Navigate: 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 Search Prices:

There were literally no problems with the installation of the entire system except for the somewhat counterintuitive behavior of the Promise controller. Everything else, from USB to IEEE 1394 and Gigabit Ethernet controller fell right into place on a fresh install. With a carry over of the operating system and configuration from the A7V333, the only issue was the correct recognition of the BroadCom Gigabit network controller, which took three attempts before it was recognized correctly.

Memory Configuration Issues

Above, we briefly touched upon the issue of different memory densities supported by the board dependent on memory bus frequency, i.e., support for 3 dual bank modules at PC2100 speed dwindling down to a single module in PC3200 mode

In our hands, the board was working just fine using two 512 MB PC 3200 DDR modules at PC2700 speed as long as DDR slots 1 and 3 were used, however, three double-sided modules at that speed is already pushing it. Overall, the limitations fall more into the genre of a disclaimer than anything that is set in stone or would involve bank sensing or similar mechanisms used e.g. by Intel with their i845 chipset boards since we were able to run 2 x 512 MB of Mushkin PC3200 even at 2:2:2 timing settings with only minor problems as e.g. temporary slowing down or stuttering of benchmarks. On the other hand, running at DDR400 speed, at least with a 266 MHz FSB CPU does not necessarily increase performance as we will show in the benchmark section below.

With a single PC3200 DIMM running at 400 MHz data rate and 2:2:2, there were no issues at all which shines a rather favorable light on the overall quality of both the board and the chipset, regardless of whatever demo system kept crashing at QuakeCon, which could have been caused by anything, including line-power fluctuations.

Performance

SiSoft Sandra

With the 266 MHz FSB being a memory bottleneck for streaming applications, it is really a no-brainer that we won't be able to see any differences at least if buffering is enabled. Both the DDR 333 and the DDR 400 are saturating the potential bandwidh of the FSB completely, in other words, 2100 MB/sec is the theoretical maximum and page boundaries within the DRAM will do their share to introduce a few page misses that will show up as delta between the theoretical maximum and the real score.

SiSoft Sandra2002 memory benchmark scores for Integer (grey) and Floating Point (red) at PC2700 and PC3200 setting with buffering enabled. In both cases, latencies were set to 2:2:2:6 with 1T CMD Rate. Interestingly, regardles of how often we repeated the benchmark, we saw a small lead for the PC2700 mode compared to the PC3200 mode. The only possible explanations are that either PC3200 mode also changes some chipset latencies or else, the greater mismatch between FSB and memory bus frequencies adds an aditional small penalty. Note that the PC2700 Integer scores are higher than the theoretical maximum of 2100 MB/sec. We checked and found that at default setting, the A7V8X actually runs a 134.5 MHz FSB instead of 133 MHz which could explain the higher score.

If buffering is disabled, the overall bandwidth drops and the FSB is no longer the prevalent bottleneck. Therefore, we would expect a small lead for the DDR 400 (PC3200) mode.

SiSoft Sandra2002 memory benchmark scores for Integer (grey) and Floating Point (red) at PC2700 and PC3200 setting with buffering disabled. In both cases, latencies were again 2:2:2:6 with 1T CMD Rate and, as expected, we see a small performance edge for the PC3200 mode.

next page:    => PCMark2002, 3DMark2001SE, Comanche4 =>