The Ram Question: 40MHZ or .5 less CAS Latency

[thesurge]

What would be more beneficial... running at 194MHZ 2.5-3-3-7, or 235MHZ 3-3-3-7? Both are 1T



UPDATE
I just got my RAM stable at 2-3-3-7 (193MHZ)... is that better than 235MHZ 3-3-3-7?

(I can safely say my RAM is stable if it runs 3 loops of Memtest and completes one hour of SPrime Large FFT, right?)

Is that normal for Corsair Value Ram?[/quote]

 

[thecoolnessrune]

first option i believe.

 

[crispy2010]

235MHZ 3-3-3-7

 

[sparkie34]

235MHZ 3-3-3-7


no doubt

 

[lopri]

Agree with 235MHz. For 20MHz difference, maybe it can go either way, but 40MHz?

 

[thesurge]

Alright thanks.

UPDATE
I just got my RAM stable at 2-3-3-7 (193MHZ)... is that better than 235MHZ 3-3-3-7?

(I can safely say my RAM is stable if it runs 3 loops of Memtest and completes one hour of SPrime Large FFT, right?)

Is that normal for Corsair Value Ram? Should I tweak any of the other timings (RAStoCAS, RAS Pre, or Tras)?

 

[nealh]

I have heard so many things on stability..personally I run memtest86 from boot screen for 6hrs..10 passes

small fft only stree cpu, large fft stress ram and cpu, blend puts more stress on ram

for Prime95..I like 8hrs plus(well really I shoot for 15hrs..many feel 24hr) for stablity...but you must decide what is acceptable for you

what feels faster to you..I felt when comparing ram 250@3-3-3-7-1t was very good for mysystem vs low latency 2-2-2-5-1t that did not overclock well

 

[Furen]

I've had errors happen like 10hrs into prime95, so I'd say it's doubtful that 1hr is enough to be sure. I think you should try both settings and see which one you like the best (which one feels snappier, etc). If both feel the same and all then just stick to the one with the lowest vcore/vdimm requirements.

 

[thesurge]

Alright. Thanks. I ran both at 2.8Vdimm, which is acceptable I presume. I'm going to stick with the lower latency one... run memtest86 over night. Although, I could probably try using 2.7Vdimm with both. Is it much of a damage difference between 2.7 and 2.8? I have not been experiencing any high temps.

 

[lopri]

While I like to have my system setting capable of 12hrs of Prime95, I have to admit that most of errors will pop up within the first hour of Priming. So yeah, for quick check-up, 1 hour is enough. Hell, for initial OC'ing, I don't even wait for 10 mins. Who has that much time? Run it for 5 mins or try Super PI 4M/8M, then up the HTT! Of course once I find my limit, I feel comfortable only after 8~12 hours of stable Prime95 run.

 

[Leper Messiah]

2.8vdimm is fine for any ram, after 3.0 I'd say one would want some kind of cooling on there. 193 2-3-3-7 should be a bit faster than 3-3-3-8 235, the A64s like the lowered latency.

 

[NewBlackDak]

If you can run both with stability then why don't you run some benchmarks? A64 Rigs like low latency, but with a 42MHz difference that might not be the case.
I don't mean synthetic one either. Run something you do all the time, and see what is faster. I have a feeling the MHz would win for encoding, and latency would be better for games.

 

[lopri]

For that I'd say CAS 2 would be faster. However, now that you got it stable @CAS2, why not try CAS2.5 @235? Maybe with some tweaking, who know.

 

[jonspd]

Sence you only going from 2.5 cas to 3.0 I would say the 235 fsb is going to give you better performance.

 

[haelduksf]

Umm...why not try all three?

It's not like it takes a long time. Run SuperPi 1M, change clocks with A64Tweaker, run SuperPi 1M. Wash, rinse, repeat as desired. Should take under 5 minutes. Whichever gives you a better time, keep it.

 

[BitByBit]

Latency (/ns) = CAS / Frequency

Your first configuration would give a latency in nanoseconds of:
2.5 / 194e06 = 12.89ns

The second:
3 / 235e06 = 12.77ns

So therefore, CAS 3 RAM at 235MHz is slightly lower latency than CAS 2.5 at 194MHz.

CAS 2 at 193MHz has a latency of:
2 / 193e06 = 10.36ns

Although the last configuration gives the lowest access latency, the additional bandwidth of the 235MHz configuration would probably offset that advantage.

 

[Shenkoa]

Originally posted by: crispy2010
235MHZ 3-3-3-7

And BitbyBi, by e do you mean epsolon.

 

[BitByBit]

Exponent to the base 10.

By epsolon, do you mean epsilon?

 

[Fox5]

I believe the 2nd one will have lower latency than the first one, so it may be the better choice.

 

[PingSpike]

Lets see some benches dammit! I've been wondering about this same kind of question. Since overclocking on the A64 only has a limited number of ram dividers...I want to know if I can get some or all of my performance "back" when I'm forced to run the ram at somewhat undesirable settings into order to up clock speed. (Think, 186mhz for instance...on cheapo value ram.)

Is latency better for games then clock speed?

 

[thesurge]

so far im runing it 2-3-3-7 @ 193 mhz, but to tel you the truth i cant really tell the difference between anything, even photoshop cs2 is still as "snappy" as ever . games are running the same as the 235MHZ 3-3-3-7. ill try 2.5 w/ 235 mhz but that's really iffy....i didnt even know i could push my value ram to 2, so who knows.

 

[PingSpike]

Sounds like you got a couple of good sticks. My corsair value ram shats the bed at like 215 default timings.

 

[GuitarDaddy]

Originally posted by: thesurge
so far im runing it 2-3-3-7 @ 193 mhz, but to tel you the truth i cant really tell the difference between anything, even photoshop cs2 is still as "snappy" as ever . games are running the same as the 235MHZ 3-3-3-7. ill try 2.5 w/ 235 mhz but that's really iffy....i didnt even know i could push my value ram to 2, so who knows.

Run Sandra memory bandwidth benchmark and I believe you will find 2-3-3-7 @ 193 will beat 3-3-3-7 @ 235. I have found is you can run 2-3-3-7 you can probably run 2-3-2-7 witch will give you a little better score.