"P4 sucks" Heard a lot of that. Any benchmarks?

[Cretin]

How bad is the P4? Please leave price away from this.

I want benchmarks, facts, information. What's so bad about the P4? Isn't it good because it'll be clocked so much higher? I am under the impression that it is slower clock for clock, though.

Please, someone clear it up for me.

Cretin

 

[chiwawa626]

look at the latest issue of PC WORLD, even their benchmarks show the athlon is far more superior

 

[Cretin]

I don't have one at hand (or anywhere at all). Again, anybody care to enlighten me?

Cretin

 

[chiwawa626]

wheres the intel zealots when u need them

 

[Cretin]

Well I have to go. I will check this thread next time I am here, if I can find it. If anybody can answer me, please private-message me. Thanks.

Cretin

 

[Ben88]

How can you leave price out of the equation?

 

[Priit]

Leaving the price absolutely out, P4's performance fully sucks compared to Alphas, Sun's, MIPS'es if that was what you wanted to know...

 

[mazdarx7]

At http://www.cpuscorecard.com/ you'll see that the Icomp3 scores, a benchmark score for a battery of benchmarks places the fastest 1.5 Ghz P4 second place to an AMD T-bird @ 1.2 Ghz. Icomp3 is Intel's creation of a weighted score based on 6 benchmarks to assess a cpu's strength. They (Intel) chose the tests and lost, although, without those benchmarks quantifying the performance and demonstrating it, we would be hardpressed to tell a difference. Kinda like when I run my cpumark99 and Sandra benchmarks on the two K6 III + 550's (ALI 512 vs VIA 1024 chipset/cache). The Tyan (VIA) is the winner, but the margin of victory isn't worth mentioning and even with cost factored in makes no difference.

 

[fkloster]

#1 P4 Review & Benchmarks: Click.

#2 P4 Review & Benchmarks: Click.

#3 P4 Review & Benchmarks: Click

#4 P4 Review & Benchmarks: Click.

In summery, Intel is hoping the P4's improved fpu will enhance 3d apps. It shows... 2d is going to need some SSE2 optimizations.

The P4's 20 stage pipeline with improved branch prediction coupled with double pumped math units running in essence @ 3Ghz helps the P4 along quite nicely in the 3d department. 2d apps will continue to prove underwhelming until sse2 optimizations are abound.

 

[JellyBaby]

In all fairness this same "controversy" occurred when Intel release the first Pentium. At 60 MHz it was narry faster than a 486/40 -/100 and more expensive. Later compiler optimizations did help but the biggest advantage, as P4 will have someday, was the fact the clock could be scaled to ludicrous speeds. Eventually we saw P1s hit 233 Mhz, a speed no 486 could ever hope to acheive.

Would I buy the first P4? No way. Not nearly enough bang for the buck, Scambus memory and no future (different pin-out on the next P4). Many of you will indeed have a P4 in your life -- it just might take a while.

 

[ragiepew]

<< In summery, Intel is hoping the P4's improved fpu will enhance 3d apps. It shows... 2d is going to need some SSE2 optimizations >>


umm improved FPU??? The pentium 4 only has one FPU unit compared to 2 on the p3... how is that improved? If you mean that they added SSE2 (or should I say, they are depending on SSE2 to take over FPU instructions...) then maybe you are right... maybe you should have said &quot;2d and 3d is going to need some sse2 optimization&quot; instead... that makes a lot more sense...

other than that, the p4 is a good chip, dont let zealots tell you otherwise... It has its place, just as the athlon does. in my opinion, the athlon is better, but ill admit it, i favor AMD so take that w/ a grain of salt. read the reviews fkloster put up and make your own decision... dont let the masses steer you in a certain direction... after all, its you who will be putting out the cash and using the chip... not them.

 

[fkloster]

ragiepew, it is my understanding that while Intel reduced the # of FPU's on the chip, they improved fpu performance. The benchmarks do show improved 3d performance...who knows! The math units operate very quickly as well, but are small in #.

 

[ragiepew]

ahh, ok well i can live with that ... just didnt understand the scope of your reply... my fault.

 

[pm]

Pritt:



<< Leaving the price absolutely out, P4's performance fully sucks compared to Alphas, Sun's, MIPS'es if that was what you wanted to know. >>



Alpha, possibly. But Sun? MIPS? Which benchmarks are you looking at? TPM/C, SPECInt, SPECFP These are the big cross-platform benchmarks and neither the best from MIPS nor the best from Sun come very close to a 1.4GHz Pentium 4. Far from &quot;sucking&quot; the Pentium IV often beats them significantly (>20%). Pritt, which benchmarks (and which processors) from Sun and MIPS were you specifically thinking of when you made this statement?

I called you on this statement, because this is the common opinion: that systems from Sun cost vastly more than Pentium 4 workstations and thus must significantly outperform them. But this is definitely not the case. Sun sells total-solutions and they have an extremely competent sales force, and so they have been doing very well. But the performance of their thier mid-range systems is easily beat in most cross-platform benchmarks by systems that cost 80% less using processors from Intel (and AMD).

Edit: looking around a little more... throw ViewPerf (OpenGL benchmark for 3D graphics workstations) into this as well. (Note that the SGI workstation in the viewperf results is using a 1GHz Pentium III processor - not a MIPS CPU).

Here's links: SPECInt and SPECFP
ViewPerf
tpm/c

 

[josphstalinator]

i read an article over at firingsquad bout a week or two ago and from the benchmards they ran it showed the p4 falling short on certain aspects but excelling in other areas. the q3 benchmarks showed a huge advantage (as much as 25%) for the p4, but quake3 is highly optimized for sse so thats to be expected. other benchmarks such as winbench and office bench (not sure if i got the names correct) showed a significant advantage for the 1.2ghz athlon. another benchmark which tests internet performance (i forget the name) gave an advantage to the p4. so supposedly the p4 outperforms the athlon when running web based scripts or what not. go over to firing squad and do a search, im to lazy to provide the ilnk.

i think the bottom line is the p4 doesnt suck, it performs well, and creams the athlon when apps take advantage of sse. if price wasnt a factor id go with a p4 but since im poor my next upgrade will likely be a tbird.

 

[Osangar]

<< The pentium 4 only has one FPU unit compared to 2 on the p3... how is that improved? >>



This has more to do with the way the FPU is described then any real difference in the FPU itself. The PIII has separate add and multiply units, but both are issued instructions from the same port. As a result the PIII is limited to a total of 1 FP instruction per cycle. The mult unit is not fully pipelined and can only execute 1 instruction every two cycles. The add unit is fully pipelined and can execute 1 instruction per cycle.

The P4 has a single fully pipelined FPU that performs both add and mult instructions, however they are performed within separate pipelines in the FPU. The add pipeline has a throughput of 1 while the mult pipeline has a throughput of 2 just like the PIII. Unlike the PIII the P4 is limited to 1 FXCH instruction per clock (or is that every second clock)

IMHO the PIII and P4 X87 units are nearly identical in function even if they are described and built a little differently.

Hardware prefetch, and high FSB bandwidth should give the P4 a big advantage on most FP intensive code in the long run.

(mult should be read mult/divide/complex add is add/subt)

 

[MikeyP]

The P4's FPU is generally accepted as weaker than the P3's and much weaker than the K7. The Quake3 benchmarks are so good because Quake3 is bandwith limited, a place where the P4 architecture is excellent. fkloster, the double pumped ALU's will have no impact on 3d performance. They are the integer portions of the chip, 3d is nearly totally reliant on FPU. It will be interesting what the future holds for the P4, if it can get the SSE2 optimizations it needs, it has the potential to be a screamer. Until then, I prefer the Athlon Isn't competition grand?

 

[Bakwetu]

It never ceases to amaze me that people seems to be unaware of what hardware site that is connected to this bbs. Anand's article

 

[BFG10K]

fkloster, your arguments are completely wrong.

In summery, Intel is hoping the P4's improved fpu will enhance 3d apps. It shows... 2d is going to need some SSE2 optimizations.

The Pentium 4 does not have an improved FPU. Clock for clock it's weaker than a Thunderbird's and a Pentium 3's. Intel is relying on SSE2 optimisations to boost FPU speed. It's a similar situation to AMD's K6-x CPU.

The reason why Quake 3 is receiving such a performance boost is because of the extra memory bandwidth provided by a quad pumped FSB and RDRAM.

double pumped math units running in essence @ 3Ghz helps the P4 along quite nicely in the 3d department.

The double pumped ALU is for integer calculations only. It has no effect on floating point calculations, hence it has no effect on 3D applications.

2d apps will continue to prove underwhelming until sse2 optimizations are abound.

Once again, SSE2 optimisations are more critical to boosting FPU calculations rather than the other benchmarks.

There are just some tasks where the Pentium 4 will always do worse than a Pentium 3/Thunderbird because it has a much longer pipeline and hence a larger execution latency.

 

[noxipoo]

of course fkloster only posts reviews from crappy reviews site that are intel friendly. try Anand's (you are here already) or tomshardware.com's reviews.

 

[fkloster]

Sorry about missing Annands review. I'll go read it now...