MSI K8T Neo2-FIR: Overclocking and Stress Testing

FSB Overclocking Results

Front Side Bus Overclocking Testbed
Processor: Athlon 64 FX53 Socket 939
2.4GHz
CPU Voltage: 1.5V (default)
Cooling: Thermaltake Silent Boost K8
Power Supply: Antec TruePower 430W
Maximum OC:
(Standard Ratios)
200FSB x12
2600MHz (+10%)
Maximum FSB:
(Lower Ratio)
263FSB x 9 at 1:1 Memory

While the K8T Neo2 reached the next performance level of 2.6GHz at a 13 multiplier, it was not the CPU frequency overclocker that we see in the nVidia-based K8N Neo2. No matter the multiplier, the top frequency that we could reach with stability was 263, much lower than the 290 on the K8N Neo2, or the 280 reached on the Abit AV8.

Memory Stress Test Results:

The memory stress test is very basic, and simply tests the ability of the K8T Neo2 to operate at its officially supported memory frequency (400MHz DDR), at the best performing memory timings that our Mushkin PC3500 Level 2 or OCZ PC3500 Platinum Ltd Modules will support. Memory stress testing was conducted by running RAM at 400MHz with 2 DIMM slots in a Dual-Channel configuration with a Command Rate setting of 1T.

Stable DDR400 Timings - 2 DIMMs
(2/4 DIMMs - 1 Dual-Channel Bank)
Clock Speed: 200MHz
Timing Mode: N/A
CAS Latency: 2.0
Bank Interleave: N/A
RAS to CAS Delay: 2T
RAS Precharge: 10T
(10T for Best Performance)*
Precharge Delay: 2T
Command Rate: 1T
*Several memory tests have shown that memory performs fastest on the nVidia nForce and VIA K8T800 chipsets at a TRas (RAS Precharge) setting in the 9 to 13 range. We ran our own Memory Bandwidth tests with memtest86, with TRas settings from 5 to 15 at a wide range of different memory speeds. The best bandwidth was consistently at 9 to 11 at every speed, with TRas 10 always in the best range at every speed. The memory bandwidth improvement at TRas 10 was only 2% to 4% over TRas 5 and 6 depending on the speed, but the performance advantage was consistent across all tests. Since best performance was achieved at 2-2-2-10 timings, all Athlon 64 benchmarks were run at a TRas setting of 10.

The K8T Neo2 had no problem at all with 2 dimms at aggressive 2-2-2-10 timings and a 1T Command Rate. We were frankly surprised at how well all the boards handled our BH5 memory, since it has sometimes been a problem memory on Athlon 64 boards. We do have to wonder if this is really the memory, however, or just the truly excellent memory controller on our replacement FX53.

Filling all four available memory slots is more strenuous on the memory subsystem than testing 2 DIMMs on a motherboard. We really expected to have to make concessions to memory timing with 4 DIMMs on all the 939 motherboards, but in fact, every board in the roundup had no problem with 2-2-2-10 timings at 1T.


Stable DDR400 Timings - 4 DIMMs
(4/4 DIMMs - 2 Dual-Channel Banks)
Clock Speed: 200MHz
Timing Mode: N/A
CAS Latency: 2.0
Bank Interleave: N/A
RAS to CAS Delay: 2T
RAS Precharge: 10T*
Precharge Delay: 2T
Command Rate: 2T

The MSI K8T Neo2 was able to run with all 4 DIMM slots at the same aggressive 2-2-2-10 settings used for 2 DIMMs. As we have seen on the other 939 boards running all 4 DIMMs, the Command Rate must be reduced to 2T when filling both Dual Channels.

MSI K8T Neo2-FIR: Features and Layout Performance Test Configuration
Comments Locked

83 Comments

View All Comments

  • thebluesgnr - Tuesday, July 20, 2004 - link

    "Our FX53 topped out at about 3.59 GHz on the ECS KV2, which is slightly below the 3.6+ achieved on the top 939 boards."

    Is this ECS a P4 board? :P

    This was a great article. I agree with other readers, CnQ should definately have been tested, as well as audio and IDE subsystems.

    btw Wesley, will there be reviews of KT880 socket A mobos in the future?

  • TrogdorJW - Tuesday, July 20, 2004 - link

    Just a few comments. It's a little (*little* mind you) unfair to compare FX-53 to P4 560, given the price advantage of the P4. Then again, comparing it to the P4EE is a little unfair in the other direction. It might have been nice to include one or two other systems in the benchmarks, though, like a 3400+. Sure, we can cross-reference other articles, but if you have all the data already it would be a lot cleaner. I'm especially interested in seeing AutoGK benchmarks with the "lesser" Athlon 64 processors (3500+ and 3400+ would be good, or maybe even 3200+ - not everyone has $400+ to spend on a CPU!)

    Of course, while it might be less fair to Intel, I would like to get AutoGK numbers using Xvid as well. That's how I use it, as I feel the quality is a little better than DivX. Oh, and while you state that you used 2-pass encoding, what was the target resolution? 640x360, or 720x408, or something else? And did you specify a target size, or was it on unlimited quality? All those are important questions, I think.

    One final request: I truly appreciate the memory stress testing benchmarks. However, I would like it taken a little further. All of the boards claim that they can support up to 4 GB of RAM. I would love to see some tests showing this configuration. After all, 64-bits is really about breaking that memory barrier. Even if the boards need to run 4x1GB at DDR266 or DDR333, it would be good to know. (Too bad there simply aren't many good 1 GB DIMMs available yet.)
  • Wesley Fink - Tuesday, July 20, 2004 - link

    #14 & #20 - Perhaps tests with the X800 XT on nF3 compared to the 6800 Ultra will at least shed a little light on where the efficiencies lie - in the nF3/nV Video combo or in the nF3 itself.

    #16 - We will make an effort to talk a bit more about Cool'n'Quiet in individual board reviews, but in a roundup like this it is difficult to explore that level of detail, and still hold the article length to anthing you might want to read. We try to do more with features in individual reviews.

    #19 - We report the full range of vCore in our board charts for people like you who are interested in umdervolting. If you notice some boards begin vCore at default, while others make a wide undervolt range available as well as overvolt. We try to report this range as accurately as possible for this reason.
  • Pete - Tuesday, July 20, 2004 - link

    Wesley, very interesting numbers. Halo is supposedly limited by some inefficient DX9 layers/commands, so I at first thought maybe nV had somehow optimized or bypassed some DX9 calls. The office and Content Creation benchmarks advantage is more puzzling, though. Could nV's performance edge be the result of some intelligent caching or either the HD or the CPU?

    Testing an X800 for reference is a good idea for Halo. Just be sure to retest the office and Content Creation suites, too, as the performance boost there is equally curious, IMO.

    I found one typo, on the system specs page. It's Wolfenstein: Enemy Territory, not Wolfenstein: Enemy Within. :)
  • JKing76 - Tuesday, July 20, 2004 - link

    Sure there's Cool'n'Quiet, but how about adding manual undervolting capabilities to the review? A lot of mobos only allow upping the vcore, but undervolting is a great tool for creating a truely cool and quiet system without losing performance.
  • XRaider - Tuesday, July 20, 2004 - link

    Love that FX53 and the MSI K8N Neo2 together. Sure is purdy nice!! ;) But must...hold...out...until...price..drops...some..more.. ;o)
    Great article BTW!
  • XRaider - Tuesday, July 20, 2004 - link

    Love that FX53 and the MSI K8N Neo2 together. Sure is purdy nice!! ;) But must...hold...out...until...price..drops...some..more.. ;o)
    Great article BTW!
  • jojo4u - Tuesday, July 20, 2004 - link

    I miss information about Cool'n'Quiet. It's a shame that anandtech.com only is insterested in overclocking and speed.
  • esSJae - Tuesday, July 20, 2004 - link

    Nice, another article touting the non-existent MSI K8N Neo2.

    Sure, you can go to MSI's Taiwan site and download the manual and BIOS, but doesn't seem to be much point in that.
  • DAPUNISHER - Tuesday, July 20, 2004 - link

    "We never expected the nVidia nForce3-250 Ultra to be a better performer in Winstone benchmarks than the VIA K8T800 PRO. However, both the nF3-250 boards are outperforming the VIA boards by a significant percentage. Since the nVidia 6800 Ultra video card was used for all benchmarking in the roundup, we plan to verify these results with an ATI X800 XT as soon as that board is available to the Motherboard Lab for testing"

    Is this to determine if it's a result of forceware opts that is responsible for the difference observed?

Log in

Don't have an account? Sign up now