OCCT test stable, unstable in Linpack (not overclocking)

[adriangb]

My machine has been unstable since day 0 (everything points to the RAM, which hasn't ever worked right), but I recently sold one of my two sticks of RAM and since then have reduced it from 1600MHz to 1066MHz, to see if I can get it a little more stable. At first I tested it, and it seemed to have worked (I could get threw 1-2 hours -at first I didn't test more- of OCCT/P95, whereas I used to be able to do <10m), but then I ran Linpack (both IBT and the OCCT version), and they both failed in less than an hour. So I went back to OCCT Large Data Set (previous to the RAM changes I had been using small data set, which worked fine for as long as I ever tested) and that's been running for 13h straight, without issues. One of the things that are wierd is that my vCore seems very high for it being on "auto" in the BIOS: it runs at <1v when EIST kicks in @1.6GHz, but when it goes to 2.8GHz it jumps up to 1.2, then 1.23 and even 1.25 for a few secs, although it usually just sits at 1.23.

My Build:
i7 920 @stock w/ stock heatsink and AS5
ASRock x58 Extreme BIOS v2.5
PNY GeForce 9800 GT EE
OCZ GameXStream 850W
OCZ Gold 4GB kit (now 2GB) DDR3 1600 LV

So does failing Linpack mean I'm still unstable? I had gotten my hopes up with the first few OCCT tests...
Is my vCore too high?

Thanks for the help .

 

[wolfram23]

You could run Memtest86+ to check the RAM.

Also, you might consider getting that stick back and RMA them for new ones...

 

[adriangb]

You could run Memtest86+ to check the RAM.

Also, you might consider getting that stick back and RMA them for new ones...

The RAM should be good, since I already did an RMA and have tested the new sticks in a frieds P55, and they ran fine. Also, if you check ASRock's web site and OCZ forums you will see that there have been lot's of issues, although mostly with the 6GB kit. ASRock seems to have fixed those in a BIOS update, but I guess that it doesn't apply to the 4GB kit, I know, weird huh? I just can't think of a better explanation .

Now that you mention it, I haven't run memtest yet (in this last session of testing, lol, I've run it one million times before -always errors ). I'll give it a go and report back.

EDIT: One pass of memtest went fine, but since I'm running @ 1066 it will probably take multiple passes to get any errors. I don't have time to do them now, but I'll leave memtest running when I go to bed, and I'll post the results tomorrow.

 

[adriangb]

memtest failed on pass 37, so I guess I'm only slightly unstable, right?
why OCCT doesn't ever fail, I don't know, but it seems strange...
and what about the vCore?

Here are my memtest results:

 

[wolfram23]

Hmm. This is a bit of a doozie. You said you already RMA'd the RAM, so I would think the problem isn't there, although I guess it could be... if possible, maybe try a totally different brand of RAM and see what happens. Otherwise it could be the mobo's controller or else even something wrong with the CPU's IMC, although that seems doubtful.

I'm assuming you have the updated BIOS with your mobo. I think right now all you can do is take your problem to OCZ and/or Asrock. One of them needs to give you a refund lol.

 

[adriangb]

Hmm. This is a bit of a doozie. You said you already RMA'd the RAM, so I would think the problem isn't there, although I guess it could be... if possible, maybe try a totally different brand of RAM and see what happens. Otherwise it could be the mobo's controller or else even something wrong with the CPU's IMC, although that seems doubtful.

I'm assuming you have the updated BIOS with your mobo. I think right now all you can do is take your problem to OCZ and/or Asrock. One of them needs to give you a refund lol.

I am going to change the RAM, but living in Argentina I can't just look for a deal and stroll out to Fry's or something and get RAM, andk go back the next day and return it if it doesn't work, lol. I'm going to change it around christmas, when I should be in the US and will be able to buy it cheaper. The only issue is I can't really take the whole PC over there, so I won't be able to test in until I get back, and hence can't return it until the next trip.

As far as the problem goes, I think it's related to the MB, since the RAM seems to be fine, and this CPU used to at least run this same RAM in dual-channel (not without errors thou), until I got the new MB (I RMA'd it too, trying to fix this). I asked OCZ to send me a different kit, be it the triple-channel one (I said I'd pay for the difference) or some other better documented (aka "known to work with my MB") kit. Of course they said no .
So my conclusion is it's either just compatibility (I hope) or the MB, in which case I'll find out when I get new RAM. I'd have to have some REALLY bad luck thou for it to be the MB. I actually have to check the warranty period on it, if it's a year I'm over that, YIKES!

 

[Shadow703793]

First off, OCZ RAM is notorious for having compatibility problems (esp. the Gold, I had them a long while back with the first rev of the X58-UD3R and I hated it; I had to relax the timing a LOT and drop speed).

My recommendations:
1. Try increasing RAM voltage and see if Memtest86+ is stable.

2. Get rid of the OCZ RAM and grab some GSkill or Corsair RAM.

3. That's not the best of boards, but it should be OK. Try it on a different board (as in test with Memtest86+).

 

[adriangb]

First off, OCZ RAM is notorious for having compatibility problems (esp. the Gold, I had them a long while back with the first rev of the X58-UD3R and I hated it; I had to relax the timing a LOT and drop speed).

My recommendations:
1. Try increasing RAM voltage and see if Memtest86+ is stable.

2. Get rid of the OCZ RAM and grab some GSkill or Corsair RAM.

3. That's not the best of boards, but it should be OK. Try it on a different board (as in test with Memtest86+).

I'm afraid to go past the "stock" (it's already over stock for an i7) 1.65v, but if you guide me thru it I'll try.

I wish I could do Nº2 right now, but unfortunately I can't. It it my long term plan thou.

I knew I was getting a budget board when I bought it, but I didn't expect something like this. Next time I'll go Gigabyte . On a friend of mines P55, it ran thru three passes of memtest86+ without issues @stock. On mine, it won't go past the first pass @1600Mhz. We also ran Linpack on his board, no issues.

 

[Shadow703793]

Ah... you are already at 1.65v....

I knew I was getting a budget board when I bought it, but I didn't expect something like this. Next time I'll go Gigabyte . On a friend of mines P55, it ran thru three passes of memtest86+ without issues @stock. On mine, it won't go past the first pass @1600Mhz. We also ran Linpack on his board, no issues.

If that is the case, is there any way you guys can "exchange" the RAM with each other? I'm 99% sure the problem is the cr@ppy RAM. Your board is probably fine.

 

[adriangb]

I think we did try that, I'm not sure thou. I'll ask him if I can do another testing session.

 

[randomizer]

memtest failed on pass 37, so I guess I'm only slightly unstable, right?
why OCCT doesn't ever fail, I don't know, but it seems strange...

OCCT may use a test that fits in the CPU cache, thus avoiding issues with unstable RAM. FWIW, any test failure in Memtest86+ is not good enough. While a slightly unstable CPU really needs to be stressed to cause issues, unstable memory (whether system or video RAM) can cause the most incredibly random problems when you're doing hardly anything at all.

 

[adriangb]

OCCT may use a test that fits in the CPU cache, thus avoiding issues with unstable RAM.

That would be OCCT small data set, right? I've been using large.

FWIW, any test failure in Memtest86+ is not good enough.

Knew that. Memtest86+ should run "indefinitely" without errors.

While a slightly unstable CPU really needs to be stressed to cause issues, unstable memory (whether system or video RAM) can cause the most incredibly random problems when you're doing hardly anything at all.

Didn't understand that part, lol. Do you mean to say that unstable RAM can cause random problems that affect CPU stressing?

 

[randomizer]

Unstable RAM can cause just about anything to go wrong, whether the system is under stress or not. CPU instability, providing it is only minor, often doesn't show symptoms until there is a lot of stress on it.

 

[adriangb]

Gotcha. My system is perfectly usable: it doesn't consistently get blue screens (I have gotten them thou, but very randomly and usually after I did some action, so it's hard to blame them all on RAM), and I can use it 24/7 without issues. But I know that the instability is there, since Memtest fails and Linpack fails.

 

[adriangb]

Update: I'm kind of occupied latley, and haven't been able to get together with the guy who's got the mems. I'll post when I have.

 

[adriangb]

I'm back
I'm with the guy who has the memos, he has a pair (as in 2x2GB) Corsair DDR3 2000. I put them in my system with the same results: in the right slots, I get 2047MB & "Single-Channel", aka it only finds the first module.If any of you guys suggest any tests (in the next few mins) I can try them.

 

[Brian_09]

A little late but just my 2 cents here... I had a problem- Asus P6T7 980x OCed to 4.2Ghz with Crucial Tracer 1600s actually ran better at 1.64v rather than 1.65v very small margin but the difference in being stable and unstable! OCCT runs very stable on all tests set to highest data rate possible.

My Build:

Asus P6T7 SuperComputer -- Koolance Full Coverage Water Block
Intel Core i7 Extreme Genuine Intel(R) CPU 000 @2.4Ghz (ES)-- OC @ 4.2Ghz -- Swiftech Apogee Water Block
Crucial Ballistix Tracer 3x2GB 1600 8-8-8-24-1T -- Koolance Ram 33x3 Water Block
Asus GTX 480 875/2025 1.125v 35c idle/up to 70c load -- EK Nickle Acetal Water Block + Nickle Backplate
Asus GTX 460 800/2005 1.075v 28c idle/up to 56c load mainly for physx -- EK Nickle Acetal Water Block + Nickle Backplate
Agility 2 60GB Boot Drive
RAID 0 C300 128GB
RAID 0 Raptor 300GB
Antec TPQ 1200
RP-450x2 Double Bay Resevoir with 2x Swiftech MCP 655-B controlled with sunbeam Rhoebus
All Koolance 3/8 compression fittings with Blue Fesser Tubing illuminated by three 8 inch ccfl black lights
Haf-X case with 2x triple rads stacked inside case and 1x single 120mm rad outside for combined 2400watts of heat disapation!!!

Oh yeah did i mention AWSOME!!!---Why because Im a noob and i did it all myself!!!