TPU's F@H Team

[thebluebumblebee]

On my main machine, I let WCG have all the cores, so WCG and FAH are kinda fighting over one. My 660ti is stuck with a core x17 at 21k.

You MUST give F@H a thread when using a Nvidia GPU or you will see PPD just like you're seeing. Core_17 on a 660Ti should be north of 50K PPD.

You guys can have all the x17s... the x18s run better on my hardware than the x17s lol.

No, your configuration is causing the results you're seeing.

@Norton , I vote for the AM3 system. AMD does not take that much CPU. But, the 2600K and giving F@H one thread would work nicely as well. (but I don't have a high end AMD GPU, basing my recommendation on my experience with my 7770) Past discussions about PCI-e lanes were a long time ago, but I think you'd be okay with x4.

 

[Norton]

@Norton , I vote for the AM3 system. AMD does not take that much CPU. But, the 2600K and giving F@H one thread would work nicely as well. (but I don't have a high end AMD GPU, basing my recommendation on my experience with my 7770) Past discussions about PCI-e lanes were a long time ago, but I think you'd be okay with x4.

Actually I forgot I still had a s1155 i3-2100 in the parts box so I'm thinking about going with that and this Asrock B75 s1155 board rather than picking up a cpu for the GB AM3+ board:

http://www.newegg.com/Product/Product.aspx?Item=N82E16813157365

http://www.asrock.com/mb/Intel/B75M-GL R2.0/

$55 and crossfire support (1xPCIE 3.0 x16, 1xPCIE 2.0 x4)

Gives me the option of going with a mATX case and can decide later about swapping the cards and/or the cpu to the Z77 setup at a later date.

Thoughts?

 

[BUCK NASTY]

Maxwell/Linux assignment server issue has been fixed! Sucks we lost 45 hrs of Maxwell production.
@mstenholm

 

[mstenholm]

Maxwell/Linux assignment server issue has been fixed! Sucks we lost 45 hrs of Maxwell production.
@mstenholm

Sort of being fixed. I got a 9201 and a core 15.

 

[revin]

Can I put the 5870 back in without the bridge, n see if it can add to F@H?
Has anyone actually tried 2 different Gen's of GPU's thru PCIe lanes only on older {z68} boards ?

 

[hat]

You MUST give F@H a thread when using a Nvidia GPU or you will see PPD just like you're seeing. Core_17 on a 660Ti should be north of 50K PPD.

No, your configuration is causing the results you're seeing.

@Norton , I vote for the AM3 system. AMD does not take that much CPU. But, the 2600K and giving F@H one thread would work nicely as well. (but I don't have a high end AMD GPU, basing my recommendation on my experience with my 7770) Past discussions about PCI-e lanes were a long time ago, but I think you'd be okay with x4.

I did free up a core, actually. PPD showing south of 20k now on a core 17.

 

[mx500torid]

@revin I had a HD 7770 folding under a 290x

 

[revin]

I think mu GPU WU is Stuck ????????????
It Still shows same as screenie 99.99% 21 min to complete
It's been well over 12 hrs......................................

 

[stinger608]

I think mu GPU WU is Stuck ????????????
It Still shows same as screenie 99.99% 21 min to complete
It's been well over 12 hrs......................................

I had this issues many times as well.

Shut FAH off and completely exit it. Then fire it back up again. That usually solved the problem.
Many times it would show 99.99% when in fact it was actually folding at a much less percentage completed.

 

[revin]

I had this issues many times as well.

Shut FAH off and completely exit it. Then fire it back up again. That usually solved the problem.
Many times it would show 99.99% when in fact it was actually folding at a much less percentage completed.

Well Crap !! Lost 12 hrs., your correct, it only gained 2%
I'll have to keep n eye on GPU usage lot closer !

 

[Norton]

Well Crap !! Lost 12 hrs., your correct, it only gained 2%
I'll have to keep n eye on GPU usage lot closer !

It happens sometimes- there's a few things you can try as others suggested but once in awhile it gets stuck to the point where you may need to accept the loss, delete and add back the folding slot, and get a fresh job.

 

[mstenholm]

It happens sometimes- there's a few things you can try as others suggested but once in awhile it gets stuck to the point where you may need to accept the loss, delete and add back the folding slot, and get a fresh job.

That is the best way to go about that - delete the GPU in "Config", "Slots", save and add a new GPU same place, save, exit. You lost 12 hours folding on a WU that give you more point for less runtime.

 

[revin]

In the 1 hr I gained from 66% to 76% with 4 hr.tc, is it still worth saving at 4 hrs time and credit of 30k ?
The WU 13001 was at 65k, now 30k in 4 hrs Should I get new WU?

I think I'm understanding we'd get 65k running a new, vs 30k for this in 4 hrs[20+ total?] close analogy ???

 

[mstenholm]

In the 1 hr I gained from 66% to 76% with 4 hr.tc, is it still worth saving at 4 hrs time and credit of 30k ?
The WU 13001 was at 65k, now 30k in 4 hrs Can it and get new WU?

I think I'm understanding we'd get 65k running a new, vs 30k for this in 4 hrs close analogy ???

I was under the impression that you only reached 2 % in the past 12 hours. Just leave it then. Depending of your core clock (default 900 MHz?) yes from 65k to 75k (1150 or there about) but more importantly 100k to 145k PPD (points per day).

Edit: If you have a 99.99 % WU then OC is not the best idea. Most 99.99 is because of OC. During your run you should switch to advanced control and watch the log to see if the log and the status view has the same percentage. As soon as they are different (forget the first 3 %) then look at the GPU load (GPU-Z or similar) to see if your GPU is actually working. 0 % load, no heat and difference in percentage - stop the client completely (kill it), re-start and watch again. HW not up to par is a pain to run.

 

[revin]

@revin I had a HD 7770 folding under a 290x

Wait, I think those are almost the same family.......7850= R280 7770=R260/270
Are my two, too far part in Generation familys ?
EDIT
I guess only thing that would happen is nothing, OR could it short out the PCIe Slot ??????????

 

[mstenholm]

Wait, I think those are almost the same family.......7850= R280 7770=R260/270
Are my two, too far part in Generation familys ?
EDIT
I guess only thing that would happen is nothing, OR could it short out the PCIe Slot ??????????

Some people have been have been folding with a NVidia and an AMD in the same rig. No sweat to run two different AMDs. I have a 7770 and a 7970 in one rig. No problems at all other then they have to share PCI lances which again is a very small decline in performance.

 

[mstenholm]

You MUST give F@H a thread when using a Nvidia GPU or you will see PPD just like you're seeing. Core_17 on a 660Ti should be north of 50K PPD.

In pure happiness over the return of the 9201 last night (my time) and perhaps because of the ONE glass of redwine I forgot to change my WCG setting (100 % CPU in use). I know that I for one have been saying that core 17 and 18 uses 1 thread per GPU (which I can see it does) the result was a plus 3 second on each of the two WUs. WCG is good at adapting but I'm sure that not programs will let the Fahcore_17.exe do its work. I'm a bit surprised that the penalty was that small.

 

[thebluebumblebee]

In pure happiness over the return of the 9201 last night (my time) and perhaps because of the ONE glass of redwine I forgot to change my WCG setting (100 % CPU in use). I know that I for one have been saying that core 17 and 18 uses 1 thread per GPU (which I can see it does) the result was a plus 3 second on each of the two WUs. WCG is good at adapting but I'm sure that not programs will let the Fahcore_17.exe do its work. I'm a bit surprised that the penalty was that small.

AMD or Nvidia GPU's?

 

[mstenholm]

AMD or Nvidia GPU's?

I would not get happy to receive 9201 on my AMDs which I did during the 45 hour of Maxwell dryout. On my 970s.

 

[johnerz]

In the past I have ran a 570 and 670, Nvidia of course in the same machine but my experience is a couple of years old on this

 

[hat]

In pure happiness over the return of the 9201 last night (my time) and perhaps because of the ONE glass of redwine I forgot to change my WCG setting (100 % CPU in use). I know that I for one have been saying that core 17 and 18 uses 1 thread per GPU (which I can see it does) the result was a plus 3 second on each of the two WUs. WCG is good at adapting but I'm sure that not programs will let the Fahcore_17.exe do its work. I'm a bit surprised that the penalty was that small.

So you mean to say running WCG on every core alongside FAH doesn't impact much?

Maybe I should let WCG have the 4th core on my Q6600 and watch PPD and see what happens... but it's only a GT720 so it wouldn't show well probably.

 

[mstenholm]

So you mean to say running WCG on every core alongside FAH doesn't impact much?

Maybe I should let WCG have the 4th core on my Q6600 and watch PPD and see what happens... but it's only a GT720 so it wouldn't show well probably.

Its worth a try. Let us know what you find. I know that a GT720 will not do a lot of WUs/day but watch two to be sure of your findings. For sure the TPF will increase but where the limit is I guess is up to each Folder&Cruncher to decide. Some will lean to optimize folding (I do, my entry to this world) others....

 

[Nordic]

Its kind of complicated really. You need to test it out. My celeron g1620 needs both cores to feed two folding 750ti's. If I have 1 core dedicated to WCG and 2 750ti's I lose 1/3 folding output on one 750ti. If I have one 750ti and have both cores crunching, I don't lose any folding output.

 

[mstenholm]

Its kind of complicated really. You need to test it out. My celeron g1620 needs both cores to feed two folding 750ti's. If I have 1 core dedicated to WCG and 2 750ti's I lose 1/3 folding output on one 750ti. If I have one 750ti and have both cores crunching, I don't lose any folding output.

There is no easy answer since it apparently depends of more then just you PCI configuration and CPU. Each folder for them self but if we get some feedback we might learn something. Some WUs have more or less the same processing time for each frame (9201 is one of them) but others varies a lot so just watching a few frames could be misleading. I tried with 1300Xs and 1047Xs but the variance between frame is greater then the CPU factor. It would take a ton of time to figure it out but a few sample runs (100 %) with no or some CPU support could give a pointer.

 

[hat]

Well, enabling the 4th core did drop my FAH output... by how much, I'm not sure. But the fahcore exe was still hogging 25% cpu anyway, and WCG tasks were split up across 4 cores at <25%. I conclude that by running both you are 'robbing peter to pay paul' as it were. I guess it's up to the individual to decide whether or not they want to take a CPU core away from WCG in order to run FAH on the GPU.