Athlon64 3400+: Part 2

by Wesley Fink on January 12, 2004 2:59 PM EST

Gaming and Media Encoding Performance



One of the biggest questions is how the $430 3400+ compares to the $750 Athlon64 FX51 in gaming. With both running at the same speed and the same 1Mb of cache, we can finally see what impact the different memory architectures have on performance. The only real difference in the 3400+ and the FX51 is, after all, the Dual-Channel Registered memory used in the FX51 versus the Single-Channel used for the 3400+. While the FX51 is still faster, the real difference between the 2 chips is next to nothing in X2 and Unreal Tournament Flyby. The Dual-Channel memory does make a larger difference in Media Encoding, Halo, and UT Botmatch. It is fair to say that at about 60% of the price of the FX51, the 3400+ performs very well indeed, coming very close in many gaming benchmarks.

It is interesting that all of the Athlon64 chips outperform the Intel Pentium 4 EE in gaming, except for the comparable performance in the older Quake 3. This again demonstrates the superiority of the A64 as a gaming platform. The one area where Intel dominates is Media Encoding, where the 3.2EE, 3.2 and even the 3.0 lead all Athlon64 family chips. While the Dual-Channel FX51 does compare much better to Intel in Media Encoding, the Intel P4 family is still the top choice for Media Encoding. Also, don't overlook the fact that the 3400+, in particular, and the Socket 754 A64's, in general, have made giant strides in Media Encoding performance compared to the older Athlon XP processors.

The Pentium 4EE does perform much better in gaming than the lower-priced 3.2 and 3.0, and the gap is much smaller when the Athlon64 is compared to the P4EE. However, unless your budget can handle $1000 for the CPU, the Athlon64 is your best choice for a gaming rig. In the Athlon64 family, the FX51 is the best performing processor in gaming, but it is also a pricey $750. The 3400+ provides most of the same gaming performance for about $430.

As we have seen in other recent GunMetal 2 tests, the benchmark seems to be video-card bound and it tells us very little about the performance differences in processors or systems. While GunMetal 2 may be useful for Video Card comparisons, it is not a good tool for processor or system performance measurements and will be dropped from future system testing.

Content Creation and General Usage Performance Workstation Performance
Comments Locked

20 Comments

View All Comments

  • atlr - Thursday, January 22, 2004 - link

    Anyone seen any performance comparisons of 32-bit versus 64-bit versions of software and o/s on the A64?
  • milehigh - Tuesday, January 20, 2004 - link

    I'd like to 2nd #13's reply to include some older CPU's in these reviews. I've got a Barton 2500+ and seeing how it stacks up can help in not only help in upgrade decisions but I think it can help illustrate just how much faster these new CPU's are...

  • KingofFah - Thursday, January 15, 2004 - link

    I would be careful with most 350's, but, like #15 said, most FSP's (no matter which brand is relabeled on it) are marked much lower than what they are capable of doing. THG did a psu round up a while back showing that the FSP-300 was really capable of being completely stable at 390W consumption and the 350 (of which I am a owner) was capable being completely stable at 454W. I have not seen a review of the FSP-400 yet, but I am sure it follows the same trend as its predecessors. Most PSU's run very little over (or even under if it is a cheap one) their specified values, but Sparkle goes well over them.
  • TrogdorJW - Tuesday, January 13, 2004 - link

    As for dropping Quake 3, how about checking out this, first:

    http://speedycpu.dyndns.org/opt/

    I've read (from X-bit Labs) that the optimized DLLs boost Athlon XP/64 performance by 13 to 18 percent. Wow!

    See:

    http://www.xbitlabs.com/articles/cpu/display/athlo...

    One last word of caution, though, is that if the DLLs in question are binary compiled as opposed to interpreted code, then id software's Jon Carmack says they are more open for cheats. In addition, there is the fact that a binary compiled DLL is already said to boost performance by up to 20%. Not sure about all this, but here's a last link if you want:

    http://techreport.com/onearticle.x/336
  • TrogdorJW - Tuesday, January 13, 2004 - link

    #15, as the review states, they could not get system stability with a 350W quality power supply and the 3400+. Maybe you have a better PS than their 350W, but I wouldn't count on that.

    There are those that claim the Prescott will be a flame-thrower. Maybe. What we know for sure, though, is that the 3400+ has raised the bar in power requirements. Looks like 450W PS will become the norm in the next year....
  • sprockkets - Tuesday, January 13, 2004 - link

    You can't go wrong with a 350 watt FSP-350PN power supply, from either Sparkle or Forton Power Source, with it's 12cm fan. Works fine for my system I built with a 3200+ and gf4 4600, soon to be 9800 ATI card.
  • rms - Tuesday, January 13, 2004 - link

    Congrats on the great article. Please STOP USING QUAKE3-BASED GAMES AS A CPU BENCHMARK. It doesn't recognize athlons as SSE-enabled, and is worthless for cross-platform comparisons.

    rms
  • clv101 - Tuesday, January 13, 2004 - link

    What I'd really like to see in reviews like this are some slower systems - I'm fed up with seeing graphs showing 6 cpu with only a few % performance difference.

    I'd like the see the A64 3400+ and P4 3.2 benchmarked against a XP 2500+, a 2.4GHz P4 and my old 1.33GHz TBird. That would be useful to see.

    Seeing that the A64 3400+ is a little bit faster than a A64 3200+ is no good to anyone!
  • Wesley Fink - Tuesday, January 13, 2004 - link

    PrinceGaz -

    You are correct, but I had to return the initial 655TX and just received the shipping version of the P4S800D-E the day we left for CES. So I did not have the board available for the full roundup.

    I did run the P4EE through the 655TX to check benchmarks and it is faster by a small amount in almost every benchmark. However, it does not change any of the positioning or conclusions.

    #10 - I could not find the list either, since it looks like AMD has stopped the PS list for the Athlon64 and replaced it with "Athlon64 Power Supply recommendations" which are just general guidelines. The best source of info on compatible PS then, will likely be Power Supply reviews by AnandTech and others.
  • PrinceGaz - Tuesday, January 13, 2004 - link

    Its nice to have a clear comparison of how the A64 and A64FX compare with the top P4's including the P4EE.

    One question though, shouldn't an article which "tests the top CPU's from Intel and AMD on the top-performing motherboards that we have tested for each platform" use an SiS 655TX rather than Intel 875P mobo for the Intel chips when your own review last month found the 655TX to be faster than the 875P in every single test?

Log in

Don't have an account? Sign up now