Message boards :
Number crunching :
GPU Requirements
Message board moderation
Previous · 1 . . . 8 · 9 · 10 · 11 · 12 · 13 · 14 . . . 15 · Next
Author | Message |
---|---|
Send message Joined: 16 Nov 09 Posts: 7 Credit: 4,901,656 RAC: 0 |
It is strange; I did not understand; however they are in my validations. Yesterday I did not see them on the Bonic Manager screen I am looking for them now, still no sight. These are the (.25-cpu) that shows the GPU crunching on the main screen. When I look on MW web page, under my account, I did see that there were GPU task running? Thinking this is crazy. I OC'ed my machine went to bed . When I got up I rushed to the rat hole computer room. Looking at the on line MW my account status I started looking at work that had been validated and there they were; GPU task issued and validated the near most of them issued and crunched in one day. It is kinda like a tail sticking out of a closed door, could be a great puppy or a angry Boar. I am just not going to pull on the tail, because with my luck it would be a bunch of Boars and Sows in heat, and that would make a bad day. So I will not inquiring about the door and tail, or even acknowledge their existence. Thanks for letting me ramble. This stuff will make you crazier than a sh... house rat! |
Send message Joined: 30 Sep 09 Posts: 211 Credit: 36,977,315 RAC: 0 |
The GTX 6nn and GT 6nn cards have some double precision capability, but not as much as similar cards in the GTX5 5nn or GT 5nn series. I haven't seen anything yet on whether the MilkyWay@Home workunits are capable of recognizing this difference and using only the double precision capability that is actually there. If not, they may insist on running on the CPU only. |
Send message Joined: 22 Oct 11 Posts: 9 Credit: 21,438,521 RAC: 0 |
Michel, The mobile HD 7970M aka 'Wimbledon XT' also supports DP and works with MW. Technically it is similar to a desktop HD 7870 aka 'Pitcairn' but GPU clock is only 850 instead of 1000 MHz. Here's a current valid result: http://milkyway.cs.rpi.edu/milkyway/result.php?resultid=374577906 and from stderr: Using device 0 on platform 0 Found 1 CL device Device 'Pitcairn' (Advanced Micro Devices, Inc.:0x1002) (CL_DEVICE_TYPE_GPU) Driver version: 1084.4 (VM) Version: OpenCL 1.2 AMD-APP (1084.4) Compute capability: 0.0 Max compute units: 20 Clock frequency: 850 Mhz Global mem size: 2147483648 Local mem size: 32768 Max const buf size: 65536 Double extension: cl_khr_fp64 (...) Estimated AMD GPU GFLOP/s: 170 SP GFLOP/s, 34 DP FLOP/s Warning: Bizarrely low flops (34). Defaulting to 100 |
Send message Joined: 15 Dec 10 Posts: 1 Credit: 695,037 RAC: 0 |
So it´s normal when milkyway uses only 5 - 10% of my ati 7770? |
Send message Joined: 26 Jun 09 Posts: 18 Credit: 469,876 RAC: 224 |
Before I go out and spend the money, I would like to know whether an Nvidia GeForce GT 610 would be a suitable GPU for this project. According to this page, it is CUDA-Enabled, with a Compute Capability of 2.1 https://developer.nvidia.com/cuda-gpus My computer only has a 305W PSU, so I can't really put in anything much more potent than this. If it is unsuitable, maybe it could be used for other projects? |
Send message Joined: 30 Sep 09 Posts: 211 Credit: 36,977,315 RAC: 0 |
The GT 6xx series doesn't do double precision as fast as the GT 5xx series, and this project uses double precision heavily. You might want to check whether there's anything in the GT 5xx series that fits the power supply limit. It will probably be able to run the applications anyway - but the question is whether it will run them fast enough. Also, don't confuse the GT 610 with the GT 610M; one won't fit in a socket for the other. |
Send message Joined: 26 Jun 09 Posts: 18 Credit: 469,876 RAC: 224 |
It will probably be able to run the applications anyway - but the question is whether it will run them fast enough. My new GT 610 is turning in the current work units in about 70 minutes each. Compared to my slow CPU, this seems very quick to me! |
Send message Joined: 4 Sep 12 Posts: 219 Credit: 456,474 RAC: 0 |
The GT 6xx series doesn't do double precision as fast as the GT 5xx series, and this project uses double precision heavily. You might want to check whether there's anything in the GT 5xx series that fits the power supply limit. The GT 610 card is based on the GF119 (Fermi-technology) chip, so it has more in common with the 5xx ranges that with its bigger GK (Kepler) siblings in the GTX 6xx ranges. Comparison of Nvidia graphics processing units |
Send message Joined: 13 Dec 08 Posts: 3 Credit: 23,340,710 RAC: 0 |
Hi, I use a HD6990 and a HD7950 But I am not able to use them both on the same MB using WIN7 or WIN8. When I have them on one machine the work units are done at the same speed as having only the HD7950 in on its own. I have removed all the AMD drivers and re installed them but the result is the same. with or without crossfire. Is there a trick to using different GPU cards on one machine? Thanks for your reply. |
Send message Joined: 1 Apr 10 Posts: 49 Credit: 171,863,025 RAC: 0 |
Greg Try plugging both cards into seperate monitors, this should cure the problem Regards |
Send message Joined: 8 May 09 Posts: 3339 Credit: 524,010,781 RAC: 0 |
Greg Sometimes you need a file text called cc_config.xml that tells Boinc to use all the gpu's it sees. Here is an example of one: <cc_config> <options> <use_all_gpus>1</use_all_gpus> <skip_cpu_benchmarks>1</skip_cpu_benchmarks> </options> </cc_config> Use NOTEPAD and just copy and paste the file in the Boinc directory saving it as a txt type file. Stop and restart Boinc and it should download and run the units on each card separately. One thing though, you will LOSE EVERY UNIT you currently have if you do this while you have units in the cache. So it is better to run your cache down before doing it. Then as John G said you MUST either make a 'dummy plug' or plug you 2nd card in a monitor on Windows startup for it to use each card separately. A 'dummy plug' how to can be found here: http://www.overclock.net/t/384733/the-30-second-dummy-plug It REALLY is as easy as they show and it works just fine. I buy my resistors at Radio Shack and they are a couple of bucks for a pack of 5 of them. |
Send message Joined: 13 Dec 08 Posts: 3 Credit: 23,340,710 RAC: 0 |
If BOINC is only using one graphics card, but it should be using more than one. It may be necessary to follow a few steps. 1. Connect at least one monitor to each graphics card. 2. Place cc_config.xml into the correct folder. 3. Restart BOINC Thank you for the help from this Message board. I was able to solve the problem with the following points. Not all operating systems let you change a file from .txt to .xml so copy any .xml file from the /ProgramData/BOINC folder onto the desk top, change the name of the file on the desktop to cc_config.xml use notepad to change the contents of cc_config.xml that is on the desktop to: <cc_config> <options> <use_all_gpus>1</use_all_gpus> <skip_cpu_benchmarks>1</skip_cpu_benchmarks> </options> </cc_config> Save the file into the correct folder (/ProgramData/BOINC) delete the file that is on the desktop. |
Send message Joined: 22 Dec 07 Posts: 35 Credit: 18,433,204 RAC: 0 |
The GTX Titan should do well here; it's FP64 performance is 1/3 that of it's FP32. 'NVIDIA's GeForce GTX Titan, Part 1: Titan For Gaming, Titan For Compute' by Ryan Smyth, anandtech.com |
Send message Joined: 26 Feb 13 Posts: 2 Credit: 1,400,786 RAC: 0 |
The GTX Titan should do well here; it's FP64 performance is 1/3 that of it's FP32. I have tested Titan last day on milkyway and it works ok. When doubleprecision is enabled it can run 7 tasks in parallel (~97% GPU, 80% of TDP=200W) and can produce cca 300k credits per day. |
Send message Joined: 11 Dec 08 Posts: 8 Credit: 14,977,931 RAC: 0 |
Thanks for sharing your results. Could you try running one task at a time with the 1/3 DP option enabled and post the runtime per task? I wanted to see how the Titan runs compared to the 7970 with a single task running. Also, do the overclocking and power target options work with DP mode enabled? |
Send message Joined: 26 Feb 13 Posts: 2 Credit: 1,400,786 RAC: 0 |
Times for one task is the same as for 7 tasks in parallel = ~500s. msiasfterburner 3.0.0 beta 5 allows me to oveclock card with dp enabled, but I didn't test it. |
Send message Joined: 25 Feb 13 Posts: 3 Credit: 4,795,765 RAC: 0 |
have a Geforce 6600gt will it work on this project ?? i seem to be crunching w/u hope i'm not wasting my time |
Send message Joined: 30 Sep 09 Posts: 211 Credit: 36,977,315 RAC: 0 |
have a Geforce 6600gt will it work on this project ?? See the first post in this thread. The 6600gt does not support double precision, and therefore will not work on this project. A few more astronomy-related projects you might want to try it on: Einstein@Home http://einstein.phys.uwm.edu/ SETI@Home http://setiathome.berkeley.edu/ |
Send message Joined: 14 Feb 09 Posts: 999 Credit: 74,932,619 RAC: 0 |
have a Geforce 6600gt will it work on this project ?? That video card is not capable of crunching at all, the first Nvidia card with CUDA was the 8xxx series. |
Send message Joined: 25 Feb 13 Posts: 3 Credit: 4,795,765 RAC: 0 |
ohhh? i have credit for compleated w/u and 2 awaiting validation |
©2024 Astroinformatics Group