Test system setup and notes:

• CPU: Intel Pentium 4 630 @ 4GHz
• Motherboard: Asus P5WD2 Premium @ 266MHz FSB
• HDD: Seagate ST3250823AS 250GB SATA
• PSU: Be-Quiet P5-470W
• VGA: Albatron PCX5750 PCI-E 128MB
• OS: Windows XP SP2

With the above configuration we tested the capabilities of the modules under different timing settings and received the following results:

dont have flash installed? then click here for jpeg

All of the tests were run with the voltage set at 1.9V in the BIOS, except the last row where we set the voltage at 2.2V. Over 1000MHz our system would randomly error regardless of the voltage and timing settings, so we concluded that the modules may have even better overclocking capabilities, but unfortunately our system does not. The stability of the modules was insured by running Memtest for at least 30 minutes. In order to keep our results comparable, we set the FSB at 266MHz which means that our CPU is running at 4GHz and used the DRAM multipliers to alter the speed of the modules. Unfortunately this method does not allow us to run the RAM modules at their limits, but at a specific speed close to them nevertheless.

667 mhz

800 mhz

890 mhz

1000 mhz

By using the 4:5, 2:3 and 3:5 dividers we had the modules running at 667MHz, 800MHz and 890MHz accordingly. In order to run the modules up to 1000MHz, we had to use the 1:2 divider and reduce our FSB speed to 250MHz. Unfortunately since the speed of the RAM is directly connected to the FSB and CPU speed and both were significantly reduced, performance was severely degraded and so we decided not to include test results in our charts as comparison with the rest of the achieved results would not be a true reflection of the systems performance.

dont have flash installed? then click here for jpeg

 

dont have flash installed? then click here for jpeg

Sisoft Sandra is a popular synthetic benchmark, commonly used when testing the speed of RAM modules. When testing the RAM buffered, the overall speed is heavily affected by the CPU and FSB speed. Testing the RAM unbuffered will provide much more accurate results, but beware that Sisoft Sandra is a synthetic benchmark and does not represent real-world performance. We see that despite the lower timings, the considerably higher speeds always increase RAM performance by a substantial percentage.