Jump to content

CUDA 418.163 is out! [UPDATE May 10, 2019]


fantomas
140 posts in this topic

Recommended Posts

CUDA 9 now official !

Now you could try nbody (also pascal), DL  from post #266.

PS: As always, because CUDA-Z (an cuda info tool) stops working with CUDA 9 - needs an update, its much to old.

So you can use the nbody for example to check if cuda works.

  • Like 1
Link to comment
Share on other sites

Just of curiosity : How about the size of cuda 8 vs 9  ???

Will be metal "2" supported within the new web drivers, what do you think?

With eGPU there are very interesting times coming considering deep learning/openCL and CUDA related programming, I think. -> https://github.com/AaronJackson/vrn (be blown away clicking on the "here"link....)

Link to comment
Share on other sites

You're using vanilla driver and it does not support this new cuda version. Web driver should.

Ah duh.  :blush: I forgot that I'm on High Sierra now  :rofl:

  • Like 1
Link to comment
Share on other sites

@Regi Yassin

 

Was the same thing in Sierra debut: http://www.insanelymac.com/forum/topic/314814-cuda-support-in-sierra/?p=2277984

those method work in high sierra

libcuda_378.05.05.25_mercury.dylib > libcuda_378.10.10.10_mercury.dylib

/Library/Frameworks/CUDA.framework/Versions/A/Resources/Info.plist > add new key 378.10.10.10

 

for now only tested with nbody, gonna try with premiere

 

thank you for the info @fantomas1

  • Like 5
Link to comment
Share on other sites

Great you got nbody (comipled with pascal support, some page before) to run.

Did you run in 60 FPS Limit with your fast card? (even with key + for more bodies?)

You win against my GT 1030 (and it won to my older GTX 650 much :) )

Link to comment
Share on other sites

those method work in high sierra

libcuda_378.05.05.25_mercury.dylib > libcuda_378.10.10.10_mercury.dylib

/Library/Frameworks/CUDA.framework/Versions/A/Resources/Info.plist > add new key 378.10.10.10

 

for now only tested with nbody, gonna try with premiere

 

thank you for the info @fantomas1

 

it working :D. Thanks you

Link to comment
Share on other sites

those method work in high sierra

libcuda_378.05.05.25_mercury.dylib > libcuda_378.10.10.10_mercury.dylib

/Library/Frameworks/CUDA.framework/Versions/A/Resources/Info.plist > add new key 378.10.10.10

 

for now only tested with nbody, gonna try with premiere

 

thank you for the info @fantomas1

Nice, thanks!  :thumbsup_anim:

Link to comment
Share on other sites

Hello, here is the content of my "/Library/Frameworks/CUDA.framework/Versions/A/Resources/Info.plist"

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
<dict>
	<key>CFBundleDevelopmentRegion</key>
	<string>English</string>
	<key>CFBundleIdentifier</key>
	<string>com.nvidia.CUDA</string>
	<key>CFBundleInfoDictionaryVersion</key>
	<string>6.0</string>
	<key>CFBundleName</key>
	<string>CUDA</string>
	<key>CFBundlePackageType</key>
	<string>FMWK</string>
	<key>CFBundleShortVersionString</key>
	<string>9.0.197</string>
	<key>CFBundleSignature</key>
	<string>???</string>
	<key>CFBundleVersion</key>
	<string>9.0.197</string>
	<key>MaxSupportedNVDAResmanVersion</key>
	<string>37805</string>
	<key>NSHumanReadableCopyright</key>
	<string>© NVIDIA, 2017</string>
	<key>NVDAResmanVersions</key>
	<dict>
		<key>355.10.05.05</key>
		<string>355.10.05.05</string>
		<key>355.10.05.15</key>
		<string>355.10.05.15</string>
		<key>355.10.05.35</key>
		<string>355.10.05.35</string>
		<key>355.10.05.45</key>
		<string>355.10.05.45</string>
		<key>355.10.05.55</key>
		<string>355.10.05.55</string>
		<key>367.15.10.05</key>
		<string>367.15.10.05</string>
		<key>367.15.10.15</key>
		<string>367.15.10.15</string>
		<key>367.15.10.25</key>
		<string>367.15.10.25</string>
		<key>367.15.10.35</key>
		<string>367.15.10.35</string>
		<key>367.15.10.45</key>
		<string>367.15.10.45</string>
		<key>378.05.05.05</key>
		<string>378.05.05.05</string>
		<key>378.05.05.15</key>
		<string>378.05.05.15</string>
		<key>378.05.05.25</key>
		<string>378.05.05.25</string>
	</dict>
</dict>
</plist>

Can you tell me which value do I have to change ?

 

I tried to change:

        <key>378.05.05.25</key>
        <string>378.05.05.25</string>

to:

        <key>378.10.10.10</key>
        <string>378.10.10.10</string>

But still CUDA doesn't work and ask for an update. Please can you help me? Thanks.

Link to comment
Share on other sites

use this terminal command because you're changement is broken

 

sudo cp /Library/Frameworks/CUDA.framework/Versions/A/Libraries/libcuda_378.05.05.25_mercury.dylib /Library/Frameworks/CUDA.framework/Versions/A/Libraries/libcuda_378.10.10.10_mercury.dylib

 

Thanks fabiosun, you're great ! :)

post-680637-0-80682700-1506631381_thumb.png

Link to comment
Share on other sites

DaVinci and Premiere are CUDA OK :-)

I done everything you said and now cuda is active in cuda-z and not asking me for update, but it don't work in premiere and after effects. Can you please help me and tell me is there any additional step?

Thanks in advance.

 

J.

Link to comment
Share on other sites

If you tried the small cuda test apps ( i uploaded last page) then CUDA works.

But it can be, always with newer CUDA versions, that some CUDA accell software may not detect CUDA because they check for some special CUDA version or similar probs came up.

So irs good, if you want to compare / find out why your Premiere not use CUDA and other told premiere OK that you compare software versions of the used Adobe products.

The software version of the apps can matter in this case - CUDA works or not.

Link to comment
Share on other sites

The only thing that is different is my graphic card. I have GTX 650 Ti. Worked perfectly in Sierra, but it seems it has a problem in High Sierra. 

I will try to find out if there is some additional steps for older cards. 

Thanks a lot my friend.

 

J.

 

p.s. here is my cuda information in attachment

post-883335-0-38510900-1506665393_thumb.png

Link to comment
Share on other sites

Yep , but its not only newer CUDA drivers for such probs, also the CUDA accel App must handle /know newer CUDA compute units. If app is i bit stupid and checks compute cabability to easy it may not see newest CUDA card! Happend even to CUDA-Z in the past with newer cuda drivers = CUDA-Z cant see the gpu .

But i think Adobe will look for that and if necessary updates the cuda check part of their  sw. If an official (buyed) user i would also contact support about the cuda /+ pascal problem,,,

  • Like 1
Link to comment
Share on other sites

Official cuda driver has already been released - Check post #1. It is the same one that you can find on Nvidia (en) web-page: http://www.nvidia.com/object/macosx-cuda-9.0.197-driver.html

The official web said that it only supports 10.12.x, but here I'm running Cuda 9.0.197 with Nvidia WebDriver 346.03.15f09 under El Capitan 10.11.6 (15G17013) w/o any incompatibily warning, nice!  :)

post-826765-0-70213900-1506684707_thumb.jpg

Link to comment
Share on other sites

Sorry, i not use HDMI Audio :)

But what i can say thats much much faster (3-5 times!) in CUDA, OpenCL Benches than my GTX 650 - which wasnt slow. Also OpernGL is up to 2 times faster

But most speedup by overholed pascal compute cores vs Kepler cores. I am sure even an highend Fermi gpu would be bet by the 1030 in gpu compute tasks.

Sure GT 1050+ are even more faster - but i like small 30 Watt card very much. Much faster than GTX 650 and less watts. 

  • Like 1
Link to comment
Share on other sites

×
×
  • Create New...