What are the combinations that you might be using today that work best with the latest projects?

For instance, a Series 30x NVidia GPU paired with an Intel 13x CPU running Red Hat works better with Amicable then a AMD 780M and Radeon 7600 running Windows 11? What about Docker/Podman using WSL? Details like memory configurations, de-bloating the OS, pagefiles, resource sharing settings? How do you have your cc_config.xml set?

The devil is often in the details! Slight differences in configuration settings can yield better results for certain projects and applications so can you share what your experiences are?

Especially projects like PrimeGrid or Einstein that can use all the different manufacturer and OS combinations - share with us the secrets to maximizing WU's!

I want the best performance I can get from the new and old rigs that I have and it seems difficult to find decent empirical experience.

Happy crunching!

Dear m0laki,

My rigs are a mixture of good, solid, older business-grade machines and newer off-the-shelf gear. The vast majority are configured to work solely on BOINC, while a wee handful are also pressed into service for daily chores.

Every single machine has been stripped of unnecessary processes so that as many resources as possible are free for optimised BOINC operation. For Windows, this involves a manual muck-through of the Registry, chucking out any unnecessary bulk-ware, and keeping a close eye on any sneaky processes that creep in with updates. The Linux installs are done raw, with a weeding out of non-essential programmes, and performance is tuned for the best possible use of the collective resources for BOINC.

Furthermore, a sharp eye is kept to ensure no unnecessary malware is inadvertently added during updates. On top of that, all PCs are checked several times a week to ensure they're running optimally, and there is a daily check to confirm they still have contact with the BOINC servers.

I hope this has shed a bit more light on the optimisation work I do.

PCs Registered on call of BOINC:
Intel(R) Core(tm) i3-3217U CPU @ 1.80GHz - GPU Intel 4000
Intel(R) Core(tm) i3-2100 CPU @ 3.10GHz – GPU:Nvidia GTX 1050 Ti
Intel(R) Core(tm) i5-2400 CPU @ 3.10GHz – GPU Nvidia GTX 1050 Ti
Intel(R) Core(tm) i7-3770 CPU @ 3.40GHz – GPU Nvidia RTX 3050
Intel(R) Xeon(R) CPU W3550 @ 3.07GHz – GPU Nvidia RTX 3050
Intel(R) Xeon(R) CPU E3-1271 v3 @ 3.60GHz – GPU Nvidia RTX 3050
Intel(R) Xeon(R) 2X CPU E5-2667 v4 @ 3.20GHz, 3201 Mhz – GPU Nvidia M4000 – Offline
Intel(R) Core(tm) i5-13400F – GPU: Nvidia RTX 3050
Intel(R) Core(tm) i5-14400F – GPU: Nvidia RTX 4060

Yours aye,

MacCoinneach.

Dear m0laki,

It's been a right doddle to find a final solution to this conundrum, I'm afraid. I've been poking around for a formula for a fair while now, hoping it would guide me to the correct choices.

Now, it's my firm belief that all the different projects have their own wee preferences, and that's why each one needs a truly individual solution to get the best out of it. What works a treat for one project will often give ye a poor show for another, even if the PC has the exact same specifications.

It's been my experience that ye simply hae to properly suss out what each project actually needs. For this very purpose, I've got a 'reference computer' I use to see how a project behaves with both CPU and GPU WUs. This lets me figure out which project is getting the most out of my kit. Mind ye, this is far from a universal fix, and often ye'll need to run a good few trials, letting one setup have a go for a number of days or even weeks. Ye can judge the efficiency by looking at the individual PC's statistics on BAM.

Over time, I've come to the conclusion that there's no single final answer that fits all computers. Instead, ye simply need to have a wee experiment with each individual machine, especially after those initial trials point ye in a promising direction, giving ye a decent chance of getting a satisfactory result.

Hope this is both informative and useful for you, pal.

Yours aye,

MacCoinneach.