Questions about the deletion of WU's

[vulcansuperior]

I've been recently been getting Core17's on my laptop. I'm fine with the new core and all, allowing bonus points on the GPU, but as it is Core 17's have such a tight window I rarely can get them done, and so far have wasted (estimated) a good 3 weeks if not more folding WU's that never got finished or sent. I don't leave my laptop on 24/7 except in certain cases, usually when I just forget and the idle folding kicks in or I intentionally want to get a WU done faster for a friend. Just yesterday I got a project 13001, and it is MASSIVE.

So really my question is: If I know for a fact that I'm going to fail the WU, does this classify as a situation where it would be reasonable to delete the WU? I don't want to waste both my and the server's energy waiting on a WU that will likely never be sent in, when it would be more efficient (and scientifically reasonable) to have it assigned to someone with a stronger desktop GPU.

Current specs:
Intel 3630QM 2.4Ghz nominal(3.2Ghz boost) and a Nvidia Geforce GT650M 2GB GPU.
8Gb DDR3-1600
1Tb 5400RPM HDD
Wireless microwave internet link at 5Mbps down/up (idk, it's weird)

Just so you know, my temps are running borderline as it is and I imagine that some minor throttling might also be taking place. I doubt this is actually contributing to the completion problem as I usually miss the deadline by a good 10%(1day) of processing anyway. I have already taken steps to help with temperatures (cleaning vents practically daily, locating in a cooler room, etc).

[bruce]

I see three different problems, not one.
First, the GPUs that are found on laptops are often too slow for the Core_17 assignments. I've been trying to convince the Pande Group that there should be a class of WUs which can be done by slower hardware, even if the rate at which points are earned is lower. So far, I have not had much luck. There is no setting that can help with this issue.

Second, if the computer sleeps/hibernates, there's a good chance that the WU will be aborted due to the way Windows manages GPUs during the awaken phase of sleep. Some progress has been made on this problem but until a new version of FahCore_17 is distributed to everyone, it is still a problem. As a work-around, the V7 client disables the normal sleep processing, too.

Third, although projects like 13001 are massive, their deadlines are proportionally longer, so I don't think that's really your issue.

If you are able to meet the Final Deadline (see the psummary page), the results will still be scientifically valulable so do not dump the WU.

If you're not able, then the WU is going to be dumped anyway. Perhaps you should permanently remove or disable the GPU slot so it doesn't download WUs which cannot be completed.

[vulcansuperior]

I understand that the GPU in my laptop is rather poor when it comes to folding, and I rarely actually put my laptop in sleep mode. My real question is: Is it ethically responsible to delete WU's that I will no doubt fail 99% of the time? I'm not really willing to disable the GPU slot really as it accounts for a good half of my points on average, as long as I can get core 15's.

[P5-133XL]

Try running v6 of the client for it will never get any Core_17's. The clients are mixable so You can, if you want, run v6 for the GPU and V7 or NaCl for the CPU.

I would argue that that is the correct thing to do because then you neigher need delete or fail WU's

[PantherX]

Welcome to the F@H Forum vulcansuperior,

Since you can still fold FahCore_15 WUs on your laptop but not FahCore_17 WUs, what I would suggest is that you remove the GPU Slot in V7 and download v6 GPU Client (http://folding.stanford.edu/home/download2011#ntoc4). That way, your GPU will only get FahCore_15 WUs since FahCore_17 WUs are restricted to V7.

Assuming that you are folding on the V7 CPU Slot, you can also run the v6 GPU Client without any issues. In future, if small FahCore_17 Projects are released, you can add a GPU Slot back in V7. You can keep an eye out for such kind of announcements in this Forum (viewforum.php?f=24) or the Blog (http://folding.stanford.edu/home/blog).

EDIT -> Seems that others posted while I was composing the post. Hopefully, you find the links helpful in my post

[vulcansuperior]

Thanks to everyone with these helpful posts.

So AFAIK, my plan of attack should be delete the V7 GPU slot, keep the V7 CPU slot going and downgrade to a V6 GPU slot to get core15's? That's doable.

[PantherX]

That is correct. In case you need help with the v6 GPU GUI Installation, you can refer to the pictorial guide (http://folding.stanford.edu/home/guide/wingpuguide). You can use the same username, passkey and team number that you are using in V7.

[vulcansuperior]

I'm trying, but the V6 client is downloading a core11 and that core isn't running on my computer. Just throws 'failed to start core' messages in the log.Apparently the core 11's won't run on my Windows 8.1 64-bit system.

Always just says this:

Code: Select all

[00:46:17] Verifying core Core_11.fah...
[00:46:17] Signature is VALID
[00:46:17] 
[00:46:17] Trying to unzip core FahCore_11.exe
[00:46:17] - Couldn't open file FahCore_11.exe 
[00:46:17] + Error: Could not extract core
[00:46:17] + Core download error (#2), waiting before retry...

[Joe_H]

As I recall, you need to use the -forcegpu nvidia_fermi parameter to use the V6 client with newer nVidia GPU's. That is mentioned in the Installation Guide that was linked to in PantherX's post.

[PantherX]

Do note that the issue you are facing is that it v6 is incorrectly detecting your GPU as a Pre-Fermi one so it has downloaded FahCore_11 WU which can't run on your GPU. Thus, if you used the flag as Joe_H stated and do the following:
Step 1: Stop the F@h Client
Step 2: Delete the Work folder
Step 3: Delete the queue.dat file
Step 4: Change the Machine ID to another unique value
Step 5: Start the F@h Client

It should be assigned a FahCore_15 WU and will fold the WU without any issues.

[vulcansuperior]

Well sorry about that. I tend to skimread a lot and that managed to slip by me. Thanks for bringing it to my attention. Now working good