Memtest + RAM issues

ChrissyT88

ChrissyT88

ChrissyT88

Associate
Joined
9 Mar 2008
Posts
1,039
Hi,

I have been having some stability problems with my PC recently, and have narrowed it down to the memory. I am currently running 2x2gb sticks of corsair pc-8500 RAM in an ASUS P5Q-Dlx motherboard. I have been having some random crashes, but have only today had some proper time to investigate. I am fairly confident that the instability is not caused by the CPU - it has passed a 7.5 hour prime stress test. The RAM has been running at 1104 Mhz (the closest to the rated speed i can get using any strap) at its rated timings and volts in the bios (2.1V in BIOS - although i am told by reliable people that the p5q boards overvolt the memory by up to 0.08v, so have kept this value fairly low).

I have run lots of memtest passes today, and cannot get the RAM to complete three passes successfully. I have upped the memory voltage by 0.04V and increased the NB voltage, but memtest still fails. The passes seem to only give a single error on either test 4,6 or 7, and i have a feeling that this could be causing instability. Interestingly, the RAM passed windows built in memory tester fine - no errors were reported. Can anyone sugggest some steps that could be taken to help eliminate these errors?

Thanks, any help would be appreciated.
 
Last edited:
Cheers Sweeny. I have mostly been having problems with the folding@home software. I have had a few BSODs whilst running it, and a couple were reported by windows as being caused by Kaspersky antivirus, alhtough i do not believe this is the case. There does not seem to be a pattern - the gaps between the BSODs can be days or weeks, but when running the more memory and cpu intensive client, i get either a failure of the software or a windows BSOD. For example, i am typing this on the same machine the blue screened a few minutes ago (in between posts!).

I will try relaxing the timings (although i would be disappointed if the RAM could not cope with a meager 40Mhz overclock at stock settings!). I am also considering re-running initial stress testing software to try and find problems. As for the speeds, i am given the option of running at either 1104/1101 or 978 (it passes three passes of memtest at 978Mhz) - these are the closest speeds i can obtain on any FSB strap (i am running a rather strange value of FSB to obtain a 3.3 GHz clock speed). I will also update the BIOS soon to see if that improves things.

EDIT: sorry cob, didnt see your post. Its just two sticks of RAM - maybe i didnt make that clear, but there are 4gb in total. I have not tried removing them and testing individually, and will do this a last resort - i am not well set up at uni to remove bits and pieces!
 
Last edited:
Cheers Sweeny - unofrtunately i have the displeasure of sleeping in the same room as the thing, so an overnight run is out of the question, but tomorrow i will aim to do ome good run throughs. I was also going to try and reset the OC, and see if that makes a difference - thanks for the advice!
 
That was the plan - would be nice to know, and is probably something i should have checked to start with, but tbh i was more concerned about getting the whole system up and running. Live and learn.
 
Thread resurrection - i have had a bit more time to test and fiddle around with the inside of the PC. On the plus side, it still seems to work! Downside, however, is that no matter what stick i use or what slot is put the stick in, i still get memtest errors in either the first or second pass. I have tried resetting the BIOS options (including running everything on auto), increasing and decreasing the voltages, running at slower than rated speeds and i get the same result. Im not really sure what to do next, and i cannot really be without the PC (no second PC).

As far as i can make out, this means one of two things:
1. Both RAM sticks are faulty
2. The motherboard is faulty

I would be inclined to say (and indeed i hope it is) the first of those options. I only seem to get problems when using RAM intensive apps (F@H, lots of apps running at once etc). I would have thought that if the mobo was broken, i would not be able to use windows etc for long periods of time without something going wrong. Anyone have any opions on this? Im not really sure how to tell if its the fault of the RAM or mobo - i guess i would need to try some newer RAM?

If i need to get some new RAM, i still have a few options. The first would be to buy a cheap RAM kit to test the mobo. Secondly, i could buy a 2x1gb set of dominators, then when the RMAd sticks come back, i could fill all slots and have 6gb of memory. Thirdly, i could get a set of g-skill pc-8000 memory - its competitively priced compared to the corsair. What would you do in this situation?

Thanks, any help would be appreciated.
 
Cheers Sweeny - 6gb could be a plus. Just done some reading around and some people have problems mixing the revsions of corsair dominator memory. I think i'll give OCuK a call and see if they can confirm the memory type. Sorry to have posted quite a bit, but im a bit of a novice when it comes to pc building, and i dont want to do the wrong thing!
 
Thanks for the replies. I bit the bullet and bought some new G.Skill sticks - they should be here on wednesday. I decided against the extra corsair sticks - some people have said there are compatibility issues with different revisions, and thats not something else i want to contend with.

@BlackAle - sorry i didnt reply sooner. I have tried all the current ASUS BIOSes - 1406 is supposed to be pretty decent and solve a lot of compatibility issues, but still errored on memtest. I must admit im not into using modded BIOSes - just in case!

I tried using both sticks in all slots on the mobo at stock settings, and still got errors on memtest within the first few passes. Hopefully this is a pretty good indication that the RAM is knackered (strange thing to be saying!). Its a bizarre problem - i havent had a crash since the original post - it really does seem to only be things that really push the RAM hard that cause it to slip up. I have downloaded a new (newest) version of memtest as well in case something is wrong there. Hopefully i can get it sorted. Thanks for your contributions guys.
 
In an ideal world i would - however, i dont have another PC at hand (at uni - butchering flat mates computers would be frowned upon!). Annoyingly i recently sold off parts for a pc that would have been perfect to use as a test bed. A shame really. The combination of RAM/mobo/CPU seems pretty solid - there are loads of people on XS running the same parts, partly the reason why they were chosen. Similarly, lots of people are using the G.Skill ram, so i figured this is possibly the best way to varify if the mobo is faulty.
 
Hi all - sorry to resurrect an old thread. I got hold of a set of G.Skill 2x1GB PC-8500 sticks to run memtest with some different memory. Unfortunately memtest still fails (latest memtest), very quickly with the voltage set to 2.1v and stock timings of 5-5-5-15 (everything else set to factory defaults, except the processor voltage). Im running F@H as a stability test of the RAM now, and will probably do a prime blend run, but am worried about the motherboard (i very much doubt two sets of memory could be faulty).

However, the errors i get on memtest are quite strange. It reports the bad memory as 000000000MB, and identifies the good and bad 'err bits' as the same thing (last run, good: dfdfdfdf, err buts: dfdfdfdf). This seems a little counterintuitive to me - could it be that memtest is not fully compatible with the memory controller on the p45 chipset? I have scanned the changelog on the memtest website, and cannot see any specific entry for p45, although it does have changes such as 'update for x38, x48...' ect listed. memtest also does not recognise the timings/speed of the RAM. Sound like a problem to anyone (memory gurus!)?

[Aside: Vista still feels pretty snappy with 2gb of memory installed - maybe of use to some!]
 
I heard this too - i went for the corsair/g.skill stuff because it does seem to be pretty much waht all p5q users use. Both are listed as being fully compatible (all slots filled) with the motherboard, whereas (almost) no OCZ RAM is for example. I must admit i am starting to think that it may be a memtest issue only. I have been running the F@H client for 5 and a bit hours now, and although its not using that much RAM, it seemed to find weaknesses with the corsair stuff pretty quick. Im gonna try and stick the corsair stuff back in tomorrow and give it a little more volts, then give it a prime run - hopefully this should give a better impression than memtest alone.
 
Right - update time. I think i have cracked it. I swapped the old corsair stuff out for some g.skill memory, and, as it turns out, am back to using the corsair.

Im pretty sure that instability issues in vista itself were caused by slightly undervolted RAM. I had heard that the board overvolts like a madman, and so wanted to be careful with the RAM voltage, but i guess that it was not enough. Perhaps this (fairly) widespread information needs to be looked at again. I upped the voltage a few notches on saturday morning and since have run F@H for 24hrs+. Whilst im sure this probably is not as good a test as prime, it achieves more, and to be honest the fact that this was not working in the first place was the cuase of the problem.

As for the memtest issues, this one is a little strange. I spent quite a while on google and in the end found some people using the p5q motherboards also having some issues with memtest. I cannot remeber where i found the information (on some godforsaken forum somewhere else!), but it was suggested that disabling USB legacy support/removing any USB hubs at startup would provide a solution, and that the fault was to do with the BIOSes. I did this tonight - unplugged a USB hub/keyboard and disabled USB legacy support in the BIOS, and ran memtest 2.01 from a bootable cd. The test ran for ~2.5 hours, completed three and a half passes, and did not find a single error. Since the memory could barely make it through two passes before (and the g.skill failed almost immediately) i am inclined to think that this is enough of a run to prove that the problem is a BIOS/USB issue (i'll let F@H be the judge of any long term instability). So, thats my fix, if anyone is interested. I'll see if there are any long term issues, and im keeping the G.Skill RAM in case something does go wrong, but for now i think its back to folding!

@BlackAle - this is something that may be fixed by Ket's BIOSes (im sure they are fantastic - he has clearly put in a lot of effort and i appreciate the support he has given the community) - however, since the issue seems to be memtest exclusive for me (for now), im going to stick to the official BIOSes released by ASUS.
 
No probs - everything was set to stock when running the tests over the weekend, so i cannot say for sure that the memory will be stable when overclocked. I was using BIOS 1406 for the p5q-dlx throughout - i have not tried with other BIOSes (although may revert back to 0803 at a later date). I had all voltages set to auto (with the exception of the cpu voltage, which i set to the VID), and increased the memory voltage to 2.16v (note: i did not try any other values for windows stability testing - took a i guess and it paid off). This survived the weekend running at 5-5-5-15 @ 1066 MHz (as i said, all stock). However, i am a little worried that this may be a tad too high (after overvolting), so i will look to minimise this when i have more time. However, for the moment, it works. When previously clocked to 3.3 Ghz using an FSB of 367, i had the NB voltage at 1.2v. However, i will probably increase this a little (maybe up to 1.26) to help cope with any extra strain (and from what i have read, this seems to be quite safe).

As for the memtest errors - try disabling legacy USB and removing any USB hubs. If you dont get any errors, this could be the source of your problem (on memtest at least - i have no reason to belive that this should affect windows stability). However, your problem (errors every 10 seconds) does seem a little more severe than mine (only a few errors), so i would try a good prime run with everything on auto except for the dram voltage.
 
You must log in or register to reply here.
Back
Top Bottom