• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

DDR Setting and Performance Qs

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Valid

Member
Joined
Jun 20, 2001
Location
Upstate New York
Item in question: Crucial PC2100 256meg stick.

I've had this running at 140mhz rock-solid as far as I can tell at CAS2. I've had it running much higher on the auto setting, but I think I'm getting better performance the other way. (As evidenced by the memory bandwidth in Sandra)

Current settings are 8-8-8-2-2-2-2.

My other question is... at 8-8-8-2-4-2-2-2 I seem to be able to squeeze an additional few mhz out of the RAM. What is the impact in performance incurred by this change? I'm not showing a large difference in Sandra, but this setting would be for "Tras Timing Value."

As always, any information is appreciated. :D
 
The crucial defaults to 2.5 cas latency in auto.


SDRAM tRAS

tRAS is the RAS pulse width, that is the time required for the bit-lines to build up the voltage potential necessary for restoring the data to the memory cells of origin. Setting tRAS too short will eventually cause data corruption not only in the memory array but can also cause hard drive corruption. Historically, tRAS was defined as the sum of tRCD and CAS latency, however, with the current high speed DRAMs, this equation no longer holds. As a rule of thumb, at 100 MHz memory bus speed (200 MHz data rate) a tRAS of 5 cycles suffices in most cases. At or above 133 MHz using tRAS of less than 6 is like playing Russian Roulette. tRAS has little or no impact on performance unless software is used that causes totally random accesses.

http://www.lostcircuits.com/motherboard/epox_8k7a+/6.shtml
 
I ran some tests on my system with various memory settings.
8,8,7,2,6,2,2 vs lower ( faster settings):

8,8,8,2,2,2,2 = no boot

8,8,8,2,4,2,2 and 8,8,8,2,5,2,2

There was no significant difference in SiSandra memory benchmarks.
 
Back