What timing should I run my underclocked ram?

[Jumpman23]

I have the corsair DDR32000 CMX8GX3M2A2000C9 rated at 9-10-9-27, 1.65V that I would like to use in my Trinity system. I want to run it at 1866mhz, but what timings and voltage should I use it at? How much lower should I set my timings to begin with?

 

[BonzaiDuck]

I have the corsair DDR32000 CMX8GX3M2A2000C9 rated at 9-10-9-27, 1.65V that I would like to use in my Trinity system. I want to run it at 1866mhz, but what timings and voltage should I use it at? How much lower should I set my timings to begin with?

Over the last five years or so, I've used RAM of maybe three different manufactures. Corsair briefly figured as one of those, and I had no complaints with the single kit of DDR2.

My best instincts along with those experiences suggest how you might approach this in two ways. See if Corsair makes a kit similar to your DDR3-2000's which is instead spec'd to run at 1866, and make a note of the timings and voltage for the latter. And -- check the SPD and XMP specs of the DDR3-2000 kit and see if there are settings for the lower speed.

Otherwise, I've been through this . . . trouble . . . of experimenting with tighter timings. I would guess you'd first try tightening the CAS latency by one notch. There is a relationship between CAS and between the other timings, and you should be able to find some general guidelines for DDR3 in forum and DIY posts, or some memory primer for DDR3. Tom's Hardware had been good for providing that sort of guidance for earlier RAM like DDR and DDR2.

If you have to experiment, you will experience instability. Also, Corsair's voltage spec (as I had known it) was loose enough that you could feel safe increasing the "maximum" spec to within 5% excess -- as reported to me by Corsair tech-reps. In your case, you'll want to lower the voltage or keep it the same (for tighter timings).

 

[gpse]

http://www.corsair.com/cmx8gx3m2a2000c9.html
I would try 9-9-9-24
then try 8-9-8-24

 

[Jumpman23]

Currently, I've tried setting it to 1866 8,9,9,20 1T @1.5V and ran a memtest with no problems shown. And then there's these other timings that I just have no idea what to do with like the tRC, tWR, tRFC, tRRD. Am I fine with the above settings? And what do I do with the other timings?

 

[Soulkeeper]

these are some relatively conservative timings I use for my 2133 set of mushkin on an fm1 platform at 1866 and 1.52v
I've had well over a month uptime with these timings more than once.

for me the set is rated at 10-11-10-28 so i'm essentially dropping them by 1

I'm running these timings at 1600 1.55v speed as I type this (very tight timings)

some more conservative 1600 timings that also passed memtest and benchmarks for me

your mileage may vary (not sure what chips your dominators have)
good luck

 

[Jumpman23]

Is there a technique in finding out what timings to use or just guess and check type thing.

 

[Jumpman23]

I get more errors if I run memtest a little bit longer past pass 1 at 8-9-9-24 1.6V I tried to decrease the voltage before to less than 1.6 and got even more errors. Should I be running the voltages at the stated 1.65V even though I am not running these at 2000MHz?

 

[BonzaiDuck]

I get more errors if I run memtest a little bit longer past pass 1 at 8-9-9-24 1.6V I tried to decrease the voltage before to less than 1.6 and got even more errors. Should I be running the voltages at the stated 1.65V even though I am not running these at 2000MHz?

Figure some rules of thumb as "trade-offs." A trade-off is a situation when you can have more of one thing with less of another, or vice-versa.

If you tighten the timings on RAM at the same speed setting with the minimum voltage for stability, you will most likely need to increase the voltage to sustain the stability. If you raise the speed and keep the same timings, you will likely need to increase voltage. Other corollaries should be obvious. You might raise the speed and loosen the timings to keep the same voltage. You can lower the speed, hold the timings constant, and you may be able to reduce voltage a little.

If you leave the other timings on "Auto" you should be fine. But there are two settings -- one for which I forgot the name or acronym -- which can improve things. I would verify this for DDR3, but there is a "Bank Cycle Time" (maybe ? tRC?) which should be roughly equal to tRP+tRAS+[some margin]. With DDR2, the margin was +1 or +2, the setting was always defaulted to an excessive number, and tightening it was supposed to make a significant performance improvement.

I'm sorry, but I wish I could remember the other setting.

ADDENDUM. If you can run them as DDR3=1866 and 9,9,9,24 and voltage below the spec for DDR3=2000, I don't see why that wouldn't be a good result. But I'd run HCI MEMTEST for at least 500% coverage on any settings at command-rate = 2 first, get things working and able to pass that test, then try CMD=1.

 

[Jumpman23]

Figure some rules of thumb as "trade-offs." A trade-off is a situation when you can have more of one thing with less of another, or vice-versa.

If you tighten the timings on RAM at the same speed setting with the minimum voltage for stability, you will most likely need to increase the voltage to sustain the stability. If you raise the speed and keep the same timings, you will likely need to increase voltage. Other corollaries should be obvious. You might raise the speed and loosen the timings to keep the same voltage. You can lower the speed, hold the timings constant, and you may be able to reduce voltage a little.

If you leave the other timings on "Auto" you should be fine. But there are two settings -- one for which I forgot the name or acronym -- which can improve things. I would verify this for DDR3, but there is a "Bank Cycle Time" (maybe ? tRC?) which should be roughly equal to tRP+tRAS+[some margin]. With DDR2, the margin was +1 or +2, the setting was always defaulted to an excessive number, and tightening it was supposed to make a significant performance improvement.

I'm sorry, but I wish I could remember the other setting.

ADDENDUM. If you can run them as DDR3=1866 and 9,9,9,24 and voltage below the spec for DDR3=2000, I don't see why that wouldn't be a good result. But I'd run HCI MEMTEST for at least 500% coverage on any settings at command-rate = 2 first, get things working and able to pass that test, then try CMD=1.

I ran a memtest on those settings and the first pass came clean with no errors. On the second pass on test #5, I started to get 4 errors. I'm pretty sure I took a picture of it somewhere if it helps. But is RAM suppose to be 100% error free? I mean I can run windows fine with no problems by doing everyday simple tasks so are errors really a big problem?

 

[kleinkinstein]

I get more errors if I run memtest a little bit longer past pass 1 at 8-9-9-24 1.6V I tried to decrease the voltage before to less than 1.6 and got even more errors. Should I be running the voltages at the stated 1.65V even though I am not running these at 2000MHz?

That's why you don't buy 1.65v sticks. Stick with 1.5v, but preferably 1.35v like the Samsungs, Mushkin and new Kingstons.

 

[Deders]

I wouldn't be happy if I got memory errors, it could cause system instability.

I found that Windws Memory Diagnostic picks up on more errors tham memtest, but that was a while ago, not sure if memtest has improved.

If you have your memory timings set too tight it's possible that you might actually slow performance down, if the memory has to repeat itself due to the result not matching the checksum.