levels of heat.

[Kanavit]

but clearly Prescott will not be slower than the regular P4 clock for clock. Prescott will only be slower to the EE.

[Soulburner]

Originally posted by Kanavit
but clearly Prescott will not be slower than the regular P4 clock for clock. Prescott will only be slower to the EE.
And not in all benchmarks. Remember the benchmarks for the EE shows that not all applications benefeit from the extra cache. Some are boosted immensely and others see no gain at all. The Prescott will gain in all areas.

[Poki]

Hmm, keep an eye on Fugger after he gets back from Comdex if you want to see how their Prescott performs. Check this link (http://www.xtremesystems.org/forums/showthread.php?postid=270653#post270653) out.

[drunkenmaster]

Hmm, afaik the sse3 is the ONly improvement, cept the cache right? nothing can use sse3 yet, so it really shouldn't show any improvement, and from several sources the sse3 instruction set is about as effective as shouting at your p4 that you want it to run faster. SSE2 was good, it took ages to be used though, sse3 will take jsut as long and won't provide much usefulness. Also, if its still x86, surely there must have been other registers taken out of the instruction set to fit in sse3? less sse2's? maybe why the dip in pcmark 2002?

Either way, you need to remember, INtel were talking about the planned 3.4Ghz prescott, IT will be slower than the p43.2EE, whatever the p4EE gives on average, say in gaming, 10%? even with a 6.25% clock boost AND teh extra cache, it WON't match the p4EE. I think intel have pretty much said it won't be faster except from the cache. Its a P4 AE(almost extreme) on a smaller die.

[Soulburner]

I'll quote myself again:
Originally posted by Soulburner
And not in all benchmarks. Remember the benchmarks for the EE shows that not all applications benefeit from the extra cache. Some are boosted immensely and others see no gain at all. The Prescott will gain in all areas.
With that said, here are the improvements I know of so far.

SSE3
Increased number of registers (2x at least)
64-Bit Extensions
Improved Hyperthreading
Larger L1, L2 caches
Longer Pipelines
Improved Branch Predictor
.90nm Process, 1.4v

Among numerous other architectural things I don't know about.

[skate2snow]

just think to this Prescott is still marked like an P4. And its suposed to release the P5 so probably is only an temporary P5 becose do remember the P4 socket423 and now socket478 is much better. So i think Intel only realese a new core so when ill do the P5 thats will be the real deal

[Kanavit]

as far as i know, intel said they will not use P5 name. Prescott will continue to use P4 name but with advanced HT instead. i have no idea to what intel's next generation processor 'tejas' will be called.

[Soulburner]

Originally posted by Kanavit
as far as i know, intel said they will not use P5 name. Prescott will continue to use P4 name but with advanced HT instead. i have no idea to what intel's next generation processor 'tejas' will be called.
Tejas is Pentium 5. :cool:

[Ace-a-Rue]

Originally posted by Soulburner
I would not by an 865/875 board now expecting future compatibility, becuase you are not guaranteed anything.

According to Intel Chipset designs PDF file, it says this about Prescott compatibility with the 865 chipset:

Section 1.3
The 865PE/865P chipset designed for use in a Desktop system based on a Pentium 4 processor with 512K-KB L2 cache on 0.13 micron processor in a 478-pin package and the processor code named Prescott.

[Kanavit]

these mobos have been tested with the Prescott review...

Gigabyte 8KNXP (i875P)
FIC P4-875P MAX (i875P)
ASUS P4C800 Deluxe (i875P)
AOpen AX4C Max II (i875P)

with Bios flash.