FAH servers; Am I being punished? {No}

[Agencyman]

It is a question of balance, (I always lose a lot of SMP throughout if I run my GPU3 on an ATI 5750).

But is it possible that now I am also hobbled by WU's that are configured for uniprocessors?

Three days ago this Q-8400 was doing a respectable job for its type, running at an FahMon indicated 6300-8800 PPD. Then I set it aside in the middle of a run to finish up several odd WUs that I started trying to do Moderator chores at Kit Guru dot net When I tried to resume it said no server existed that had my old Proj. number, so I deleted all the work files, F@H server noted that, and immediately got a new project started me out and here I am.

Now the best I get on console SMP, is 250 -1100 PPD. That Proj in this image 5732 was for the GPU. I printed out the instruction sheet and did it by the numbers on the SMP install. Is there a troubleshooting step I can try?

[Agencyman]

Update, Computer unexpectedly climbed back up to 4400PPD, (last night 1100 for hours with SMP the only thing running on the computer). I was truly slow by, step-percentage rate.

BTW Is there a better program for monitoring that you would recommend.

[PantherX]

Please note that the issue of yours "no server existed", I haven't come across. In future, please save the FAHlog/FAHlog-Prev and post it in this Forum so we can have a better understanding and can help you.

Regarding the SMP2 Client getting Classic WUs, it could be possible that you forgot the -smp flag and started the F@H Client and it connected to the Servers and downloaded the Classic WU and after that you placed the -smp flag and restarted the Client. Once it finished processing the Classic WU, it automatically downloaded the SMP2 WU. Do note that if CPU Cycles are taken from FahCore_a3, the slow down is in a non-linear fashion hence your PPD will drop as long as that application is running.

Since you are using an ATI/AMD GPU, I suggest that you set up Environment Variables (Details) to reduce the CPU usage.

I am using HFM.NET and haven't encountered any issues Do note that your FahMon is not up-to-date and even with the latest version, I am facing some occasional PPD issues

[P5-133XL]

When you get the no server error message the problem is at Stanfords end and you did nothing wrong. It means the server that issued the WU went down before it had a chance to inform the collection servers that it had issued the WU. All you can do is wait for the server to start running again, so it can inform the collection servers about that WU. The clients will continue contacting the collection servers on a regular basis and when the problem resolves, the WU would have been sent. It matters not now that you have deleted the work folder. Just keep it in mind if it happens again.

I encourage you to experiment with the ATI environment variables to optimize PPD and they definitely do help. But, for some with lower-end ATI GPU's but with SMP capable CPU's it may be better (PPD-wise) just not to GPU fold. The SMP client is very sensitive to other applications using the processor and for the ATI video cards the GPU client uses a significant amount of the CPU. When that happens the SMP client's PPD drops when running the GPU client. Depending on how much the SMP decrease is, and how much the GPU produces you may do better without GPU folding. Important: This is not normally a factor with Nvidia cards.

Most likely, if the smp client is running uni-processor WU's it means that you didn't add the -smp flag/parameter to the client. That is its default behavior without it.

Your PPD will vary over time. You have little control over what WU's you get assigned and different WU's give out different PPD. I know that Stanford tries to standardize PPD for all its WU's but the reality is that PPD varies and for some it varies a lot. So don't worry too much about your PPD jumping around as long as all your clients are working properly.

[Agencyman]

Thanks Panther,

You were, as before, very helpful. I'm glad you saw the humor, and didn't think of it a whiney rant. It did seem a possibility to me that a server might have handed down a lesser WU, if it "thought" I was having problems, as I'm told it will even withhold if it thinks your machine is unstable!

I'll go and read that material now.

Case closed!

Again, appreciate your taking the time to reply.
Bruce Hinton
KitguruFolders

[Agencyman]

P5-133XL,

Now I feel bad about the work I got rid of. I wrongly assumed it had gone beyond deadline, when I got that message. It looked like it was going back to work, but never went anywhere, if I can find that message source, I'll let you know. It may have been an [Fcn]4 out of

That's why I come here, to learn!

I'm going to look at those pages, but I think your advice just to stay with SMP, is good; the flag really has been there from the beginning, really

If my SMP server goes down I'll run my next best for a while. Before I installed this quad, I already had been switching GPU2 for 3 when work was hard to get.

When some of the "even newer than me" noobs had questions, I eventually got all the clients that I can run. I have an Ubuntu drive, I guess I'll have to tackle that next, ohboy.

Thanks
Bruce

[Agencyman]

I hit F4, not [Fcn]F4, right out of FahMon, I think, and it took me to fahinfo.org. I just tried it for my current WU 6070, happily folding, and it said the same darn thing! I adds "No scores appear to have been submitted for project 6070.

I found that whole thing by accident.

I will drill a bit deeper for info next time,.

Bruce

[codysluder]

The screenshot that you show above doesn't tell which WU is being processed. It does say the WU is worth 992 points but at the moment I don't see any 992 point smp projects on psummary so I'm going to make an unsupported guess.

FahMon says the ETA for the WU for 2d8h57m and the deadline is 2d9h23m. You are very, very close to exceeding the deadline. If you have been running smp less than 100% of the time, it's possible that you're seeing varying estimates for completion time and sometimes FahMon predicts a bonus and sometimes not. Until the WU is actually uploaded, everyting reported by FahMod is a prediction, and those estimated can vary considerably, especially if there's a chance you'll miss the deadline.

I see no evidence of any uniprocessor WUs.