September '98 PC100 SDRAM Comparisonby Anand Lal Shimpi on September 13, 1998 11:47 AM EST
- Posted in
With the basics explained, we are placed at the beginning of this rabbit chase, how do we compare SDRAM of nearly equivalent quality? First of all, we must pick the test system we'll use in order to illustrate any unstable operation coming directly from the SDRAM, meaning the components used in the system must be operating in a nearly perfect harmony with one another. The motherboard used would have to be one that supports virtually all available bus frequencies, and the BIOS would have to allow manual configuration of the SDRAM CAS Latency. The system AnandTech came up with? Quite simple in design, but quite stable in operation, let's take a look at the AnandTech RAM Test System:
|CPU||Intel Pentium II 400|
|Video||Matrox Millennium II PCI|
|Hard Disk||Western Digital 5.1GB Caviar UltraATA|
|Operating System||Microsoft Windows 98|
Keep in mind that the processor was only clocked at 400MHz during the tests at 66/100MHz, however for all other tests the clock multiplier was adjusted to ensure that the clock speed remained as close to 400MHz as possible without going over.
The ABIT BH6 features a 1/4 PCI clock divider when the 133MHz FSB is activated, so the PCI bus frequency remained as close to 33MHz as possible at all times eliminating that as a possible source of instability.
With the system ready, let's take a look at the contenders:
Update: 9/24/98 - The Memory Man submitted a more up-to-date revision of their Samsung (SEC) original PC100 SDRAM modules, this newer module is virtually identical to the Mushkin1 modules.
|Advanced Megatrends||Azzo||Corsair||Memory Man1||Memory Man2||Mushkin1||Mushkin2|
|PCB Manufacturer||Tanisys||Tanisys||Corsair||Lite On||Memman OEM||Lite On||Mushkin OEM|
|PCB Height||3.50 cm||3.50 cm||3.50 cm||3.50 cm||3.20 cm||3.20 cm||3.20 cm|
Running the Test
Ziff Davis' Winstone 98 was run in a repeat session, all benchmarks of the Business Winstone 98 suite were run except for the two Task Switching tests which have known issues with Windows 98 and would end up tainting the data results. Each memory module was run on a system in a continuous operation state for a matter of hours. In the end, the number of times the system crashed in ratio to the number of times the system successfully completed the tests was expressed as a percentage. Only whole number percentages were represented, and all fractional percentages were rounded up. Tests were conducted at CAS Latency timings of 2 and 3 and were represented separately.
|100MHz FSB||112MHz FSB||124MHz FSB||133MHz FSB|
|CAS - 2||CAS - 3||CAS - 2||CAS - 3||CAS - 2||CAS - 3||CAS - 2||CAS - 3|
Note: Higher Percentages = Greater Stability