Bus Speeds and Synthetic tests
The first test's on the module were to determine the maximum speed obtainable:
TwinMOS PC3200 |
Corsair XMS 3202 Platinum |
|
| Cas 2.5, 3-6-3-2T | 207 MHz |
212 Mhz |
| Cas 2, 3-6-3-2T |
201 MHz |
212 Mhz |
| Cas 2, 2-6-2-1T | 193 Mhz |
208 Mhz |
As you can see the module exceeded its specification and achieved
a maximum bus speed of 207 Mhz at its rated timings, and still managed to do
212Mhz at more relaxed timings. The Corsair XMS3200 outperformed it by a noticeable
margin, but to be fair the corsair ram is considerably more expensive and is
rated for better memory timings. The corsair may have done higher than shown,
but with only a 2/5th AGP divider I could not do anything higher than 212 MHz
without the AGP bus flaking out.
Benchmarks
To go with the 'flow', I've included the obligatory Sisoft Sandra memory bench's. I personally find this benchmark a bit flawed as it is hardly effected by memory timings, where in reality memory timings matter quite a lot and can be seen in real world gaming benchmarks.
As expected, processor speed has very little to do with this bench mark. A score of 3GB/s is one of the highest you will see on an AMD platform, the DDR400 on a synchronous 200 MHz bus really sets itself apart from the default 133 MHz bus, coming in over fifty percent faster.
A test I find more useful in gauging effective system bandwidth is the PiFast Bench. The Hexus PiFast benchmark consists of running one of the fastest PI calculators available for the windows platform, using the following settings:
Calculate PI to 10,000000 places using the Chudnovsky method, standard mode with an FFT size of 1024k.
Here you see just how dependent on system bandwidth this test really is. The 1.5ghz setup, although coming in as the slowest, was only 2.5 seconds behind the 2ghz setting with a 133 bus. This test also shows that memory timings do matter, with the fastest 2 setups having a noticeable difference based on memory timings alone.
If you think the 1.5ghz setup was close to the other's, wait until you see what happens with game benchmarks....
