Best Video Cards for DCing?

O

osrk

[H]ard|Gawd
Joined
Jan 10, 2003
Messages
2,032
I have a bunch of video cards from mining. I’m downsizing now but wanted to know which to keep as I’ll throw a few in case and let them rip on DC projects. All are nvidia 3070 and up.
 
osrk said:
I have a bunch of video cards from mining. I’m downsizing now but wanted to know which to keep as I’ll throw a few in case and let them rip on DC projects. All are nvidia 3070 and up.
Click to expand...
Welcome to our DC forum.

There are a list of projects (both cpu and gpu) that you can review them here and how to get BOINC up and running. In most projects, you need to select run gpu tasks only and uncheck cpu tasks.

From the top of my head, here are the gpu BOINC projects that Nvidia cards will perform very well.

1. Folding@home. This is a non-BOINC (medical) project. This project is popular among the [H]ordes.
2. World Community Grid. Medical. Only Covid tasks have gpu. Currently in the process of migration from IBM to Krembil but should be up shortly. That's what they have been saying for weeks. This project is popular among the [H]ordes.
3. GPUgrid. Medical. Limited work units availability. Most of the time your gpu will be idle.
4. Einstein@home. Astrophysics
5. MilkyWay. Astrophysics. Need FP64 (double precision) card for best performance. Probably not worth running with consumer cards.
6. Primegrid. Math
7. MooWrapper. Math (cryptography)
8. SRbase. Math
9. Amicable Numbers. Math.

There are few other math projects where I think they are not worth running with gpu since the codes written are poorly optimized for it.

If you need additional help, post it here.
 
osrk said:
I have a bunch of video cards from mining. I’m downsizing now but wanted to know which to keep as I’ll throw a few in case and let them rip on DC projects. All are nvidia 3070 and up.
Click to expand...
What is "best" to run really comes down to a lot of factors. Each project has different needs and the hardware that runs "best" on them can also be different. For example, some projects run better on nVidia cards. Some on AMD because of the implementation of OpenCL. Typically the projects will code their app for one card and then make the app "work" on the others. This produces a little bit of bias in the efficiency and performance. Some projects ONLY support one GPU type. Then you have projects out there that may utilize double precision. Most newer cards aren't so great for that. The older AMD cards had better double precision. However, when you start breaking it down to cost these days, I really don't know what is best there since those older cards really are getting long in the tooth. The 3XXX series cards are awesome at "most" projects. If you aren't trying to have the "top rig", 3070-3080 are probably sweet spots for efficiency but I will let others that have the cards chime in. My best card is a 2080 Super. Your cards will blow these away in performance.

The biggest thing I would look at is which sciences you want to support. If you are open to all of them, then great. But some only want certain types and that can really limit your options. For example, with BOINC, there really is only 2 projects that have Bio/Medical GPU work. One of them only supports nVidia. Folding@home is another option but is outside of the BOINC spectrum. Once you decide that, then I would ask are you wanting to be someone that tweaks and optimizes or are you looking for a simple set and forget setup? FAH is extremely limited on tweaking and optimizing your hardware compared to the multitude of options that BOINC has available to it. Some people get turned off with how flooded they get with options with BOINC. Next, not all BOINC projects run or score the same. So, if you are joining to be competitive at all (this is where the fun is) then you need to know that going in. We can certainly give direction for you if you have more incite to give.
 
Folding @ Home also added CUDA support awhile back so you'll see great results with those RTX cards you have. I assume you keep your drivers reasonably current which is recommended for folding. I noticed an increase in points when I upgraded a driver on a lesser used machine to a newer one with a higher CUDA version a few months ago. pututu's quick list spells it out quite well : GPUGrid will be mostly idle, WCG has been dormant for months and is experiencing migration issues atm. I'm mostly folding as I prefer medical or environmental projects. Couldn't care less about finding prime numbers or celestial maps, etc. Folding is pretty much set and forget too and is easily monitored with HFM.net . If you decide to join consider team 33 for the accolades and trash talk ;)
 
You must log in or register to reply here.
Back
Top