2T vs. 1T on DDR3?

[freeagent]

Fast-n-tight, end of secret

 

[Woomack]

Just tested this, changed my command rate from 2t to 1t and my physics score went from 9250 at 2t, to 9500 at 1t not to big of a diffrence, but there was a 2fps increase on the test.

Best is to download Maxxmem Preview and check settings on different timings and clock. Test is really fast and easily comparable.
What you are looking for in games/benchmarks is fast memory read.
Tight timings are not always best, sometimes it's balance between clock and timings what is clearly visible on newer platforms. However on X58 you don't have big choice so just work on timings .

On SB/IB Command Rate isn't so important as latency is already low and transfers are high ( fast cache etc ). In this case higher memory clock can give more than playing with timings and even on the same clock, results will be almost the same on 1T and 2T ( unless you bench Spi32m then you may need 1T ).
Somewhere here you have my older tests on P67 and 1T/2T so at least on this platform there is not so big difference.

 

[djscrew]

Well, I ran 3dmark 11 again three times @1t and 2t and the results are the exactly same it shows an increase in performance. While running @1600mhz 9-9-9-24 2t I averaged 8944 for three runs and @1600mhz 9-9-9-24 1t the average was 9013 3dmarks. I don't think its a margin of error since I ran it three times at each setting an saw the same results each time. I mean grated its not a huge gain but it pushed me above 9000 3dmarks which has been a goal of mine for the last few weeks. I ran my 2500k @4.9ghz my 560ti's at 920mhz each and the ram at 1t &2t. Stoked!

now just check to see if you can tighten the timings anymore... try 8.9.8.24.1T

 

[Neuromancer]

I've wondered about 1t-2T settings myself. My memory is at 2T and was looking for a simple explanation about the performance difference of 1T and 2T and also stability issues. I have 4 memory modules set as dual channel for a total of 8gb in all memory slots. Looks like I'll leave it at 2T. Thanks for the info. guys.

http://www.newegg.com/Product/Product.aspx?Item=N82E16820211409

Those your sticks?

With all 4 populated sometimes 2T is necessary for stability.

Your timings are correct otherwise 9-9-9-24 and if vdimm is set to 1.65v then you can increase the memory multiplier to 1600 effective.

You can then try setting 1T, it either works or it doesnt.


generally anymore 2T is only necessary with 8GB sticks (older 4GB modules might be affected as well) or super cheap RAM that is marketed at higher speed or lower timings/voltages than it was designed to run at.

 

[madman7]

http://www.newegg.com/Product/Product.aspx?Item=N82E16820211409

Those your sticks?

With all 4 populated sometimes 2T is necessary for stability.

Your timings are correct otherwise 9-9-9-24 and if vdimm is set to 1.65v then you can increase the memory multiplier to 1600 effective.

You can then try setting 1T, it either works or it doesnt.


generally anymore 2T is only necessary with 8GB sticks (older 4GB modules might be affected as well) or super cheap RAM that is marketed at higher speed or lower timings/voltages than it was designed to run at.

That's the memory I'm using. I'm set at 1333 @ 9-9-9-24-33 2T with voltage set at 1.54v. I have run it at 1600 @9-9-9-25-34 2T with the voltage set at 1.66v. What's your opinion where I should run the settings?

 

[Neuromancer]

1333 with less voltage is fine. I would run faster though just to accelerate that snappiness AMD systems have.

Start with 1600 9-9-9-24-2T at 1.65v, then try 1T. If still stable you can try dropping the vdimm .1 volts at a time. When unstable go back up .1v and then do a prime95 (64bit) blend test or a memtest 4 run while you sleep to test for stability.



If you increase to 1600 then you need to increase CPUNB as well. 2400 CPUNB minimum which may also require a small voltage bump. I ran my 1090T at 2700 at 1.25 vCPUNB IIRC for 1866 mem. But it was only for testing, I did not run a long stability test on it. Someone else might have a better recommendation for average vCPUNB settings, it might take more it might take less

 

[madman7]

I have the CPUNB at 2400 already and the CPUNB voltage to 1.25 volts. I have heard/read that the Phenom II IMC doesn't like 1600 memory settings. I've also heard that the C3 stepping might handle 1600 memory better than the C2 stepping. I haven't found anything definite to verify that. The voltage can only be changed 0.02 volts at a time. That's why the 1.66v settings. Anything less than 1.66v-1.64v at 1600 and the board won't boot windows.

 

[homebuildit]

Oh nice, I just bumped my vdimm to 1.64, set timing 8-9-8-24 t1, and after a few reboots (finnicky Ridata SSS boot), started windows just fine, running on it now as I post. Personal record for me Have to try a bench.

Tough to type, cut my pinkie on my cpu cooling fan pretty good haha.

 

[Neuromancer]

I have the CPUNB at 2400 already and the CPUNB voltage to 1.25 volts. I have heard/read that the Phenom II IMC doesn't like 1600 memory settings. I've also heard that the C3 stepping might handle 1600 memory better than the C2 stepping. I haven't found anything definite to verify that. The voltage can only be changed 0.02 volts at a time. That's why the 1.66v settings. Anything less than 1.66v-1.64v at 1600 and the board won't boot windows.

There was something specific about the 1600 divider yes. I read that too.

The way around that was to drop HTLINK to 1600/1700 set CPUNB to 2000 set memory to 1333 and drop your cpu multi to 17. Then increase HTREF clock from 200 to 240. You might need a little more voltage on CPU and NB. CPUNB should not be as affected I think.

Oh nice, I just bumped my vdimm to 1.64, set timing 8-9-8-24 t1, and after a few reboots (finnicky Ridata SSS boot), started windows just fine, running on it now as I post. Personal record for me Have to try a bench.

Tough to type, cut my pinkie on my cpu cooling fan pretty good haha.

Delta fans? lol. Glad your running well. Dont forget to stability check

 

[Archer0915]

No questions! Run everything as fast as you possibly can! Run like a banshee with your hair on fire!

Timings will or will not work simple as that though I guess the worst you could do is botch your win install.