Darktable 2.2 NVIDIA OpenCL Benchmarks

[phoronix]

Phoronix: Darktable 2.2 NVIDIA OpenCL Benchmarks

With this weekend's release of the Darktable 2.2 RAW digital photography workflow software being out and it having OpenCL improvements among other advancements, I've been carrying out some fresh benchmarks for this popular open-source, cross-platform program...

http://www.phoronix.com/scan.php?page=news_item&px=Darktable-2.2-OpenCL-NVIDIA

[bug77]

Wth is making those 4 cards so much faster than everything else? If 980Ti wasn't among them, you'd think it was some Pascal voodoo.

[cj.wijtmans]

Wth is making those 4 cards so much faster than everything else? If 980Ti wasn't among them, you'd think it was some Pascal voodoo.

perhaps they based pascal off the 980 ti?

[Jumbotron]

Precisely my question. Why is it that the 1050 Ti which is supposed to be a Pascal based GPU does so poorly and the 980 Ti which is NOT supposed to be Pascal based right up there with the best of the Pascal based cards ? I know that the 1050 Ti shouldn't be up near the 1060 much less the 1080. But if it were truly Pascal and not some hot rod Maxwell then it should be a little closer to its Pascal breathren and not stuck in the middle of the pack with the rest.

[illwieckz]

Michael are you paid to do Nvidia benchmarks on early hours after significant third party software release then optionally mixed benchmark (or more rarely AMD benchmark) several days later if the moon is correctly aligned? Or is it just a matter of personnal taste? If it's just a matter of personnal tast (something readers can interpret as a kind of fanboyism and it will be hard to convince hem it's not), I greatly recommend yourself to being paid for that bias (or use that bias to ask AMD to being paid to fix that bias, whatever you prefer). There is no shame to being sponsored, except when it's not clearly advertised. And if you are not sponsored, be honnest with yourself: do not give for free what you must be paid for.

[starshipeleven]

NVIDIA is making some tricks in the driver

[sp82]

Hi Michael,
can you compare this results with the CPU only?

[bug77]

Michael are you paid to do Nvidia benchmarks on early hours after significant third party software release then optionally mixed benchmark (or more rarely AMD benchmark) several days later if the moon is correctly aligned? Or is it just a matter of personnal taste? If it's just a matter of personnal tast (something readers can interpret as a kind of fanboyism and it will be hard to convince hem it's not), I greatly recommend yourself to being paid for that bias (or use that bias to ask AMD to being paid to fix that bias, whatever you prefer). There is no shame to being sponsored, except when it's not clearly advertised. And if you are not sponsored, be honnest with yourself: do not give for free what you must be paid for.

If you've truly been around since 2008, like your profile says, then you know everything here is about page hits. That's what pays the bills, that's what drives Michael. To insinuate anything else is just childish (to put it mildly).

[Erathostenes]

Great work as always.

I ran the test on fc25 and a slightly higher version of darktable and is somewhat pussled by the results on my 1060GTX. Is the test written/optimized for a certain version of darktable or is it a regression within darktable? Or something else?

http://openbenchmarking.org/result/1...ZUCH-161226875 (Geforce 1060GTX, darktable-2.2.1-1)