Linux Version for Folding

[Dead Things]

The reason I use Server rather than Desktop is that I can configure Server quicker than I can Desktop, and since I don't use the GUI anyway, there is no reason to install it. I do not use Server because it performs better in any way.

One difference I have noticed on multiple rigs between Server and Desktop is that with a Desktop install, clock drift will begin to set in after a few weeks of continuous folding. PPD will then slowly start to creep downward. After a couple months, a 200k PPD machine might only be doing 185-190k PPD on average until it is rebooted. Server installs never seem to do that to me, so I've always chalked that up to the kernel timer interrupt rate of 100Hz being more ideally suited to a high-load environment.

Since you've run both Desktop and Server flavours as well, I was wondering if you've noticed anything similar? Everytime I've mentioned this in the past, folks look at me strangely (as much as you can look at somebody strangely in an online forum).

 

[rmdashrrootsplat]

Since you've run both Desktop and Server flavours as well, I was wondering if you've noticed anything similar? Everytime I've mentioned this in the past, folks look at me strangely (as much as you can look at somebody strangely in an online forum).

Normally installing vmware tools on the client fixes this issue (If you're using vmware)...

 

[musky]

One difference I have noticed on multiple rigs between Server and Desktop is that with a Desktop install, clock drift will begin to set in after a few weeks of continuous folding. PPD will then slowly start to creep downward. After a couple months, a 200k PPD machine might only be doing 185-190k PPD on average until it is rebooted. Server installs never seem to do that to me, so I've always chalked that up to the kernel timer interrupt rate of 100Hz being more ideally suited to a high-load environment.

Since you've run both Desktop and Server flavours as well, I was wondering if you've noticed anything similar? Everytime I've mentioned this in the past, folks look at me strangely (as much as you can look at somebody strangely in an online forum).

While I have never noticed this, that is not to say that it doesn't happen. PPD is not actually decreasing, but it very well may get reported as decreasing in HFM. I guess I never cared enough to even notice it, let alone look into it.

One thing you can do is set up ntp updating, which is probably a good idea anyway.

 

[Dead Things]

One thing you can do is set up ntp updating, which is probably a good idea anyway.

Good idea - thanks!

 

[014]

What hardware are we talking about? If it is bigadv capable then there is no question that Linux will be better. Even if it isn't we have seen solid gains on the same work units. I would suggest popping in a different drive and installing linux per the Musky guide and see how she runs. That way is you have issues you can just swap back and be back to where you were.

It's a Dell PowerEdge R710 with dual Xeon E5630 @ 2.53GHz running VMWare ESXi 5.0. There are no virtual machine guests except the one I have folding, which is Windows Server Standard 2008 R2 configured with 6GB of RAM and 2 quad cores. In order to try out Linux, all I have to do is create another VM with matching specs and shut down the Windows VM. It's even easier than using a different HDD as you suggested. And yes, I do have the Windows machine running bigadv. The point calculator estimates about 15K PPD for project 7504 at the moment.

I'm in the process of creating a Linux VM with two quad cores and 6GB of RAM. Maybe it will be ready to fold sometime tomorrow.

 

[014]

The Linux client turned out to barely be faster. It's getting about 800 (<1K) PPD more than the Windows client. Running both VM's at the same time produces about 14K PPD combined. If I run one alone, it gets 17K.