Does Folding detect your hardware?

Tim

When you sign up with Folding, and the computer is getting / sending work units, does the Folding server know what kind of hardware you have, like memory amount, CPU speed, etc? And adjust the size of the work units you get to best match it?

Or does it just give you something, and only look to see if you have large work units enabled or not?

GHoosdum

I believe that the client knows, and not the server. The client merely requests units based upon how it's set up, and in the end just gets what the assignment server gives out.

edcentric

The only hardware that it checks is memory, I beleive.
Other than that it only knows what you entered.

Not real efficent, but very secure.

shwaip

It checks for extended instruction sets, IE: SSE/2, 3DNOW!

edcentric

I thought that it was the local software that checked for that.
I don't think that the client or server know if you can run and extended instructions. If they did I shouldn't get the same units on both my TBird and Barton powered boxes.

muddocktor

It all depends on the work units, what the assignment server checks for. On some wu's, the assignment server will check to see what performance fraction (pf) the client is averaging and on others (like the BP work), it checks for a minimum memory requirement. A lot of the wu's being assigned don't have any checks being performed on them though.

You can check to see if there is a minimum performance fraction requirement for a particular server by checking the serverstats page and looking at the server(s) in question under the "Min PerfFrac" column and if there is a performance minimum, it will be listed there. Checking right now, servers VSP46, VSP57, and VSPMF81 have wu's requiring a minimum pf of .95 and servers VSPMF86, VSPMF82, and VSP77 have wu's in the .7-.85 pf range.

edcentric

I forgot about the pf.
It is sort of a raw deal. If you are on dialup and run multiple clients you will automaticaly have low pf's.
But as long as I keep turning points it is OK.

mmonnin

If you check, it does a benchmark of your system, just of your CPU and it only doesnt it when the client is first started. Other than that, the memory amount is the only thing checked for I think.

muddocktor

Marc, Stanford got away from using that system benchmark the client does when starting some time ago (like maybe a year ago). The only bench done now is the pf, which you see when a wu finishes up in the fahlog. Attached is a pic of the client on my lappy here at work showing the pf:

Medlock

muddocktor wrote:

Marc, Stanford got away from using that system benchmark the client does when starting some time ago (like maybe a year ago). The only bench done now is the pf, which you see when a wu finishes up in the fahlog. Attached is a pic of the client on my lappy here at work showing the pf:

None of my five clients has that in their logs. It goes from

CoreStatus = 64 (100)
to
Sending work to server

primesuspect

you might need to enable verbosity 9

Tim

muddocktor - I don't see those highlighted lines in my folding stats. I'll look again after the next work unit finishes.

muddocktor

Tim, you might need to use the -verbosity 9 switch, like prime said to TheGr81. If you are running the console as a service, you add it to the startup config in regedit just like adding the -advmethods switch in the registry. If you are starting the client with a shortcut, you just add that to the target line in the shortcut. Just be sure to leave a space between each individual switch added in either place.

mmonnin

The clients do still do a benchmark.

muddocktor

Yeah but Stanford doesn't use it to control anything now. I think it's because if you notice what your benchmark numbers are when you reboot the machine and have the client start as a service or automatically start when booting, the benchmark numbers at boot are much lower than just starting the client while the machine is already running. I've noticed a couple of thousand point difference between the 2 benchmarks.

Genesis

Just been looking at your site and webmail, wont have the courage to strip down a car like you did. I am very impressed. Just thought i will let you know.

Genesis

Medlock

Thanks, prime and Mudd. I'll see if it lists the pr later when my computer finishes another WU.

Straight_Man

edcentric wrote:

I thought that it was the local software that checked for that.
I don't think that the client or server know if you can run and extended instructions. If they did I shouldn't get the same units on both my TBird and Barton powered boxes.

Yes, you should get same WUs. Servers nor clients check for the advanced stuff in the Barton, so to Folding a Barton "looks" like a T-Bird. What you get is the better GHz, not better WU to client customizing.

What they have started doing, and this is exchanged also, is to take and start implementing efficiency of folding as well as --advmethods use by client to get more advanced WUs. These can be almost any type. After a bout a month or so of client working, you will get more effective WUs than non-effective WUs for the hardware, when this is fully implemented.

I've been urging a more full classfying of machine types, voluntary to begin with, in the folding community forums, and if you get REAL ineffective WUs in a long row of WUs that do poorly on one machine, would suggest mentioning that in the Folding Community forums.