Overclock 550BE and graphics card blows?!!

[ingodwestream]

I built a PC this weekend from some parts I had laying around with a deal I got on a AMD Phenom II x2 550 Black Edition. Ran AMD Overdrive App and while Prime95 testing with 200x18.5 (3700mhz)@ 1.375v the screen goes to multi-color horizontal swiggly lines 15 minutes into the stress. reboot and same thing even during post. remove card, no issues, put card back in, same error effect. Mind you that it was a lowly HD2600 PRO 256mb ddr2 card. But I was wondering was just an bad card i had or was it something I did while OCing?

Phenom IIx2 550 BE
Kingwin XT-1264 HSF
Mushkin 4gb DDR2 (2x2gb) 4-4-4-12
ECS A780GM-A Black Series
80MM NB fan
HD2600 PRO 256mb ddr2

 

[erdinger]

hmm the only thing you could have done is increasing the PCI-e speed which is unprobable. If you havent change the PCI_Express frequency (100 is standard) then it was just a bad coincidence I would say.

Another possibility is that you PSU couldn't handle the extra power draw

 

[aberchonbie]

card could be overheating from excess heat in case

 

[stridervm]

I think it's just a dying video card.

 

[Kortana]

Might be a bad card, CPUs drawing too much power, or the mobo has problems with OCing and distributing voltages stably.

 

[JofaMang]

While not as extreme of a situation, I experienced a massive drop in my stable GPU overclocks with the upgrade of my CPU from a 710 x3 to a 965BE. I had to lower my 4870 to 775/1000 from 795/1075 after installing my new CPU yesterday, with no other changes at all.

I'm not claiming that we have the same issues with the same causes, but giving an immediate example of a similar situation, if only different in severity.

 

[ingodwestream]

Thanks guys. I didnt increase PCIe freq.
1st - moved x15 ---> x18 @ stock 1.5v [ Prime stable 20 min]
2nd - moved x18 ----> x18.5 [Prime BSOD 10min]
3rd - x18.5 @ 1.5 ---> 1.75v [Prime 15min, BAM! rainbow screen!]

Case is APEX 3620 http://www.newegg.com/Product/Product.aspx?Item=N82E16811154098
front 120mm rear 120mm, 120mm hsf blowing directly to rear fan, and 80mm fan between gfx card and hsf blowing down on nb heatsink.
PSU is Etasis ET750 750watt (the OEM maker of the Silverstone Zeus 750w ST75ZF)

Funny thing is i bought this card back in 2007 and never used it. Since this was going to be a non-gaming build I thought I'd finally put it to use. I think it's like you all said.. just a coincidence and/or bad card.
I'd say if this card had a personality it would have said "you leave me laying around for 3 years and NOW you want me to work? F you!" lol
now where's that old x1650 at...

 

[ingodwestream]

While not as extreme of a situation, I experienced a massive drop in my stable GPU overclocks with the upgrade of my CPU from a 710 x3 to a 965BE. I had to lower my 4870 to 775/1000 from 795/1075 after installing my new CPU yesterday, with no other changes at all.

I'm not claiming that we have the same issues with the same causes, but giving an immediate example of a similar situation, if only different in severity.

what do you use to monitor your gpu stability? I'm on the beginner side of overclocking. I figure I'd learn the rope by starting with CPU OC (reason for purchasing x2 500 BE) then move onto RAM (Mushkin CL 4) and then GPU (future card purchase). I'm not deterred from future OC because this was an old card i bought years ago, though I never like to see money gone down the drain like that so fast. But I sure would hate to see this happen again once I get a 4 or 5 series Radeon card. If anyone has any good read links for stuff like this or preventive measures to take...i'd appreciate it.

 

[JofaMang]

You have to understand the limits of your components as far as heat and voltage go. I have not changed the voltage on my GPU, so it is mostly heat and artifacting that I am watching for when OCing my card.

I run and monitor Furmark for at least 10mins. Furmark causes 48xx cards to shoot flames, so I don't leave it alone. It will often just crash with a VPU recovery event if the clocks are too high. If it passes 10mins in furmark without too insane of temps or crashing, I will run benchmarks of 3dmark06 and 3dmark vantage, in back to back sets of 3 each. I will do all of these without rebooting and little to no downtime between. If after all that I am still running smooth, I load up H.A.W.X., a fighter pilot game, but only for the menu screen, as there is a rotating 3D background that for some reason runs my GPU at 100%, with comparable temps to furmark, if only a little lower. The big benefit to HAWX to me, is that it will show artifacting and other graphical glitches long before it actually crashes, giving me a better idea as to how much the card can realistically take, by seeing how long before it artifacts, if at all. I consider my card stable if it can run HAWX menu screen for at least 30mins. I have never had a setting fail in any other game or benchmark after it passes these accumulated tests.

I use Catalyst Control Centre (CCC) to change my clocks, and GPU-Z to monitor the Core, memory, and shader temps as well as the actual GPU load.

 

[ingodwestream]

Thanks JofaMang! I found my old backup x1650 card and used that until my new GTS250 1GB arrived. I think this premature card failure was just Him letting me know that I deserved a better card for my rig. I will adopt your monitoring practices and make changes with Nvidia's Smart Doctor app.
God Bles

 

[ingodwestream]

Best answer selected by InGodWeStream.