** The Official Nvidia GeForce 'Pascal' Thread - for general gossip and discussions **

[wunkley]

The reference Nvidia coolers have always been quite poor in my experience. I've had 670 ref design and 780Ti ref design ( the 780Ti used the 'best' vapour chamber one like the Titan, far as I know).

They have always gone up to their max temp limits of 83 degrees and throttled, unless I set up a custom fan curve in afterburner or similar. Usually a 1:1 curve, but doing that was far too noisy for me (eg at 75 degrees fan is at 75% etc etc).

I now run 2x 980Ti's both custom H55 AIO watercooled, and I would never buy anything else now, unless the next round of cards have completely new re-designed blower/ exhaust technology, which I very much doubt.

And, I thought the whole die shrink / lower TDP of pascal would produce much much cooler running cards anyway? What a disappointment the 1080 is, on many levels.

 

[tyler_jrb]

Is it in a case or open air? Has it been run for 30 minutes to see what the boost clock settles at?

The article I'm quoting did both of those things to make it as real world as possible.

That's kind of pointless. At which you might aswell be comparing a 1075mhz 980Ti to a bone stock 1080. So both endure a worst case throttle scenario.

The 1080 was overclocked to 2050mhz for every test and kept at that speed. The 980Ti was also overclocked to max aswell.

No point in saying well the founders throttled down to 1650mhz on stock profile therefore is 4% faster than a 1450mhz custom 980Ti.

Either compare stock for stock or max OC to OC. That review shows both.

 

[melmac]

So i take all this ASIC thing with a grain of salt...

Funny you say this, cause back when i had a pair of 7950 in crossfire(same model from sapphire) the one with lower ASIC overclocked a hell of a lot better than the other(both of air).

People are reading too much into the ASIC value, in my opinion of course. On another forum, years ago, people were putting their ASIC in and showing their overclocking results, and at the end of the thread the conclusion reached was that the ASIC value was no guide to overclocking.

 

[Jono8]

That's kind of pointless. At which you might aswell be comparing a 1075mhz 980Ti to a bone stock 1080.

The 1080 was overclocked to 2050mhz for every test and kept at that speed. The 980Ti was also overclocked to max aswell.

No point in saying well the founders throttled down to 1650mhz on stock fan profile therefore is 4% faster than a 1450mhz custom 980Ti.

Either compare stock for stock or max OC to OC. That review shows both.

Indeed. Everyone seems to be comparing custom cooled 980Ti's to this reference blower...and now even without changing the fan curve!

 

[Seanspeed]

Anyone else thinking the GTX 1060 might be the card for them? Even prices for the 1070 look a bit rich for me, so hopefully the 1060 is decent and pulls ahead of a 970 for 1080p gaming.

1060/Ti is going to be the best non-enthusiast card, definitely. There's no reason it shouldn't be able to beat a 970.

I think this will go in direct competition with Polaris 10, though. So in this range, I'd also be checking out AMD cuz I think they're going to have to come out swinging hard in this range if they cant compete at the higher end.

 

[D.P.]

The reference Nvidia coolers have always been quite poor in my experience. I've had 670 ref design and 780Ti ref design ( the 780Ti used the 'best' vapour chamber one like the Titan, far as I know).

They have always gone up to their max temp limits of 83 degrees and throttled, unless I set up a custom fan curve in afterburner or similar. Usually a 1:1 curve, but doing that was far too noisy for me (eg at 75 degrees fan is at 75% etc etc).

I now run 2x 980Ti's both custom H55 AIO watercooled, and I would never buy anything else now, unless the next round of cards have completely new re-designed blower/ exhaust technology, which I very much doubt.

And, I thought the whole die shrink / lower TDP of pascal would produce much much cooler running cards anyway? What a disappointment the 1080 is, on many levels.

The 1080 does produce much less heat than the 980ti when both set up with equal cooling. Thete was one review that put the 1080 under water and it was far cooler than a 980ti with similar water cooling.

The problem with the 1080bis the ref cooler is not very good and the default fan profile is too conservative. It pulls 50w less so should be much easier to cool and keep quiet. Nvidia just did a bad job with the cooler and then put egg on their face by claiming how good it is.

 

[harney]

Funny you say this, cause back when i had a pair of 7950 in crossfire(same model from sapphire) the one with lower ASIC overclocked a hell of a lot better than the other(both of air).

Should be a good card and on the heels + of the 970 so depends on the price but the way things are at the moment and with Nvidia's at this rate even the budget mainstream cards maybe £250 £300

 

[Deleted member 87287]

.

 

[harney]

The reference Nvidia coolers have always been quite poor in my experience. I've had 670 ref design and 780Ti ref design ( the 780Ti used the 'best' vapour chamber one like the Titan, far as I know).

They have always gone up to their max temp limits of 83 degrees and throttled, unless I set up a custom fan curve in afterburner or similar. Usually a 1:1 curve, but doing that was far too noisy for me (eg at 75 degrees fan is at 75% etc etc).

I now run 2x 980Ti's both custom H55 AIO watercooled, and I would never buy anything else now, unless the next round of cards have completely new re-designed blower/ exhaust technology, which I very much doubt.

And, I thought the whole die shrink / lower TDP of pascal would produce much much cooler running cards anyway? What a disappointment the 1080 is, on many levels.

Yes they would create less heat but it seems nvidia are pushing the pascal core as it is re an Nvidia OC....Basically to me nvidia have played the safe bet with pascal architecture and looks like a maxwell on speed shrunk with a few things bolted on...and i bet most features can be driven by the driver not hardware ...

Now if the 1080 was at lower clocks say 980 clocks that would be results i would like to see ..i bet it would be far lower power less heat and still would it perform well against the 980 who knows ... So to me Maxwell on speed

 

[Phixsator]

Yes they would create less heat but it seems nvidia are pushing the pascal core as it is re an Nvidia OC....Basically to me nvidia have played the safe bet with pascal architecture and looks like a maxwell on speed shrunk with a few things bolted on...and i bet most features can be driven by the driver not hardware ...

Now if the 1080 was at lower clocks say 980 clocks that would be results i would like to see ..i bet it would be far lower power less heat and still would it perform well against the 980 who knows ... So to me Maxwell on speed

all it takes is for someone to run a 980 at 1500 and a 1080 at 1500 core and then try to the best of their abilities to match the memory. Would be interesting to see the lead the 1080 has then..

 

[Rroff]

all it takes is for someone to run a 980 at 1500 and a 1080 at 1500 core and then try to the best of their abilities to match the memory. Would be interesting to see the lead the 1080 has then..

Try it with a future DX12/Vulkan title as well as DX11 heh be interesting to see what architecturally the 1080 is good for.

 

[harney]

all it takes is for someone to run a 980 at 1500 and a 1080 at 1500 core and then try to the best of their abilities to match the memory. Would be interesting to see the lead the 1080 has then..

Yes would make great results terms of power heat ect


re mem speed would be good to test match the bandwidth to the 980 as well ...

 

[straxusii]

Indeed. Everyone seems to be comparing custom cooled 980Ti's to this reference blower...and now even without changing the fan curve!

Of course we are because the only 1080 data at the moment is what we see in the reviews, we can't compare a 980Ti to something that we have no benchmarks for.

Computerbase are one of the few reviewers that actually test in a case, most are open so it's far easier to swap out cards.

You can stick your head in the sand if you want to but the stock 1080 is a poor choice if you are going from a TX or 980Ti and probably a Fury X.

When we see reviews for partner cards with aftermarket coolers this may well change.


Also I don't think you'll see much better clocks on a 1080 unless the partners add more than a single 8 pin connector, the card seems to be close to it's power limit around 2.1Ghz.

Don't get me wrong 1080 is a good card and if you've got a 980 or less it's a great upgrade but I think we've been riding the hype train after seeing the 2.1ghz 67c on air demo by JSH.

 

[straxusii]

all it takes is for someone to run a 980 at 1500 and a 1080 at 1500 core and then try to the best of their abilities to match the memory. Would be interesting to see the lead the 1080 has then..

Aside from acamdeic interest in the differences between the processes this would be largely pointless, part of the proposition of the 1080 is it can run much faster clock speeds than previous gens.

 

[Phixsator]

Aside from acamdeic interest in the differences between the processes this would be largely pointless, part of the proposition of the 1080 is it can run much faster clock speeds than previous gens.

The point would be to see the clock for clock performance which makes or breaks the whole WOOOOOW factor of the 2.2ghz on air that the 1080 has hanging over its head right now.

A lot of reviewers these days are all playing it safe and fast and dont go the extra mile doing perhaps more "academic" tests which would be of interest for many of us nerds. It's quite boring actually reading reviews these days in most cases.

 

[.Trooper.]

Everyone saying it's no surprise the reference cooler isn't great because it's been the same in previous generations, seems to be forgetting that they're charging $100 extra for this generations reference, because it's supposedly meant to be better.

 

[Telecaster]

Indeed. Everyone seems to be comparing custom cooled 980Ti's to this reference blower...and now even without changing the fan curve!

Well we've nothing else to compare it with as that's the only 1080 that has been reviewed. Until we see the silly mark up priced, properly cooled, non throttling version we can't be sure how it will perform.

But one thing is for sure, the reference version looks pretty bad from the realistic, in case reviews done so far, unless you plan on water cooling them I'm not sure who would want such a gimped card for £630. Maybe deaf folks who run their GPU fans at 100% wont care

 

[harney]

The point would be to see the clock for clock performance which makes or breaks the whole WOOOOOW factor of the 2.2ghz on air that the 1080 has hanging over its head right now.

A lot of reviewers these days are all playing it safe and fast and dont go the extra mile doing perhaps more "academic" tests which would be of interest for many of us nerds. It's quite boring actually reading reviews these days in most cases.

Yes this indeed ....In terms of power / heat and performance what would it really be up against clock for clock texel n pixel n bandwidth against the 980 this would more or less show what really has been achieved from 28nm to 16nm.

For us NERDS at least

 

[D.P.]

The point would be to see the clock for clock performance which makes or breaks the whole WOOOOOW factor of the 2.2ghz on air that the 1080 has hanging over its head right now.

A lot of reviewers these days are all playing it safe and fast and dont go the extra mile doing perhaps more "academic" tests which would be of interest for many of us nerds. It's quite boring actually reading reviews these days in most cases.

Clock for clock performance is irrelevant becuase if the processor is designed with a different IPC and clock speed target.

The only thing that matters is Instructions per second. Whether that is achieved by IPC or clock speed is irrelevant

 

[Born_2_Kill_83]

This is a great slide:

)

Just gonna keep quoting this cause my TX runs 1500mhz for 3 hours without dropping a hz which must mean 15% increase over a 1080 @ 2050