Question How well did I do on my overclock for GTX 1080

Da Big Gohunay

Reputable
Feb 26, 2019
17
0
4,510
Specs:
Ryzen 5800X (just upgraded from 1800X)
MSI GTX 1080 DUKE 8G
16Gb DDR4 2833mhz

Alrighty decided to give overclocking a go for my card, because I am going to be stuck with a while due to current gpu prices.
While slowly adjusting the overclock, I had furmark running simultaneously.
I achieved a 110+ mhz on the core clock, and 850 mhz on the memory clock. It was stable while running furmark for 10 minutes straight. Which should be a stress test enough, but I may be wrong. I am about to give 3Dmark timespy a go, and I may regret making this post if it doesnt hold up. Wish me luck fellas!
 

Da Big Gohunay

Reputable
Feb 26, 2019
17
0
4,510
Update: Overclock was way unstable, note to self do not use furmark for oc application. Was able to get a stable 620+ memory, and 60+ on core.
 

Phaaze88

Titan
Ambassador
The OC Scanner tool will let all this go more smoothly.
Run Msi Afterburner, open OC Scanner, click 'scan' and wait for it to finish.
It will find new settings and apply them to the gpu right after. Save 'em to one of the numbered profiles. Once that's done, and you're done using the gpu, click the reset button on the main HUD to restore the gpu's default.
EZ overclock on top of the OCs already present(Gpu Boost 3.0 and Msi factory OC).


To simulate the all those undervolting guides out there, apply the OC profile from earlier and simply move the power limit slider down as desired(92%, 90, 85, etc.). Click apply, and save it to a different numbered profile.
Don't forget to click reset when you're done using the gpu.
 

Da Big Gohunay

Reputable
Feb 26, 2019
17
0
4,510
Alrighty, I got some results. I used Unigine valley benchmarks for the scores.
Stock-
Score 2778
FPS 66.4
min: 29.3
Max: 129.5

OC Curve- (as you suggested)
Score 2793
FPS 66.8
min: 34
Max: 128.2

My OC (620+ Memory, 60+ core)
Score 2931
FPS: 70.1
min: 28.8
Max: 131.2

Something to note, OC Curve was truly unstable. I tried multiple different voltages, but every time I would run the benchmark, it would crash the graphics driver right after it ends, every time.
 

Eximo

Titan
Ambassador
On my GTX1080 I was only ever able to add about 200Mhz on memory for 400 effective increase (10410Mhz I recall). Core clock tapped out around 2100Mhz, but that was very much dependent on the silicon lottery. On super demanding games I would occasionally have to decrease these numbers, but only slightly.
 

Da Big Gohunay

Reputable
Feb 26, 2019
17
0
4,510
Not trying to insult you, but you do realize OC came about because people were trying to squeeze out as much performance as they can as the system got older, and they did not have the funds to replace or upgrade the parts? Also you can look up hundreds of videos on how OC your card will literally do zero to longevity. The software will literally not let the card hurt itself (temp limits). Its why throttling exist. If a persons card was going to brick itself 4 years down the line, it was going to brick itself regardless of the OC.
 

Da Big Gohunay

Reputable
Feb 26, 2019
17
0
4,510
-Edit
You are right from the standpoint of old video cards (pre 2011) where some would not slow down from temperature increases and brick themselves. You can see that in the cpus from early 2000s era. Where I believe it was AMD processor would catch on fire, if the heatsink was removed. While the intel would throttle itself so slow, as to not produce enough heat to damage itself. I might have the companies switched around. I know one of them did. Its been a while.
 

Eximo

Titan
Ambassador
I would say overclocking came from getting the most performance out of a given budget, not necessarily overclocking after the fact to retain parity. You could have always overclocked from day one. Back in the day nothing separated one SKU within a generation from another but clock speed. That started changing with the integration of cache, and then the model mattered a little more, but clock speeds were still the big separator. i7-920 was identical to the 930, 940, and 950 only the max multiplier was different.

Often you could pay less and get the same performance with a little extra heat generation. AMD K6 166Mhz, simple jumper flip to run it at 233Mhz with zero negative effects. Now though, usually only one or two chips that are separated only by clock speeds, and with Intel you have the artificial K SKU to drive purchasing decisions (oddly funny that the K SKUs are actually the rejects, the very leaky chips that can effectively overclock, and the good ones end up as mobile or locked parts)

I would say that overclocking of GPUs and CPUs has reached a practical limit. They overclock themselves as long as they have headroom. I actively underpower my GPU and have a locked CPU after decades of overclocking.
 

Phaaze88

Titan
Ambassador
On my GTX1080 I was only ever able to add about 200Mhz on memory
That's what OC Scanner applied to the 1080Ti's memory after I ran it - several times, never went over 200mhz. The core varied a little(63~76mhz), depending on temperature, but I expected that one. It's not like OC Scanner is aggressive either. There's just not much room left. So I took it and set a 90% power limit.
It all made me skeptical to some of the 500+ memory OCs out there.



Something to note, OC Curve was truly unstable. I tried multiple different voltages, but every time I would run the benchmark, it would crash the graphics driver right after it ends, every time.
I can take this as a clean driver install? Hmm, I'm gonna have to rerun it myself later and see if it does that.
FYI, the built in Gpu Boost may ignore the voltage slider setting. While you can increase it to 'suggest' to the gpu that it can use more voltage for stable clocks, but it may not do it. Thing has a mind of its own...


As Eximo posted, the hardware dynamically OCs itself these days, leaving casual users with crumbs, or even nothing to play around with. Takes some of the fun out of it, huh?

The Duke's memory clock got bumped up 1% from Nvidia spec. The advertised base and boost clocks can mostly be ignored, since the dynamic boost/OC will push over that if the temperature/power/voltage headroom is there anyway.

One or two benchmarks won't confirm stable OCs either. Run as many different games as possible. Apps like Valley and Timespy are not a one size fits all.