Client Setup Optimization Questions

[dawhippersnapper]

I wanted to reply back here with some results for others reference.

If I lower the core count for the dual e7-4830 Xeon server to 30 during highest load hours it doesn't drop below 40k ppd. If I raise priority manually during low to no load hours it will achieve nearly 60k ppd. During high load hours without FAH running on this server it is doing high amounts of transactional computations (SQL Server with Enterprise SSD Drives http://www.intel.com/content/www/us/en/ ... eries.html and lots of memory). It was using many cores but very fast peaks.

I also have a pair of dual L5520 Xeon servers that have very low work load (Directory Services, File Servers (Attached to Dell 3200i SAN), Update Management, DNS Servers) that get from 20 to 24k PPD. They don't really require any tweaking.

My desktop with the i7 3770 and nvidia gtx 680 SC+ is doing nicely (50k to 205k ppd with 8057 units), but it's always going to be sporadic on performance because I use it heavily and run VMs on it for testing.

I wanted to thank you all for the tips. Since tweaking I have gone from an average of about 150,000 ppd to about 230,000 ppd, with a one day peak of 289,000 ppd.

[Joe_H]

In trying out core counts on your dual-E7 Xeon server there are some even number settings you may want to avoid. Those would be numbers like 26 and 28 which have prime number factors over 5. There are known issues with some projects with a factor of 7, and primes 11 and higher usually do not work. A couple anecdotal reports have come in about 5, but not enough to determine a consistent pattern. So if you do run into issues with any specific WU's or projects with the setting of 30 please do report them.

But it sounds like your tweaking has resulted in better F@H throughput overall. Hopefully you will not run into any issues.