WUs for GPUs
Message boards :
Number crunching :
WUs for GPUs
Message board moderation
| Author | Message |
|---|---|
|
Send message Joined: 24 Jul 22 Posts: 1 Credit: 961,485 RAC: 408    
|
Hi. I want to know whether it's planned to generate WUs for GPUs, NVIDIA, Intel or whatever. Thanks. |
 Jesús Carro Send message Joined: 18 Mar 15 Posts: 317 Credit: 3,636,673 RAC: 9,755   |
Hi AkasPDM, Not in the short/medium term. We have an idea for a new type of simulations that would greatly benefit from GPU power, but it's a big project. Best, Jesús. Jesús Carro Universidad San Jorge @InSilicoHeart |
|
Send message Joined: 9 Apr 15 Posts: 207 Credit: 1,573,819 RAC: 137
|
Hi AkasPDM, 1 year later, our gpus are still ready :-P |
|
Send message Joined: 9 Apr 15 Posts: 207 Credit: 1,573,819 RAC: 137
|
Hi AkasPDM, Any news?? |
|
Send message Joined: 21 May 15 Posts: 10 Credit: 30,804 RAC: 94 
|
I believe DENIS-fiber could be highly efficient on GPUs. If there's interest from the admins, I could help with porting it. |
|
Send message Joined: 9 Apr 15 Posts: 207 Credit: 1,573,819 RAC: 137
|
I believe DENIS-fiber could be highly efficient on GPUs. If there's interest from the admins, I could help with porting it. This could be another big help from you. Thank you in advance!! |
|
Send message Joined: 9 Apr 15 Posts: 207 Credit: 1,573,819 RAC: 137
|
I believe DENIS-fiber could be highly efficient on GPUs. If there's interest from the admins, I could help with porting it. But, maybe, it's better if the app will be marked as stable and not beta... |
|
Send message Joined: 21 May 15 Posts: 10 Credit: 30,804 RAC: 94
|
The code where most of the time is spent is generated, and if those models change with each version, rewriting it efficiently for a GPU could be challenging... As boboviz mentioned, the app (and the server) isn't stable yet, so it's a bit too early to start thinking about GPU versions. |
 Send message Joined: 5 Jul 15 Posts: 44 Credit: 6,852,754 RAC: 8,843
|
The code where most of the time is spent is generated, and if those models change with each version, rewriting it efficiently for a GPU could be challenging... Party Pooper!! But you are still right of course!! |
|
Send message Joined: 26 Jan 23 Posts: 9 Credit: 340,615 RAC: 539
|
Take the MacBook Air Late 2020, for example: it has an SoC integrating the GPU and CPU onto a single chip, but lacks active cooling. When I run the GPU on a configuration like this – as I experienced running Einstein@home (GPU 1 core/WU x 2 WU) – it gets scorching hot, almost enough to fry an egg on, reminiscent of the old Intel NetBurst days. Until countermeasures are added, like the ability to set a maximum temperature limit for the SoC, I'd really prefer it if the GPU and CPU workloads were kept balanced, as there's a potential risk of shortening the hardware's lifespan otherwise. 
|
|
Tamagoch Send message Joined: 4 Aug 15 Posts: 5 Credit: 1,227,140 RAC: 1,800 
|
A shorter lifespan of laptops is a longtime vendors' strategy to keep you buying new laptop every couple of years. You obviously cannot fight that with the distributed computing. The 'power' of newer devices is designed only for a peak usage to make you think that laptop is operating smoothly and damn fast. They have learned that from smartphones, which I hate even more. I believe it all have started from Intel's 8th Gen. Comparing to i5-6xxxU having 65-67C under a constant BOINC workload, those newer i5-8xxxU was melting to 98-100C and since then it became 'normal'. That's my conspiracy theory for a modern hardware ) |
|
Send message Joined: 19 Jun 24 Posts: 23 Credit: 359,569 RAC: 3,840
|
I believe DENIS-fiber could be highly efficient on GPUs. If there's interest from the admins, I could help with porting it. i need an aarch64_cuda version :) |
