OTOY Forums

passive
http://www.isycorp.com/Home_I.ASP?id_co ... cat=HHPGPU

active
http://www.isycorp.com/i/mkg/skp.asp?id ... &origine=1

"Tesla K20 GPU Computing Accelerator – Designed for double precision applications and the broader supercomputing market, the Tesla K20 delivers 3x the double precision performance compared to the previous generation Fermi-based Tesla M2090, in the same power envelope."

At more than 5000 NZ dollars each I don't think I'll bother thanks.
Wasn't there supposed to be a Tesla variant tuned to single precision?

http://www.fudzilla.com/home/item/27196 ... -by-nvidia

- why is it so difficult to get news from nvidia official site - so any news on when are the gaming (read: cheap) equivalents coming out

so now we wait for 2496 cuda cores x2 gaming cards -

All in all, the GK110 looks like a pretty impressive GPU, but we honestly doubt that we'll see a Geforce graphics card based on the GK110 anytime soon.

So, for someone who would need to buy a GPU for Octane right now, which family would be a better choice: 5xx Fermi or 6xx Kepler. My 460 died a few days back and I'm not really sure what would be a smart buy right now (or in the next months). Fermi is more performant right now, but it might become less performant in the future when Octane moves entirely on Kepler. But who knows when Otoy will produce a functional Kepler build...

As it seems that waiting for the next GTX line could not be an option, what would you guys suggest?

I would wait for a gamer GK110.
Octane use single precision, so much i know. There is also no need for a Tesla...

face

It seems that 5xx will still be the better option for some time, as Marcus explained in this post.

face wrote:I would wait for a gamer GK110.

Is it clear yet when that might happen? (sorry, I don't follow hardware news much )

I will test a K5000 at the end of this month with Octane for Max and then the K20 when available.

gabrielefx wrote:I will test a K5000 at the end of this month with Octane for Max and then the K20 when available.

K5000 wasn't supposed to be used for calculation, so i really doubt it will do any good in octane. For that reason nVidia introdused maximus technology so Quadros could be paired with teslas..

ooops! I posted it in the wrong index

btw, interesting link:

http://www.theregister.co.uk/2012/09/18 ... enchmarks/

"By tuning the granularity of the simulation with the granularity of the data across space and time, you will get better results and do less work (in less time) than you might otherwise with fine-grained simulation in all regions and timeslices."

I think the K10 has better single precision performance for less money. Might want to check that out too.