This is the so called club of "participants requesting a new app".
As the availability of the Pascal GPUs increasing, its price is reducing bit by bit, it is time to have a GPUGrid app which supports these GPUs.
I think the current app would work fine if it would include this "nonbonded.cu" module for CC6.1 cards.
(So it won't have to be a CUDA8 app, perhaps if it would be CUDA8 it could be even faster.)
Everyone who already owns such a GPU, or placed an order for such a GPU, or planning to buy one:
Please post the number and type of your Pascal GPU here
to help us to assess the amount of computing power this project could have, and the amount of the energy could have been saved.
I hope this will give sufficient motivation for the staff, at least to communicate more about their plans & progress.

I have a GTX 1080 up and running.

Great idea for the "Wish list"
I'm planning to buy a GTX 1070 (upgrade from GTX 660) as soon as the new app for Pascal GPU is ready.

You can count me in...

I'm planning on getting the mini-ITX card of the 1070 made by gigabyte as an upgrade from my GTX 650Ti BOOST. I've already started making a spare stash of money for it, although the price is still unknown... doesn't hurt to save, though.

I have set aside 1k for the Pascal Titan, which last I heard was rumored to come out end of summer early/fall.

Keep saving - the Pascal Titan will cost ~$1200 (due 2nd Aug).

---

FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

Keep saving - the Pascal Titan will cost ~$1200 (due 2nd Aug).

That's obscene. Will it be as slow as the last Titan?

Yeah, crazy money for a GPU. Quite obscene and not something I would consider forking out for.

That said scientists will be able to build a 44TFlops CUDA research rig for $6000 to $8K. When I was working in various labs lost of research equipment cost tens if not hundreds of £K and some >>more than that. So, for R&D it's good, but not for DC.

For us here, scaling was an issue with the GTX 980Ti so I'm expecting it to be an issue with anything roughly that big with Pascal too (1080 upwards).
By scaling I mean actual throughput performance doesn't increase linearly with the GFlops of bigger cards.
In theory the Titan P will be ~24% faster than the 1080 but despite the favourable Bus width and Bus/GFlops ratio, scaling is likely to be a factor.
Then there's the silly price tag and the fact that two MSRP 1080's would cost the same as one Titan P Founders Ed and offer 60% more GFlops.
Finally, we don't yet know how the Pascal's perform here relative to the GTX 900 series or each other.
IMO the whole NV GTX1000 range is too costly at present and will likely stay high for some time, mostly due to the lack of availability. For most people here the 1060 might be the most attractive, when the price is right and when the drivers support the apps use of CUDA8...

---

FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

Any Chance that support for my 2 Titan Vs will be added? Right now all I get is a computation error.

---