NVidia Coolbits Test failure...help?

[wolfman140]

Hey now,
Got a 7900GT, I'm maxed out at 1600 mem core, and climbing past 585 Mhz core. However, even though my temps are only hitting a max if 67C during 3Dmark, when I "Test" setting, I can't raise my core past 590 Mhz. So...its not cuz of temps. I'm thinking I need to do the couple volt mods on the card I've read about that let it OC higher. (I say this because Coolbits won't let u accept the OC changes without "testing" first) Would volt modding solve this problem? Thx!

 

[slicessoul]

IMO, Coolbits test are nearly the same as Memtest86. On simple way to say, it test the bits 1 and 0 are send/receive without error.
Memtest doesn't tolerate any errors, same story as Coolbits. If it received errors, then the test wouldn't passed.

Maybe try to use some software to raise the memory voltage. I heard some are exists such as Nibitor etc. If you want to go extreme, then do physically the volt mod.
Last word, BE VERY CAREFULL on doing that, i hope you've got some budget to spend.


*******PpPpPpPpPpPp$$$$$$$$$$$*****************
I'm not the programmer of Coolbits or Memtest. What i've said is just my opinion and some experiences on using them.
*******PpPpPpPpPpPpP$$$$$$$$$$$$$$$************

 

[wolfman140]

Aye there's 3 mods you can do...I believe the first two are totally safe because the chip on the 7900 is the same as its bigger brother, just modded down voltage. So if you raise it, its perfectly safe (using caution of course)
You mentioned that Coolbits acts like Memtest? Meaning Coolbits tests the gfx card memory right? Just asking because after 20 cycles of Memtest it DID have an error in one of my modules...Just wondering if that would affect gfx card OC performance. (I didn't think it did)
That wasn't Memtest86+, it was just Memtest, the one you run in Windows. So...I think today I'll bootup Memtest86 and let it run all night...if I get another error I'll take out one module and test one at a time. Off topic, but oh well.

 

[slicessoul]

It's not recommended to run memory test in Windows, that's what they said. It's better that you run it by using a Memtest86+ bootup floppy or CD.

It doesn't mean that your PC are totally bad when you find errors.

If the error are really exists, then yes it affecting your overall performance.
Such crash, hang, corrupted data would occur on error.

 

[wolfman140]

Aye I ran Memtest86+ all night last night with no errors...So...I imagine my memory is fine.
Well anyway, I guess the next step is Riva tuner for my card and to disable the Testing phase. Do you know of a program to test video IN A LOOP? I have the trial of 3DMark06 but I don't believe you can loop it. THanks

 

[13thmonkey]

Aye there's 3 mods you can do...I believe the first two are totally safe because the chip on the 7900 is the same as its bigger brother, just modded down voltage. So if you raise it, its perfectly safe (using caution of course)
You mentioned that Coolbits acts like Memtest? Meaning Coolbits tests the gfx card memory right? Just asking because after 20 cycles of Memtest it DID have an error in one of my modules...Just wondering if that would affect gfx card OC performance. (I didn't think it did)
That wasn't Memtest86+, it was just Memtest, the one you run in Windows. So...I think today I'll bootup Memtest86 and let it run all night...if I get another error I'll take out one module and test one at a time. Off topic, but oh well.

I know it is different, but the 7900GTO is a downclocked GTX and there are reports of some dying when the memory is increased, so just because it is a G71 (or whatever) core doesn't mean that the rest of the GPU card can cope with higher speeds.

Just a warning...

 

[wolfman140]

When you say "dying"/..you mean with signs of course yes? Not just outright "DAH!" dying. You'd see artifcating and/or crashes before you KILLED your video card with overclocking corect?

 

[slicessoul]

I never said that Memtest86+ is to test VGA card Memory.

I just said that Coolbits is "nearly the same" as Memtest86. What i mean "nearly the same" here is : the way it run the bench or the execution of bit/bits or other way to say it that i don't know. This is just my opinion or the way i think.

I would agree with you mpilchfamily that Memtest86+ is specially designed for testing only the PC RAM, while Coolbits is specially designed for testing only the VGA Card (core and memory). This one i know the differences, and i was supposing that wolfman140 knows the differences too because he said that he was using Memtest in Windows.

Or maybe, i got confused ?

 

[13thmonkey]

I've had artifacting on mine and so I knocked it back a notch or three. But I have heard of others, perhpas less patient, who pushed to GTX and it didn't survive for long enough for them to see the problems. These could be urban myths though.

 

[Mousemonkey]

They are not Urban Myths, they are sessions that end in or close to tears, so far (but not at the same time) 1 6600GS AGP, 1 6600GT PCI-e and almost a 7900GTO have been pushed over the edge by myself just using Coolbits, all the cards still work (to a fashion) but so long as you are aware of just how far off from the reference clocks you have gone (450/630 or 450/1260 in GDDR speak) then volt mod away to your hearts content, just bear in mind that graphic cards like everything else in your rig relies on the magical blue smoke to work and a graphic card will let that smoke escape just as quick and with as little warning as any other part.

 

[13thmonkey]

no need to tell me, I've given up on that game to a degree, I'm at a minor OC for now, precisely because of those fears.

 

[niz]

.
.. Just asking because after 20 cycles of Memtest it DID have an error in one of my modules...Just wondering if that would affect gfx card OC performance. (I didn't think it did)
That wasn't Memtest86+, it was just Memtest, the one you run in Windows. So...I think today I'll bootup Memtest86 and let it run all night...if I get another error I'll take out one module and test one at a time. Off topic, but oh well.

No but you've exceeded your ram capabilty if memtest has found errors. even 1 in 20 cycles is bad. The reason you don't want that at all is becase corrupted memory makes your machine behave unrealiably, which in turn can cause corrupted files or even screw up the formatting of your hard drive making the whole thing unreadable.

You need to get your memory back to settings where memtest86 never fails, otherwise you risk any/all files you have on your HD such as your windows install and pr0n collection.

 

[wolfman140]

Yeah well now you guys scared me lol. Maybe now I'll just get Rivatuner, disable the NVidia test, up it till I get artifcating and then back it down a few....forgetting about the volt mod. Doing that cautiously, even though Im getting rid of the NVidia Tester...I really have a slim chance of killing the card right?

 

[slicessoul]

I really have a slim chance of killing the card right?

Slim or not it's depend on your VGA card quality and other peripherals.

The people who replied aren't scaring you. They've just sharing their experiences on overclocking. It's a patience, and courage thing to do. When you've done it, you'll be proud of yourself.

 

[Mousemonkey]

At 585/1600 speeds you are already pushing that card way beyond what Nvidia intended, a standard 7900GT runs at 450/1260, both of mine now run @549/1652, be happy with what you've got and try to stay as far below the 'artifacting' level as you can bear, as for how long your card will last when it's OC'd?, how long is a piece of string?



@13thmonkey My previous post was directed at the OP, It was just the Urban Myth thing that I felt I had to share, I know your pain *sniff*, I've been the executioner of more than a few bits of hardware in my time. :twisted:

 

[wolfman140]

Aye...as far as I knew...the OC was safe for the card until A) The temp gets over 80C or so and B) artifacting began or a crash. If neither of those were the case, you could OC it to your heart's content...Or until you lacked the voltage. NVidia cards have a throttle on it that will crash the computer/card should temp levels get to high and may cause damage.
So...I will get Rivatuner, OC the Mem till artifcating begins...I'll record it...and reset it. After that I'll bring the core till artificating begins, record it and bring it back down. Then I will set both of them about 10Mhz-15Mhz below what their limits were. And as far as I know, if that setting doesn't artifact during 3dmark, then you should be good to go. Someone correct me if this is not the case.

 

[13thmonkey]

my old 6600Gt was stable at stock at 100C, nvidia's limit was 140C, but when OC'd it'd black screen and reset the PC at maybe 85-90C (dual screens helped with pinpointing this number, only if you are looking the right way at the right time)

So the temps may be a red herring, the thermal limit for stablity appears to drop very quickly when OC'd.

I've been artifact free in 3dmark and in ati tools on my GTO, but generating artifacts that are consistent within a level on BF2, i.e. for one load, that flagpole is always wrong. So it sounds like memory loading was the issue there, as opposed to memory reading 'in game' where i'd expect texture artifacts to move around randomly.

But I've dropped it 10-15 below that and now I'm fine.