F@H Web Client Dark Skin with PPD comparison by GPU / WU

[Mihle]

Any reason why so many AMD cards are lumped together as one while every single Nvidia card with only the slightest difference get its own?

[cbigfoot]

3 hours ago, Mihle said:

Any reason why so many AMD cards are lumped together as one while every single Nvidia card with only the slightest difference get its own?

My guess is it is F@H being Lazy as the driver is 100% specific but they dont want to seperate

[LAR_Systems]

4 hours ago, Mihle said:

Any reason why so many AMD cards are lumped together as one while every single Nvidia card with only the slightest difference get its own?

@cbigfoot is correct.  It’s how F@H groups them on the gpu list.  So if they don’t separate them unfortunately there is no way for me to do it in the client.

[Chord]

My 1070 is ranked "0 of 119 GPUs" and the slot performance is based on an average of 0 work units. Any ideas on how I might be bale to fix this?

[LAR_Systems]

4 hours ago, Chord said:

My 1070 is ranked "0 of 119 GPUs" and the slot performance is based on an average of 0 work units. Any ideas on how I might be bale to fix this?

Hi Cord,

 

Can you look at the bottom of your window and let me know what version of F@H it says you're running.

[Sumgaivg]

Hi, I'm getting this when using chrome beta as well as regular chrome and Edge. It works fine when not using the extension. Any ideas as to what could be the cause?

[LAR_Systems]

9 hours ago, Sumgaivg said:

Hi, I'm getting this when using chrome beta as well as regular chrome and Edge. It works fine when not using the extension. Any ideas as to what could be the cause?

Looking at the power slider you have it set to light, so it's paused and not running cause your using the computer, it's "waiting for idle" to do stuff while you're not there.

The update system, chart logging, PPD performance etc. only happens when you're folding (status is running), try putting the power slider to medium or full to have it run while you're looking at the window.

Edit:  It also looks like a mobile GPU, so I assume a laptop, it might not fold on battery, so this might happen if you're not plugged in.

[Sumgaivg]

Sorry, should have been more clear. I've already run through a full project, this is all that I've seen since I installed it, no charts or points per day etc. just those dashes. I usually pause  when I'm not at home or able to monitor due to temps on drives getting out of hand(known issue with my particular model)

Yes I am using  a laptop but I'm not running on battery.

[LAR_Systems]

5 minutes ago, Sumgaivg said:

Sorry, should have been more clear. I've already run through a full project, this is all that I've seen since I installed it, no charts or points per day etc. just those dashes. I usually pause  when I'm not at home or able to monitor due to temps on drives getting out of hand(known issue with my particular model)

Yes I am using  a laptop but I'm not running on battery.

Could you please provide a screen shot of it in the running state, there are some labels etc. that change when it's running that can help me diagnose where a problem may lie that I can't see when it's waiting.

Ideally set it to full, wait a minute for it to do all the calls it's going to do then screen shot that.

 

Thanks,

 

[Sumgaivg]

As requested,I let it run for about 4 min on full then took the screen shot.

[LAR_Systems]

Thanks, looks like your GPU is not getting added.

I have made a change on my side for what I'm guessing the issue is.

Please clear your browser cache and try reloading the window. 

 

Please let me know when you have done this so can check the database to see if your GPU has been added.

[Sumgaivg]

1 hour ago, LAR_Systems said:

Thanks, looks like your GPU is not getting added.

I have made a change on my side for what I'm guessing the issue is.

Please clear your browser cache and try reloading the window. 

 

Please let me know when you have done this so can check the database to see if your GPU has been added.

Not seeing a change, did everything waited a few minutes and also tried the same in Edge.

[LAR_Systems]

23 minutes ago, Sumgaivg said:

Not seeing a change, did everything waited a few minutes and also tried the same in Edge.

Please try again now, cache clear, refresh... I have cleared the global cache on the content delivery network in the event something is stuck between my server and your browser.

 

Thanks in advance, know this is a pain, but it helps a lot to work out these issues when they only show for some users.

[Sumgaivg]

20 minutes ago, LAR_Systems said:

Please try again now, cache clear, refresh... I have cleared the global cache on the content delivery network in the event something is stuck between my server and your browser.

 

Thanks in advance, know this is a pain, but it helps a lot to work out these issues when they only show for some users.

Done, but no change. Happy to help.

[LAR_Systems]

2 hours ago, Sumgaivg said:

Done, but no change. Happy to help.

A few more changes has been made to try and identify your issue.

 

I have pushed them to the build if you would like to try and clear cache and reload again to see if anything changes.

If this does not resolve it, I will have to try another approach, but it will need to trickle out across the network, so it could take a couple days before we can retry.

Please let me know what you see.

[LAR_Systems]

3 hours ago, Sumgaivg said:

Done, but no change. Happy to help.

I also just checked the server logs for the past 2 days... your GPU has not been passed to the log aside from my testing.  This suggests that something on your system could be blocking the API calls.   Please check if you have a firewall / firewall rule that could be blocking stuff.

[Sumgaivg]

2 hours ago, LAR_Systems said:

I also just checked the server logs for the past 2 days... your GPU has not been passed to the log aside from my testing.  This suggests that something on your system could be blocking the API calls.   Please check if you have a firewall / firewall rule that could be blocking stuff.

Is there anything in particular I should be looking for in the firewall besides browsers and fahclient? Those have full access.

Did everything mentioned but no change.

I did notice that a GTX 1060 mobile is now showing up in the GPU PPD database.

[NetoriusNick]

Am I being really dense I'm struggling to find a table on @LAR_Systems DB that can be ordered by cost per 1M

 

[LAR_Systems]

2 hours ago, NetoriusNick said:

Am I being really dense I'm struggling to find a table on @LAR_Systems DB that can be ordered by cost per 1M

 

Power consumption estimates are only on the GPU profile pages right now.

 

My development cycles have to go in phases due to my time available, web client then website, repeat.   

 

So the client just got finished getting a major update last week and fixes etc. 

 

Now I'm working on a complete UI update to the database site to bring new screens, sorts, rank lists etc. by CPU, GPU, Projects, Atoms and power information tools.   

 

Power is a big pet project of mine with this as I would like to enable power management for people that have peak hour rate usage billing to the web client.   So that info of cost per million could be used by an individual to choose what cards run etc. at what times of day to help you do research, but not go broke doing it.  The key being, the client can switch WUs to finish automatically (not pause) so they stop as close to the end of your cheap power window, which for most is overnight to early morning.

[NetoriusNick]

2 hours ago, LAR_Systems said:

Power consumption estimates are only on the GPU profile pages right now.

 

My development cycles have to go in phases due to my time available, web client then website, repeat.   

 

So the client just got finished getting a major update last week and fixes etc. 

 

Now I'm working on a complete UI update to the database site to bring new screens, sorts, rank lists etc. by CPU, GPU, Projects, Atoms and power information tools.   

 

Power is a big pet project of mine with this as I would like to enable power management for people that have peak hour rate usage billing to the web client.   So that info of cost per million could be used by an individual to choose what cards run etc. at what times of day to help you do research, but not go broke doing it.  The key being, the client can switch WUs to finish automatically (not pause) so they stop as close to the end of your cheap power window, which for most is overnight to early morning.

would you consider more indepth local recording of temps and power specs on each local machine when you switch this to a python script? That is if I have understood previous comments, that in the long run this is to become a merger of HFM, babysitter.

[LAR_Systems]

2 hours ago, NetoriusNick said:

would you consider more indepth local recording of temps and power specs on each local machine when you switch this to a python script? That is if I have understood previous comments, that in the long run this is to become a merger of HFM, babysitter.

TLDR
While I never say never.   The first round of the python version is only for PPD and folding machine management (run, stop, finish, etc.)

 

 

The Development Issue:

When you get in to the local access to hardware like CPU / GPU temps, power settings, and local recording of data etc. in python you're moving in to software for the local machine that will require Admin permissions to run vs. something that is just massaging the F@H API data that their service is already providing and add information / perspective with my APIs and Server side logic.

The issue is the hardware access layer to get CPU/GPU hardware level in Python is done in part usually though addition of DLLs, frameworks and other fun 3rd party API stuff that that is OS specific.  This means different builds with possible different feature sets / limitations possible for Windows, Linux and OSX in co-ordination with 3rd party libraries and software I do not control.

As this adds various barriers to development, testing and requiring users to run some random stuff I need to bundle in but don't get to see the source code, as admin, I'm reluctant to focus on local hardware access for temps and power settings etc. until I have a basic plain text, platform agnostic python script that can provide what the web extension does first.

 

 

The Lack of Community Value Issue:

On a side note, once I have the basics in place I'm not sure I will ever focus on local hardware monitoring features in general, as there are lots of solutions already providing great hardware monitoring logging tools cross platform with giant dev teams to keep on the moving target of os drivers and hardware.

Temps and power targets might be of interest to an individual local user on their hardware, but globally these types of data points don't really change the average PPD by Project calculations (highs, lows, all in-between) for a GPU as what I provide are averages for a GPU on a Project so that you can see if you are above or below the community average if you tweak your system to be the best at PPD, or best at power / thermals.   

 

The data I provide is an indicator to know if your GPU on a project is close to average compared to everyone, or wildly out of spec above or below, and if out of spec, you need to tinker depending on what you're trying to accomplish.   

 

The average PPD for your GPU and a Project is always changing unlike regular GPU synthetic benchmarks (which are consistent) as the work units and folding cores are always fluctuating variables in terms of performance which is why I show the averages over the past several weeks on the site as F@H changes stuff up.  Todays hardware specific insights, overclocks etc. might be different or useless tomorrow if F@H changes the WUs or cores.

 

This is why I recommend using what's already out there for hardware info, I do this looking at the PPD average +/- indicator and charting, then using MSI afterburner to dial in my GPUs until I'm stable and above the average for my GPU and that project.  Then adjust system fans until the thermals are where I'm comfortable, which is subjective.   I do this weekly as I see the PPD trends change for a GPU model globally.

 

Everything else is really more on the Individual user to figure out what they are trying to accomplish be it more PPD, or better power usage, lower temps and all the environmental (room temp, cooling with air or water/aio, LM2  and and computer settings, silicon lottery, silicon degradation etc. that play in to that which are not captured in software and are too numerous to account for and provide meaningful feedback to the community in terms of the overall database, as that level of detail becomes data tailored to a single users configuration and specific physical hardware settings, environment and issues.  

 

[LAR_Systems]

Perfect example case came up on my one machine for showing the difference by WU while everything else about the system folding is identical which is why averages are used in the database as this is not benchmarking on a 1:1 basis by Machine config, GPU and Project.

Same exact 2060s cards, same production batch, clock and memory settings, folding machine, GPU temps, Room temps, etc.

The important part, Same Project... the 200K difference in PPD is specific to the WU PRCG that is being folded and/or a silicon difference in the GPUs.

 

[SodaDog]

Part time lurker here.
Love the interface so far, and how CPU stats are being implemented. Basically having it running to add to the database.

May I suggest a way to schedule folding tasks? We have solar and time of use rates where electricity costs basically double so I’d love the ability to have my computer go all out when the sun shines and shut down when it gets pricy.

Probably a long shot but I’d love to see a way to look up  and compare( Recorded PPD per GPU/CPU)  over (Manufacturer’s stated Wattage) in a sortable table. 

[Favebook]

1 hour ago, SodaDog said:

May I suggest a way to schedule folding tasks? We have solar and time of use rates where electricity costs basically double so I’d love the ability to have my computer go all out when the sun shines and shut down when it gets pricy.

That would require additional scripts which would probably require additional program downloads or google chrome admin access to your device. Not feasible as of now.

 

1 hour ago, SodaDog said:

Probably a long shot but I’d love to see a way to look up  and compare( Recorded PPD per GPU/CPU)  over (Manufacturer’s stated Wattage) in a sortable table. 

That would be more realistic to make but not sure it's worth it. Current tables have more than enough info, however down the line, I expect @LAR_Systems will probably make this one too.

[A-J-M]

Just noticed the new "WU reward". The "OOF" on one of my slots drew my attention...