Windows: Quadro P600 or GTX 750 Ti SC

[TTAlex77folder]

Building a machine from spare parts while I'm stuck at home.
Which graphics card is better for Folding? The P600 is newer, but I think the GTX 750 TI has more CUDA cores (the SC in the 750 Ti card I have identifies it as "Superclocked" with what I believe is a marginally faster base clock).

Rest of system is pretty standard i7-4790 and 24 GiB of RAM.

[ajm]

This site can be helpful: https://www.techpowerup.com/gpu-specs/g ... 0-ti.c2548

[bruce]

Both will work.

I'd go for the 750 Ti

In fact, if your Mb has two slots, you can run both of them as long as your P/S is big enough and your airflow can provide adequate cooling.

[TTAlex77folder]

Both will work.

I'd go for the 750 Ti

In fact, if your Mb has two slots, you can run both of them as long as your P/S is big enough and your airflow can provide adequate cooling.

Ended up doing a bit of what Bruce recommended -- running the GTX750 in a separate machine with only one PCIE slot.

But realized that I could put the P600 in my main machine along side a GTX1070. I managed to put the P600 in to TCC mode too - not sure if it helps or not.

The P600 has only 384 CUDA cores vs. the 1,920 on the 1070. Every little helps, right?
Anyway, reinstalled FAH, and it immediately picked up all COVID jobs on both cards and the CPU. No heat required tonight.

[foldinghomealone2]

Sorry to say it like that but I wouldn't fold with any of them.

Right now there is a shortage of work units.
When you fold with a 750Ti which is much slower than a 1070 but your 1070 is idle because slow GPUs take 'forever' to return the WUs, then I hope you understand my point.

FAH is a relay race. Only once a WU is returned, a new one can be generated and distributed.
When you're the slow runner your team will lose.

You might consider to postpone folding with slow GPUs until there is no shortage in WUs anymore.

[Nathan_P]

there is not a shortage of WU, there is a bottleneck in getting WU out to clients. the 2 are not quite the same thing, more servers are being added - 9 are currently in the process of being spun up. This should ease the bottleneck somewhat

[foldinghomealone2]

NathanP, being too overly smart doesn't help either.

Let me quote from the main page:
"The Folding@home team know about the work unit shortage.
It's happening because of an approximately 20x increase in demand.
..."

Edit:
But maybe explain you point in more detail.
I don't understand your point. You're implying that WU stocks pile up and can't be distributed fast enough. But how can that happen when a WU can only be uploaded once it has been downloaded/distributed/assigned (and processed)?`
--> In my understanding you can't create more WUs that you assign.

[Nathan_P]

Every project has a different number of runs and each run has several clones, therefore a an individual project could have several hundred or thousand WU ready at any time, these runs and clones are generated when the project is set up by the researchers. You can see this if you look at your log, you will see a Project, followed by a run, clone gen number such as this example:- Sending unit results: id:01 state:SEND error:NO_ERROR project:14551 run:3 clone:0 gen:2 core:0xa7 unit. Gens above zero are only created when the preceding one is returned. Have just returned Gen 2 of that particular clone & run and the servers will generate gen 3 to be distributed. Pulling up my HFM log shows project 14543 having 178 clones on run 0, 14541 having 379 and 16405 953. If there are say 10 different runs of this those numbers go up 10x, so while there may not be many projects, there are lots of WU.

According to server status there are currently 509k WU available and they are currently being handed out at a rate of 126k per hour. In the banner at the top of the page is a link saying server stats, this will take you to the server status page where you can see how many servers, assignments and connections there are.

Each server can only handle so many connections per hour, this ranges from 28k down to 2-3k depending on network and I/O bandwidth. As more servers get added the assign rate should go up which will each the bottleneck in getting WU out to the donors

[foldinghomealone2]

Now I understand.
As long as there are enough WUs available which were generated when the project was set up (Gen 0) the assignment is the bottleneck (in this current situation with 'too many' folding CPUs/GPUs).
Thank you for your explanation.