|
Advice Beginners BIOS Guide CPUs Links Mainboards Memory Network Storage Video/Sound Cards Contact Forum SiteMap Sponsors WebNews Home |
. | . |
|
Prices: CPU Intel P4 2.4C-800 P4 2.6C-800 P4 2.8C-800 P4 3.0-800 P4 3.2-800 AMD AthlonXP XP 1700+ XP 2000+ XP 2400+ XP 2500+ XP 2700+ XP 3000+ XP 3200+ Athlon64 Athlon64 3200+ Athlon64 FX-51 Opteron Opteron 240 Opteron 242 Opteron 244 Opteron 246 Memory Corsair Crucial Kingston Mushkin OCZ |
LOSTCIRCUITS |
|
|
| AMD MPX Production Boards ASUS A7M266D vs. MSI K7D-L vs. Tyan Tiger MPX: Dual Power at 1900+ | ||
| (Review by MS, March 11, 2002) |
Stability and Overclocking
Both the MSI K7D-L and the Tyan Tiger MPX were working flawlessly at stock settings. In fact, both boards were burnt in over a 4 week period running RC5 in the background while being tormented with any variety of application. The big surprise here was the ASUS A7M266D which initially showed a variety of glitches. That is, applications like MDK2 deveolped errors and 3DMark2001SE would crash every single time. After beating my head against the wall for a few days, including a few format c: and reinstalling all applications, it turned out that the AGP aperture on the A7M266D is by default set to 32 MB, causing texture jams in the AGP pipeline. Increasing the AGP aperture to 64 MB and setting AGP mode to Uncached Speculative Write Combining (USWC) solved all stability problems as well as the issues with 3DMark2001 and MDK2, even when Turbo Mode was selected.
Overclocking
Both the Tiger MPX and the K7D-L offer bus speed settings up to 150 MHz with the Tiger MPX giving the better granularity of the two boards. Still, both boards maxed out at 140 MHz FSB (using registered DIMMs), with unbuffered DIMMs that should have a higher margin, we found that the highest FSB to warrant stable operation was 138 MHz. The fact that both boards started dying at the same clock speed, independent of the memory used, points to the chipset as the limiting factor. The ASUS A7M266D offers 1 MHz increments starting at 100 MHz which is somewhat of a nuisance since changing the FSB has to be MHz by MHz, that is without the possibility of scrolling. In the end, the board would clock up to 141 MHz but that's as high as we could get.
Performance
In most standard applicactions, and I think I mentioned this earlier, the MPX chipset, even in an SMP configuration, will not match the performance of a high end single CPU system. The main reason is that, e.g. the VIA KT266A chipset have a more powerful memory controller and, thus, offer higher memory bandwidth that, after all is the bottleneck in standard desktop applications. Also mentioned earlier, there are applications, though, particularly 3D rendering that will take full advantage of dual CPU configurations, my favorite is Caligari TrueSpace 5.1. A detailed description is available here
To start with the obvious, we are going to take a look at some synthetic SiSoft Sandra benchmarks:
Memory Bandwidth
For those who are not familiar with the modus operandi of SiSoft Sandra Memory Benchmark, here is the short run-off. Most modern chipsets and CPUs feature prefetch buffers that can be used to load data into the chipset buffers in anticipation of a coming request. Using this algorithm based on locality and other "branch prediction" parameters, the top performance in streaming memory applications can be as high as 95% of the peak bandwidth. This scheme is used by SiSoft Sandra when "buffering is enabled". Prefetching can also be turned off within the benchmark itself and results in scores more realistic for non-streaming applications.
Memory benchmarks at the "By SPD" setting as well as after pushing CAS latency to 2. The Tyan Tiger MPX does not offer any timing adjustments and, therefore, cannot compete with the ASUS and MSI boards. Note that enabling Turbo instead of Normal on the A7M266D does not buy too much.
next page: => more benchmarks from SiSoftware =>