Pascal 1080 Benchmarks

[railven]

I want to see max overclocks. It looks like the reference cooler is not good enough for overclocking. 55% on the stock blower is freaking loud. Plus, that single 8pin connector might be a limiting factor when it comes to max overclocking. I think anything beyond 2000mhz will require better cooling and power delivery. We need to see AIB cards with better cooling and power delivery.

This was my opinion from the start. The ref PCB and power circuitry is most likely set to deliver efficiency.

I'd like to see what partners got with more power and better circuitry. And if the price is cheaper, well then that's just icing.

Aight I got my fix for some benchmarks, time for more DOOM!!!

 

[Silverforce11]

OC 980TI vs OC 1080..OC 1080 is 13% faster than OC 980TI

That's because it's a reference OC, 10-12% headroom so far. Basically this was the best case scenario demo by NV themselves for the OC potential of the reference card.

Vs custom 980Ti that does 20-25% OC.

So that's why the gap shrinks.

To really have a nice leap vs OC 980Ti, you need a custom 1080 and it better do a 20% OC or more. 5 VRM is limiting it's TDP, even if you put it on water, that reference board will throttle or go unstable. Need a custom board with more power phases.

 

[96Firebird]

These cards look to be getting very good OC's, I wonder what aftermarket $600-$650 cards will be getting, and if they will allow for even higher clocks with more power connectors. Is the BIOS limiting OC potential?

Also makes me wonder how the $380 1070s will performance at stock and overclocked... If an OC'd 1070 can match a stock 1080, that might be my next card.

 

[Grooveriding]

I want to see max overclocks. It looks like the reference cooler is not good enough for overclocking. 55% on the stock blower is freaking loud. Plus, that single 8pin connector might be a limiting factor when it comes to max overclocking. I think anything beyond 2000mhz will require better cooling and power delivery. We need to see AIB cards with better cooling and power delivery.

http://www.computerbase.de/videos/2016-05/nvidia-geforce-gtx-1080-lautstaerke-im-test/

Load starts around 35 seconds, it's sort of loud, but the reference cards are always noisy.

 

[linkgoron]

Unless I misunderstood, computerbase said that their 1080 couldn't keep a 2GHz oc for long.

From Google translate:
The 2.0 GHz sound barrier was the test sample so that not to break permanently. In the short term are indeed at greater than 2.0 GHz, but not later than five minutes, the frequency is lower. In Anno 2205, the GeForce GTX 1080 works finally with about 1,870 MHz instead of 1720 MHz and in Star Wars: Battlefront with 1,970 MHz instead of 1780 MHz. Depending on the game, the speed increases in order to ten to eleven percent.

 

[Silverforce11]

These cards look to be getting very good OC's, I wonder what aftermarket $600-$650 cards will be getting, and if they will allow for even higher clocks with more power connectors. Is the BIOS limiting OC potential?

Also makes me wonder how the $380 1070s will performance at stock and overclocked... If an OC'd 1070 can match a stock 1080, that might be my next card.

Not the bios, tear down shows 5 power phases. It's at the power limit, if pushed too high it gets unstable and throttles.

A GTX 980 has a 4 power phase for example. A reference 980Ti has 6 power phases.

Custom boards have 8+ power phases. It's no comparison, for real OC you need good PCBs.

Seeing this, which AIB is gonna make a BETTER custom 1080 and price it for less than the reference board that's TDP limited?

 

[tential]

or

Are people even running benches or just making up numbers?

 

[Borealis7]

TPU review really makes this card look glorious. 33% up from 980TI, low power usage, just the temps are not so good.

i'm sold! just need a beefy cooler from an AIB partner and they can have my money. bad cooling really screws up the whole concept of the Boost 3.0 technology.

Are people even running benches or just making up numbers?

probably different benchmark scenes and rigs.

 

[S.H.O.D.A.N.]

Those reviews are based on the 'founder' cards, right?

 

[railven]

Yerp, I'm signed up for third party custom. I might just say screw it and pay from the start for EVGA's hybrid kit.

Woof!

Dem temps be hot!

 

[torlen11cc]

Those reviews are based on the 'founder' cards, right?

Yes.

 

[Borealis7]

Those reviews are based on the 'founder' cards, right?

Yes, guess we'll have to wait 2-4 weeks for OEM card reviews.

 

[CakeMonster]

With that OC potential, a custom and more silent cooler is definitely the way to go. From what I saw in one review, the temperature and watt usage was exactly the same as the 980. Looking good to me.

 

[Silverforce11]

TPU review really makes this card look glorious. 33% up from 980TI, low power usage, just the temps are not so good.

i'm sold! just need a beefy cooler from an AIB partner and they can have my money.


probably different benchmark scenes and rigs.

Yeah and there's nothing wrong with TPU's review, it's accurate because they have a skew towards newer games (a good thing IMO), whereas other sites have a lot of older titles.

You'll notice sites with older games, the gap is smaller, 20% or so, lots of new games, gap grows.

Pascal blows away Maxwell when it comes to GCN-optimize console ports that we've seen. The trend will grow, compute heavy games will gimp Maxwell while Pascal's new features allow it to steam onwards.

 

[Silverforce11]

Yerp, I'm signed up for third party custom. I might just say screw it and pay from the start for EVGA's hybrid kit.

Woof!

Dem temps be hot!

Wouldn't do it, even if you slap water on it, that reference board is out of juice with the weak 5 phase design.

Wait for a proper OC enthusiast grade custom 1080. Seriously paying a $100 premium for the worse version is to me, very stupid. lol

 

[railven]

Interesting...

https://www.youtube.com/watch?v=2wBDt9tN5-c

So expect 3/4-Way SLI with Pascal.

EDIT:
OMG I should have watched to the end. I hope that "key" doesn't cost money. NV is milking this to the moon! haha.

 

[Paul98]

So how does OC 1080 compare with OC 980ti, or even a stock high clocked 980ti?

 

[railven]

So how does OC 1080 compare with OC 980ti, or even a stock high clocked 980ti?

So far, if you had a ref 980 Ti, OC is great.

If you have a custom balls to the wall 980 Ti, OC is decent, but not as exciting.

Conclusion (mine): wait for custom third party solutions. WTB robust board+power.

 

[Spanners]

Wait for a proper OC enthusiast grade custom 1080. Seriously paying a $100 premium for the worse version is to me, very stupid. lol

The PCB is pretty sparse, looks like they could have shortened it somewhat. Also there looks like space for another phase and another power connector facing rear-ward? Quite a few non-populated components, theory?

 

[Aristotelian]

@Paul98 from Anandtech's preview, the reference 1080 OC'd performance is about 30% higher across the board than a reference 980Ti OC'd.

I guess at this stage, if EVGA puts 1080s out in hydrocopper soon, buying two of those should be pretty good for 4k if SLI scaling is good.

 

[Mopetar]

With crazy high OC of Pascal, I think Polaris or Vega can reach 2K , why not?

I don't think the GCN architecture clocks as highly as Maxwell/Pascal does. We know from last generation that AMD typically didn't do well past 1 GHz and they were using the same node as NVidia so it's not down to a difference (which will exist between Polaris and Pascal and further complicates the comparison) in fabrication process.

It seems unlikely that this is the result of a single bottleneck in GCN that would be easy to fix in order to see massive clock speed gains or AMD would have already fixed it. It's more likely that there are a series of constraints that prevent it from clocking as high as Pascal and while implementing a better design for whatever was the biggest constraint would yield better clock speeds, it doesn't open the door to insane clock rates that we're seeing from Pascal.

Not that it really matters though. Clock speed is only part of the performance equation. If AMD makes major IPC improvements in the architecture, then even a 30% increase in clock speed yields massive performance gains.

 

[Silverforce11]

Interesting...

https://www.youtube.com/watch?v=2wBDt9tN5-c

So expect 3/4-Way SLI with Pascal.

JayZ just summed it up: NV isn't officially supporting 3/4 way, but they allow enthusiasts to unlock it with a special code.

Does it mean their drivers will focus on 3/4 scaling? Doubt it. 3/4 is effective dead, unless DX12 devs want to focus on it... yeah who's gonna do it? lol

 

[selni]

or


Are people even running benches or just making up numbers?

Those screenshots already show different settings. Not surprising the ultra + 4xMSAA numbers are lower than VHQ?

 

[showb1z]

Unless I misunderstood, computerbase said that their 1080 couldn't keep a 2GHz oc for long.

From Google translate:
The 2.0 GHz sound barrier was the test sample so that not to break permanently. In the short term are indeed at greater than 2.0 GHz, but not later than five minutes, the frequency is lower. In Anno 2205, the GeForce GTX 1080 works finally with about 1,870 MHz instead of 1720 MHz and in Star Wars: Battlefront with 1,970 MHz instead of 1780 MHz. Depending on the game, the speed increases in order to ten to eleven percent.

Pretty much yeah.
Next paragraph they said the power target on these reference cards is too low to allow a serious overclock.

 

[Krteq]

With crazy high OC of Pascal, I think Polaris or Vega can reach 2K , why not?

~ 300MHz isn't "crazy high OC" at all.