Posts by ronny

1) Message boards : GPUs : Replacing electrical ovens with BOINC computers. (Message 56373)
Posted 2 Oct 2014 by ronny
Post:
Update: May have found my future supplier of heatpipe equipment. Didn't find cooling solutions that do what I require them to do, so I'm going to make it myself with heatpipes and other heatpipe parts.
Here's what I have ordered so far: http://www.aavid.com/products/standard/heat-pipe-discovery-kit
Going to underclock a small quad core I have and then put on this to get an idea of how substantial the heatpipe and heatsink has to be to cool something as substantial as multiple GPUs.
2) Message boards : GPUs : Replacing electrical ovens with BOINC computers. (Message 56362)
Posted 2 Oct 2014 by ronny
Post:
I think I was the one talking about it, I have talked about it a while back. But had no space for it.

There's 580gtx and 590gtx and tons more for sale for pennies. I think finding the hardware for cheap enough is not the issue, but actually putting it together in such a way that it don't scream like an airplane is going to be an issue.
When it comes to the business aspect of it, I already have the concept drawings for how it would have to be made and assembled for it to be viable (it can't take more time and effort to install than underfloor heating and wall-mounted heaters). And the owner of the computing hardware would buy it for the purpose of the computing power, then installing it somewhere that requires heaters at no cost for the building owner(s). Then the building owners pay for the electricity for the computing power.
3) Message boards : GPUs : Replacing electrical ovens with BOINC computers. (Message 56326)
Posted 30 Sep 2014 by ronny
Post:
Here in Norway its now getting cold, and it will stay cold for about 9 months or so. I am therefore trying to put together enough hardware to heat most of the house with BOINC machines. If I don't get close enough I'll have to overclock and overvolt to get as much as I can from the hardware, both in terms of heat and computing. The termostat-controlled normal electrical ovens will make up the small difference between the heat produced by the machines and the heat that escapes the building (more heat will escape the building than what the machines will output, so the temperature would be for example 16-18 degrees C with the machines themselves running, then the termostat ovens make up the difference, that way windows don't have to be opened when temperatures are relatively high outside).

I've contacted a few dozen people who have put up used 280x/7970 cards for sale, and have also contacted a large seller of such hardware (they might have some demo-products and returns etc). And I'm looking for cheap ways to get as many GPUs as possible running within the budget (both money budget and electricity budget, since I don't require 10kw heating, more like 3kw).

I'll update here to let you guys know what happens.

First point, I've found this marvelous thing called "PCI express backplane": http://www.trentonsystems.com/backplanes/bpx6806-pci-express-backplane
4 - x8 PCI Express electrical / x16 mech. connectors
12 – x4 PCI Express electrical / x16 mech. connectors
1 - x4 PCI Express electrical / x16 mech. connectors
1 - x4/x1 PCI Express electrical# / x8 mech. connector
Do you have any idea of how for example 7970/280x would run milkyway@home on such a setup? Its probably enterprise-expensive but if possible it would be nice to just have 1 system to maintain.
Also, do you know of any competitors that make similar products?
4) Message boards : BOINC client : My Wish List (Message 46144)
Posted 28 Oct 2012 by ronny
Post:
To be able to manually suspend the current gpu project so the other gpu project starts up until such time you suspend that, and/or resume the first gpu project. It is such a pain to have to suspend gpu work completely just to surf the web because for example "help conquer cancer" tasks in the world community grid project make the fps drop to about 10 intermittently, when gpugrid would work fine for that period.



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.