Help with setup of i5 4590 plus nvidia GTX960

Edit your app_config.xml file and insert the line I gave you in the previous message as an extra line, immediately after the line. Insert enough leading space characters on this new line so that it lines up correctly with both the line before it and the line after it.

Save the modified file and then force a re-read of config files. There's no need to stop BOINC for any of this.

When you read the event log excerpt, you can see it is insisting on having a plan_class as it can't interpret a blank value (the non-existent stuff between the '' - two single quotes). So, we will need to pay attention for when the Devs announce that the beta cuda55 app is being replaced by the final version as the beta plan_class probably wont work after that.

A second solution (probably a better one for you) is to still use app_config.xml but to use the alternative method as is shown in the application configuration section down near the bottom of the BOINC client configuration page. Here is an example based on that alternative method that should do the same job and doesn't require a plan_class, so will continue to work after the beta test app is replaced.

The disadvantage of this method is that if you had two different GPUs in your machine, they would be forced to run tasks using the one set of settings. This may be not what is needed in other cases but shouldn't affect you unless you wanted to start using your internal Intel GPU. There doesn't seem to be a reliable working driver so it's not an option at the moment.
[pre]

einsteinbinary_BRP6

0.33
0.34

[/pre]
You could copy and paste the above into notepad and save it as app_config.xml. Replace the previous version with the above, re-read config files and the desired changes should occur.

Very good! Yes, we wait until we can accurately judge the change. The smaller the change, the longer you should wait to be sure.

I'm a nutter that craves an overheated working environment :-).

You say you find space absolutely amazing. I understand the feeling. I want to contribute as best I can to discovering what's out there and working out really how it all evolved. Doing the grunt work in the background to help true scientists work it all out is what works best for me. All my machines essentially do nothing but crunch. I try to make that as efficient as possible.

I know my limitations ... I think I'll pass on that one thanks ;-).

Three years ago, when I decided to get into GPU crunching in a big way, I reckoned the most important factors would be (1) keeping things cool, and (2) using electricity as efficiently as possible. Because I had concerns about multiple GPU configs - would they be less efficient, harder to cool, more prone to crashes and lockups, have less durability, etc - I made the decision it would be 1 GPU per host. If I were starting again now (which I'm not - well not yet :-) ), I would probably go for a dual GPU design. Things always evolve over time - a shorter time these days than previously.

Here is a list of some things that had a big influence on my decisions:-

• * I wanted to make a big contribution so power usage would be high. It would all end up as heat so cooling would be an issue.
  * I wouldn't be prepared to pay for aircon for an enclosed environment so there would need to be a good flow of external air to remove heat.
  * I had an existing big supply of 300W 80+ efficient PSUs (270W at 12V) that I had already purchased extremely cheaply as a job lot. This put an upper limit on power consumption per machine and on how much power the GPU could use.
  * I wanted to support both CPU and GPU applications.
  * All machines needed to be in open frames on open racks for good airflow and ease of maintenance.
  

With these constraints on the design, I chose to go with single PCIe slot budget motherboards with low power CPUs and the best possible compromise between work output and power use for the GPU - which basically pointed to the GTX-650 at that time.

I chose Pentium dual core and i3 with HT quad core CPUs and budget single x16 slot motherboards. In a second stage about 12 months later, I had the opportunity to acquire around 30 AMD HD7850s (2GB) with a 256bit interface, very cheaply. This last factor (256 bit bus width) was really a deal clincher because it had become apparent this had a big bearing on crunching efficiency at that time. I hasten to add that it's no longer so critical since the Devs have optimised the apps to pretty much remove the problem.

So, those were things that influenced my decisions. Without knowing all considerations and factors that affect you, it's pretty hard to give a single "best" option. You really need to put down on paper and think through all the factors you consider important, rather than taking what someone else did in a different era, for different reasons, and with different design constraints. I still think that power efficiency and heat dissipation are key considerations. But if you're looking at a single machine, perhaps in a quite different environment, it might be a quite different answer from what someone else might choose. So, make a list of what you want to achieve, the important factors to you, where you intend to house it, what other uses it might have, etc, and I could probably be a bit more specific about what might work best for you.

If anyone browsing this thread wants to tout their particular favourite design, feel free to do so provided you at least make a basic attempt to lay out the various factors and considerations that your design is supposed to be addressing. That way, the reader has some chance of evaluating each design and deciding if it is fit for his/her particular purpose.