Review: Mushkin EMS PC3500 Memory

by Tarinder Sandhu on 10 April 2003, 00:00 4.0

Tags: Mushkin

Quick Link: HEXUS.net/qaqw

Add to My Vault: x

Test setup and initial impressions

A quick rundown on the test setup.

  • Intel Pentium 4 1.7GHz Mobile CPU capable of 215FSB+

  • ABIT BH7 i845PE motherboard capable of 215FSB+

  • 256MB Mushkin PC3500 Enhanced (EMS) memory

  • ATi Radeon 9800 Pro (380/680) with 6307 drivers

  • Samcheer 420w PSU with tweaked voltage lines

  • Memtest86 v3.0

  • SiSoft SANDRA 3.1

  • Pifast v4.1

  • 3DMark 2001SE v330

How to test a module that natively is specified to run at 217MHz (DDR-433, PC3500) ?. To eliminate the testing MHz ceiling, I ran a mobile 1.7GHz Intel Pentium 4 CPU on an ABIT BH7 i845PE motherboard. Both the motherboard and CPU have seen 215FSB with stability. Further, the ABIT's FSB-to-DRAM ratios dictate that testing can be undertaken up to 286MHz (DDR-573) through the use of the 3:4 FSB:RAM ratio at 215FSB.

Initial stability testing was done at the weakest timings the BH7 could offer of 2.5-7-3-3. The CPU was run at 162.5FSB with a 3:4 ratio, giving DDR-433 speeds. As expected, the Mushkin module ran flawlessly. That's a relief when we're dealing with high frequencies.

To then test the RAM's potential, I simply tightened up my timings to the preferred 2-7-2-2. Raising the voltage to an acceptable 2.75v, the module managed to run at exactly the same frequency. This may not sound impressive, but with tightened, performance-enhancing, low latency timings, it is.

To show what kind of benefit tighter timings can make, I'll run the Mushkin at both 2.5-7-3-3 and 2-7-2-2, at DDR-433. That's on a mobile P4 1.7GHz CPU that defaults down to 12x multiplier on a desktop motherboard. So 1950MHz (12 x 162.5FSB, DDR-433 through the 3:4 ratio). Running SiSoft SANDRA, Pifast and 3DMark 2001SE should show a benefit for the lower latency timings.