what does all the numbers mean in ram timings?

[frankierx]

I'm OC'ing my 2500+ to 3200+ ; 200mhzx 11 =2.2Ghz at 1.75 Vcore, 1 gig corsair xms pc3200 at 3-3-3-8
Epox 8rda mobo
Enermax 350W ; 12 V rail 26A; only 2 hds; fx 5900 128mb

under 14 mins prime95 1 error.

Do I need to change ram timings? What do these timings mean? Do I need to increase Vcore to 1.8?

1.75 Vcore with 39C idle 42C load.

 

[Dethfrumbelo]

Ram Timings Explained

Basically, the most important number is the first one, which is CAS latency. You're already at 3, I wouldn't recommend any more loosening there. Maybe 3-4-4-8?

You might try raising the voltage, but I don't know the exact specs on that mem.

 

[Navid]

First, run memtest, which stresses the RAM more than the CPU. Make necessary adjustments until you can run memtest with no errors overnight.

Then, run prime95, which stresses the CPU more. Make necessary adjustments so that you can run it overnight.

 

[eyeballkid]

Originally posted by: Dethfrumbelo
Basically, the most important number is the first one, which is CAS latency. You're already at 3, I wouldn't recommend any more loosening there. Maybe 3-4-4-8?.

That is a market myth. The most important timing is not CAS latency. Reducing the RAS to CAS delay (Trcd) and the RAS precharge (Trp) has higher effect on memory performance. If one were to choose between a timing of 2.0 - 4 - 4 and 2.5 - 3 - 3, the latter would be better.

Originally from PC Perspective
"So if you need to loosen RAM timings in hopes of achieving a higher clock, it is recommended and accepted that you increase the value of CAS first, then tRP, and then finally tRCD."

I recommend the OCZ Memtest86 with which you can change memory timings on the fly. It works for most chipsets, including i915/i925 and the AMD64.

 

[frankierx]

Originally posted by: eyeballkid

Originally posted by: Dethfrumbelo
Basically, the most important number is the first one, which is CAS latency. You're already at 3, I wouldn't recommend any more loosening there. Maybe 3-4-4-8?.

That is a market myth. The most important timing is not CAS latency. Reducing the RAS to CAS delay (Trcd) and the RAS precharge (Trp) has higher effect on memory performance. If one were to choose between a timing of 2.0 - 4 - 4 and 2.5 - 3 - 3, the latter would be better.

Originally from PC Perspective
"So if you need to loosen RAM timings in hopes of achieving a higher clock, it is recommended and accepted that you increase the value of CAS first, then tRP, and then finally tRCD."

I recommend the OCZ Memtest86 with which you can change memory timings on the fly. It works for most chipsets, including i915/i925 and the AMD64.

Is that ocz memtest86 the same memtest86 at www.memtest86.com?

 

[frankierx]

Okay, RMA'd the bad corsairs xms pc3200 ram and poped in the old samsung pc2700 ram. no memtest86 errors whatsoever now.

 

[BonzaiDuck]

Eyeballkid --

Thanks for the tip. I didn't know they had their own revision of the software.

 

[BonzaiDuck]

I came to the seemingly inexorable logical conclusion that the Row Precharge can overlap the CAS latency. Diagrams I have seen show the overlap only under the Burst-Length cycles, so I never considered the first possibility.

Is it true?

 

[BonzaiDuck]

Here. This is a useful article.

http://www.tomshardware.com/howto/20030701/index.html