eVGA GTX 1080 SC Overclocking Results

[Golgatha]

Ran eVGA OC Scanner using their version of Furmark with around 5.5GB of video RAM usage to get the card nice and toasty, and started playing with settings yesterday.

So far I'm able to get just over 2 Ghz boost clock (+150 Mhz to base clock on an already +100 Mhz factory OCed card) and about +300 Mhz on the memory by just raising the power and temp profiles to 110% and 87°C linked. Ran it for 8 hours overnight without a single pixel error and the temperature stabilized at 65°C under full load. eVGA's ACX 3.0 is very effective, as I have large (92 or 120 mm), but generally lower airflow, fans in the system.

 

[guskline]

Ran eVGA OC Scanner using their version of Furmark with around 5.5GB of video RAM usage to get the card nice and toasty, and started playing with settings yesterday.

So far I'm able to get just over 2 Ghz boost clock (+150 Mhz to base clock on an already +100 Mhz factory OCed card) and about +300 Mhz on the memory by just raising the power and temp profiles to 110% and 87°C linked. Ran it for 8 hours overnight without a single pixel error and the temperature stabilized at 65°C under full load. eVGA's ACX 3.0 is very effective, as I have large (92 or 120 mm), but generally lower airflow, fans in the system.

Sounds great. I have a "plain jane" Zotac GTX 1080 FE With EK waterblock and it runs fine.

 

[fastamdman]

Done any overclocking yet guskline? numbers?

I haven't gotten 2.1 100% stable yet but 2088 is and the card just sits there.

 

[guskline]

Done any overclocking yet guskline? numbers?

I haven't gotten 2.1 100% stable yet but 2088 is and the card just sits there.

Done a bit. I hard locked up by setting the vcore to +300 and memory to +400 using MSI afterburner(my dumb fault).

I've let it run stock but will do some OCing today and report back. Truthfully. the card is so fast stock I'm amazed.:thumbsup:

 

[guskline]

I'm now using Zotac Firestorm, OCing software that came with my card. Without altering memory but unlocking power, I had the vcore with boost to 2059 and it ran solid on Heaven 4 maxed out. Decided not to go any higher.

 

[Deders]

I'm now using Zotac Firestorm, OCing software that came with my card. Without altering memory but unlocking power, I had the vcore with boost to 2059 and it ran solid on Heaven 4 maxed out. Decided not to go any higher.

When you say unlocking power, are you just raising it to the max the software allows or can you bypass the power limit completely?

@Golgatha

is 110% the max power target for your card?

 

[guskline]

Firestorm shows 120% as the maximum power levels but I touched NOTHING but the core clock. Was able to get mine to 2100. Of course my card is "under water" with an EK waterblock.

 

[Deders]

You should really make use of that power target, especially with watercooling.

It's the first place I go when overclocking and usually accounts for the largest part of my overclock.

Just Max it out, ask anyone here.

 

[guskline]

You should really make use of that power target, especially with watercooling.

It's the first place I go when overclocking and usually accounts for the largest part of my overclock.

Just Max it out, ask anyone here.

OK I'll try it.

 

[guskline]

Deders I did try it. On Firestrike regular with vcore upped to 1973 (1607+366) which shows boost at 2100 exactly and power target at 100% the score was 23185. When I raised the power target to max of 120% the score jumped to 23401.

I'll fool around more with OCing later.

 

[Deders]

You may even be able to raise your clockspeed more with more power available.

 

[Termie]

Deders I did try it. On Firestrike regular with vcore upped to 1973 (1607+366) which shows boost at 2100 exactly and power target at 100% the score was 23185. When I raised the power target to max of 120% the score jumped to 23401.

I'll fool around more with OCing later.

Fantastic numbers, guskline. You obviously aren't temp limited, and with 120% power you probably aren't all that power limited. Did the boost go above 2100 when you made this adjustment? Given the FireStrike score, I assume it did, at least a bit.

In the end, it appears the GTX 1080 is limited by the GPU itself - it runs out of overclocking headroom before you use all available power or hit the temperature ceiling. This is why custom cards with dual power inputs aren't performing any better in reviews. And that's why I'm waiting on one of EVGA's ACX editions to finally show up in stock.

By the way, one small tip: I think when you say vcore you mean core clock. Vcore is the voltage!

 

[fastamdman]

These cards can be volt modded to allow more voltage through them as an fyi. I will probably be doing it later on as the card I have now only hits 2088 100% stable, with occasional spikes to 2100. I want 2100+ 100% completely stable, I don't like "uneven" numbers like that lol. Darn gpu boost.

 

[Deders]

I used to be a bit ocd like that but then I realised is that the computer doesn't mind whether it is a nice round number or not.

 

[guskline]

Fantastic numbers, guskline. You obviously aren't temp limited, and with 120% power you probably aren't all that power limited. Did the boost go above 2100 when you made this adjustment? Given the FireStrike score, I assume it did, at least a bit.

In the end, it appears the GTX 1080 is limited by the GPU itself - it runs out of overclocking headroom before you use all available power or hit the temperature ceiling. This is why custom cards with dual power inputs aren't performing any better in reviews. And that's why I'm waiting on one of EVGA's ACX editions to finally show up in stock.

By the way, one small tip: I think when you say vcore you mean core clock. Vcore is the voltage!

Thanks Termie, my mistake. You are correct, it is core clock. BTW, I used the Firestorm software to measure the boost when it is set. I'll have to try MSI Afterburner to get a log of the highest boost clock.

 

[thxdd]

I can't wait for my MSI 8G to arrive, getting impatient! These are some decent results, I was not expecting to see over 2 GHz - hopefully mine isn't a crappy binned dud!

 

[LTC8K6]

It's probably limited by the need for the 1080ti to be faster.

 

[thxdd]

It's probably limited by the need for the 1080ti to be faster.

Funny cuz it's true.

 

[guskline]

It's probably limited by the need for the 1080ti to be faster.

LTX8K6: There might be a grain of truth in that; however, from what I read and God knows there was a lot, when the GTX 1080 was undergoing testing the original clocks were @1325 to 1400 at first and obviously were not able to surpass the GTX 980TI. Apparently a lot of work went into bumping up the core clock. Also remember going from 28nm? to 16 so it was a big change.

I think Nvidia tuned this 1080 much higher and thus reduced the overclock room, but not by a ton.

No doubt the 1080TI will be faster, even if the clock speed is near the 1080. It will also be more expensive.

I have my GTX 1080 under water, it is a stock pcb with 1607 core speed. I've been able to up the core by 375 without artifacting.

 

[sxr7171]

Why do I keep thinking about Pentium 4?

 

[Deders]

Why do I keep thinking about Pentium 4?

That's probably a question only you can answer after some deep soul searching.

 

[guskline]

Golgatha, how is that GTX 1080 SC running now?

 

[Golgatha]

Golgatha, how is that GTX 1080 SC running now?

I can overclock it with +130 to +150 on the core and it doesn't make a difference in the boost clock; it stops and holds at 2088 boost. Somewhere between +160 to +200 (remembering there's a +100 stock OC since it's a SC version) on the core, it starts screen locking on occasion regardless of memory clock or power target adjustments.

So, I can get it to about 2.1 Ghz if I want to run a benchmark quickly, but for day to day usage I just put it about +130 and the boost clock never lowers from 2088 Mhz with 110% power target. I leave the power and temp target linked, as the card still won't break 65°C. Raising the power target above 110% doesn't seem to help stability, but around +130 on the core I did have one screen lock at 100% power target. Leaving it at 110% power target keeps things stable at 100% utilization overnight still, so that's where I'm leaving it. One the memory side of things, it locked up somewhere between +350 to +500. I leave it set at +300 and don't feel the need to tweak memory anyway; the core speed is where you get almost all your perceivable gains.

I'm running this on a Dell G-Sync 27in, 1440p monitor. It's purdy!


Basically, I'm just shy of the core overclock and right in line with the memory overclock Hothardware got with their FE card. I'm thinking my overclock is limited by the core and memory chips themselves rather than the cooling. The ACX 3.0 cooling keeps it from ever getting anywhere near a throttling temperature though, so my overclock should be a max boost throughout any gaming sessions, which are typically less taxing than running Furmark constantly anyway.

http://hothardware.com/reviews/nvidia-geforce-gtx-1080-pascal-gpu-review?page=8

 

[fastamdman]

Remember that with pascal you should only be upping the clocks by increments of 25. So go from 100, to +125 to +150 to +175 etc. You wont see a change from +125 vs +149 to my knowledge, or +100 to +124. It needs to go in increments of 25 from everything I have read.

I am also stuck at the 2088 boost clock with an occasional spike to 2100 here and there on my 1080 gigabyte g1 edition.

 

[LTC8K6]

LTX8K6: There might be a grain of truth in that; however, from what I read and God knows there was a lot, when the GTX 1080 was undergoing testing the original clocks were @1325 to 1400 at first and obviously were not able to surpass the GTX 980TI. Apparently a lot of work went into bumping up the core clock. Also remember going from 28nm? to 16 so it was a big change.

I think Nvidia tuned this 1080 much higher and thus reduced the overclock room, but not by a ton.

No doubt the 1080TI will be faster, even if the clock speed is near the 1080. It will also be more expensive.

I have my GTX 1080 under water, it is a stock pcb with 1607 core speed. I've been able to up the core by 375 without artifacting.

NV compares the 1080 to the 980.

Logically, the 1080 is the 980 replacement.

I'm fairly certain the 1080 was intended to soundly beat the 980, and not the 980ti.

The 1080ti is the 980ti competitor, imo.