Folding Forum

The problem with 1060 is that it also uses twice to trice the amount of power. Not only initial cost, but I do pay my electric bill, and folding isn't like Bitcoin where I can recoup the money over time.

I will, in the future, buy a new card when a newer generation comes out that is affordable.

What I don't understand is that many sites say that those PCIE risers use a USB interface, but it won't work when plugging in an external graphics card via a laptop's USB port.
It would be interesting, because I do have a laptop that folds, and can take it to my job to fold there. If I could add an external gpu via USB it would currently triple my PPDs @ work, as I can only CPU fold on the administrator locked work PCs.
I wish there was more research done on folding via USB3 ports.

I mean 1x gtx1060 will take 100 watts and 2x gtx1050 will take 2x50watts=100 watt, so there is no difference in power usage. And you can use the gtx1060 for gaming with double performance.

There is even some coins you can get with folding to pay electric bill see
https://www.reddit.com/r/FoldingCoin/
https://www.reddit.com/r/curecoin/
but no support in this forum.

The USB on pcie risers is not a real USB, you cannot plug it in any PC but it only used as cheap reliable wire to connect pcie x1 and x16 board.
There are indeed some external GPUs for laptops but only if laptop supports these like from alienware or razer.
This guy managed to build an external GPU for a laptop using laptop mPCIe slot. https://www.youtube.com/watch?v=bP_8EYQ-2RA

ProDigit wrote:It'll be a dual boot system.
Windows 10 home 64 bit, is for my own projects and games,
And Linux, I'll have to decide which one.

I have a dual-boot system and there's one aspect of a system like that which you have not considered. You're asking a lot about maximizing FAH performance but the factor you have not considered is the loss in throughput wherever you boot the "other" system. You can expect that the active OS will be folding one WU per slot and you'll be shutting that system down. When you boot the other OS, it will abandon those WUs and will download a new WU for each slot which will start from the beginning. The same will happen when you switch back.

Of course you can set the active WUs to "Finish" so nothing will be lost, but all slots will not finish at the tame time, so they'll be a perdiod of time when some slots are working and others are idle, wasting resources in another way. Bottom line: you'll lose a lot of productivity from (A) WUs you abandon or (B) Slots you have to suspend on the departing OS. I rarely switch OSs because it's a hassle if you try to remain productive.

I won't be running FAH on both operating systems; other than in the testing phase.
When I'm working on my server, I need a lot of CPU crunching power, and will more than likely close FAH anyway.
If there's a large speed difference between Linux and Windows, favoring Linux, then I'll be running FAH on Linux nearly indefinitely; pausing when I switch over to WIndows.
In between tasks on WIndows, I can easily use the NaCl to process some jobs in between; but as soon as my work is done, I'll boot back in Linux, and continue work loads.
Realistically I won't expect to switch operating systems once or twice, for more than a few hours a week.
In some cases, I might even stay in Linux, if the work required can be done via Linux (eg: file copying, BD burning, 7z compressing).

While I try maximizing the workload, I don't mind if I don't reach my previously month PPD record. The server gets about as many points in a day, as all my PCs together in a week.
And even more soon, as my second graphics card is on the way.

When you boot the non-FAH OS, the WUs assigned to the FAH preferred OS will be sitting idle on the non-booted OS. This causes the bonus points to decay as the time you hold the WU between download and the uploading of the results is extended.

The NaCl client has spoiled you by issuing only very short WUs where the loss of a half-processed WU isn't much of a problem.

Linux is at least 10% more PPD on gpu's, not sure what it currently is on CPU but probably 20-25%

bruce wrote:Many years ago, FAH initiated a development effort to see if it could run under BOINC. That project was not completed because FAH and BOINC have entirely different concepts of priority.

* In FAH, rapid completion of WUs is essential to the assignment process, to the scientific value of the work, and consequentially to the points awarded. (Unnecessary running of WUs twice is a bad thing.]

https://foldingathome.org/faqs/high-per ... -on-boinc/

In January 2006 we launched an initial release BOINC client which we alpha tested in a small group, but we ran into some significant issues with the client. In April we updated much of the code, but we had to deal with a staff turnover in the BOINC part of the development team, which slowed development. As of June 2006 we are putting this platform on hold, as until such time as our staffing situation changes, and the incompatibilities on both sides are resolved, further development has been shelved.

BOINC is Better.

Mod edit: changed Code tags to Quote on quote from FAQ, moved URL to separate it from quote of bruce's post - j

Nathan_P wrote:Linux is at least 10% more PPD on gpu's, not sure what it currently is on CPU but probably 20-25%

There's a way you can even out the losses in Windows, I found.
You can add to up to 10% of PPD count, by simply disabling services and programs in Windows, that you don't need, and setting thread priority.

All can be done in Taskmanager.
Windows has too many background activity, like automatic updates, aero, 3D effects, widgets, and the craps, that can be disabled; services you don't really need in a machine dedicated to folding (like printer spoolers if you don't use a printer, audio drivers, .. I mean, there are plenty of guides on what services you can disable and which not).

But if you're not into tinkering too much, you can very easily do:
Set the thread priority to high, on both the CPU and GPU application in taskmanager,
And closing (terminate) 'explorer'.
The latter will disable your desktop.

You can re-enable your desktop by pressing CTRL + SHIFT + ESC at the same time,
Taskmanager pops up, and start a new task (File>>Run a new Task), "explorer" (without quotation marks), and your desktop will boot again.

You should be able to gain about 10% more CPU PPD just by setting the tasks to high.
Disabling the desktop should get you a few percents.
My taskmanager says my GPU is encoding at 95% all the time.
I guess the last 5% is for running a desktop.

perhaps it helps running a clean desktop, with no icons, and a simple black background (no images, and especially no Aero or 3D effects; you can disable themes, or google to optimize your computer for performance).

The gain for the latter is a few percent at most, but it's free. Doesn't cost you any extra electricity.

Additionally you can just create a user profile in Windows that has all these settings optimized for FAH; and switch users when you prefer to run a nice looking desktop with all your apps, widgets, and aero effects enabled.

ProDigit wrote:But if you're not into tinkering too much, you can very easily do:
Set the thread priority to high, on both the CPU and GPU application in task manager...

I believe you'll find that your Priority change will not persist across WUs or reboots. In BOINC one can make that change permanent in the cc_config.xml file.

I have not found any increase in performance by increasing the Priority.

I do. Mostly on CPU, and lower powered computers.

Setting priority doesn't really get any more work done, it just changes which of the pending tasks get the use of the processor next. This may be what you want to happen but it may not be. e.g.- if a high-priority task keeps all of your threads busy, you may find that you can't move your mouse effectively because the tiny bit of processing that it takes to actually move the pointer has to wait for one of the high-priority CPU-bound tasks to pause briefly. Leaving a thread or 2 out of the high priority CPU-hogging list may leave enough resources for your mouse to work normally.

I would have wished that by default the client under full load, would set thread priority to above average. Setting thread priority just reduces latencies, and on thousands of calculations they do add up.
But the more threads your computer can handle, the less the effects of this change.
My Xeon rig, changes are there, but not really measurable.
I still haven't measured GPU performance changes, mostly CPU over NaCl that benefit from this.
In fact, minimizing the browser increases performance as well.

Saw this post in Puget Systems HPC blog and thought Dr. Kinghorn’s method for reducing the number of PCIe lanes available to a GPU with Post-It Notes was a neat solution for testing throughput and with a longer Post-It Note could be extended to x4 or x1

I tried that Post-It Note thing but system did not recognize my GPU anymore when taped as x4 or below, maybe I did it wrong.

gordonbb wrote:Saw this post in Puget Systems HPC blog and thought Dr. Kinghorn’s method for reducing the number of PCIe lanes available to a GPU with Post-It Notes was a neat solution for testing throughput and with a longer Post-It Note could be extended to x4 or x1

For 1x folding there are risers available.
Some Motherboards allow you to set pcie speeds, but most would automatically adjust the speed depending on how many ports are occupied.