Message boards : GPUs : 4x 295x2
Message board moderation
Author | Message |
---|---|
Send message Joined: 27 Mar 15 Posts: 5 |
Hello, i got 4x 295x2 on my mainboard LGA1150 mainboard. I didn't buy any CPU yet, i want to ask which GPU would be enough to run those beasts ? I will NOT play any games :) it will only run for BOINC. If i buy something like a Intel Core i5-4430S will it handle the power of 4x 295x2 Cards ? |
Send message Joined: 27 Mar 15 Posts: 5 |
Okay.. i want to give some extra power, i payed for those GPUs 2800€ and here is not even one person that could help me ? I know how PCs works but i have no idea how much CPU BOINC uses. So if i invest 3000€ in the BOINC project there should be someone that would be nice enough to help me... If you dont need 60TFlops just say... i sell 3 295x2 and i have my money back and i will find a better hobby... |
Send message Joined: 2 Jul 14 Posts: 186 |
Just get the i7-4790K. Then you eliminate the uncertainty of "if it has or if it hasn't enough power". It is bigger than i5. http://www.cpu-world.com/CPUs/Core_i7/Intel-Core%20i7-4790K.html You asked about how much CPU Boinc uses. That depends very much on what projects and what applications you decide to run. Some projects have applications which load very much data into GPU and then it crunches that data largely by itself. Then it doesn't need much CPU power or bandwith between CPU and RAM. Then there are GPU applications which transfer very much data between motherboard (including memory controller and CPU) and GPU. In this case, RAM speed and CPU resources can significantly have an effect also on the overall GPU speed... or productivity. And that would be the case especially if there are 4 of those GPU's having a constant data traffic with motherboard. If you run any kind of "CPU application", then of course the CPU speed will have a direct effect on how fast the CPU tasks can complete. So, If I had those GPU's already, I wouldn't try to save on the processor. I would also mentally prepare myself: I will propably face some moderate problems when I'm trying to get all the 4 GPU's to be recognized in the system and to get them all running simultaneosly in BOINC. |
Send message Joined: 27 Mar 15 Posts: 5 |
I'm run only GPUGRID (Medicine) not useless math like Collatz Conjecture... when if you run this CC Project and after 20 Years there is still nothing... its like calculating to the endless... in Medicine you can help finde a cure. How Collatz Conjecture can help the humanity ? But that's not what the thread was made for. My i7 930 is used 10% max the most time 2-5%. I will only run GPUGRID.. |
Send message Joined: 16 Apr 15 Posts: 1 |
i have i5-3.4Ghz and 2xhd7850, intel hd, geforce210 in seti@home ang lhc@home : one amd card use 33% of one cpu core. nvidia, intel use 1% of one cpu core. |
Send message Joined: 23 Feb 12 Posts: 198 |
Okay.. i want to give some extra power, i payed for those GPUs 2800€ and here is not even one person that could help me ? I know how PCs works but i have no idea how much CPU BOINC uses. So if i invest 3000€ in the BOINC project there should be someone that would be nice enough to help me... It's called having patience as these forums may not be as heavily used as some of the project forums. Since you want these for one project only, it would have been best to ask in their forums as they would be the experts on their project. However, the advice given above is acceptable for what you requested. If you are only running the GPU's then I would suggest getting minimum of 1 core/thread per GPU and find the fastest within your budget. More cores/threads if you really feel the need. However, if you are running any CPU work (yes GPUGrid has CPU work at times as well), then I would recommend extra cores/threads based on your desires to fulfill your needs. However, as a rule of thumb (basically a blank statement to cover all projects) you should leave one core/thread per GPU free to feed it. |
Copyright © 2024 University of California.
Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU Free Documentation License,
Version 1.2 or any later version published by the Free Software Foundation.