Folding Forum

You don't have to test with live work units. That's silly.

Make a copy of the entire fah folder. Then test with the copy. You can reset the WU back to frame zero any time you want. Also configure the test copy to prompt before sending, no need to waste your bandwidth quota on a WU that won't be accepted anyway. And also set the client to ignore deadlines so the WU never expires.

Eno wrote:I agree overall- but if you buy a race horse you don't exactly put it on a treadmill either... so pushing the chip to its very limits was / is a challenge. What is daunting is the prospect of actually being penalized because the WUs are returned unfinished... It's tough to test the OC when they only thing that makes it fail is actual WUs.

I understand the "challenge" part of it, and as long as you don't waste downloads (see previous post), you're welcome to tweak to your hearts content.

As far as "daunting" is concerned, I'm sure you can see it from the Pande Group's point of view as well as your own. (A) Work which is completed quickly is better than work that is completed more slowly. (B) Incomplete work has little or perhaps no value -- All work must be completed by someone if it's possible, at all. (C) Any work which is downloaded and never returned causes bigger delays than the slowest of machines that do complete work.

The points system does a pretty good job of reflecting both of those priorities.

I'm not sure I understand this.

Do I CNTL-C while a job is running, then copy the directories (home+work) to another location and launch from there?
Or do I do this between jobs? (-oneunit)

I have crashed a few SMP's (less than 1%) when experimenting, but haven't crashed any bigadv's yet. (knock on wood).

I'd much rather do this without risking a crash.

From my bigadv Guide:
If you want to experiment with your Overclock, you can "capture a WU" and let your system fold it. Do remember that you will not get any points for it. You can capture a WU by copying the entire SMP2 directory while the WU is being folded. You will have to reconfigure the Client and set this option (in advance configuration):

Ignore any deadline information (mainly useful if
system clock frequently has errors) (no/yes) [no]?

Set it to "Yes". You can then hold this WU indefinitely and use it to test your system. In the Work folder, you can keep only wudata_0X.dat and wuinfo_0X.dat files and delete the rest so it will fold from the start. (This may vary from version to version and I tested it with v6.30). I also choose to block the Client from internet connection to ensure that it won't connect to the Servers.

To answer your question, you can copy the files while it is folding. No need to stop the Client. YMMV but I haven't encountered any issues with it so far.

So the server "knows" that job has already been assigned, and will ignore duplicate data from another instance?

If the Server assigns a WU to you and you return the wuresult successfully, any further uploading of the same wuresult from the same system will be ignored.

So the best option is to copy the install directory, including the WORK directory, not long before a WU is completed. Then you can complete it AND experiment on the copy. (If you experiment BEFORE you complete the WU, you stand a chance of getting partial credit for a failed WU that could have been completed for a bonus.)

If you keep both a working copy of FAH-SMP and a test copy, you can disable the possibility of a connection by several methods, but probably the easiest is to intentionally configure a proxy that doesn't exist, while leaving the good installation fully functional. As long as you're sure you'll never download a fresh WU (or waste time uploading a duplicate), that may not be necessary, but I still do it.

Not all cpus are created equal either, my first 980x took more volts for the same mhz than the one I have now. The first one would only do 4Ghz with 1.325v, the one I have now will do 4.2Ghz with 1.325. I notice when the heat is to much it will error out on folding. Of course the first one was a lot better clocker under LN2 but thats a different topic...