GeForce GTX 1080 Founders Edition Overclocking Preview @ [H]

FrgMstr

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,630
GeForce GTX 1080 Founders Edition Overclocking Preview - We take the new GeForce GTX 1080 Founders Edition video card and find out what is the highest sustained real world gaming GPU clock speed that can be obtained. We also find the combined highest GPU and memory overclocks. If you want to see how high and how cool this video card can be pushed, we have the answer.
 
Really liked the article, one of the best so far for looking at this in a more analytical approach.
Nice work.
Some may complain about the fan speed setting, but TBH if looking to find the limits possible this makes sense (and considering the temp it shows a nice balance can be found), more so than the OC we have seen in quite a few reviews where it has been left on default fan speeds/power target/etc.
Cheers
 
Needs more juice! With the way the clocks dropped a bit with the memory overclock, I'm excited to see cards from AIB partners with an increased power limit.

When you overclocked the memory by 1100mhz the average clock on the core dropped by about 30mhz. With this change, you gained about 2fps... Makes me curious about the core to memory scaling. Like what's the cutoff? +550 on the memory (1100mhz effective) > +30 on the core. At what point does this stop being the case?
 
Last edited:
Yes another No BS approach to testing on [H] nice one guys.
I especially like how you do real world testing with games for 30 minutes, this is what I usually do myself :)
 
Called it, save for the temps. I was not expecting it to do so well on a blower type cooler. This clenches it too; I'm waiting on a model with a 5lb heatsink, 10 heatpipes and 40phase VRM bolted on because with as much voltage dumped into them as possible and kept cool, those are gonna be absolutely ridiculous overclockers.
 
I think you can download dB apps for your phone. Obviously it's not calibrated but perhaps you can check it vs your shitty meter. Might be better than nothing.

62C was a little higher than I expected but you do have 20C of headroom (basically 50% more thermal room). IIRC my 980 used to start to throttle at 57C until I BIOs modded it to lock the frequency.
 
Dayaks said:
I think you can download dB apps for your phone. Obviously it's not calibrated but perhaps you can check it vs your shitty meter. Might be better than nothing.
Click to expand...
I think calibration is pretty important as it could be a little or a lot out depending upon frequency.
Cheers
 
Excellent, this is what I was waiting for, thanks both Brent and Kyle because this will be the article that will shut up a lot of hhhmmm, "concerned" people (whiners and occupational moanful).. so 2.08ghz at 62C? This just confirm what I have been saying for days: very low temps and over 2ghz clock can be easily achieved and sustained by just ramping up the fan speed, now should be good to find the point of diminishing returns, check how much difference in temps can offer a 80% fan speed versus 100% and the direct effect in the final overclocked result, but well I think i'll have just to wait until find one of these puppies in a couple of weeks for testing and scientific purposes only :happy:..

Im just happy with that result 2ghz clock sustained on reference air cooling..
 
Dayaks said:
62C was a little higher than I expected but you do have 20C of headroom (basically 50% more thermal room). IIRC my 980 used to start to throttle at 57C until I BIOs modded it to lock the frequency.
Click to expand...

drop in temps on maxwell cards start at 67C with about only 13mhz or so.. 62C is great, it will keep a lot of people mouth shout'd forever :)
 
Looking forward to reading the full benchmark article.(y)
I hope it'll give us an indication of what to expect from high end custom design cards (ie. will it be worth the wait)
 
Araxie said:
Excellent, this is what I was waiting for, thanks both Brent and Kyle because this will be the article that will shut up a lot of hhhmmm, "concerned" people (whiners and occupational moanful).. so 2.08ghz at 62C? This just confirm what I have been saying for days: very low temps and over 2ghz clock can be easily achieved and sustained by just ramping up the fan speed, now should be good to find the point of diminishing returns, check how much difference in temps can offer a 80% fan speed versus 100% and the direct effect in the final overclocked result, but well I think i'll have just to wait until find one of these puppies in a couple of weeks for testing and scientific purposes only :happy:..

Im just happy with that result 2ghz clock sustained on reference air cooling..
Click to expand...

I do remember that from when I had a reference blower. There definitely are steep diminishing returns. I bet the noise wouldn't be too bad and you'd get 90% of the cooling at 70-75%.

I am really curious how these scale with voltage. We need someone to hard mod that power limit away. :)
 
yeah, that's exactly my point. But I think the issue is not only voltage scaling, but the really lack of power delivered. wait until you see the monsters with 2x8pin connectors and I think 2ghz will be the common starting place for those cards.
 
It clearly wasn't temp-constrained, hitting only 62C, so the question remains what the actual power or architecture constraints really are. Needs voltage testing next.
 
I really hope that this is put against a high clocked 980 Ti with 1500/2000 as clocks as a minimum and not stock 980 Ti.
 
Didn't Maxwell have issues were adding voltage actually caused the clocks to drop due to bios constraints on TDP? I would assume NVIDIA would have done the same for Pascal. If so i hope the Lightning or Kingpin edition remove the constraint.
 
Interesting report. Adding a pci-e connector might be the way to go, although pricey. I wonder if disconnecting the fan cord and running it off a molex would save enough watts to help any.
 
This is pretty impressive for a first try. If the shipping cards can hit around 2ghz (or dare I suggest, higher) that will make this a solid enthusiast choice.
 
I'm pretty hyped about the overclocking piece you guys will be doing. Regarding voltage settings, I believe HardOCP has always included them in their in depth overclocking reviews.
 
Awesome Write up. Looking forward to the Overclocking Game Performance follow up coming. Good find on the Memory overclock. It seems the TDP Wall got hit. I'm sure we'll figure out a way to get more Juice once Bios Modding is available for these cards.
 
That's actually a pretty decent overclock all things considered. I wonder what 65% fan speed would achieve as that usually seems to be the noise/performance sweet spot.
 
What I find interesting is that there is more headroom to lower the fan speed and still possibly achieve this clock. At 62-64c we can lower the fan till it warms up to around 80c and possibly operate at the same overclock but with a less annoying fan noise. This is a test I will make in the full review.
 
Definitely-- 100% speed on that blower is unacceptably loud.

Also would like to see micromanaging voltages, could be some gains there.

And finally, the impact of overclocking the GPU versus VRAM independently on actual gaming performance.
 
Brent_Justice said:
What I find interesting is that there is more headroom to lower the fan speed and still possibly achieve this clock. At 62-64c we can lower the fan till it warms up to around 80c and possibly operate at the same overclock but with a less annoying fan noise. This is a test I will make in the full review.
Click to expand...

yeah, that's exactly my thoughts as pointed in one of my above post, the effect of diminishing returns with fan speed/temps/clocks.. so yes, I would appreciate that extra job testing you are making with the full review..
 
Brent_Justice said:
What I find interesting is that there is more headroom to lower the fan speed and still possibly achieve this clock. At 62-64c we can lower the fan till it warms up to around 80c and possibly operate at the same overclock but with a less annoying fan noise. This is a test I will make in the full review.
Click to expand...
Very good information in this article, thanks again for your hard work.
120% power is still only 215W of the rated 225W available, right? (not to mention many power supplies can give more than 150W on the 8pin.
If only that slider went to 11.
I wonder how much power would be freed up by having the fan running lower ~5W? (I guess that comes from the ~10Ws of headroom) Maybe NV had the ambient temps in the 60s in the room where the computer was...

One thing I am curious about is the scaling from pure memory over clocks. Given that the SPEC for G5X targets 6GHz I expect that over the lifecycle of the 1080 cards will offer dramatically higher memory clocks but might not be able to scale the GPU clocks in a similar manner.
 
I am willing to bet that the power delivery is the limitation here. Curious to see how aftermarket designs and upgraded PCBs with 2x 8pins will perform.

As-is the power delivery is limited to 225w.
 
Fixall said:
When you overclocked the memory by 1100mhz the average clock on the core dropped by about 30mhz
Click to expand...
where are you seeing this. article says "This is the highest obtainable GPU-only overclock we could manage. We are only overclocking the GPU while memory is left at default."

as far as noise goes, you say that its mostly the sound of the rushing air causing it. would opening up the io bracket or removing it make any difference? the little slots look pretty damn restrictive. this is also why I think they should ditch the dvi on these blower style cards.
 
pendragon1 said:
where are you seeing this. article says "This is the highest obtainable GPU-only overclock we could manage. We are only overclocking the GPU while memory is left at default."

as far as noise goes, you say that its mostly the sound of the rushing air causing it. would opening up the io bracket or removing it make any difference? the little slots look pretty damn restrictive. this is also why I think they should ditch the dvi on these blower style cards.
Click to expand...

Ummm, did you read Page 2 of the article, with the combined OC results?
 
Very nice. Looking forward to the full article. Very interested in seeing the effects of adjusting the voltage curve.
 
Rizen said:
I am willing to bet that the power delivery is the limitation here. Curious to see how aftermarket designs and upgraded PCBs with 2x 8pins will perform.

As-is the power delivery is limited to 225w.
Click to expand...

Will NVidia even allow AIB to add a 8pin to it and mess with the power supply? I wonder if they will save that for 1080ti/Titan instead.
 
All in all, if you turn the Founders Edition fan up, you should not expect the card to be quiet. It is probably something that is livable with, and even moreso if you wear headphones while gaming.

I have heard this before with this type of cooling solution.

We (as in most of us) here have had years of cards with the turbine style cooling units as Nvidia and AMD both moved to them.

I have never felt they were "quiet" even at stock speeds you could always pick them out over a properly designed quiet case setup. They always whined to life when the GPU heats up and tend to get rather annoying real quick.

I haven't experienced a single unit with this type of fan assembly that I could live with. Maybe I'm too picky, but my wife gets tired of yelling to talk to me from only 7 feet away.

I ran watercooling on my 290X's and Now have 2 Fury's that are quiet with 3 fans each. I am sure that the 1070 and 80 will get custom cooling from board partners very very soon that will be far more livable then the default garbage that this coolers noise will produce.

Don't be duped into believing the turbine cooler suddenly got quieter.... they never do.
 
What is not possible is achieving a sustained 2.1GHz at 67c using the default 55% fan profile

Thank you for coming right out and saying it. That is the sort of down-to-earth, verified reporting that needs to be more widespread.

After a *real* review, I'm slightly surprised in a good way about the 1080. I expected the sustained OC to be right at 2 GHz, but maybe that's where it will land with a more manageable 80% fan speed. A 12% stable OC where your card sounds like your central AC instead of your vacuum cleaner and still is handily better than anything in the price range now...that's something to consider.
 
This card should be really nice from a partner like Asus, with more than an 8 pin connector
In a custom loop
mmm

Though I wonder
Overclocking the memory seems to do quite something for the 1080, fps wise

Wonder how much improvement it could get on a ti version with a wider bus and more mature RAM
As in GDDR5x is only clocked conservatively on the 1080

Come on AMD, NVIDIDA is only holding a ti back to counter you're high end part :)
 
Wonder if Afterburner is ready yet for the 1080?

I'd like to see what a watercooled 1080 with an 8+8 or 8+6 power connector could do.

This card seems to be a real beast.

My plan is to wait for the partner boards and then get a couple in August or September......maybe a 1080 TI will be out??

Right now my 980 Tis are still really good.
 
I still game at 1600x1200, and I'm still on a 670 as I work my way through FarCry3. It's an Asus DC2 4GB, so for my behind-the-curve gaming life it does well. After that, I picked up an el cheapo copy of Crysis 2 as a Christmas list item. I'm hoping by the time I finish those two titles, 1070's will be plentiful and at sub-$400 prices. If it can OC like the 1080, I think I'll be ready for a better monitor.
 
  • Like
Reactions: DF-1
like this
Wonder if EVGA is going to roll out an ACX 3 cooler for this card?
 
Of course they will. Only question is how long we have to wait to get that second power connector.
 
A custom PCB and some water should make for some interesting numbers.
 
  • Like
Reactions: N4CR
like this
You must log in or register to reply here.
Back
Top