nForce2 - High Tras = Better?

[Techy]

I've been reading a lot of people have been getting better performance running their RAM at 11 Tras on nforce 2 mobo's. So I've done a little experiment, here's the results:

Sandra 2003
------------------------------

217FSB 2-2-2-6 DC

Test1 3325/3134
Test2 3332/3135
Test3 3330/3133

------------------------------

217FSB 2-2-2-9 DC

Test1 3357/3169
Test2 3356/3170
Test3 3355/3170

------------------------------

217FSB 2-2-2-10 DC

Test1 3356/3175
Test2 3357/3172
Test3 3356/3173

------------------------------

217FSB 2-2-2-11 DC

Test1 3361/3175
Test2 3358/3171
Test3 3358/3174

------------------------------

Diminishing results after tras 11 so I didn't bother going any higher. Now onto 3DMark testing!!! *stops* Radeon 8500 . Hehe, I did sum tests and they were pretty much all the same, which isn't a surprise there.

Heres some prime results however:

Prime95 version 23.4 [ 2-2-2-6 ]
Best time for 384K FFT length: 24.653 ms.
Best time for 448K FFT length: 27.829 ms.
Best time for 512K FFT length: 30.483 ms.
Best time for 640K FFT length: 39.991 ms.
Best time for 768K FFT length: 48.835 ms.
Best time for 896K FFT length: 57.236 ms.
Best time for 1024K FFT length: 65.839 ms.
Best time for 1280K FFT length: 85.114 ms.
Best time for 1536K FFT length: 103.190 ms.
Best time for 1792K FFT length: 124.485 ms.
Best time for 2048K FFT length: 139.141 ms.

Prime95 version 23.4 [ 2-2-2-11 ]
Best time for 384K FFT length: 24.305 ms.
Best time for 448K FFT length: 27.352 ms.
Best time for 512K FFT length: 29.959 ms.
Best time for 640K FFT length: 39.311 ms.
Best time for 768K FFT length: 48.260 ms.
Best time for 896K FFT length: 56.462 ms.
Best time for 1024K FFT length: 65.040 ms.
Best time for 1280K FFT length: 83.672 ms.
Best time for 1536K FFT length: 101.422 ms.
Best time for 1792K FFT length: 122.197 ms.
Best time for 2048K FFT length: 136.602 ms.

I ran the Prime95 Benchmark 5 times on each setting and tras 11 came out on top everytime.

Anyone else wanna try?

[Drisler]

Good Job..


Strange

[HauZDauG]

I dont have any pics but i can confirm that my tests did seem to go up very little but not as significant as some... im using 9-3-3-2.5 right now

[notdrugged]

It raised mine very noticeably. How strange. 11-3-3-2.5

[Techy]

http://www.mushkin.com/mushkin/pop-up/latencies.htm

"Memory, in many ways is like a book, you can only read after opening a book to a certain page and paragraph within that particular page. The RAS Pulse Width is the time until a page can be closed again. Therefore, just by definition, the minimum tRAS must be the RAS-to-CAS delay plus the read latency (CAS delay). That is fine for FPM and EDO memory with their single word data transfers. With SDRAM, memory controllers started to output a chain of four consecutive quadwords on every access. With DDR, that number has increased to eight quadwords that effectively are two consecutive bursts of four.

Now imagine someone closes the book you are reading from in the middle of a sentence. Right in your face! And does it over and again. This is what happens if tRAS is set too short. So here is the really simple calculation: The second burst of four has at least to be initiated and prefetched into the output buffers (like you get a glimpse at the headline in a book) before you can close the page without losing all information. That means that the minimum tRAS would be tRCD+CAS latency + 2 cycles (to output the first burst of four and make way for the second burst in the output buffers).

Any tRAS setting lower tRCD + CAS + 2 cycles will allow the memory controller to close the page “in your face!” over and again and that will cause a performance hit because of a truncated transfer that needs to be repeated. Along with those hassles comes the self-explanatory risk for data corruption. That one is not a real problem as long as the system is kept running but in case it is shut down and the memory content is written back to the hard disk drive, the consequences can be catastrophic. For the drive, that is."

So in theory running 2-2-2-5 is slower than 2-2-2-6. The ideal timings according to Mushkin works out to be:

CAS+tRCD+2 = tRAS

Examples:

2 CAS + 2 tRCD + 2 = 6 (2-2-2-6)

2 CAS + 3 tRCD + 2 = 7 (2-3-3-7)

2.5 CAS + 3 tRCD + 2 = 7.5 (2.5-3-3-8) *Round up*


Heres another thread, but its in French.

Here

Heres some pics from that thread:







According to that thread, it helps especially when your in Dual Channel mode.

[TheOtherDude]

I've spent a lot of time investigating this phenomenon. It is related to frequency also.

Here are some benchmarks I ran.

200 FSB

6-2-2-2 2404/1593
7-2-2-2 2420/1604
11-2-2-2 2425/1626

166 FSB

6-2-2-2 2031/1361
7-2-2-2 2035/1381
11-2-2-2 2018/1357

133 FSB

6-2-2-2 1625/1103
7-2-2-2 1633/1111
11-2-2-2 1615/1075

100 FSB

6-2-2-2 1224/837
7-2-2-2 1223/836
11-2-2-2 1207/811



This is the thread I originally posted them in. As you can see, we came to a conclusion, however, I was never really satisfied with it.

[TheOtherDude]

The ideal timings according to Mushkin works out to be: CAS+tRCD+2 = tRAS

The two clock cycles that they factor into thier equation is actually a finite time. Therefore, 2 clock cycles at 100 MHz is equal to 4 clock cycles at 200 MHz. Obviously, that doesn't explain everything though.

[Techy]

We'll just leave it as a mystery for now.

Gonna stay at 2-2-2-6. Its pychologically hard to raise timings

[TheOtherDude]

We'll just leave it as a mystery for now.

I wish I could...

So far I have concluded as a supposed fact that CAS and tRCD effect the optimal tRAS setting. I say "supposed" because I would like other people to confirm my results. Also, through the relationship between clock frequency and optimal tRAS setting, I propose the following equation for getting a range of tRAS values:

tRCD + CAS + 30ns to 40ns = Optimal tRAS (always round up)

Ex.

For 2.5-3-3 @ 190 MHz

3+2.5+5to8= 11 to 14


For 2-2-2 @ 150 MHz

2+2+5to6= 9 to 10


For 2-2-2 @ 200 MHz

2+2+6to8= 10 to 12


For 2-2-2 @ 100 MHz

2+2+3to4= 7 to 8

I would really appreciate it if someone could tell me what they think of the equation. Did it prove true in your guys' cases?

Thanks

}EDIT{ I corrected an error in my arithmetic.

[AZTEC_GIZMO]

Well, my Sandra memory bandwidth scores did increase (by about 20 points) when switching from 2-2-2-6 to 2-2-2-11. So, I think it helps to switch the settings.

[danny5]

Most interesting thread. I kow I will be going into the bios tonight and checking this out. Jeez, I just changed my setting from 7 to 6 thinking tighter settings will make it run faster. I see that this may not be the case at all.

[TheOtherDude]

Originally posted by AZTEC_GIZMO
Well, my Sandra memory bandwidth scores did increase (by about 20 points) when switching from 2-2-2-6 to 2-2-2-11. So, I think it helps to switch the settings.

Could you please try to apply my formula, and tell me if it works?

[JaccDude]

Yes it's true.. I have this 11-2-2-2 settings for a couple of months now...

just look at my sandra mem at the bottom with 223fsb
and also it raise your 3dmark quite more check bottom again

AND I just realised... is this a new MEMORY forum????

[TheOtherDude]

Originally posted by JaccDude
Yes it's true.. I have this 11-2-2-2 settings for a couple of months now...

just look at my sandra mem at the bottom with 223fsb
and also it raise your 3dmark quite more check bottom again

AND I just realised... is this a new MEMORY forum????

Could you please varify that a value of 12 doesn't work better than 11. 11 is not the magic number. It will not work best for all people. That is contrary to the understanding of those french people in that thread Techy linked to. From the thread:

vive 11

[JaccDude]

Originally posted by TheOtherDude
Could you please varify that a value of 12 doesn't work better than 11. 11 is not the magic number. It will not work best for all people. That is contrary to the understanding of those french people in that thread Techy linked to. From the thread:

I did 12 and 13 but 14 wouldn't boot ..and the best result for my system was 11 . the lowest I could get was 4-2-2-2 but that was far worst than the 11.
I don't have the link that "11" is the magic number, I benched it myself cause that's where I will know more about my settings by experience,
if you don't believe me , try it with your own experience. could be worst or could be not, give it a try
only a couple of resets