Final Words

With the 570J being the last desktop Pentium 4 launched in 2004 (once again, not counting the Extreme Edition), it looks like AMD will close out this year on top, which is fitting considering how strong AMD has been throughout the entire year.

Looking at the performance of the Pentium 4 570J it's clear that had Intel launched the 4GHz Pentium 4 things would have been much more competitive than we first thought. AMD would still hold the crown in gaming performance, but Intel would have been able to pick up a lot of lost steam in other areas and continue to solidify leads in content creation, 3D rendering and encoding applications.

We're still being cautiously pessimistic about the types of performance gains we'll see from the upcoming 600 series of CPUs from Intel. As you may have already read, Intel is planning on doubling the L2 cache of Prescott and launching a new 600 series of CPUs next year. With twice the L2 cache Intel will attempt to get most of the benefits of an on-die memory controller, mainly reduced memory access latencies, without actually implementing one. We've seen the positive impact this can have with Intel's Extreme Edition chips, but even then, it may not be enough. Raw clock speed is what the Pentium 4's architecture was designed for, and only that will give Intel a commanding lead - unfortunately for them 3.8GHz does seem to be the end of the road for quite a while.

The performance paradigm will eventually shift to being more depending on multithreading capabilities, but that transition is far from being complete, especially on the desktop. It may end up being that Longhorn in 2006 is when we start to reap the benefits of more than just clock speed with every processor release.

Right now we couldn't be happier with AMD, they are more on top of their game today than they ever were with the Athlon XP and the Athlon 64 platform is by far the most attractive platform AMD has ever had. We've seen AMD offer leading performance in the past, but never have they commanded such a strong lead for such an incredible length of time. If AMD could have repeat of 2004 next year the few companies that still don't take them seriously enough may finally come around.

Power Consumption Comparison
Comments Locked

42 Comments

View All Comments

  • Anand Lal Shimpi - Sunday, November 14, 2004 - link

    ThePlagiarmaster

    Thanks for your post, whew it was a long one, I'm going to try to respond to all of it but forgive me if I missed something.

    1) The issues with SYSMark were with SYSMark 2002, SYSMark 2004 is an AMD sanctioned benchmark. AMD had input in the development of the benchmark and had equal voting rights to Intel in the benchmark as well. You can email AMD to ask whether or not they support the use of SYSMark 2004 as a benchmark. AMD also uses SYSMark 2004 in their benchmark suite for determining the model numbers of their Athlon 64 and Athlon 64 FX processors, so the benchmark is clearly valid as a comparison point as it is used in the creation of AMD's model numbers.

    2) For DivX and XviD encoding benchmarks we chose Gordian Knot as it appears to be the preferred tool by the video encoding community. This is the very reason we dropped Xmpeg as it was a decent benchmark but not a tool most users actually used. If we see that DVD2AVI offers the features and tools that the community wants, we will use that as a benchmark. That's not for us to decide, that's for the users of the applications to decide.

    3) The Roxio test is a part of the WorldBench 5 suite, which you mentioned that you would rather us use instead of anything BAPCo produced. It would not make sense for us to only use specific benchmarks in a suite that favored AMD processors.

    4) The Nero test is also a part of WorldBench 5, that specific application was not one we chose but one that is a part of the WorldBench suite. See the rest of my comments above as you seemed to be fine with us using WorldBench.

    5) AMD won the same test in 3dsmax, I mentioned in the review that the two tests are identical, just different APIs. Once again, this is a part of the WorldBench suite, not my decision to include two different API tests.

    6) The reason we didn't report the individual scores that make up the rendering composite score is because all of the charts show the same thing - the rendering composite score is a geometric mean of the four rendering tests, so it is representative of the individual tests. If you are interested, here are the Athlon 64 4000+ and P4 570J scores:

    A64 4000+: 20.968 34.766 169.562 232.516
    P4 570J : 20 29.515 144.969 197.391

    Those are render times in seconds for the following sample files included with 3dsmax 6: 3dsmax5.rays, CBALLS2, SinglePipe2, UnderWater. As you can see, the P4 570J is faster in every single test. You should not immediately assume that my reasons for doing things a certain way is to make one company look better, I assure you it never is. I could really care less about which company does better so long as the recommendations I make are the best for the reader :)

    7) Comparing application versions in SYSMark to Winstone isn't the most scientific comparison of which applications run faster on either platform. Although the tests may use the same applications, how they use them (and weight the results) may be completely different. Regardless how the weighting is done (BAPCo is the only company that makes this information public, we do not know how the final score is calculated in Winstone), these are real world uses of the various applications. You could argue that WorldBench shows AMD performing much better than Intel than Winstone does and thus Winstone is written by Intel and WorldBench is the only fair benchmark. Personally I don't put much faith in SYSMark 2004 but it does provide a real world usage model and it is supported by *both* AMD and Intel so I include it in our reviews for a more complete comparison. This is also why I include as much information as I can (whenever it's provided to me) about what the benchmarks are actually doing, so users can determine for themselves whether or not the benchmark is representative of a real world usage environment as far as they are concerned.

    I understand your concerns but I assure you that I'm not here to make Intel look better than they are - nor am I interested in making AMD look better than they are. I hope I've laid to rest all of your concerns, but if not let me know what I can do to explain my reasons for doing things further.

    Take care,
    Anand

  • Hans Maulwurf - Sunday, November 14, 2004 - link

    I want to underline what #12 said. A credible source like Anandtech should not use benchmarks like sysmark. If you don´t believe him use his links, its quite obvious sysmark is a cheater benchmark.

    #17 Call it cheating or not, but valid benchmarks should represent software people use. Of course you can use SSE3 in a benchmark, but you cannot use this benchmark to show performance differences between P4 and A64 in the real world. Although it is certainly usefull to show if using SSE3 can make sense.
  • SLIM - Sunday, November 14, 2004 - link

    #17, you missed the very long and ranting-like point of #12. Bapco is using the exact same software (MS office 2002, winzip, etc) as winstone and pcbench, but their benchmark scripts somehow completely reverse the results by exploiting p4 advantages. The connections/relationship between intel and sysmark has been questioned repeatedly over the years, and many questions were never answered. Those facts as rehashed by ThePlagiarmaster have led me over the years to not put any faith in the sysmark suite (especially as it ballooned from 2 graphs/benches to SIX).

    At any rate, I don't do 12 excel sorts in a row while encoding divx files and repeatedly checking for viruses. I write papers, do research and game on my system, so the answer is still quite clear for me:
    cheaper processor, memory, motherboard + higher performance, free 64 bit upgrade next year= AMD64
  • mlittl3 - Sunday, November 14, 2004 - link

    #14, Plagiarmaster doesn't care if you read his long post. He is hoping that Anandtech reads the post, considers the info and rethinks their benchmarking strategy. Although a bit harsh, Plagiarmaster is using this forum for its purpose, constructive criticism. Yes, it is AMD favored criticism. However, it is Anandtech's job to set the favortism aside and glean any validity out of the comments in order to provide us with more accurate journalism.
  • bofkentucky - Sunday, November 14, 2004 - link

    Is it cheating that Bapco utilized all the instructions on the chip? This is a purely synthetic benchmark we are talking about, raw performance. If Intel has written an instruction that makes multipling matrices together they should get credit for it. AMD's 3DNow instructions back in the K6 days helped one of my professors get a Top 500 supercomputer built. Some MMX, 3DNow, and SSE instructions are useful, if you program, use them to help speed your app up.
  • Myrandex - Sunday, November 14, 2004 - link

    lol @ 14
  • skunkbuster - Sunday, November 14, 2004 - link

    toasty
  • IamTHEsnake - Sunday, November 14, 2004 - link

    #12


    Your post is too long. I won't bother reading it.
  • MasterYoda - Sunday, November 14, 2004 - link

    I wonder how hot it runs.
  • ThePlagiarmaster - Sunday, November 14, 2004 - link

    I have to wonder about the benchmarks and some of the statements here. Lets start with the fact that Intel appears to only win tests from bapco. If you look at the same types of apps from the winstones you see a completely different picture (even when later versions of the apps are used, dreamweaver mx etc). Let us not forget that Bapco lives on INTEL OWNED LAND (cheaper rent for favored benches eh?). Let us not forget that INTEL OWNED BAPCO's DOMAIN NAME (until too much press caused them to hide this). Let us not forget that INTEL WROTE CODE FOR BAPCO! Let us not forget that BAPCO didn't respond to PROOF that they cheated on sysmark 2002 (and 2001 as Van Smith showed ages ago regarding SSE cheats) in Intel's favor (AMD PROVED IT!). I could go on and on here but you can read all about it here: http://www.vanshardware.com/reviews/2002/08/020822...

    A quote from the article on AMD's take on sysmark: "AMD reluctantly admitted that due to BAPCo's nature as primarily a meeting facilitator, Intel itself has been providing software engineers for the development of the SysMark products. When pressed further, the AMD representative admitted that it is likely that all SysMark development so far has been conducted internally at Intel by Intel."

    The Inquirer (www.theinquirer.net) covered this heavily also. Even asked for Bapco to respond to the evidence of the cheating and hounded the subject to death. Bapco never responded one word. Why? They couldn't. There was no excuse, they were just caught. PERIOD.

    Why do most of the bapco tests have you scanning your system in the background? Nobody scans they whole system all the time. I haven't done it in years. Norton scans my email as it comes in, Instant messaging when its used and I personally scan what I download (when downloaded, and just that folder) and thats it. I still have a red X telling me I haven't scanned my whole system and have had a red X for years (over multiple OS installs on new pc's). Never had a virus yet, and yes I do a lot of dangerous stuff (newsgroups anyone? Eh, nefarious sites...LOL). You shouldn't be using known cheater benchmarks (written by Intel engineers no less). They should be dumped in favor of either your own benchmarks using popular apps and fully revealed scripts, or other benchmarks by companies that don't live on Intel land :rolls eyes:

    Next the encoding tests. Quit stating Intel things like "but Intel would have been able to pick up a lot of lost steam in other areas and continue to solidify leads in content creation, 3D rendering and encoding applications.". They only lead in Divx/Xvid because of your choice of frontends (autogk). If you use DVD2AVI as the front end and run your same clip through it you'll suddenly show AMD dominating by 20% or more (Check out this link at hardocp written only a few weeks ago showing AMD dominating- http://hardocp.com/article.html?art=NjgyLDM=). Your use of Roxio software is also suspect. Note they live in Santa Clara, CA (uh, Intel's there too). Roxio is on Intel's P4 presskits quotes page TWICE! Apparently it's the only video editing software Intel wins in eh? From their Financial page at roxio : http://investor.roxio.com/EdgarDetail.cfm?CompanyI... "We have also established strategic relationships with Intel and Palm, whereby we have optimized our software or created unique products to run on their hardware platforms and, in certain instances, engaged in unique marketing". Translation, Intel pays for their ads. Perhaps you should be using any ONE of the top 7 at pcmag.com shown here: http://www.pcmag.com/category2/0,1738,4835,00.asp Anyone else see how far down the list ROXIO is? It's the crap of the heap. So is their burning software (remember win2k systems dropping like flies out of the box with EZCD creator?...LOL - Long live NERO!). What made you pick their software (subpar audio and you use this?)?

    Your comments in the render stuff don't make a lot of sense either. Claiming Intel rules here (is that AMD I see winning Maya?). AMD wins 3d Studiomax 5.1 TWICE. Then you go to v6 and say "For the next 3dsmax test we used version 6 of the program and ran the SPECapc rendering tests to truly stress these CPUs. Since there's not much new to report here we're only going to report the Rendering Composite score." So I guess v5.1 doesn't stress the cpu's at all huh? Only going to report the Rendering Composite score? Is that because it's the only part of the v6 tests that Intel won? Report the rest of the test so we can decide whether there's "not much new to report" or NOT. Slighting AMD victories where you can or what? I count 3 victories for AMD (2 smax 5.1's and maya) and one for Intel (smax6). CC winstone (which includes a lightwave test mind you) shows AMD winning. Sysmark (blah!) shows Intel winning Content Creation (but we know who writes these tests eh?). Why claim Intel the leader in content creation when at best it's even (CC Winstone for AMD, Sysmark for Intel), at worst sysmark is lying BS artists that we all know about (Intel, heh) and AMD is the winner? While AMD wins in almost everything, you seem to be blowing up the Sysmark stuff with a 5000x microscope for all to see. Each sysmark test even comes with a detailed explaination. Are you hoping we'll all forget the BS of the past (and present, futuremark anyone?..ROFL) and buy this crap wholesale? You seem to be selling us sysmark results heavily. Even breaking it down for us bit by bit.

    Let me break down their results further. Lets see, MCC Winstone 2004 (note that its MULTITAKSING also!) shows us AMD winning a suite of Photoshop 7.01, dreamweaver MX, Premier 6.5, DirectorMX etc even includes LIGHTWAVE and Media encoder v9! When bapco writes the test (ICC Sysmark 2004) Intel wins 3dSmax 5.1+DreamweaverMX combo and come vector deal. We know worldbench shows AMD winning 3dsmax5.1 TWICE! We know MCC winstone shows DreamweaverMX/DirectorMX with AMD winning. Hmmm.

    On to 2D stuff we see MCC shows the exact version of Photoshop (7.01) and Premier exact version (6.5) with AMD winning. But when Bapco writes the test (2d Sysmark 2004) Intel dominates.

    Look at the Web Sysmark 2004 test. Cache shows no difference on the A64, nor the P4. A64 I understand (on die memory controller), but the P4EE should show some difference over equal clocked NON EE chips. So clearly the script is about keeping the bandwidth/long pipelines of the P4 full. This reminds me of the 12 repeated excel sorts in sysmark 2002 etc. You spend most of your time in excel entering data, not sorting it repeatedly.

    Granted you say AMD is the way to go in the end. But nearly every Intel victory is can be proven as the complete opposite by another website (hardocp and encoding), or your own tests when written by someone other than bapco (worldbench and Winstones both show opposites to bapco stuff). Yet you claim Intel leads these areas. Again I say, at best a tie, or at worst Bapco's full of crap and AMD's the winner in these areas too. Maybe it's just the way I'm reading this and I'm completely off base (hard for me to believe that). But clearly you can show AMD winning in encoding with DVD2AVI (as hardocp did, 20% victory). Clearly all Bapco stuff can be contradicted easily. Any statement about Intel leading a category, when I can show you results that show the complete opposite is misleading your audience. I wouldn't even have written all this stuff if you would have at least said you can show the complete opposite and given examples of it. At least you should have said you wonder about these results when knowing all of this, instead of claiming them leaders in these areas. I hope you retract your 'Intel leads the this or that' comments and restate this stuff. I just showed these statements to be flawed at best. I could go on and on with showing other results, but I think I made my point already.

Log in

Don't have an account? Sign up now