Use for 2p/4p servers?

[Sneevly]

I'm looking around the threads trying to figure out what people are doing with their 2p/4p set ups that used to be used with -bigadv in F@H. I've got two sitting that can not connect to F@H servers and am waiting on answers on the folding forum.

My plan is to reach a billion with F@H, which is only about ~40 million away, and maybe quit.

I know most projects may not be optimized for these rigs any more but I'd like to use them for SOMETHING soon. I'm leaning towards the idea of WCG. But I'd prefer to do any health type research that can actually use my two 48 core set ups at least some what efficiently.

So I guess I'm just asking how does 4p work with WCG, or are you still all folding with the high core count computers?

 

[Gilthanis]

WCG works fine with 2P/4P. However, the difference is that they will run a single work unit on each core/thread. So, you need to either verify you have enough RAM to support that many work units, or you will need to limit the number of cores/threads you let BOINC use.

 

[Gilthanis]

References can be found in this Guide for the links of various installs. All Inclusive DC Project list v.2
Just make sure to setup your account using the right team, which this link should do for you. World Community Grid

 

[Sneevly]

WCG works fine with 2P/4P. However, the difference is that they will run a single work unit on each core/thread. So, you need to either verify you have enough RAM to support that many work units, or you will need to limit the number of cores/threads you let BOINC use.

How bandwith intensive would 48 instances be? Sounds a little much. I've got 32GB of RAM so I have no idea. Guess I'll give it a shot soon and see how it turns out.

References can be found in this Guide for the links of various installs. All Inclusive DC Project list v.2
Just make sure to setup your account using the right team, which this link should do for you. World Community Grid

Thank you very much!

 

[Gilthanis]

Typically not bad as far as bandwidth goes. If you run several CEP2 work units at a time, downloading those can be rough. However, by default your account should have it set to only download 1 per computer at a time. Those are also the highest RAM requirements and highest IO intense work units. Once your account is set up at WCG, you can choose which sub projects you want to run from your account at WCG's website.

 

[Nathan_P]

There is currently no work available for 48 thread machines or any for the v6 client. You need to install the v7 client and configure the client to use only 24 threads, you can have 2 cpu:24 slots running to max out the machine and it will net you about 200k per day depending on the cpu's installed

 

[ChristianVirtual]

And adding few GPUs in ... Lots of PCI lanes

 

[[Ion]]

How bandwith intensive would 48 instances be? Sounds a little much. I've got 32GB of RAM so I have no idea. Guess I'll give it a shot soon and see how it turns out.

Thank you very much!

Depending on which sub-projects you choose, not much. I have an 88-thread Xeon E5 system (see sig) that gets by running WCG's OET WUs on 8GB RAM (and actually has 2GB free). CEP2 is a RAM hog, but really even that should be fine on 32GB. I have all my 48C systems running on 8GB each and that's more than enough...hell I have a 24C system running on 3GB.