memtest stable/prime fail

[obscurehero]

I'm having this problem where I can run my ocz vc for hours (excess of 40 now) and not have any errors. Not even a hint of an error. THEN, when I boot into windows and run stressprime...I get errors within 3 minutes on the blended test. The other tests work great (or error out much later) but the blended dies right away. I thought it might be a voltage issue with my cpu (overclocked by a mere 100mhz right now), however, after giving it 1.45 volts I saw no reduction in errors. I underclocked it as well and it didn't do anything for it either. Am I doing something wrong??

:confused:

[MrSeanKon]

:rolleyes: If I understand well maybe your CPU is unstable.
Give us more info.
Timings??
Vdimm???
FSB???

I do not understand why you are seeking for RAM limits first.
You should seek your CPU limits first.
I posted some tutorial posts click on my sign.

[obscurehero]

I'm not seeking for ram's limits first...at least I didn't think so. The ram is running at 250mhz 2-2-2-8 (Trc=2, trfc=auto, trrd=2, twr=2, twtr=1, trwt=2, tref=3120, twcl=auto, dram ds=lvl 8, data ds=lvl4...) which are spec timings. The CPU is at 10x250 with a "1:1" divider ratio. The vdimm is 3.3v. The CPU is at 1.4v with a special vid of 1.04 (the board undervolts so its actually about 1.4 with a multimeter).

I thought maybe that the cpu is unstable but I can't see any performance increase when underclocking it or giving it more voltage. I don't know...

[Major]

1. Test with memtest b4 entering into windows, when passes several loops of test #5 then ur usually safe to enter windows.

2. Test with memtest (for windows) after loading windows, you should test "all unused memory" the default. If you get errors, adjust timmings using A64 tweaker from within windows, I like to lower memtest to 64meg so it loops fast. leave it running while you adjust settings.

3. Use Prime95 and SuperPi to test cpu stability, both will fail with too low or too high Vcore.

[MrSeanKon]

Remember SuperPi stretches CPU+RAM so if errors occur you cannot know exactly what is going on.
Moreover I do not use old programs; LOL it is written 10 years ago..
Set dividers and relax timings.
Run S&M step by step or something else for small time.
I follow for years this method and I do not spend time.
I gave you directions above.

Good luck. :)

[obscurehero]

Major. I did do the memtest thing. It came back with no errors whatsoever. I'm now trying memtest for windows...didn't know about that. I'll let you guys know what happened. Do you think its a cpu issue?

[hovo73]

Well from my recent experience with Memtest is that it is not a good indication of windows stability. I dont' know if it is the new version of memtest or the new platforms, but I am constantly getting hours and hours of stable Memtest runs (test 5, 8, 4, 7, all) but when windows loads it could be as bad as an immediate BSOD or as "mild" as failing P95 under a minute.

Until something changes my mind, i don't trust Memtest any more.

EDIT: Just wanted to add that when running those tests I could be certain that it is not my CPU that was the problem, because it was running at speeds far below at what it had been tested. For example, I know my CPU can do 9x280 with 1.55V. It had passed many hours of Prime95 (20+ hours) with those settings. But when I was running above mentioned tests, CPU was running at only 9x250 to 270, and even at 8x250 to 270.

[obscurehero]

Yes...same for me. The windows memtest comes up with errors right now. How should I go about burning it in so that these errors slowly go away if the bootable memtest isn't picking it up?

Its definitely a memory issue, so I'm glad I got that sorted out...now what to do about the memory??

[B5I8]

Try setting your DRAM Drive strength to Level 5 and your DATA Drive Strength to Level 2 and report back.

[obscurehero]

Ok...but why?

[B5I8]

That made all the difference in the world in getting my setup stable. Before that I could not go above 240 on my RAM. Now I'm 24 hour Prime stable at 250 with only 3.2V. Also have you tried the Hellfire modded BIOS?

[Magnj]

for VX I use drive strength of 8 Data of 4. VX is touchy. But in my )short) experience. Metest Stable + Prime unstable, escpecialy on a divider = Bad mem controller on the chip. however on a 4K+ Diego I doubt thats the problem

[B5I8]

Well it worked on my VX as well. I think it might work seeing as how we have very similar systems. We both are running the DFI SLI-DR and a 4000+ San Diego.

[hovo73]

for VX I use drive strength of 8 Data of 4. VX is touchy. But in my )short) experience. Metest Stable + Prime unstable, escpecialy on a divider = Bad mem controller on the chip. however on a 4K+ Diego I doubt thats the problem

Data Drive Strength goes only up to 3 in the newer BIOS.

[obscurehero]

I'll try the new bios. I did get errors with the windows memtest...does that mean I've still got a bad mem controller?

[cupra]

aua, 3 months ago exactly the same problem happen to me (tccd) , and the bad news - it will be worse !

and it was my rams, so check it up .

[obscurehero]

seriously?? I just got this memory back from an rma. I sure hope this isnt bad memory again. I still haven't gotten an ocz support person to answer this question so I guess I'll ask them when they get to my question.

[Rave]

I had this exact same problem (albeit at much slower speeds;)) with my 8RGA+ and 2x512MB TwinMOS CH-5. It would memtest fine at 225MHz and boot into Windows, but Prime95 was not stable at anything over 185MHz. Since it would prime at 220+ with my old 2x256 sticks, I've always assumed it was the NForce2 memory controller struggling with 2x512 with CPC-on. I've never got round to trying a CPC-off bios- TBH I know this is heresy for XS, but it's fast enough for me at 185MHz, and at some point when prices come down I'll get a S939 system to give the CH5 the thrashing it's begging for:slapass::D .

[obscurehero]

So...after a week of waiting. Everything is rock solid stable. Both prime and memtest. I basically let it sit for 4-5 days and everything works great...whats up with that?

[Magnj]

computers are silly

[Jomiy]

Sort of related....

I did Memtest86 100 passes #5, 30 passes #8, 5 passes full test at DDR600.
Enter windows and did Memtest 3.2 in Windows.

It errors at 120% and then 126% so I stopped it. How much percent coverage is recommended for "stable" RAM? 1800% error free?

[brandinb]

Sort of related....

I did Memtest86 100 passes #5, 30 passes #8, 5 passes full test at DDR600.
Enter windows and did Memtest 3.2 in Windows.

It errors at 120% and then 126% so I stopped it. How much percent coverage is recommended for "stable" RAM? 1800% error free?

i would say even higher to let the heat build up and achieve max thermal load to see if its gonna be stable in summer in a stuffy garage at a lan party!!1

[ben805]

if you can passed memtest with flying colors but crapped out in Windows, you have two options, 1)try to raise the vdimm 0.1~0.2v, or 2)relax some of these parameters: Trfc , Max Async Latency, Read Preamble, Trp, and Twr. Keep RWQB to 8x and bypass max to 4x, set drive/data strength to either 5:2 or 8:3, and disable Dynamic Counter. Personally, I prefer to spend the extra time to tweak then simply raising voltage unless I really have to. I'd also found something interesting... some time my system can prime 24hr but other time it failed within 30min (under same ambient temperature), it was very inconsistant overclock, but raising my 12v rail from 12.3 to 12.4v stabilize the system and the problem went away.