So I started having problems with my desktop about a week ago. While playing games, I started seeing lots of artifacts and glitchy video, and soon after the glitchiness started happening while not in games (while surfing, etc.).
I replaced the video card, but I'm still having problems: in Skyrim, shapes momentarily become pure blue or red; in Civilization V, after playing for an hour, the whole screen turned to video nonsense.
I tried running some diagnostic tools. I ran the older version of memtest that was included with my linux distro, and it turned up no errors in 1 pass. This was 2-3 days ago.
Today, I tried to run some of the UBCD cpu diagnostics, and all failed to load with kernel panics. I booted to Windows (XP) and ran prime95, but for some strange reason, it wouldn't run tests with the minimum fft size as 8mb, I had to set it to 10mb. I ran the combo test for 3 hours and it showed no errors or warnings.
I tried to boot back into the locally-installed memtest86, and the computer restarted after showing me the program for a split second. I'm now running memtest86 off of the UBCD with no problems, I'm going to leave it overnight, but it's run 2 passes and is showing no errors.
Any ideas, or thoughts on how I could further diagnose to find the problematic part? I'm thinking it's either the PSU, which probably killed the old video card and is causing the memtest restarts; or it's the cpu or motherboard, but the prime95 tests passed fine once I raised the
My setup right now is: MSI 975x powerup edition motherboard; intel e6400 core2 duo cpu; 4gb ddr2-800; evga gtx550 ti video 1gb video card; 2 sata disks.
The Prime95 test I did was blended, but for whatever reason I had to change the minimum fft from 8mb up to 10mb to get the test to run. The large FFT ran fine. I googled to see if anyone had this same problem but didn't find anything.
The power supply is a seasonic 430 watt passive cooled. The passive cooling is a big part of why I suspect it, but it's worked for the last couple years without a hitch.
I remember I initially had OCZ ram that required raising voltage, but this RAM requires just regular stock voltage. I haven't upgraded the BIOS in a while, and I always left it at default settings. Nothing is overclocked. I'll try a BIOS upgrade tomorrow, if there is one available, and resetting to factory.
Memtest is running fine off the UBCD with no errors 4 hours in. I'm going to leave it overnight, but I suspect it's not RAM.
Do not update the bios while the system seems unstable. Do not want it to crash while doing that.
How hot does this power supply feel?(do you have a link the the exact power supply you have?)
Do you have a rear case fan to help move some air under the power supply?
Do you get an error when you try 8k(I see you list megabytes, but is it K instead?) under custom prime? If so, i have never seen that one.
Ha, I just checked the psu again. It's not passively cooled at all. There's a 12cm fan on the underside, just no fan at the rear. It doesn't feel hot at all, feels room temperature. The model is SS 430hb active. Here's a link to specs: http://www.newegg.com/Product/Product.aspx?Item=N82E168...
I just noticed that the video card box says it needs 24 amps on the 12v rail, but the specs of my psu only claim 15. Could that be the problem?
I do have a rear case fan, a big 12cm one.
And the error I get from the program is "no fft lengths available in range specified." I just found a forum thread elsewhere that says it might just be a problem with the version of prime95, so I'll try another one tomorrow.
The 12 volt amperage you need is the total combined. In many cases you can not just add the 2 rails or more rails(on this power supply you can), but as you see it shows 348 as the total 12 volt wattage.
You are good on that front.
When you changed cards, you got another identical one?
Are you running the latest drivers?
Have you tried the Beta's?
It may be worth monitoring the video card temperatures(should not be,but if it takes a while to crash) with something like GPU-Z since it can log to a file so even after a crash, it should still have logged the temperatures for you.
Yep, updating drivers was one of the first things I did. I'm running the newest beta drivers.
The card I got was not the same-- the old one was an amd/ ati hd 4830, the new one is an nvidia gtx 550 ti. I have been monitoring temps with speedfan, and they get into the 60s (Celsius), but no higher. My old card ran a little hotter, into the 70s.
Thanks for all your help, nukemaster.
I'll post an update tomorrow when memtest and prime95 have run for 8-12 hours each.
Downloaded an older version of Prime 95 which allowed me to run 8mb fft size, ran blended test for 14 hours, no errors, no warnings.
I do have another computer I can put the video card into, but I don't think it has the 6-pin video connector, so I'd have to install this computer's power supply in the other computer with it. Since it doesn't seem to be the motherboard, cpu, or memory, doing that wouldn't remove anything further from the equation.
Is there a good way to test the video card itself, to somehow separate testing the psu and the card?
Though, it seems like 14 hours of stress testing would probably make a bad psu show itself with a system restart or shutdown, right?