Nvidia Pascal Reviews [1080XP, 1080ti, 1080, 1070ti, 1070, 1060, 1050, and 1030]

[Gipsel]

According to NVIDIA, async is already disabled on Maxwell.

In the literal sense there is no way to "enable" or "disable" async, as the independent engines are basically a core concept of DX12 and it is also allowed to freely mix graphics and compute within a single command queue. Any DX12 compatible GPU has to have a way to deal with that (otherwise it would not be compatible). But MS didn't specify how a GPU or its driver has to deal with it. A complete serialization into a single command stream is perfectly fine, as long as the explicit synchronization done by the developer is obeyed.
Hence, the only thing nV (or AMD for that matter) could do, is to change the way their driver (or the GPU itself) handles this. But there is no feature flag for "Async" which you could switch off with a driver update and it can't be used anymore.

 

[CarstenS]

You can choose between complete serialization and static partitioning of course. If you wanted.

 

[CSI PC]

Now, this is interesting...
https://twitter.com/dankbaker/status/732670985409826821

When I change the settings.ini to AsyncComputeOff=0 it should be enabled? Or is it not enabled somewhere else?

It is disabled (or off if you prefer) in the actual code, specifically to the DX12 render path using a VendorID specific to NVIDIA.
This will target all Nvidia cards, until they modify the associated DX12 render path.
So changing any settings will not do much with regards to async compute, however it does have an effect on some other DX12 aspects with regards to NVIDIA drivers.
Check the two quotes I did awhile ago from Kollock at Oxide.

Edit:
https://forum.beyond3d.com/threads/nvidia-pascal-announcement.57763/page-36#post-1912816
And regarding the Feb 2016 Kollock clarification: http://www.overclock.net/t/1590939/...-async-compute-yet-says-amd/370#post_24898074
Where he says:

Async compute is currently forcibly disabled on public builds of Ashes for NV hardware. Whatever performance changes you are seeing driver to driver doesn't have anything to do with async compute.

I can confirm that the latest shipping DX12 drivers from NV do support async compute. You'd have to ask NV how specifically it is implemented.

Cheers

 

[CSI PC]

Now, this is interesting...
https://twitter.com/dankbaker/status/732670985409826821

When I change the settings.ini to AsyncComputeOff=0 it should be enabled? Or is it not enabled somewhere else?

But that tweet does sort of make it messy compared to the clarification Ryan received.
Cheers

 

[CarstenS]

But that tweet does sort of make it messy compared to the clarification Ryan received.
Cheers

That's what I was trying to get at.

 

[A1xLLcqAgt0qc2RyMz0y]

Any idea when we can expect AnandTech and "The Tech Report" GTX 1080 reviews?

 

[Razor1]

one reviewer (youtube reviewer) stated they have, or being sent AIB partner cards two weeks ago, so pretty sure it should be soon.

 

[Ryan Smith]

Any idea when we can expect AnandTech and "The Tech Report" GTX 1080 reviews?

Thursday or Friday, for our part.

 

[Razor1]

Thursday or Friday, for our part.

Do the AIB cards have have more power delivery?

 

[Ryan Smith]

Do the AIB cards have have more power delivery?

No idea. Sorry.

 

[Razor1]

darn!

 

[Clukos]

Nice video from Jay, the partner parts shouldn't have too many problems.

 

[lanek]

The problem seems really to be the temperature, crazy how it hit this 83°C limit immediately .. ( well on the video we see it pass over the TDP limit and temp, so dont know what to think about it )

 

[pharma]

....the GeForce GTX 1080 SuperClocked from EVGA, this time in ACX 3.0 flavor
While the leaker refused to give us any more details, we can make few assumptions based on those photos. Same as ACX 2.0, the new cooler would use LED illuminated logo on the side. It is also split into two heat-sinks, just like the ACX 2. The custom board designed by EVGA has two power connectors, which means that theoretically more power can be applied to the GPU, as long as power limits are not a concern.

http://videocardz.com/60297/evga-geforce-gtx-1080-sc-acx-3-0-pictured

 

[pjbliverpool]

Has the TPU review been posted yet? As usual it's excellent with some really interesting summary's. Just look at that INSANE performance/watt:

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_1080/27.html

 

[SimBy]

Nice video from Jay, the partner parts shouldn't have too many problems.

So much for 'premium components and thermals'. Partner cards will obviously be better in every regard. So the question why would they be priced lower is very legit.

 

[CSI PC]

So much for 'premium components and thermals'. Partner cards will obviously be better in every regard. So the question why would they be priced lower is very legit.

Well I guess Nvidia can say technically they are 'premium', Jay mentions the fan is much quieter than the previous models, still a cheek I agree of Nvidia

Nvidia dropped the ball by not setting the Boost3/fan profiles to be more aggressive, same way AMD dropped the ball on the Hawaii 2xx products by not giving them a better fan (a worst situation because the Nvidia throttling can be resolved with tweaks to the Boost3 algorithm or some settings).
In the video notice Jay mentions just changing the power and temp targets and that was enough to give the Boost3 algorithm a buffer to not calculate it needed to throttle, he did not even touch the fan profile which he could had made a touch more aggressive so it kicked in at 80degree rather than 83degree for temp ceiling rpm (68% according to Jay when left on variable).

Cheers

 

[Clukos]

Has the TPU review been posted yet? As usual it's excellent with some really interesting summary's. Just look at that INSANE performance/watt:

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_1080/27.html

Might pick up an aftermarket 1070, Tx perf at a fraction of the price and great perf/watt, although I don't really need it yet

 

[I.S.T.]

Thursday or Friday, for our part.

Why was Anandtech not given a review card? Y'all are amongst the biggest sites in the biz...

 

[SimBy]

Well I guess Nvidia can say technically they are 'premium', Jay mentions the fan is much quieter than the previous models, still a cheek I agree of Nvidia

Nvidia dropped the ball by not setting the Boost3/fan profiles to be more aggressive, same way AMD dropped the ball on the Hawaii 2xx products by not giving them a better fan (a worst situation because the Nvidia throttling can be resolved with tweaks to the Boost3 algorithm or some settings).
In the video notice Jay mentions just changing the power and temp targets and that was enough to give the Boost3 algorithm a buffer to not calculate it needed to throttle, he did not even touch the fan profile which he could had made a touch more aggressive so it kicked in at 80degree rather than 83degree for temp ceiling rpm (68% according to Jay when left on variable).

Cheers

If they changed the profile (increase power limit and fan speed) power consumption and noise would obviously be less favorable.
68% fan speed seems quite high to me. Don't have much experience with blowers though.