Folding Forum

I have a bunch of questions. I'm comming over with this dual 7950 GPU and cpus unit from Boinc and some of the complexities of different multi-configurations that are possible that you have worked to run concurrently I still have questions about.

Ok I have a engineering sample 8/16 core Intel processor. Running dual HD 7950's in Boinc PrimeGrid it was not stable. So I might find it necessary to turn off my Zero GPU and run on GPU 1, unless you have better knowledge on the subject. (On Boinc, WCG, Help Conquer Cancer, the runs were shorter and it behaved well).
So far I have been running F@H for 2 hours 45 minutes with no errors that I can see in the log.

Next questions, at end of story. On WCG, Help Conquer Cancer, it was multi threads in the GPU and high processing usage with multi-cpu threads, (when that project was done). ... While doing DC I would sometimes split projects, ie GPU to one DC project and reserve 2 cores for the GPU's then select 14 threads for cpu work of another DC project. ... Do the GPU's reserve 2 threads in F@H? Are there high cpu thread processing usage for the GPU's or is it low usage for AMD HD 7950's? ... Also what about Nvidia 660 Ti's?, Boinc GPUGRID is Nvidia is low processor usage but on my computer I wish to not turn on the extra cpu threads because it creates to much heat. How do I run just 2 Nvidia GPU's and not the other 6 cpu threads? threads if I put this rig on F@H?

Also I didn't see where to say "accept no new work units" to clear wu and power down for whatever reason for extended periods.

Also I read something a long time ago about some projects worth more points depending on hardware or commitment or something. In Boinc you could select a project, and then sometimes a specific sub-project could be isolated and committed to. How do I investigate further to see if it suits me?


Bruce G.

Welcome to foldingforum.org, StitchExperimen. ( and welcome, Bruce)

Windows, Linux, or MacOS X? V7.3.6 or an older client? Using the FAHControl GUI or WebControl?

FAH supports both AMD and NVidia GPUs as well as multi-threaded CPUs. Like BOINC, there are multiple projects which use a variety of analysis codes (called FahCore_xx.exe) so you won't get a single set of answers. Some GPU codes requires a full CPU support thread but that also depends on the drivers. 14 threads devoted to CPU processing and 2 left for GPU support is a good place to start.

Unlike most of BOINC, minimizing the time between download and upload is an important component of the points awarded -- not just "Was it completed?"

Assuming you're running FAHControl of V7.3.6, you can right-click on the Folding Slot in the middle of the screen and choose "Finish" to complete the current WU(s) and upload the results without downloading a new assignment ("accept no new work units" for that slot.)

You cannot select a specific project, although at times there's only one or two projects running in a category that you CAN select. Currently, projects for FahCore_17 are limited to client-type=advanced but if none happen to be available, you'll get another assignment. Selecting a specific disease might work once it completes the roll-out. (The client is ready; the servers are not yet.)

Windows V7.3.6 Advanced Control
Win 7 Pro 64-bit
Intel Socket 2011 engineering sample 8/16 core
16 GB memory
Two GPU Gigabyte HD 7950

I had a graphics driver crash and have a question. From another article it had a different problem of an overclock and after not finishing 4 times didn't finish the wu and sent another one. On a graphics driver crash what happens? I presume it resets to the previous 15 minute mark and starts from there but after 4 times of graphics driver crashes does it cancel the wu and send out another wu and "all is lost" or what?

I tried terminating (deleting) GPU zero in F@H "Advanced Control" but the program starts up on GPU zero anyway and runs but not on GPU 1 which is supposed to be still active.

Got more prior questions but will ask later.

Bruce G.

Configure, Slots, select GPU, Edit, change GPU index from -1 to 1, assuming 1 is the second card. Try it. If it still folds on the first card, change the setting from 1 to 0 and try again.

Too much overclocking can lead to data corruption. What was a safe OC for Boinc may not work for FAH. FAH is known to work the hardware much harder than most any DC project so we get the most science done out of your donations.

I was not overclocking at the time. My problem occures becauses of the motherboard and the Intel cpu engineering sample running 2 GPUs. I don't have $500+ to spend on a dual cpu motherboard that will run the engineering sample correctly and then purchase a second processor and power supplies and 2 more GPU 7950s and cooling and noise from having to use 5,000 rpm fans mid case.
I have another machine but heat is WAY to much for 3 video cards of ~200 watt.

The other Bruce

So is your CPU crashing or your GPU? (You didn't post the log.) The eng sample for the CPU might cause one kind of crash and the GPU overclock or drivers might cause the other type.

Of course cooling alone might be the problem. Commercially available computers supposedly have their cooling engineered and tested. Home-built may not. Some folks, even those who report "reasonable" temperatures, have taken the side off their case and directed a room fan into it and the crashes went away. This is often the case with a dual GPU or with a pair of GPUs in adjacent slots. (Unfortunately the temperature probe may not find the hottest place when two GPUs heat each other.)

The log is erased from shutdown and restart. Graphics driver crashing. But the engineering sample is known to have issues on certain motherboards. Supposedly a non-engineering sample will run on this motherboard. As for it being the GPUs there was two 120 mm 9 blade fans placed lengthwise in a push configuration over the GPUs and ran at 2,750 rpm. Temps were in the low 70c range with cross ventilation, vents and exhaust. The processor 150 watt is external water cooled.
I turned off the first computing GPU so the graphics driver crash would stop.

Next question different computer.

I have a o/c 3770K it gets to hot to run all the cores/threads. It has two Nvidia 660 Ti's that use 2 threads. How do I set it up to NOT run cpu work units?

The other Bruce mohaha!

StitchExperimen wrote:I have a o/c 3770K it gets to hot to run all the cores/threads. It has two Nvidia 660 Ti's that use 2 threads. How do I set it up to NOT run cpu work units?

0) Open Advanced Control (aka FAHControl)
1) Click Configure.
2) Go to the Slots tab.
3) Select/highlight the CPU slot.
4) Click "Remove".
5) Hit Save.

How hot is "too hot" for you anyway?

It's living with the byproduct of waste heat.
Hot is when it's 100+ outside and the cooling won't keep up with the room and your running 14 inch fans to keep yourself cool.
The problem is living comfortable with all the equipment and it's convinces ... a 40" HDTV monitor still puts out 100 watts which is okay in the other 3 seasons.

Also this computer is under someones desk but most of the time their gone and wear shorts a lot.

Bruce

Couple more questions.

I found a bunch of programs for monitoring F@H, but am clueless what I need, any help?

Is this new program really going to run for 38 days (Nvidia) (660 Ti) or is the card going to shave a bunch of time off? Or a miscalculation for this first start up?
Also 38 days and ~27,000 points (total???) or is that per day???

Bruce G.

It takes 3 completed frames to get a solid points estimate. Startup estimate will be slightly off. It's labeled PPD if it means that.

What about the programs about monitoring of F@H at this page? of multiple machines.

foldingforum.org/viewtopic.php?f=14&t=52#p189

HFM seems to be popular.

The best program to monitor all your folding systems is HFM.net: http://code.google.com/p/hfm-net/

Instaledl on one computer it is able to monitor different Client-Versions (V6 and V7) on different OS (Win XP/7/8, OS-X, Linux) in your network.