GPU......again.

Arthur.S wrote on 11/13/2014, 10:13 AM
After a long time of relative stability with V12 and GPU, out of the blue I get crash after crash when editing in multicam. Normal editing is still OK. Error message: unmanaged exception. C:Windowssystem32nvopencl.dll

Turning off GPU sorts it, but I'm baffled as to why it's misbehaving now. No Windows updates, and no software installed or updated recently. (I keep my editor PC off line) Tried a system restore just in case, but that only offers me one from 3 days ago (which didn't work) or a saved one from February. Don't really want to go back that far, but will do if I can't find anything else that works. Any advice appreciated - including why I can't see any other restore points!

Should also add that I tried re-installing the driver. Card Is Geforce GTX 770.

Comments

Tom Pauncz wrote on 11/13/2014, 11:55 AM
Check to see if perhaps the card is overheating.
Tom
Arthur.S wrote on 11/13/2014, 2:00 PM
How do I do that Tom?
Tom Pauncz wrote on 11/13/2014, 3:52 PM
I use a free tool for that called GPU-Z, from TechPowerUp.
Tom
Byron K wrote on 11/13/2014, 4:51 PM
You can try and re-seat the card in the slot?

My graphics card was acting up and found the motherboard slot went bad and by moving the card to the 2nd SLI worked.
Arthur.S wrote on 11/15/2014, 6:41 AM
OK, I've installed updated 334.89 drivers, and GPU-Z. GPU-Z shows a normal temp of about 50C. When rendering this rose to about 56C. On a short Multicam edit though it rose to 61C. There were no crashes though. Are these normal temps?
Tom Pauncz wrote on 11/15/2014, 10:58 AM
Those temps seems to be in the same range as mine. I have an old GeForce 9800GT card in the system.
Tom
OldSmoke wrote on 11/15/2014, 11:31 AM
Perfectly normal temps, I had two GTX570s. Gamers will run those cards in the mid to high 80s for hours. I have been for a long time on driver 296.10 but about 4 month back had some issues after a Windows update and switched to 334.89 with my OpenCL memory tweak and that works very well. Also the 360xx driver works very well and is slightly faster.

Proud owner of Sony Vegas Pro 7, 8, 9, 10, 11, 12 & 13 and now Magix VP15&16.

System Spec.:
Motherboard: ASUS X299 Prime-A

Ram: G.Skill 4x8GB DDR4 2666 XMP

CPU: i7-9800x @ 4.6GHz (custom water cooling system)
GPU: 1x AMD Vega Pro Frontier Edition (water cooled)
Hard drives: System Samsung 970Pro NVME, AV-Projects 1TB (4x Intel P7600 512GB VROC), 4x 2.5" Hotswap bays, 1x 3.5" Hotswap Bay, 1x LG BluRay Burner

PSU: Corsair 1200W
Monitor: 2x Dell Ultrasharp U2713HM (2560x1440)

Arthur.S wrote on 11/16/2014, 8:17 AM
Yes, I've also done that memory tweak OS to 1024. Overheating isn't the issue then.
Arthur.S wrote on 11/19/2014, 6:29 AM
Not fixed. Latest multicam project, crash, crash, crash. Turn off GPU. 100% reliability. Weird thing is, the preview is still as smooth!
farss wrote on 11/19/2014, 7:08 AM
[I]"Turn off GPU. 100% reliability."[/I]

Why anyone would turn GPU on escapes me.
SCS should change Vegas so it's Off by default, that'd save Support a lot of work I think.

[I]"Weird thing is, the preview is still as smooth! "[/I]

What's weird about that?
The only thing I use the GPU for is when working with vector graphics, I'm not a gamer so that's limited to the times I play around with 3D apps and CAD. Even there it's recommended to not use the GPU for final renders as the quality is not as good as CPU Only.

Bob.
OldSmoke wrote on 11/19/2014, 8:51 AM
Oh, another GPU hater.
I never had issues with the 2x GTX570s I had previously in my system, nor with the 2x GTX580s after that or the current 2x R9 290.
But I do agree that GPU acceleration should be off by default as too many users don't have the right hardware for it anyways.
I also don't see any noticeable difference between GPU or CPU rendered videos but I don't really render at bit rates below 6500; if anything lower I used HB and of late frameserve to HB if I need something for the Internet.
I would rather say that the issue is with the GTX770 not being fully supported hence there isn't much difference between ON or OFF. Does this happen with a particular multicam project or any?

Proud owner of Sony Vegas Pro 7, 8, 9, 10, 11, 12 & 13 and now Magix VP15&16.

System Spec.:
Motherboard: ASUS X299 Prime-A

Ram: G.Skill 4x8GB DDR4 2666 XMP

CPU: i7-9800x @ 4.6GHz (custom water cooling system)
GPU: 1x AMD Vega Pro Frontier Edition (water cooled)
Hard drives: System Samsung 970Pro NVME, AV-Projects 1TB (4x Intel P7600 512GB VROC), 4x 2.5" Hotswap bays, 1x 3.5" Hotswap Bay, 1x LG BluRay Burner

PSU: Corsair 1200W
Monitor: 2x Dell Ultrasharp U2713HM (2560x1440)

Hulk wrote on 11/19/2014, 9:46 AM
I also pulled my GPU long ago. The GPU in my 4770k provides Best/Half preview at full framerate for pretty much anything I do and the GPU doesn't assist when I frameserve>render so for me there was no point in dealing with the instability since the "return on investment" was so low. Now if the HD4600 in the 4770k didn't work with Vegas that might be another story and I'd still be struggling with it. My younger self would still be trying to make it work "just because." But my older, wiser self just moved on and is getting work done...
Arthur.S wrote on 11/19/2014, 1:36 PM
No, it's all multicam projects. I'll repeat though; It's been flawless for almost a year now on everything, inc multicam.
farss wrote on 11/19/2014, 2:09 PM
Possibly failing capacitors.

Bob.
Arthur.S wrote on 11/20/2014, 2:32 AM
NOW you've worried me Bob! ;-( A google on that produces this set of 'clues':

Difficulty turning on your computer, frequently having to hit reset or restart
Instabilities (hangs, seeing the “Blue Screen of Death,” etc.), especially when the symptoms get progressively more frequent over time
Memory errors, especially ones that get more frequent with time
Spontaneous reboots
In case of on-board video cards, unstable image in some video modes
Failing to complete the POST (Power-On Self-Test) or rebooting before completion
Never starting the POST; fans spin but the system appears dead
Power system malfunctions, fluctuations in system voltage, possibly with an increase in CPU temperature as the core voltage rises
In the case of Dell Optiplex GX270s, often a “Thermal Event” is brandished in white on a very dark computer display when rebooting

I'm not suffering any of those. All completely normal...apart from this multicam thing.
farss wrote on 11/20/2014, 4:06 AM
Electrolytic capacitors are one of the few electronic components that can degrade over time. As they lose the ability to do their job the symptoms are highly varied, after all they're in every part of electronic devices. Typically though the problem will start to show up when the computer is working hard unless of course it's say the capacitors in the power supply that have simply blown up, then probably the PC will not boot. You'll usually know If that happened due to the really bad smell.
From the guys that are into the "recapping" thing around 80% of the electronics that ends up on the kerb is there because of capacitors going bad.

The other thing that can cause weird sudden onset problems is connectors, try reseating [I]everything[/I]. A few years ago my DVD duplicator seemed to have two drives fail. The local agent was happy to sell me new drives but was pretty sure the problem would be fixed by reseating all the ribbon cables and he was right.

Bob.
Arthur.S wrote on 11/20/2014, 6:20 AM
Here's another strange one that's popped up today; A render with a cocktail of 4 FX, including Neat video, was showing (once it settled) at 21hrs to render. (CPU) I was expecting it to take a while because of the Neat video, but nowhere near that! Cancelled it and changed to GPU......went up to 32hrs!!!
Never had a render that long - EVER. All of the FX were at track level, so deleted them, then reapplied EXACTLY THE SAME FX in the SAME ORDER by applying them to first clip, then copy/paste to rest of track, and the render (CPU) has dropped to around 5hrs, which is about what I was expecting. WTF?? Is it normal for track FX to take longer than event FX to render? Can't say I've ever noticed it before. Maybe I should start a new thread with this one.