Difference : Timeout and Expiration date

[HugoNotte]

Are you running Windows? On your hardware it might be worth a try to run Linux, e.g. Ubunu, LUbuntu, Linux Mint Mate or XFCE edition or even another light weight distribution such as Puppy Linux. Since FAHClient runs at very low priority, operating system tasks have higher priority. WIndows can keep a low end CPU relatively busy and take away resources from FAH.
Linux, with a lightweight or lighter weight desktop environment might just be less demanding as an OS itself on the CPU freeing up more resources for FAH. Just a thought...

[rhavern]

From the current FAH FAQ:

Are there any limits to how long my machine can take to finish a work unit (WU)?

Yes. Work Units are serial in nature. When a completed WU is sent back, a new work unit is generated from those results. This must happen many times over within each project (group of work units). A generation 1 work unit must be turned in before a generation 2 work unit is created and sent out.

To keep these generations moving along, we have to set expiration deadlines in the event a work unit is not uploaded in a timely manner (lost, deleted, whatever). These unfinished work units “expire” and are reassigned to new machines. You will still receive credit for all WUs completed and uploaded prior to the Timeout (formerly preferred deadline). However, after the Timeout, your contribution is not as useful scientifically because another copy of that work unit had to be sent out to another contributor. Even if you eventually complete the work unit, that other contributor still had to process duplicate work to assure the science moves forward. And it would be unfair not to also credit that second contributor.

Even so, full credit is given up until the Deadline (formerly Final Deadline). After the Deadline has expired, the client will automatically discard the work unit and download new work. If you have trouble completing work units before the Timeout (formerly Preferred Deadline), it is recommended to either run the FAH client more hours each day, or to run the client on a faster computer.

[famasoft]

Hello, I have followed the 3D up to here and I would like to present my real case.

On October 3, Server 128.252.203.11 assigned Work Unit 18201 (2896, 2, 7) to the NVIDIA GPU of my Notebook with an Estimated TPF of 46 mins 05 secs.
This Work Unit has TimeOut on October 5 and Expiration on October 8.

At this moment we are at October 6 and there are still 4 hours to finish the processing that lasts more than 3 days with the PC always on and operating almost exclusively.

If, like all normal people, I kept the PC on for about 10 hours a day, it would take at least 7 days to finish this work and therefore would exceed the Expiration date, I believe, with the risk of throwing away all the calculations made but not completely finished.

On the basis of the above, because it is not the first time it happens, I cannot understand with which logic, or particular calculation, are assigned such heavy Work Units to a PC so slow.

I hope someone can give me an answer and maybe some suggestion, but I don't have a faster PC and this one is already working 24 hours a day and maybe it's better if I stop folding and I donate the electricity saved to charity.

Thank you very much.


(translated with DeepL.com)

[JimboPalmer]

I kept the PC on for about 10 hours a day, it would take at least 7 days to finish this work and therefore would exceed the Expiration date, I believe, with the risk of throwing away all the calculations made but not completely finished.

F@H works best i you fold 24/7, it is possible you are not an ideal folder,

I cannot understand with which logic, or particular calculation, are assigned such heavy Work Units to a PC so slow.

While there are projects to change this, currently F@H assigns work on the capability of a GPU, not it's performance, A GT 1030 might get the same WU as a GTX1080ti

[gunnarre]

F@H works best i you fold 24/7, it is possible you are not an ideal folder,

I think famasoft folds 24/7, but is pointing out that if they had instead folded 10 hours per day, completing the WU would be even more hopeless.

[famasoft]

That's exactly what I said, so much so that I think maybe I'd better give up F@H and get on with other things.

[gunnarre]

One problem with gaming notebook GPUs is that although they might have a "1650" or something inside, a notebook 1650 is not the same as a desktop 1650. Possibly your GPU could be re-classified to a lower performance tier and make timeouts in that, but at the moment GPUs are stratified by capability in FAH, not performance. There is a project to change this, but it's on the back burner to the science.

And some notebooks are just not up to folding - at least not on the GPU and CPU at the same time, and some can't even fold on one of them, due to lack of adequate cooling. Putting it on a notebook cooling stand with fan might help. If it's an old laptop, it might be possible to clean out the cooling ribs or perhaps even re-paste the cooling solution. But I have tried folding on several notebooks that were just too old or temperature-compromised to make timeout or be energy efficient.

[famasoft]

.....

I understand and agree with what you said but I still do not explain why in 2020 I folded from 70 to 80 Work Units per month with scores from 750,000 to 1,000,000 and in 2021 I fold at most 20 Work Units with just 500,000 points credited.
The notebook, equipped with an NVIDIA GEFORCE 940 MX card and purchased in 2017, is always the same, is constantly connected to the mains, rests on a cooling unit with two 120mm fans. and is set up to work with maximum performance.

[gunnarre]

The higher performance of Covid Moonshot sprints on your card might account for a higher PPD in 2020 than now, so perhaps a 33% slower performance now, but I wouldn't expect it to be fully 50% slower. (ref https://foldingathome.org/2020/09/28/fo ... a-support/ )

Could it be that your card is running on OpenCL instead of CUDA? It would help if you could post some of your log, especially the part where it says GPU X: Bus:X Slot:X etc and the part where it starts folding work unit saying something like "0x22: Using CUDA and gpu 0" then we could eliminate issues with the OS or drivers before starting to look at hardware issues (like cleaning out dust). Edit: Guides and parts for replacing cooling fans can be found on sites like IFixit (other suppliers also available).

[famasoft]

Maybe it's the translator's fault if I can't get what I'm saying across, so I'll try again.

F@H uses OpenCL in both year 2020 and year 2021.
The Notebook is connected to the power supply and F@H works 24 hours a day every day of the week using only the GPU because the CPU was overloading the system with not appreciable results.
The video card drivers are updated to the latest version available.
The Operating System was, in 2020, and is, in 2021, Windows 10 Home.
From January 2021 until June the notebook has been in service at ASUS for power supply related problems, it has been repaired and tested so I exclude further hardware problems.

At this moment F@H is processing a WU Assigned on October 18 with Timeout on October 20 (today) and Expiration on October 23, it has calculated only 60% and there are still 1.25 days to go.
Probably all the work done so far on this WU by my PC will be lost and will have to be reassigned to others.

Added this to the previous posts, I still don't understand with which criteria the Work Units are assigned.



Translated from Italian with www.DeepL.com/Translator (free version)

[BobWilliams757]

Maybe it's the translator's fault if I can't get what I'm saying across, so I'll try again.

F@H uses OpenCL in both year 2020 and year 2021.
The Notebook is connected to the power supply and F@H works 24 hours a day every day of the week using only the GPU because the CPU was overloading the system with not appreciable results.
The video card drivers are updated to the latest version available.
The Operating System was, in 2020, and is, in 2021, Windows 10 Home.
From January 2021 until June the notebook has been in service at ASUS for power supply related problems, it has been repaired and tested so I exclude further hardware problems.

At this moment F@H is processing a WU Assigned on October 18 with Timeout on October 20 (today) and Expiration on October 23, it has calculated only 60% and there are still 1.25 days to go.
Probably all the work done so far on this WU by my PC will be lost and will have to be reassigned to others.

Added this to the previous posts, I still don't understand with which criteria the Work Units are assigned.



Translated from Italian with http://www.DeepL.com/Translator (free version)

From my understanding the criteria used to assign work units is somewhat flawed, and based on architecture rather than ability. For that reason it's not common for GPU's to reach a point where they aren't meeting the timeouts or expiration dates, even if they have in the past. The best I can figure, with so many larger work units not suited to certain GPU's within a series, the greater the chance it will happen.

I do understand what you are saying, and I've had more work units assigned that could not meet the timeout in the last few months than I have in the last year and a half or so. There are a couple WU series in particular that won't finish in time, but have been assigned multiple times.

I've also noticed that over time, my PPD averages have dropped more in recent months. It's not directly related to the number of atoms, but seems to be more common.

I think in general the demands of GPU folding have increased, probably due to the fact that so many more higher end cards are available for use. If not it might just be a coincidence that these harder WU's are coming out now, and forcing more hardware into being obsolete sooner.

[gunnarre]

From my understanding the criteria used to assign work units is somewhat flawed, and based on architecture rather than [performance].

It would be great if the project to fix that would start progressing, but it seems that just getting the science done takes up most of the time.

[famasoft]

Using GPU-Z on my Notebook I could see that the NVIDIA 940 MX graphics card also has the CUDA platform but this is strangely not mentioned in the F@H log and in fact the OpenCL platform is used.

Extract of the most recent log:
.............................................................................................................................................................................................
08:44:06:WU00:FS01:0x22:No -opencl-device specified; using deprecated -gpu argument as an alias for -opencl-device.
08:44:06:WU00:FS01:0x22:Please consider upgrading your client version.
08:44:07:WU00:FS01:0x22:There are 3 platforms available.
08:44:07:WU00:FS01:0x22:Platform 0: Reference
08:44:07:WU00:FS01:0x22:Platform 1: CPU
08:44:07:WU00:FS01:0x22:Platform 2: OpenCL
08:44:07:WU00:FS01:0x22: opencl-device 0 specified
08:44:12:WU00:FS01:0x22:Attempting to create OpenCL context:
08:44:12:WU00:FS01:0x22: Configuring platform OpenCL
08:44:42:WU00:FS01:0x22: Using OpenCL on platformId 0 and gpu 0
08:44:42:WU00:FS01:0x22:Completed 0 out of 2500000 steps (0%)
08:44:43:WU00:FS01:0x22:Checkpoint completed at step 0
09:07:15:WU00:FS01:0x22:Completed 25000 out of 2500000 steps (1%)
.............................................................................................................................................................................................

Maybe that's what's causing the poor performance of F@H ?



In the log I noticed this note:
08:44:06:WU00:FS01:0x22:Please consider upgrading your client version

and I updated F@H to the latest version on the site hoping that it would improve the situation and in fact now the CUDA platform is mentioned in the log and used but the performance remained similar to before and the WU assigned now requires 4.10 days of work while the Timeout is in 3 days.

In addition, the 'F@H Control' icon that I used to use to verify operation no longer appears in the SysTray.




[ Translated from Italian with http://www.DeepL.com/Translator ]

[Joe_H]

Code: Select all

08:44:06:WU00:FS01:0x22:No -opencl-device specified; using deprecated -gpu argument as an alias for -opencl-device.
08:44:06:WU00:FS01:0x22:Please consider upgrading your client version.

What version of the F@h client are you using? The current version is 7.6.21, if you are not using that I would strongly suggest downloading and upgrading to that.

[gunnarre]

The video card drivers are updated to the latest version available.

Which Nvidia drivers are you using? "Latest" might mean Windows-provided, but these don't always include everything you need for CUDA.

Edit: What does GPU-Z say about the driver version?

Could you try the latest "Game ready" drivers from Nvidia. (Don't use the "Studio" drivers - these don't always have the latest required hooks that the newest folding core needs to take advantage of CUDA.)
Try using this search tool with your info: https://www.nvidia.com/en-us/geforce/drivers/ and it should suggest

GEFORCE GAME READY DRIVER - WHQL
Driver Version: 496.13 - Release Date: Tue Oct 12, 2021