980ti OC runs hotter with no memory OC

Laskoir · 20 Mar 2017 at 09:27

Hi All,

I have trawled old posts and cant find this particular question anywhere. Would appreciate some help please.

If i OC my 980ti windforce x3 (OC edition) core without the memory OC the core gets up to about 1425 and it runs at 70 degrees.

If i OC core by same number but also OC memory it runs at slightly lower core 1375 and the temp drops to 65 degrees. FPS change negligible. Running between 45 and 60 FPS on 4k depending on game and what is being displayed obviously.

I've played with various offsets anywhere from core +150 to core +250 with and without memory OC and it's always running that 5 degrees hotter without a memory OC.

Obviously this is all fiddling round and 5 degrees doesnt matter and a few MHz and a few FPS don't matter.

I've left volts at +87 mcrv as anything less than that also makes it runs hotter.

Again not an issue as +87 is perfectly safe.

However i would appreciate your thoughts to better understand my GPU about why the absence of a memory OC or increased volts makes it runs hotter. And why memory OC pushes the core clock down a bit. Possibly the memory supports the core so the core doesn't have to work as hard?

Thanks so much,

Paul.

new boy · 20 Mar 2017 at 13:18

By overclocking the memory, you will be increasing power draw. Your card is probably power throttling and by using some more power on the memory, less can be put into the core, hence lower core clocks and core temperatures.

Laskoir · 20 Mar 2017 at 13:22

Yes makes sense
Just realised the power limit can go to 110% but i was using on old MSI afterburner skin so it wasn't available. All sorted thanks. Although now running at a cool 67 degrees with a hefty 1450 7900 which is still cooler than the 70 degrees on just a core OC. I can see why people get obsessed with this.

Disco_P · 21 Mar 2017 at 12:04

It's important to benchmark before and after with memory clocks cause it's possible to push them to instability but due to error correction you won't crash however performance will suffer.

new boy · 21 Mar 2017 at 13:35

To add to that last point, there are also straps with different timings. For example one strap might be 1201mhz to 1200mhz, and the next strap might be 1201mhz - 1400mhz. As you go up straps the timings get looser, so being at the start of a strap, say 1210mhz in our example, will often give you lower performance that being at the end of the previous strap (1200mhz in our example)

geordiejc · 22 Mar 2017 at 08:59

You can check if your GPU is hitting a power limit via MSI afterburner or EVGA Precision X