?tRAS? 6/7: Best performance on nForce4 boards!!*updated*

[MADMAX23]

Memtest is not sensitive to tRAS variations.

 

[Zebo]

Originally posted by: MADMAX23
Memtest is not sensitive to tRAS variations.

Sure it is try it.

 

[MADMAX23]

I did tried it, same speed (bandwidth Mb/s)
I tried Memtest86 v3.2 and Memtest+ v1.51
No variations with tRAS 5/6/7/8

 

[Zebo]

Yes it does. Try 10 if you don't believe me then your prefered 7.

 

[MADMAX23]

Originally posted by: Zebo
Yes it does. Try 10 if you don't believe me then your prefered 7.

Of course, if I set the tRAS to 10 or up, bandwidth varies but if you set 5/6/7/8 nothing changes, it is unable to detect little variations. RightMark is the only tool well-known because of this and other capabilities.

 

[Zebo]

I am slower at 7 than 8 with memtest..that's all I needed to convince me, on my board, with Ballistix and the timings I was using 8 was superior. This right mark just validated what memtest already told me. I am using NF3 ultra though which may account for some variations. When I get NF4 board I'll be sure to keep your study in mind.

 

[MADMAX23]

You should have told me that first, man. I didn't know you were using an nForce3 Mobo. This thread was intended for nForce4 chipset mobos.
Yeah, with your nForce 3 you may get better results with tRAS of 8/10.
Make tests with RightMark following my test method, maybe you will find a better (?) performance with a different tRAS (it is said a tRas of 10 performs best on nForce3 boards, although I haven't tested it).

 

[Zebo]

First thing I said I depends on the board Sorry about the confusion. Still I bet depends on the NF4 boards too... Anand said 10 was good in NF3, I found it worse, found 8 to be best. Good testing BTW

 

[MADMAX23]

Sure, it depends on every system config.
Ok

 

[EndGame]

On my main A8N-E, I run Corsair XMS XLL 2-2-2-5 ram I had before I switched this system over from an Intel 478 system.

In my tests with this board and ram with a 3700+ San Deigo, running my ram at anything less than 5 tRAS results in lower scores concurrently.

In my other A8N-E system with OCZ rated at 2.5-3-3-7, lowering the tRAS to 5 gives negative results as does raising it above 8 much as the OP stated. Levels of tRAS between 6, 7, and 8 give results which are close enough to make little appreciable difference.

IMHO, results are much dictated by the ram the system is running and it's given rated speed.....

 

[MADMAX23]

So, what's your system now? Mobo and ram timings?

 

[EndGame]

Originally posted by: MADMAX23
So, what's your system now? Mobo and ram timings?

Two of mine are with the A8N-E, ram mentioned above also, 1GB in each. Timings with the Corsair are as they boot @ 2-2-2-5, likewise with the other ram. The system with the faster ram outscores the other substancially.

 

[MADMAX23]

For both of your Ram timings I suggest you to use a tRAS of 7. Best performer of all. Read my tests at the beginning of the topic.

 

[EndGame]

Originally posted by: MADMAX23
For both of your Ram timings I suggest you to use a tRAS of 7. Best performer of all. Read my tests at the beginning of the topic.

LOL! I ran the SAME tests on my systems after reading this thread....that was the point of my post! Ram with default timings of 2-2-2-5 score lower on all B/M's by lowering the tRAS to 6, 7, or 8. Ram with looser timings, like the OCZ I use in the other system seem to be within a margin of error and vary B/M to B/M using 6, 7, or 8 tRAS but do give markedly lower scores by lowering tRAS to 5 or raising it above 8.

If you have any low latency ram laying around, I suggest testing these modules similarly because my point is/was, ram that is low latency spec scores best when left at lower tRAS (5) and will score lower with each successive move to a higher tRAS (6,7,8,9+) while ram with loose timings will score higher in the 6,7,8 tRAS range and lower at 5 tRAS timings.

 

[MADMAX23]

Did you follow my Test method?

 

[EndGame]

Originally posted by: MADMAX23
Did you follow my Test method?

Ran all the same tests the same way.....stands to reason, ram with tighter timings and lower latency will test and score lower with looser timings. Ram with already loose timings will score worse if you try to tighten it or loosen it further.

 

[MADMAX23]

Don't think so...A little example; Guitardaddy runs the OCZ PC3200 EL Platinum R2 (2-2-2-5) at 2-3-2-7 because he gets the best performance with a tRAS of 7 than with the stock tight 5. As he does, lots of people with 2-2-2-5 Ram modules...Sorry.
Anandtech also stated 7 was the best performer ,on the whole, for nForce4 boards. Of course there will be exceptions depending on the rest of your settings, board, memory...etc but, as a generic rule, suitable for most of us, 7 and ,in some cases, 6 are the best tRAS values.

 

[EndGame]

Originally posted by: MADMAX23
Don't think so...A little example; Guitardaddy runs the OCZ PC3200 EL Platinum R2 (2-2-2-5) at 2-3-2-7 because he gets the best performance with a tRAS of 7 than with the stock tight 5. As he does, lots of people with 2-2-2-5 Ram modules...Sorry.
Anandtech also stated 7 was the best performer ,on the whole, for nForce4 boards. Of course there will be exceptions depending on the rest of your settings, board, memory...etc but, as a generic rule, suitable for most of us, 7 and ,in some cases, 6 are the best tRAS values.

Yes, and is he O/C'ed with those settings? Has he run the tests?

If he's overclocking while running the tests with that memory, all results will be skewed. If you simply want to see what setting your ram performs best at, test an un-overclocked system with ram settings at stock, what they are rated for or boot at and then adjust the tRAS settings and rerun with each setting.