Message boards :
Number crunching :
What means Vulkan API to crunshing ?
Message board moderation
Author | Message |
---|---|
The_Matrix Send message Joined: 17 Nov 03 Posts: 414 Credit: 5,827,850 RAC: 0 |
http://www.chip.de/news/Die-Grafik-der-Zukunft-OpenGL-Nachfolger-Vulkan-kostenlos-downloaden_89813815.html Thats new, i guess. Anything exiding on this are known ? |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
http://www.chip.de/news/Die-Grafik-der-Zukunft-OpenGL-Nachfolger-Vulkan-kostenlos-downloaden_89813815.html To me, It basically means that I'll be able to have a pretty potent graphics and compute option to factor into x42, that should be capable of working on just about any device with Vulkan support. Its low latency architecture, and thread scalability should help resolve A fair few things that with both Cuda and OpenCL involve wrestling. While it won't replace Cuda or OpenCL, on many devices it will be the only viable compute option, and its threaded and object-ish structure will probably be mirrored in the next generation of applications, even where Vulkan isn't used as such. ALready started work last year on a plugin architecture to make the same builds do Cuda, OpenCL, Vulkan, DIrextCompute and Renderscript. Am registered for attending the webinar. While I'm expecting them to be mostly Grpahics/Gaming targeted, I do hope they'll discuss compute to some extent at least. That part looks fairly OpenCL like though, just with different synchronisation mechanisms. "Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions. |
The_Matrix Send message Joined: 17 Nov 03 Posts: 414 Credit: 5,827,850 RAC: 0 |
can it be ? Just noticed on E@H , with the NV Vulkan drivers, a littebit less cpu usage by cunshing time on cpu are displayed. In a hour i can tell more, or impossible ? |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
Could be, but early days I'm sure. The Vulkan drivers are supposed to be lower latency, which, speculating, would mean differences that would affect Cuda and OpenCL as well, at least a little. If so, some of that could just be because of switching in some improved OpenGL calls, and in the case of Cuda maybe bypassign DirectX for some things. "Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions. |
The_Matrix Send message Joined: 17 Nov 03 Posts: 414 Credit: 5,827,850 RAC: 0 |
the lowest cputime on a workunit (cuda) of Einstein@Home i ever had. 399 seconds, before there were 475 to 600...but not always equal, because the clocking was not constand, now it were constand at 1202 Mhz on a GTX 670. The usage of the MCU of the gpu ist higher. 60 to 70 % percent. Now trying on SETI :D |
Chris Adamek Send message Joined: 15 May 99 Posts: 251 Credit: 434,772,072 RAC: 236 |
ALready started work last year on a plugin architecture to make the same builds do Cuda, OpenCL, Vulkan, DIrextCompute and Renderscript. And Apple's Metal I hope too.:) |
The_Matrix Send message Joined: 17 Nov 03 Posts: 414 Credit: 5,827,850 RAC: 0 |
u would´nd believe that, cpu time have lowered by 50% percent on Seti cude50 tasks naturally. :DDDDDDDD Is it true ? Crossing fingers they are all not invalid...we´ll see tomorrow. |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
ALready started work last year on a plugin architecture to make the same builds do Cuda, OpenCL, Vulkan, DIrextCompute and Renderscript. Yeah, making the plugin architecture potentially cope with anything that will accept being wrapped in pluggable libraries, so probably other apis that don't exist yet too. It'd make life easier if these pesky companies would stay still for a few months, lol. "Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions. |
The_Matrix Send message Joined: 17 Nov 03 Posts: 414 Credit: 5,827,850 RAC: 0 |
Astropulse running insane, got here one 3rd of the gpu time as before... all drivers dependend, as far as i can "see", but will only believe if the tasks are valid. First VALID result: http://setiathome.berkeley.edu/result.php?resultid=4735103696 |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
With Cuda50 I'm seeing a more modest reduction in CPU time (Old COre2 driving a 980 is a bottleneck), down from ~33 secs to ~27 seconds (or an ~18% reduction) Probably the latency improvements will be heavily system and application dependant (i.e. if not bottlenecking on driver Latency already, or is starved somehoe else, the driver won;t yield as much). It's going to be interesting to work out what they're doing under the hood. Old school graphic OpenGL synchronisation was faster than DirectX and Cuda style synchronisation for years, so it wouldn;t durprise me if similar streamlining sees some kruft removed over time. "Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions. |
The_Matrix Send message Joined: 17 Nov 03 Posts: 414 Credit: 5,827,850 RAC: 0 |
Anybody´s there to confirm , that the drivers work correctly ? |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
Seem Ok here so Far (though very early to tell), and also seem to work with actual Vulkan tests as well. "Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions. |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
I'm still using Cat 15.7.1. Which I expect doesn't have Vulkan. My R9 390x is listed as being Vulkan compatible, but I imagine a new driver will be required. Which will probably only be in the first Crimson 16.x for Ati/AMD GPU drivers. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
I read somewhere amidst the flurry, that AMD's Vulkan drivers were still in Beta "Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions. |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
I read somewhere amidst the flurry, that AMD's Vulkan drivers were still in Beta Maybe the 16.1.1 driver has it, but that is a beta driver. For my home systems I only use release drivers. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
The_Matrix Send message Joined: 17 Nov 03 Posts: 414 Credit: 5,827,850 RAC: 0 |
@HAL9000 Think u got already DirectX 12 , it coundn´t get better, but on DirectX11-11.1 cards there must be an advantage.I am blind, i am only guessing. |
The_Matrix Send message Joined: 17 Nov 03 Posts: 414 Credit: 5,827,850 RAC: 0 |
https://www.khronos.org/vulkan/ There is also an intel-gpu driver for linux available. Anybody´s there , who crunshes on lunartics , plus installed the Vulkan API ? Results needed , before,after and so. Pioneer work must be done, ok, i´ll the first whos running lunarics with the new Vulkan API. Hope for valid results... ok, on Astropulse the wu will take 5 min more. I set the system back. |
Ulrich Metzner Send message Joined: 3 Jul 02 Posts: 1256 Credit: 13,565,513 RAC: 13 |
Well, looking at your already crunched results using the Vulkan driver, i cannot see a significant gain? The units crunched "faster" have a higher AR and therefore alone will be ahead. :? Aloha, Uli |
The_Matrix Send message Joined: 17 Nov 03 Posts: 414 Credit: 5,827,850 RAC: 0 |
sorry, wrong card |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
Those are [were before edit...] very highly blanked APs... "Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions. |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.