Folding Forum

My rig runs about 1400 watts per the Kill-a-Watt meter. I am not having a problem getting work for my eight GPUs but I am having a problem with repeated incomplete and failed uploads of the twenty-five or so completed work units in the queue. I am on measured service and the system is wasting about 8GB of bandwidth daily on FAH to upload only a few hundred MB of results per day. I cannot continue much longer before I surpass my 250 GB limit and have to pay extra.

I'm cheering for you, servers! I know everyone is doing their best.

ajm wrote:Is it just me or are the servers becoming smoother, at last? I'm reaching some 50% of my Total Estimated Points Per Day! And most of the time all my ressources are now used.
The servers stats are now constantly over 100,000 jobs per hour: https://apps.foldingathome.org/serverstats - kudos to LTT (I suppose) for their linus1 server!
Feels great!

Not for me. Before March 30 I pretty consistently had a GPU WU processing but very seldom had a CPU WU although I had allocated -1 of my 12 threads.

However since the morning of 30 March, both CPU and GPU have pretty consistently been idle, with upwards of 30 attempts each to get a WU, then either the CPU or GPU will grab a WU process it, and the 30 attempt wait starts all over again.
Also my CPU WU typically process in 2 hours or less for 2500-5000 points whereas the GPU process for about 8-10 hours for 40,000-60,000 points (and yes I have Passkey)

BillAnderson wrote:

ajm wrote:Is it just me or are the servers becoming smoother, at last? I'm reaching some 50% of my Total Estimated Points Per Day! And most of the time all my ressources are now used.
The servers stats are now constantly over 100,000 jobs per hour: https://apps.foldingathome.org/serverstats - kudos to LTT (I suppose) for their linus1 server!
Feels great!

Not for me. Before March 30 I pretty consistently had a GPU WU processing but very seldom had a CPU WU although I had allocated -1 of my 12 threads.

However since the morning of 30 March, both CPU and GPU have pretty consistently been idle, with upwards of 30 attempts each to get a WU, then either the CPU or GPU will grab a WU process it, and the 30 attempt wait starts all over again.
Also my CPU WU typically process in 2 hours or less for 2500-5000 points whereas the GPU process for about 8-10 hours for 40,000-60,000 points (and yes I have Passkey)

Must be different levels of better for everyone.
What GPU are you using, my 1080ti cards can get up to 150k or more per work unit. And they are usually closer to 200k.
At least for me, everytime I check my systems (twice a day at least) they have work. Not that the stats are updating very often today.

My client was completely idle all of yesterday for the first time since starting this. Something is definitely up..

Well, my system has been at close to full capacity for three days straight (24/7).

ipkh wrote:Must be different levels of better for everyone.
What GPU are you using, my 1080ti cards can get up to 150k or more per work unit. And they are usually closer to 200k.
At least for me, everytime I check my systems (twice a day at least) they have work. Not that the stats are updating very often today.

My i7-6700K and GTX 1080 have had work available for the last two days, 24/7. Looking at the logs there is only one failed WU retrieval, that WU got downloaded during the next retry after a minute and didn't result in any idle time. Must admit that I'm running the client with the advanced option, perhaps that's one contributing factor.

Depending on the WU, up to 145k on the GTX 1080, just in line with what you are seeing on your 1080 Ti

try "client-type", "beta" or "advanced". I've had all my slots on each machine full 24/7 now for 3/4 days

scerbera wrote:try "client-type", "beta" or "advanced". I've had all my slots on each machine full 24/7 now for 3/4 days

that seems to have worked. Of course if everyone makes that switch then those WUs will run out as well.

pachydermus wrote:

scerbera wrote:try "client-type", "beta" or "advanced". I've had all my slots on each machine full 24/7 now for 3/4 days

that seems to have worked. Of course if everyone makes that switch then those WUs will run out as well.

Don't use the "beta"-flag unless you are a beta tester. The advanced flag can be used, if no advanced projects are available the client will try to download a normal project instead.

I have gotten GPU work (Advanced flag) continually since the beginning of the week for my three GTX 1070s, one GTX 1660 Ti and one RTX 2060 (all Ubuntu 18.04.4).
But I have a Ryzen 3950X on order, and will see if Prof. Voelz can give me some CPU work to do.

Are there delays in server-side accounting of the WUs? I have submitted dozens of WUs in the past 24 hr that don't appear to be accounted for in the foldingathome.org stats pages.

Yes. There have been variable delays in posting the credit so the stats database. I could list a number of problems that have occured, but the technical people are working on the problems and I can see definite progress but I can also see times when of the processing are very slow or when certain stats are just plain stuck. Then the process often catches up.
https://apps.foldingathome.org/credit-log

bruce wrote: https://apps.foldingathome.org/credit-log

Correct me if i'm wrong, but the numbers on this page seem way too low to me. On the Server Stats page the WU assign rate is constantly above 100.000 WU/hour since days, so i expect to see those 100.000 getting credited every hour, but instead there are only 70.000 WUs/h credited. Are there some problems in the stats system so 30.000 WU/h are not credited right now (indicating a huge backlog) or am i missing something?

Cheers,

ComputerUser

ComputerUser wrote:

bruce wrote: https://apps.foldingathome.org/credit-log

Correct me if i'm wrong, but the numbers on this page seem way too low to me. On the Server Stats page the WU assign rate is constantly above 100.000 WU/hour since days, so i expect to see those 100.000 getting credited every hour, but instead there are only 70.000 WUs/h credited. Are there some problems in the stats system so 30.000 WU/h are not credited right now (indicating a huge backlog) or am i missing something?

Cheers,

ComputerUser

Well, that would be consistent with the delays that some are seeing. And there are some big spikes on the graph now.

That would be correct only if every distributed WU came back with an hour. Most take at least 2 hours to complete so there will be a delay in the return stats.