Questions regarding comparision to BOINC

[sean0118]

Hi everyone,

I have run BOINC for a few years now I'm branching out into Folding@home as well. I just have a few questions for you...

1. Is there somewhere to view more detailed statistics like in BOINC? Statistics I'm interested in finding are the results of individual work units (valid, returned after deadline or error) and their runtime.

2. Is it possible to suspend the download of new work units? Sometimes I might want to swap over to BOINC for awhile.

3. I reduced the checkpoint interval from 15min to 5min because I turn the computer off quite a bit. My cpu is a Q6600 and gpu is ATI HD5700 and I have a solid state drive. Is 5 min likely to gain greater overall performance in this case? I don't care too much about the amount of data being written to the SSD.

Thanks!

[bruce]

Welcome to foldingforum.org, sean0118.

1) Not really. At one time FAH did have on-line transactions that reported detailed information. The server capacity to support repeated requests for large amount of detailed information was deemed inconsistent with the scientific goals of FAH so it was restricted. As forum modrators, we do have access to that information and if you're wondering what happened to a specific WU, you can provide us with the Project/Run/Clone/Gen numbers and we'll run the query for you, but we're careful not to overuse that capability.

2) If you're using the WebControl Application or the NaCl client, press Stop Folding and you'll be presented with a choice If you use the FAHControl (aka advanced control) program, the FINISH function provides that capability. If you have more than one slot, the big "FINISH" button will set all slots to finish, but you can right-click on individual slots and manage them separately.

3) That's not an easy question to answer. If the checkpoint is set to 15 minutes and you stop processing at random times, you'll have to repeat between 0 and 15 minutes on restart for an average of 7.5 minutes. If it's set to 5 minutes, you will have to repeat an average of 2.5 minutes. Every time you restart, some additional time is lost reinitializing and re-validating the current data and that time depends on both the WU and the type of analysis although with a SSD, it will not be as long as with a classic disk. On the other side of the equation, some time passes preparing for extra checkpoints and actually writing the data.

I leave mine set to 15 minutes, but you'll have to make your own decision.

By the way, some FahCores use the checkpoint interval settings while others ignore it, preferring to do what somebody else decided for you. Like BOINC, FAH consists of a variety of projects with some variations between projects.

4) You didn't mention it, so you may already have figured it out, but there is a big difference in the importance of deadlines in FAH and in many of the classic BOINC projects such as Seti where deadlines are more of a guideline that an absolute deadline. FAH has two deadlines commonly called a Preferred Deadline and a Final Deadline. If you miss the Preferred deadline, you' get no bonus points. If you miss the Final Deadline, you get zero points. The fact that you asked about the Finish function in question 2 means you've though seriously about it.

[sean0118]

Welcome to foldingforum.org, sean0118.

1) Not really. At one time FAH did have on-line transactions that reported detailed information. The server capacity to support repeated requests for large amount of detailed information was deemed inconsistent with the scientific goals of FAH so it was restricted. As forum modrators, we do have access to that information and if you're wondering what happened to a specific WU, you can provide us with the Project/Run/Clone/Gen numbers and we'll run the query for you, but we're careful not to overuse that capability.

Thanks for your reply that makes sense, so I guess the best way to check that everything is fine is by looking at the points output and log?

2) If you're using the WebControl Application or the NaCl client, press Stop Folding and you'll be presented with a choice If you use the FAHControl (aka advanced control) program, the FINISH function provides that capability. If you have more than one slot, the big "FINISH" button will set all slots to finish, but you can right-click on individual slots and manage them separately.

Thanks that makes a lot of sense as well.

3) That's not an easy question to answer. If the checkpoint is set to 15 minutes and you stop processing at random times, you'll have to repeat between 0 and 15 minutes on restart for an average of 7.5 minutes. If it's set to 5 minutes, you will have to repeat an average of 2.5 minutes. Every time you restart, some additional time is lost reinitializing and re-validating the current data and that time depends on both the WU and the type of analysis although with a SSD, it will not be as long as with a classic disk. On the other side of the equation, some time passes preparing for extra checkpoints and actually writing the data.

I leave mine set to 15 minutes, but you'll have to make your own decision.

By the way, some FahCores use the checkpoint interval settings while others ignore it, preferring to do what somebody else decided for you. Like BOINC, FAH consists of a variety of projects with some variations between projects.

I guess I'll test a few different settings then and find one that works best.

4) You didn't mention it, so you may already have figured it out, but there is a big difference in the importance of deadlines in FAH and in many of the classic BOINC projects such as Seti where deadlines are more of a guideline that an absolute deadline. FAH has two deadlines commonly called a Preferred Deadline and a Final Deadline. If you miss the Preferred deadline, you' get no bonus points. If you miss the Final Deadline, you get zero points. The fact that you asked about the Finish function in question 2 means you've though seriously about it.

Ah right, yeah I didn't know there were two deadlines, good to know. In the advanced view is 'Timeout' the preferred deadline and 'Expiration' the final deadline?

One more question I thought of is whter the client automatically keeps enough cpu free for the gpu slot? Or will I need to free a core or two manually?

Also, I'm guessing there might not be an easy answer, but is there a recommenced driver for the HD5700 series?

[Zagen30]

Yes, Timeout=Preferred Deadline and Expiration=Final Deadline.

The current client automatically uses n-1 threads if a supported GPU is detected, be it Nvidia or AMD. This is enough to keep one GPU fed. If you ever wanted to manually adjust that, you'd open Advanced Control and select Configure->Slots->double-click the CPU slot and adjust the CPUs counter. Then make sure to select OK or Save on all the windows that had opened in the interim.

I haven't seen much discussion of 5000 series cards in a while given how old they are by hardware standards. I guess go with the latest drivers. Most AMD cards saw a noticeable folding boost starting with 14.1 vs. the 13.x drivers, but people were usually posting results for the 7000 series or R7/R9 series cards.

[sean0118]

Awesome, thanks guys, I'm looking forward to giving Folding@home a go.

[Joe_H]

To add to the information on the 5000 series cards, I have run folding on a 5870 in the past year. Using older drivers it was fast enough to complete WU's using Core_17 within the preferred deadline. Discussion at the time from others with 5000 series cards was that cards below the 5700 series usually could not complete within the deadline, and the exact point in the 5700 series depended a bit on the project. Since the Mac Pro the 5870 is in usually runs OS X, I have not gone back and retested with current version 14 drivers to see if they improved processing speed. So you would have to see if your card is still usable for GPU projects.

[folding_hoomer]

If you want to fold with your HD 5700 series card use Catalyst 14.4 WHQL - it will provide you the highest PPD for Core_17-WU´s.
I´m using it under Win7 folding with a HD 5870@930MHz getting about 18K PPD - and only the former Beta-releases of the Catalyst (14.1 and 14.2) provide the same PPD - but: Beta . . .

[sean0118]

Cool thanks guys, Catalyst 14.4 WHQL is what I have now.

My estimated PPD for the GPU is only 3000, I think I just need to run it more, I get too much lag to use it much. I think I'll mod my case to make it quiet, then I can run the GPU at night as a heater.

[sean0118]

hmm is the following line normal in the log? I don't have Nvidia..

"06:47:28:WU01:FS01:0x17:Temperature control disabled. Requirements: single Nvidia GPU, tmax must be < 110 and twait >= 900"

[Joe_H]

Yes, it is normal.