Kepler test build 2.58e Standalone (for GTX 6xx cards)

[p3taoctane]

Thanks for the helpful links regarding getting this running.
I got a thunderbolt to ethernet adapter and was abel to ge to the dialog window that asks for Username and password

I placed in the values copied for the web site and it tries to authorize it but fails.

So all the pieces are in place but it does not get authoirized.

I have an ethernet port now and it is working and there may some other issue but I cant think what it might be.

I saw in one of the posts that someone else had exactly the same thing.... got all the way to this point but I did not see if there was resolution.

Any ideas
Thanks

Peter

[matej]

What is the state of Kepler Octane builds performance-wise right now? Are we closing the gap? Will Kepler have comparable performance per card as Fermi on new CUDA framework?

Also will Otoy maintain two separate Octane builds one for Fermi one for Kepler, or it will be merged at some point (possibly sacrificing performance of Fermi)? When can we expeoct new Kepler builds for Octane 3.01 and forward?

I'll need to buy a new card in the coming weeks or months and I don't know which family it would be a better choice

[abstrax]

What is the state of Kepler Octane builds performance-wise right now? Are we closing the gap? Will Kepler have comparable performance per card as Fermi on new CUDA framework?

Also will Otoy maintain two separate Octane builds one for Fermi one for Kepler, or it will be merged at some point (possibly sacrificing performance of Fermi)? When can we expeoct new Kepler builds for Octane 3.01 and forward?

I'll need to buy a new card in the coming weeks or months and I don't know which family it would be a better choice

Since we are almost done with the work in this area, I can let you know. We hope we will be able to release a new version next week, that will have good performance on Fermi and on Kepler cards in one build. Keplers will be a bit faster than in the recent Kepler builds, but still be slower than Fermis. Fermis will be as fast as usual, sometimes even a little bit faster.

In most scenes we achieve now with a 680 roughly the same performance as with one 590 GPU, which is about 10-20% slower than a 580. I can't say anything about 690s, since we didn't test it yet with the new system. One positive side effect of those changes is that you don't need to install a CUDA toolkit on Mac OS and Linux anymore.

We try to improve things further, but I have only little hope that we can gain much more without serious algorithmic/structural changes. The good news is that we didn't spend much time on performance yet, so there is a reasonable chance that when we finally get around to spend some time on thios area, we might gain some substantial improvements.

To anwer question: If you can get a 590 or 580 somewhere, then you probably get the most bang for your buck with these.

Cheers,
Marcus

[matej]

Thanks Marcus for the fast & detailed answer.

A merged build that keeps the Fermi performance and improves on Kepler is really good news for everybody!

[gabrielefx]

a pro system with 4 680s will consume less power = lighter power bill

[glimpse]

Thanks for good news, Abstrax.

One more pluss is that Keplers
do have a bit more vRam,
that might come handy too.

[matej]

a pro system with 4 680s will consume less power = lighter power bill

If a 680 is 10-20% slower than a 580, then it would need a bigger than 10-20% difference in power consumption. If this is the case, then rendering with 680 is indeed cheaper.

One more pluss is that Keplers
do have a bit more vRam,

Yeah, that's the biggest plus. 680s with 4GB are everywhere, while 580s with 3GB are hard to find and even those are second-handed. Everything less than 2GB is not an option.

[madcoo]

Hi guys,

Looking forward to the next version of Octane as always

Before I make a mistake and spend my money : can someone please confirm the following :

1) does the current 2.58e Kepler Build support multiple GTX 680 2GB GPUs ? (I'm just about to buy a 2nd one)
2) will the next version do too ?
3) will the current topic be updated with a message warning us that the new version is out and ok for download ?
(it would help, as I get an e-mail notification for every reply in this topic)

Those may sound like stupid questions, but I don't want to spend 550€ for nothing...

Cheers guys and thanks !

[abstrax]

Hi guys,

Looking forward to the next version of Octane as always

Before I make a mistake and spend my money : can someone please confirm the following :

1) does the current 2.58e Kepler Build support multiple GTX 680 2GB GPUs ? (I'm just about to buy a 2nd one)
2) will the next version do too ?
3) will the current topic be updated with a message warning us that the new version is out and ok for download ?
(it would help, as I get an e-mail notification for every reply in this topic)

Those may sound like stupid questions, but I don't want to spend 550€ for nothing...

Cheers guys and thanks !

The answers:

1) Yes, of course.
2) Yes, of course.
3) Probably not, but we hope to have a build out on Friday. Maybe today, but more likely on Friday. So check out the release candidates forum tomorrow and the day after tomorrow.

Cheers,
Marcus

[madcoo]

Thanks a lot for your quick reply Abstrax.
On to my favourite shop now, so I get another GPU !!!


I'll check out the forums towards the end of the week.

Cheers