PQI PC4000 Turbo Memory
Testing and Benchmarks Cont:
Here’s the results of all the benchmarks for each of the FSB levels attained. As you’ll beable to see with my test system, there is a point of deminishing returns when it come to dropping the CPU multiplier to attain a higher FSB. After I had to drop the multiplier to 9, I never did catch up to the scores I had with the multiplier set at 10. But the memory was more than capable of going farther.
Sandra 2005
| FSB |
memory bandwidth Int Buffered iSSE2 |
memory bandwidth Float Buffered iSSE2 |
|
10×200 2-2-2-5 |
5670 |
5594 |
|
10×250 2,5-3-3-7 |
7037 |
6931 |
|
10×260 2,5-3-3-7 |
7320 |
7221 |
|
9×270 2,5-3-3-7 |
7167 |
7036 |
Sandra showed the bandwidth falling off after the multiplier drop, but the ram was still going up in FSB.
Everest 1.51.195
| FSB |
memory read |
memory write |
memory latency (ns) |
|
10×200 2-2-2-5 |
5875 |
2021 |
45.0 |
|
10×250 2,5-3-3-7 |
7139 |
2444 |
40.5 |
|
10×260 2,5-3-3-7 |
7492 |
2678 |
39.6 |
|
9×270 2,5-3-3-7 |
7568 |
2563 |
38.2 |
Everest showed the band width coming back faster than Sandra, but still wasn’t totally there by the time the system maxed out.
PC Mark 2004
| FSB | PC Marks |
|
10×200 2-2-2-5 |
4199 |
|
10×250 2,5-3-3-7 |
5122 |
|
10×260 2,5-3-3-7 |
5297 |
|
9×270 2,5-3-3-7 |
4979 |
With PC Mark the trend follows the other benchmarks.
3d Mark 2001 SE
| FSB | 3D Marks |
|
10×200 2-2-2-5 |
20525 |
|
10×250 2,5-3-3-7 |
22870 |
|
10×260 2,5-3-3-7 |
23228 |
|
9×270 2,5-3-3-7 |
23017 |
3D Mark has just about caught back up, If I had been able to get one more increase, I would have been able to beat the old high mark.
Comments are closed.