Disagreements with Tom's P4 Benchmarks

[Guest]

I have questions upon questions... ne one feel free to join in. Firstly..I would like to know EXACTLY how the benchmark tests were run... how could he run the new P4 Intel chip on a 100MHz FSB or 115MHz when it runs on a 400MHz system bus, the RAMBUS is also 800MHz and not 400MHz as he implied, and if the P4 chip was run @ 100-115MHz system bus, WHAT multiplier frequency ratio was he using..because...to get 1.75GHz of CPU frequency running at 100-115MHz system bus you calculate for yourself.. 100MHz * multiplier rate = 1.75GHz CPU ...come on.. are you trying to tell me that the chip..let alone the board has such a vast multiplier ratio. Also..all the tests done in the gaming area, were the game written for "3D NOW" archtecture..which of course is AMD specific, because then obviously you will get some fallback on the Intel chip if the games were coded for 3d Now architecture...but as you can see...it will still run on the Intel chip...BUT..if you run a game NOT written for 3D now architecture...and run it on an AMD chip you will see problems..a drawback still present on the AMD chips..but the Intel chip will run with no problems..Also why did he fail to mention how the new architecture supports online gaming...a feat that would be MUCH enjoyed by all I think..the 400MHz system bus..the 800MHz RAMBUS support ...or even the 3.2 Gb transfer speed ..or the Intel Netburst architecture... well...personally..I think that the tests were done very biasedly...and that the floating point execution on the AMD chip still limits applications.. cos..if it ain't 3D now...then it ain't for the AMD.. Next time...if u do tests..don't go into it with a mind of putting the Intel down..

 

[Tormented]

Man did you READ also his article or just seen the pictures?

First of all the FSB is 100 mhz but it is quad pumped that means that per one clock the CPU transfers 4 times the data
It is like an Athlon is a DDR that means that Dual pumped
it is also a 100 FSb but transfers 2 timer the Data.
So when he did the 115 Fsb the Cpu actualy worked at 115*4=460 Mhz (WOW) I admit that fast, very fast...

About the RAMBUS I just don't know why bother with this expensive Memory yes it is lightning fast, 3 times faster than PC133 (can you get me a link to the RAMBUS specs?).

The boards are designed for Future processors too so ofcourse the MB had such high Multiplier I wont be suppriesed if the board supports upto 20x!

About the 3D NOW hmm... you know tom should have specified if the games were 3D now patched or NOT ...

About The FPU the Athlons FPU is much better than the P3
and because the P3 had 2 Fpu units and one was removed in the P4 you can see that Athlon if he Kicked p3 in Fpu performance he will Slay the P4!

<b>-----------------------</b>
-<font color=red><b>R.K.</b></font color=red>

 

[Guest]

Dude all reviews are biased. To counter bias, one must read a great many reviews. Tom's review is on par with many others I have read. His article is not deviating ,that much, from the norm.
I have read Dr. Tom's reviews for several years now and he is usually right on the money. He is arrogant but usually on the money.


"in the end its all about holding your breath"

 

[Guest]

I'd have to agree with the previous post. I find it kind of amusing how people can get so emotionally charged when they read such a review. I'm pretty confident that Tom would not publish falsehoods. He receives to much airplay and to much scrutiny from the AMD fanatic crowd as well as the loyalty to Intel crowd, both of which, like I mentioned before, get quite emotional about the latest and greatest in computer hardware.

 

[Guest]

Okay...I did read...but I am first to admit fault..and i admit mayb i did not read as well as i should have.. but my point is this..i've also been on here for years.. but first time in the community..and not saying that AMD is a bad chip..cos back in the day I had an AMD DX4 100..and it used to fly..but my point is that AMD is still restricted by its floating point system..limiting it to software specific for the AMD.. maybe Intel have over-rated their chip..who knows..but until the chip is benchmarked properly..utilising all its potential..how can we say that the tests reveal true data... true...not much can handle SSE right now.. but I just feel that everything here is beginning to be based strictly on how much it can b overclocked...and maybe because all the new Intel chips come multiplier locked.. it is not a good over-clocking chip... Emotional...no... wanting real facts with all Intel's new P4 chips potential.. yes

 

[Guest]

Are you running SSE2-enabled software currently? I'm not. Will you be running SSE2-enabled software in a year? Will I? Probably. Are you buying a system today to run software that won't be out until next year? I wouldn't. Will the Hammer family of AMD chips support SSE2? Far as I know, yes. So please, do explain, why would I want to buy a P4 NOW, when it performs worse than T-Bird - and costs more? When SSE2 applications become widespread, then P4's - and Hammers' - support of them will become a factor in choosing a new CPU - but until then, it's FPU unit vs FPU unit, and this is an extremely uphill battle for Intel, as the 3DSMAX test results show.

 

[Guest]

You need to format this message a little better. My vision was gettin blurry after the first few sentences.

 

[Guest]

As far as I am aware the games used(unles a specific patch has been added) are optimised for the Intel instruction sets(SSE I think) as well as the AMD 3D Now, and the Fact that the Benchmarks don't reflect the true potential of the P4 is mentioned in the article.

M

 

[Guest]

You referring to gaming...
There was a fair amount software probs with earlier AMD chips ie Multiple Windows Protection Errors when addressing memory (Crazy Pointers pointing to protected memory used by your OS). However reading lots of reviews in the
past 2 years this is no longer the case in a majority of situations. It appeared that Progs written in Borland C++ V3.1 suffered from this naturally as it was originally designed for 16 bit win 3.1 and written specificaly for intel hardware. The current intel 1Ghz and 1.13Ghz have had a lot of teething problems that the AMD chips haven't ie not working when the original batch was sent to reviewers (oops). However these are now being ironed out with those
microcode releases with new MB bioses (basically tells the chip which transistors not to use)(what is the collective for a group of bios anyway?). Personally being of the knowledge that software written for the intel architecture is mostly written with Programming languages designed to intels specs on intel machines i would spend the extra on intel. Whats the chip for a server? 1Ghz is going overboard for games etc the 800MhzEB using CAS2 latency Ram is a more economic solution and this chip is tried and tested,
let others have the problems and benefit from their pain.
So as you can see... written for intel..on intel...by intel... Do you think that the AMD can NON-3d Enhanced games as well as Intel can play games designed AMD-specific..have you even tried OLD games...and by that i'm meaning..e.g NFSIII or whateva... tell me if the AMD floating point system can handle non 3D enhanced games well...if you can say that...then i bow down.. MMX is here to stay..

 

[Guest]

Nothing is 'written for intel..on intel...by intel...'. Software is written for 80x86 platform, by developers. If you look back, you'll see Intel releasing new instruction sets for x86 platform every so often, and every time, it took a year or so for software developers to catch up - and by that time, when the consumer actually NEEDED those instructions, competitors (AMD, Cyrix) already had those instructions in their chips, nullifying Intel'a advantage. It has happened before with MMX, it will happen again with SSE2. As for floating point performance needed by older games, 3DSMAX is PURE FPU performance, it doesn't use consumer 3D accelerators for the final renderings, and look what happens:

http://www.tomshardware.com/cpu/00q4/001120/images/image011.gif

P4, IMHO, was born already dead. Within next 6 months or so, Intel seems to have nothing that can compete directly with Athlon. AFTER SSE2 sees widespread use in software, Intel might catch up on AMD - but until then, they are far, far behind. There is absolutely no reason to buy an Intel CPU at the moment IMHO (unless you want to spend $3000+ so you can run Quake at 300FPS in 640x480x16).

As for Itanium, I'm not very impressed by what we know about it so far. Being natively 64-bit, it will VERY likely have HORRIBLE performance with 32-bit applications, same as any other CPU that tries to emulate another platform (Crusoe anyone?). On the other hand, Sledgehammer will be a natively 32-bit CPU with a set of 64-bit instructions added on top - which means it will run old 32-bit software just as well as anything else, AND it will run new 64-bit software - a perfect CPU for the 2-3 year transitional period.

 

[slvr_phoenix]

Man, what are you smoking talking about 64 bit architecture having to emulate 32 bit architecture? Remember way back when the 386 came out with a 32 bit architecture? Before that everything was 16 bit. Yet did a 286 EVER run 16 bit software faster than a 386? Did it? NEVER!

Why? Because you don't have to emulate a lower bit architecture. That's absurd and anyone who believes that doesn't know jack. Just because a register holds 64 bit numbers doesn't mean that it can't hold a 32 bit number just as easily.

I do think that Intel is going wacko with their Itanium design though. As a programmer I don't even want to think about trying to write software that doesn't have to execute sequentially. Ugh. As if my job isn't hard enough already. But if Intel can just make a 64 bit version of the P4, AMD's hammers won't stand a snowball's chance in silicon hell.

And I agree that Intel was pretty stupid for castrating their floating point processing ability. And that any software written specifically for that good old standard will run better on an AMD chip. But any software that does get written for SSE2 standards is going to make AMD's chips all look like crap except for their hammers. And since SSE2 is going to support floating point operations, except for purely scientific software and supreme graphic rendering software that require exceptionally accurate numbers, people will be using the SSE2 for most floating point operations. Which means that the majority of future software will scream on the P4 because it won't be stuck using that crappy floating point processing Intel put into the P4.

Which means that in the future AMD will really have a very small market indeed for places where their non-hammer chips will perform in any way better than P4 chips.

As I see it, Intel's P4 has struck AMD hard because ALL of AMD's striving right now is to break into the market against Intel's P3s and Celerons. And Intel's P4 is going to smack AMD's plans down hard because the P4 has sooooooooo much more potential than the P3(which is already at it's limits).

All AMD has left to stand on now is their hammer chips. Right now both chip companies are left with their last cards played. Intel has the P4s. AMD has their hammers. And the true test of who will come out on top is going to be processor speed. So unles AMD's current roadmaps are a hoax, AMD is screwed because Intel is finally getting their act together and providing their products on-time and in-mass. And Intel's products are going to leave AMD in the dust according to roadmaps.

I think that the decline of Intel is finally reversing and we'll see the chip giant proving it's superiority once and for all against AMD.

And I would like to say that Tom's review of the P4, just like virtually everyone else's, is biased for AMD. Maybe they didn't mean to, but it's true. Tom's has proven that SDRAM has lower latencies than RDRAM. And anyone knows that software written to use optimizations is going to run faster on the chip it's optimized for than it will on a chip that doesn't have full optimization support yet.

So it's obvious given just those two things that Intel's P4 should do worse under the testing conditions that were used. I was surprised the P4 did so well in Quake3 and can only guess that it is because of SSE1 optimization code.

Now, I know it isn't fair to blame Tom's for this, since there is no DDR SDRAM P4 support yet and there is no SSE2 optimized software yet. So finding a FAIR test right now is next to impossible.

But Tom's DIDN'T make this obvious to the reader. They DIDN'T repeatedly express this so that the reader KNOWS that any current tests done on the P4 just aren't going to show it's true potential.

It was mentioned, but I don't think it was mentioned enough and expressed thoroughly enough to really give Intel their due. So I do blame them for that and say it was indeed a biased review if for that reason alone.

Frankly, Tom's has been biasing a lot of their reviews lately. They USED to provide unbiased reviews. But for some reason they changed. Maybe it isn't much of a bias, but it's still there.

But I still like thier reviews because they do provide a lot of information and it's easy to negate the bias in which the reviews are written to pick out the facts. So I'm still an avid Tom's fan. But I wish they'd go back to being completely unbiased.

And if anyone thinks that a P4 sounds cool now. Just think what it'll be like when it upgrades to a 133MHZ bus, a .13 micron engraving, and has DDR SDRAM support. (And even better, if Intel also improved the floating point processing to at least two channels, if not four dual-pumped channels.)

- Anything can be fixed with duct tape, a swiss army knife, and WD-40.

 

[LTJLover]

Wow all these long responses. All I have to say is you can spend $1100 on your "top of the line" P4 and I'll spend less than $500 on my AMD and still leave you in the dust. Unless you can find a bunch of SSE2 software, but good luck.

Jon
"Water-Cooled CPU Runner"

 

[Guest]

The whole reason that the Athlon does well in 3D games and rendering is because it has the most powerful x86 floating point unit ever built. This has nothing to do with 3DNow! optimisations as these are handled by a different part of the chip. With or without SSE(2) and 3DNow!, most FP intensive programs will give the Athlon a sizeable performance lead over the P3 and P4 - at least until it can get to very high clock speeds.

As for the new IA-64 (Merced/Itanium) architecture, that is irrelevant at the desktop level for a few years yet. Its 32bit performance is hopeless because it is NOT x86 compatible. It has to perform software emulation in the same way that an Alpha or UltraSPARC 64bit system would to run PC programs. It is a TOTALLY different situation from the move from 16bit to 32bit that occurred when the 386 was introduced. Fundamentally it was still compatible with the 286 and had a 16bit hardware mode. The parallel today is that the Hammer will be a 64bit processor with a 32bit hardware mode whereas Itanium will not.

 

[Guest]

Wow, I'm glad everyone here knows everything.

A 64-bit processor CAN handle 32-bit input IF it is designed to AND the processor can decode the 32-bit input (i.e. x86). The problem with the Itanium is that Intel wants to move away from x86 (probably to remove other companies from competition as nobody would be able to clone the new processor without a long legal battle again - see AMD vs. Intel). AMD plans to take x86 and extend it to 64-bit. A much more simple solution and better for everyone involved until 64-bit becomes a standard. Very similar to the switch from 2x86 to 3x86. I agree with AMDs move. Why complicate something that is already complicated enough?

 

[Rop]

AMD Athlon has a much better FPU then any other x86 procesor out there currently. It does not need 3dNOW! nor any other especific intructions to take the crown. You might be refering the AMD's past perhaps? as in the K62, K63 etc? Since those are procesors with a subpar almost none existant FPU, they tried to help them with an intruction set known now as 3DNOW!.
Sure the Athlon has the instruction set as well but it does not need them to execute excelence in FPU performance.

AMD is not to worried about the P4 since it can beat it with a chip at a slower clock rate. I do belive that it is not a smart move on AMD to wait and see what happense with the normal joe consumer since he/she does not know anything but Mhz.

Then again who knows what the future might bring. But as of now K7 Thunderbird Athlon is the best x86 procesor in the market period.


Why do I use LINUX ? Cause its the BEST OS
Why do I use Windows? Cause its the BEST Nintendo..

 

[razvan]

slvr_phoenix,


"Remember way back when the 386 came out with a 32 bit architecture? Before that everything was 16 bit. Yet did a 286 EVER run 16 bit software faster than a 386? Did it? NEVER!"

You compare that to P4 ---> Itanium jump?!! Itanium has a new architecture that is not based on 80x86. Itanium doesn't have the same machine code instr for a basic ADD or JMP so it must use a program for conversion. But those are BASICS! I don't think you are a programmer!

"Just because a register holds 64 bit numbers doesn't mean that it can't hold a 32 bit number just as easily."

Alpha is on 64 bits. Try to run some 80x86 software on Alpha.

"down hard because the P4 has sooooooooo much more potential than the P3(which is already at it's limits)."

P4 has 2 advantages: SSE2 and clock speed. Like you've said SSE2 will be in the Hammer family when the software will actually use it. The trade-off for reaching high clock speeds is a lower IPC performance. IMHO this is a bad choice. For having the same performance as Athlon P4 must have a higher clock and that means more sophisticated production lines and in the end more costs to produce it.




Razvan

 

[bodhi]

Oh do shut up AMD fans please. You are getting a tad dull. Hammer this blah blah FPU blah blah...... Do you really think anyone is going to trust their multi million dollar network to an AMD server? AMD who, while may be fast, have the reliability record of Italian cars? I really dont think so.

As has been pointed out, do you think we have seen all P4 can do yet? Lets take a quick trip down memory lane shall we?
Pentium Launch - Intel's own 486 DX4-100 was faster at the time of launch. 3 months later 486 was dead.
Pentium Pro Launch - In Win 3.1/DOS/95 a Pentium 200 could rape a PPro 200. Launch up NT, and it was a whole different story.

(Interlude - either of these sound vaguely familiar?)

Pentium 2 Launch - P233 was just as quick as P2-233, could be had for a lot less. 6 months later P2 was at 350Mhz and Pentium was dead.

Pentium 3 Launch - P3-450 was NO faster than a P2-450, yet had a 100 quid price premium. 6 months later Athlon came out, and Intel "had a problem" (apparently). 3 months later we had Coppermine, Intel laughs in AMD's face.

The moral of this story? Intel processors are NEVER the dog's testicles when they come out (future proofed stuff rarely is), and believe me I know from owning a P60 and a P3-500. Give em 6, even 12 months and they shall start raping the competition stupid. So enjoy your period in the sun AMD, knowing Intel it wont last very long.


Bodhi

P.S Interesting observation at a LAN the other weekend. Everyone with an Athlon 900Mhz or over, who wasnt using extra cooling ( as in more than case and CPU fan), had to take the covers off their machines, as excess heat was causing frequent lockups. And you expect businesses to go near Athlons with that happening. Btw my P3-500 with SLI aka Heat Producers of Death, motored on fine with the case fan supplied by the manufacturer and the most irregular CPU fan on earth. 400 Mhz diff granted but have you felt inside a case with 2 voodoo 2's?

 

[Guest]

Ok, real quick, how many recalls has Intel had in the recent past? That's what I thought...and you think AMD fans are bad...

Unless you have an Intel system that outperforms every AMD system, don't try to tell me how great Intel is...I don't have a time machine, so the awesome software written for the P4 a year from now is of no use to me...

 

[Guest]

oh so now it's "FPU blah blah blah" huh! One little reminder, what's the performance of Pentium's FPU compared to the 486, and how does the Pentium Pro's compare to the Pentium? Basically every new generation of Intel's processor the FPU gets a new level of performance boost until the Pentium 4 came along. Since it's pretty safe to say software such as 3D animation package and CAD/CAM software are not going to be SSE2 optimised an year from now and very likely the year after that (after all, didn't Intel positions P4 as a WORKSTATION chip?), unless, man this statement has been repeated for two billion times already by everyone, P4 ramps clockspeeds like crazy it's simply not gonna cut it.

btw ppl with multi-million dollar network trust more of IBM and SUN than Intel anyway.

 

[8procstooslow]

If I were to build a corparate network, I would not use the P4 or the current Athlon. I would be using a 500Mhz-667Mhz Xeon. That is the chip I would like to see the P4 compared against. The reason? The majority of server applications load lots of small files a frequent intervals. Yes high bandwidth is important, but latency is the killer and the P4 with Rambus has far too high a latency.
Yes I like Athlon (have two myself) but in Business a technolgy has to be well proven to entrust corporate data with. Take IA-64 when it gets released, we will be sticking with Alphas, and PA-Risk machines for a long time before upgrading to IA-64.
Intel really should have waited till the final release of the P4 with the new pin outs and manufacturing process before releasing to market, AMD or no AMD. The people who are likly to by this chip at these high prices also look at total cost of ownership over a period of time. If they have to throw out the motherboard/CPU and maybe even the memory in nine months time+invest in upgrades to proven software just so it can run efficiently on the P4 they will probably not upgrade for a while. This will hurt Intel in the short term and give AMD a chance in the high end, with DDR+SMP without the need to upgrade software.

 

[Guest]

But the coppermine never did kill the athlon .....
oh and as for cooling on the 900 Athlon shall we look at the colling needed for the top end P3's......
Maybe I wouldn't entrust my servers to 900 Athlon box's but I would certainly not put a high end P3 or P4 into them ...
I would stick with a 700Mhz or less CPU and, at that speed it doesn't matter which one you choose......
the P4 will be pretty good in time but I don't think for a second that it will be out there by itself I even expect(Hope) the cyrix chips to become contenders with the backing of VIA......
so I suggest you try comparing like for like instead of putting you blinkers on and shouting at other people for doing the same......


M

one of the first UK T-Bird users....

 

[Guest]

BTW Sharky Extreme and Anandtech also have reviews of the P4 and they basically come to the same conclusion. The P4 doesn't give enough bang for the buck right now to be worth buying. Maybe in the summer when Intel switches to .13 process, there is some software optimized for SSE2 and has a DDR chipset, a P4 might be worth considering.
http://www.sharkyextreme.com/hardware/guides/pentium4/
http://www.anandtech.com/showdoc.html?i=1360