Damn DIMM?!

[Preacher]

Gents,
Just need to bounce a theory off you all. As mentioned in my mistaken Firefox thread (http://www.short-media.com/forum/showthread.php?t=36281), my system has been unstable over the last couple weeks. I originally thought it had something to do with my upgrade to the latest edition of Firefox, but now I think it's 1 of my 2 G.Skill DIMMs.

I know PSUs can also cause instability, but I have a PC Power and Cooling 510 SLI. Its rails seem okay at load (3.28/5.02/12.14). So I then started testing my memory....

I am not overclocking at all and am actually running the memory at 200MHz at 2.5-4-4-8 when it's rated at 2-2-2-5. With both DIMMs installed, I ran Memtest the first time overnight and got 4 errors in 7 hours. So then I decided to see which was responsible except they both have passed Memtest 3.1 as the only DIMM installed. I then ran the Prime 95 torture tests with DIMM #1. It ran test #1 and #2 for hours, but Test #3 crashed after 55 minutes. After having only it in my system for 3-4 days, I have had no crashes.

Then I installed DIMM #2 yesterday and ran Prime 95. Test #1 was okay, but Tests #2 and #3 crashed almost immediately. AND back came the BSODs..

SO I'm thinking DIMM #2 is the culprit...anything I might be missing or other way to check?

Thanks,
Jeff

[TheBaron]

if memtest hit errors, your memory is bad

prime95 is not a memory test, it is really only a test of overall system stability, and thats not what you want in this case. if you're trying to prove whether one stick or the other is bad, try running memtest with only 1 stick, then memtest with the other one. whichever one hits errors is bad

[shwaip]

alliteration aside:

it could be a problem with one of the dimm slots - try just testing the "good" stick in the second slot.

[TheBaron]

not if stick #1 works in slot #1, but stick #2 does NOT work in slot #1, it is more likely its just a memory issue and not a bad slot

[shwaip]

So then I decided to see which was responsible except they both have passed Memtest 3.1 as the only DIMM installed.

[ryko]

what voltage are you running it at? try adding some for stability... also, it has been my experience that low latency stuff usually doesn't like being run at very relaxed timings. you should try it at its rated 2.2.2.5 and around 2.7/8v and see if that helps. plus if your g.skill is tccd, it should have no problem running 250+ at cas2 timings...no need to relax the timings at all.

[TheBaron]

kk :D

[Mt_Goat]

what voltage are you running it at? try adding some for stability... also, it has been my experience that low latency stuff usually doesn't like being run at very relaxed timings. you should try it at its rated 2.2.2.5 and around 2.7/8v and see if that helps. plus if your g.skill is tccd, it should have no problem running 250+ at cas2 timings...no need to relax the timings at all.
That is absolutely correct! Mine won't run relaxed at all and at anything below 230 errors with anything but 2-2-2. In this case voltage doesn't help much either. But if one stick is behaving dirrerently then an RMA might be in order anyway.

[Preacher]

TheBaron,
I tried Memtest three times. The first test with both DIMMs inserted gave me 4 errors in 7 hours. Then I Memtested each DIMM and both passed which makes no sense. I'm no expert on Prime95, but quoting from its torture test options, "Blend (tests some of everything, lots of RAM). It's weird that the suspect DIMM can pass the Prime95 test that is easy on RAM, but fails the other two tests which supposedly stress RAM. BTW, I say fail, but it usually means BSOD.

I have the RAM voltage set to 2.9, but the BIOS reports 2.98 volts. Also, while it says 2-2-2-5 at PC3200 (DDR400) on the DIMMs themselves, CPUID and the BIOS AUTO setting are 2.5-4-4-7. Why the difference? :scratch:

The DIMMS are G.SKILL F13200DSU2-1GBLE.

[ryko]

maybe it's too much vdimm. you are at almost 3.0v. try setting it 2.6/7v and actual will be 2.7/8v. it depends on what type of chips they are. is it tccd?

also, i have seen many things (BIOS, cpuid) incorrectly report the spd timings. not a big deal. go with what they are rated for from the manufacturer. if you really want them relaxed, try 2.5.3.3.6---but it should be stable at 2.2.2.5 or it's RMA time.

[Preacher]

It is TCCD. The 2.9 setting was recommended to me on the DFI-Street Forums.

[ryko]

seems high for tccd. it doesn't need lots of voltage to get the job done. especially when trying to run stock. probably 2.9v if you want to find its max oc.