Official Statement from Otoy towards GTX 980/970

[Derya]

Can we please have an official statement regarding the new 980/970 cards from NVidia and why they aren't working as they should? After reading some threads it seems Octane doesn't recognize them properly. As Cuda fueled cards are essential for working with Octane this should have a rather high priority, shouldn't it?

[abstrax]

Can we please have an official statement regarding the new 980/970 cards from NVidia and why they aren't working as they should? After reading some threads it seems Octane doesn't recognize them properly. As Cuda fueled cards are essential for working with Octane this should have a rather high priority, shouldn't it?

We just got a 980 on Friday afternoon and we didn't have any time to do some proper testing, because we are currently trying to release 2.11. Until we have finished the testing / tweaking we can't say anything specific.

[Derya]

Thanks for the heads up abstrax, it's good to know that this is not overlooked

[Seekerfinder]

I'd be very keen also to know what Otoy can do with the latest cards. And I'm somewhat surprised that no testing seems to have been done as yet. I thought that app developers like Otoy would be first to test the newest compute capability (even virtually) before public release to optimize applications for the new hardware and even give some input / feedback. I guess it doesn't work that way.

The new cards seem to get rave reviews everywhere except around here. Really hope the lads can draw all the juice out of the new gen cards.

Seeker

[Jaberwocky]

Very Strange

Maxwell 750Ti and 750 have been supported for months,since their release in February.
Effectively the 980 and 970 are these cards but with a double width bus and more cores and memory.This means that either

1) Maxwell cards in general have been made to get running on Octane by Otoy, but have never been optimized over these last 6 months.

or

2) They have been optimized and...that's yer lot mate !

or

3) Nvidia are struggling to get an Optimized version of CUDA out the door to work with / program these cards properly and Otoy are stuck in the middle.

Time will tell which of these is correct i suspect!

Jaba

[Jonas]

Sweet. Looking forward to it.

Jaberwocky: Abit of more research on your part would have showed you that the architecture between 750
and 970/980 is not exactly the same just because it is all called Maxwell. Also the Cuda drivers
from Nvidia are already there. They just need to implement it into Octane as far as i understand.

[Vue2Octane]

Does not it make kind of sense that GTX980 is slower with Octane than GTX780.

GTX 780 has more cuda cores (and higher memory bandwidth).

http://www.geforce.com/hardware/desktop ... ifications
http://www.geforce.com/hardware/desktop ... ifications

GTX980 just does not seem to be designed for improved CUDA performance.

[glimpse]

Firsty You can't compare CUDA cores side by side from different architectures (Fermi/Kepler/Maxwell),
Secondly Maxwell IS designed for compute & as nVidia says "Maxwell: The Most Advanced CUDA GPU Ever Made"

http://devblogs.nvidia.com/parallelfora ... ever-made/

[Seekerfinder]

Firsty You can't compare CUDA cores side by side from different architectures (Fermi/Kepler/Maxwell),
Secondly Maxwell IS designed for compute & as nVidia says "Maxwell: The Most Advanced CUDA GPU Ever Made"

http://devblogs.nvidia.com/parallelfora ... ever-made/

And yet we get poor results with the 980... Also interesting is the new speed improvements with 2.1, which are significant. So I would have thought it's precisely the new 'compute capability' (5.2) that would have helped. It's a bit of a black art it seems.

The Otoy team seems very stretched and while they clearly have achieved some lofty goals, with all the great things planned (cloud, Brigade and the new plugins) there are some areas that are inevitably going to suffer. Keeping up with Nvidia seems to be one of them. Their focus is probably more on the cloud now.

Seeker

[smicha]

I was a little bit surprised that Octane TEAM hadn't received the 900 series for testing just before launch.