Possbile damage to gpu?

[Deleted member 1560910]

I was running my 960 overclocked
50+ core
500+ memory
no volt increase

And was having instability issues and I didn't know it. I just thought battlefield 4 was being buggy again. I dropped the overclock and the problem went away. Will this have done any damage to card ?

 

[Deleted member 217926]

Very doubtful. The card would have throttled itself to prevent damage if it overheated. You could try dropping the memory overclock as it's probably what's causing the problems and memory overclocks don't really add any performance.

 

[Dunlop0078]

No I highly doubt it. Whats with +500 offset on the memory? Thats likely whats causing the instability. I have found overcloking the memory does very little for performance in most games and it sometimes takes away what you could add to the core. Most of the time I dont overclock the memory at all, I have tested it and found it does little to nothing in most games. Less on the memory more on the core.

 

[Deleted member 1560910]

thanks guys I had found this overlock on a page and decided to use it. I am still pretty new to overclocking gpu

 

[JackNaylorPE]

Unless you use a hacked BIOS or physically modify the PCB, it's close to impossible to damage a 9xx series card. nVidia's restrictions, both legal and design wise, stop you and the 3rd party card manufacturers from doing anything voltage wise that would damage the card.

As for memory overclocking, here's the results I had on my twin 780s (3D Mark Vantage)
+366 = 72,201
+413 = 74,574
+460 = 74,515
+514 = 74,803
+601 = 75,021
+656 = 74,076

Tests were not done at max core clock because I wanted to make sure that I knew the memory limit and was not seeing any degradation from the core OC. Note IIRC, the card came with a +366 on the memory "outta the box"

 

[Deleted member 217926]

^ Do note that those are synthetic benchmarks that translate into about 0-1 FPS in a real gaming scenario. The core clock is where you will see gains.

 

[RobCrezz]

Very doubtful. The card would have throttled itself to prevent damage if it overheated. You could try dropping the memory overclock as it's probably what's causing the problems and memory overclocks don't really add any performance.

Vram speed increases on these cards with low bus widths actually do increase performance by quite a lot!

 

[JackNaylorPE]

^ Do note that those are synthetic benchmarks that translate into about 0-1 FPS in a real gaming scenario. The core clock is where you will see gains.

Unlike CPU synthetic benchmarks, graphics benchmarks do in fact duplicate real gaming scenarios. I use RoG Realbench for CPU because it uses real programs and is therefore indicative of real world program performance. Same w/ the GFX benchmarks... they are in fact indicative of real world gaming performance, they are just not using any game developer copyrighted material. The Unigine series as well as Firestrike , Cludgate, Ice Storm are examples.

Core certainly has more of an effect than the memory. However, how you test and what you test comes into play. For example one might conclude based upon single card testing that more or faster system memory has little effect on average fps. But add a 2nd GFX card and we see significant upticks in average fps and even more so on min fps.

OCing the GDDR increases the memory bandwidth, which does help for example when using anti aliasing and anisotropic filtering. It also will have much more of an effect when you are texture thrashing, and more so at higher resolutions. Using high resolution with high AA, any type of sync, buffering, etc., memory overclocking has a significant effect. Of course what game you are testing with ... matters just as with system memory where games like Metro 2033 show no gain and F1 jumps 11% going from 1600 to 2400. With video memory, games like Civlization V and RAGE are two games known to benefit significantly from memory OCs.


As to those results ...

Uningine Valley went from 126.3 to 131.3 fps
Unigine Heaven went from 119.7 to 12.1 fps
Ice Storm went from 199,614 to 206,438
Firestrike went from 15,956 to 16,523
Cloudgate went from 34,189 to 34,669

So yes, you could pick 3 games to test with and conclude that it has little to no effect ... you could take another 3 games and conclude that it had great effect.

And yes, tho the core OC will bring greater rewards, either way, there's oodles of room on GTX 7xx / 9xx series and no reason not to take advantage of it. it's particular advantage may range from insignificant to significant depending on the game. I may only get a significant advantage out of my car's AC 20 days per year, but i sure don't want to be driving around with it off on those 30 days.

 

[RobCrezz]

^ Do note that those are synthetic benchmarks that translate into about 0-1 FPS in a real gaming scenario. The core clock is where you will see gains.

Sorry, but thats just not true. On these low memory bandwidth cards, increasing it can really add a lot of performance.

See here, the biggest gain comes from the additional vram speed.

http://www.tomshardware.co.uk/evga-super-super-clocked-gtx-960,review-33153-4.html