1T vs 2T Command Rate - Is there a Real Performance Difference? - Overclock.net

*ncsa* · 07-07-05

The matter of 1T vs 2T gets debated frequently and often with no resolution of an outcome. Many (including myself at times

) have stated that there is a huge or big drop in performance when running 2T over 1T Command Rate timing - but how big is this difference? And for those who can run 4x Memory modules they default to 2T. Are they really missing out on much?

What I have decided to do is to try and help discover what the size of this difference could be by running a series of common benchmarks and tests, record the results and then measure the difference of each run.

The system used was:

A64 3000+ Venice
DFI nF4 DR-SLi (7/04-3)
2 x 6600GT SLI

Memory Timings were kept loose to allow for minimal changes across the DDR range and the actual value of each score is not important here but the differences between each. By keeping these changes between each run to a minimum we are reducing the influences of any other memory tuning.

CAS 2.5-4-4-7

Of the changes made between the individual runs; were the DDR ranging from 200-300 at 1:1 ration and the CPU on x9 multi. A full re-run of the tests were then made changing the Command Rate from 1T to 2T.

The Benchmark Apps used were:

SiSoftware Memory Bandwidth
Everest Home Bandwidth
Aquamark3
3D Mark2001 SE
SuperPi 1M

While I am running these benchmarks what do you think the difference will be?

EDIT:
Summary and Results

Attached is a Zipped PDF with the graphs showing the results of the test.

The Bandwidth tools (Everest & SiSoft) show a significant difference between running 1T vs 2T Command Rate - upwards of 25%, however, the Benchmark tools did not reflect such a large gap and there seems to be about a 2-3% performance hit when running 2T. It would appear that the Bandwidth Tools use some test process which is not reflective of the real world situation and as a result people quote these figures as being the reason not to use 2T timing...

Now if you look across the DDR scale you will see that if you are able to increase DDR by 7-10% this will have effectively reclaimed the 2-3% performance hit of the 2T setting. This is something to consider when trying to get that high OC where maybe 2T is stable but 1T just isn't.

I know that I will now consider 2T for these high OC and see if it helps rather than dismissing it without giving it a go .... Food for thought

SPHERE · 07-07-05

sticky!!!

im realy interested i the answer to this aswell

Ket · 07-07-05

i generally found the bandwidth difference to be around 800MB\s+ between 1T and 2T command rates, then you have the impact on memory writes as well

**bigvaL** · 07-07-05

well 0T is the ultimate best, but lots of ram cant handle that, 1T is better then 2T because it doesnt have to wait as long between actions, althought, some ram might not be able to run as efficiently at 1T as it can at 2T because it struggles and cant work as well. The best thing to do is aim for 0T, if its unstable, go to 1T, do a ram benchmark on sandra, set it to 2T and do it again, take whichever rate gives u the best results, thats a decent way to test, theyre pretty close but u might as well take the better one right?

BSman · 07-07-05

This is going to be very interesting. Thats sounds like very extensive test, nice work. How long do you thing it will take you to finish testing?

Cheers to you dude.

*ncsa* · 07-07-05

@Sphere, let's see of there is alot of interest before it can be a sticky maybe a FAQ. thxs

@Ket, yeah what I am finding is that the Everest and SiSoft tools are showing a gap but does this gap reflect as wide in the benchmarks, hence these test runs

@bigval, good point on the 0T I will try to include this setting across the whole range if possible (which tweak tool are you using to set 0T? A64 has only enable / disable)

@BSman, hopefully today or tomorrow pending 0T outcome...

Cheers

*ncsa* · 07-08-05

For a much talked about subject throught out a lot of other threads there seems little interest here... is it that the facts may spoil someone's generalised statement ...

Anyway I have completed my tests and will update post #1 with the findings soon, hopefully they will be of interest.

**The Duke** · 07-08-05

GREAT WORK ncsa!!!
Congrats on the effort

*ncsa* · 07-08-05

Quote:

Originally Posted by The Duke

GREAT WORK ncsa!!!
Congrats on the effort

Thank you very much... so the next thing is the CAS ratings which Jack and I are working on together in the background and there will be another thread on that sometime next week.

Hope it all helps.

**The Duke** · 07-08-05

Frankly... I can't wait to sell of some of my extra stuff to see how well my Ballistix actuall does on a 929 SD platform! I know Ballistix not the fastest, but it's been tested to do rather well.
On this sys I get 436Mhz @ default vdimm with 2-2-2-2-1t

View Poll Results: 1T vs 2T Command Rate - Is there a Real Performance Difference?
< 1%	1	5.26%
1-5%	7	36.84%
5-10%	5	26.32%
10-15%	6	31.58%
Voters: 19. You may not vote on this poll

		Overclock.net - Overclocking.net > AMD > AMD Memory
1T vs 2T Command Rate - Is there a Real Performance Difference?

07-07-05	#2 (permalink)
SPHERE ATI Enthusiast Join Date: Apr 2005 Posts: 6,214 Rep: 743 Unique Rep: 330 FAQs Submitted: 17 Hardware Reviews: 14 Trader Rating: 4	sticky!!! im realy interested i the answer to this aswell

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)