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

[Sequence211]

34 minutes ago, LAR_Systems said:

Sounds like you're doing it all right, sorry if it came off preachy.
I tend to warn people to be on the safe side, rather than someone brick their laptop.

Less of an issue when talking to people on this forum I have to say.

I understand why you said so, and i would say the same to others.

No harm done

[RollinLower]

@LAR_Systems

on my main system, the PPD is not correct for my second GPU.

it's only displaying 200K, while the GPU is doing over 2M. i feel like this may skew the PPD comparison list somewhat.

what should i do about this?

[Kilrah]

The numbers match exactly, looks like you're looking at the current WU points instead of the PPD...

[Favebook]

1 hour ago, RollinLower said:

@LAR_Systems

on my main system, the PPD is not correct for my second GPU.

it's only displaying 200K, while the GPU is doing over 2M. i feel like this may skew the PPD comparison list somewhat.

what should i do about this?

What you think is PPD of second card is actually estimate for the WU of SELECTED GPU before QRB. Once you select other/second GPU, you can see what PPD on it is.

 

Everything is working fine, do not worry, you are just not used to web UI.  

[RollinLower]

35 minutes ago, Kilrah said:

The numbers match exactly, looks like you're looking at the current WU points instead of the PPD...

welp, and there you have it. excuse me being dumb.

[TVwazhere]

16 minutes ago, RollinLower said:

welp, and there you have it. excuse me being dumb.

Dont worry I did the same thing a few days ago

[RollinLower]

4 minutes ago, TVwazhere said:

Dont worry I did the same thing a few days ago

it helps knowing i'm not alone on this one!

[LAR_Systems]

2 hours ago, RollinLower said:

@LAR_Systems

on my main system, the PPD is not correct for my second GPU.

it's only displaying 200K, while the GPU is doing over 2M. i feel like this may skew the PPD comparison list somewhat.

what should i do about this?

I love (no really) that @Kilrah @Favebook @TVwazhere got my back replying to people while I'm finally sleeping

 

@RollinLower you are definitely not alone in confusing WU points for PPD, new folders do it all the time and experienced ones do it when looking at a glance because of the silly way F@H web client wants you to click on a slot to make it active to see it's stats as @Favebook pointed out.

This is why I'm currently testing on my systems changing the web client more fundamentally than just skinning what is there so that you can see everything at a glance like a proper dashboard.   

 

Example of this below, you will notice each slot shows it's info at all times, and tabbing would only be to see the project description in full.  Needs some tweaks before I can deploy it but I feel your pain on the way it works by default.


 

[TVwazhere]

51 minutes ago, LAR_Systems said:

I love (no really) that @Kilrah @Favebook @TVwazhere got my back replying to people while I'm finally sleeping

 

@RollinLower you are definitely not alone in confusing WU points for PPD, new folders do it all the time and experienced ones do it when looking at a glance because of the silly way F@H web client wants you to click on a slot to make it active to see it's stats as @Favebook pointed out.

This is why I'm currently testing on my systems changing the web client more fundamentally than just skinning what is there so that you can see everything at a glance like a proper dashboard.   

 

Example of this below, you will notice each slot shows it's info at all times, and tabbing would only be to see the project description in full.  Needs some tweaks before I can deploy it but I feel your pain on the way it works by default.

Spoiler

 

Suggestion; Perhaps instead of "This GPU", you could list the broad model term? "1060's are ranked #40 out of 87 for overall PPD".

"This GPU" always made it feel like it was referring to my specific card(s) in ITNOS against other cards of the same make. 

[LAR_Systems]

30 minutes ago, TVwazhere said:

Suggestion; Perhaps instead of "This GPU", you could list the broad model term? "1060's are ranked #40 out of 87 for overall PPD".

"This GPU" always made it feel like it was referring to my specific card(s) in ITNOS against other cards of the same make. 

Agreed.  The entire text blurb may go away in favor of plotting the average and your above / below on the charting component but it needs more testing.   Want the interface to be more friendly to multi-slot setups.

The general idea being a through line on the chart "Average  GTX 1060" and it just plots a straight line so you can visualize the collective average for the card on this project against what your system is currently doing on the chart.

The chart then plots as a live update component about once a second so you can see the difference in PPD on specific folds, or a dip / rise in PPD base on changing system settings or using your system etc.

Where did all my PPD go?   Well it took a major dip when you started watching full screen 4K you tube videos and you can see that on the chart that you're way below the average line... etc.

[TVwazhere]

12 minutes ago, LAR_Systems said:

Where did all my PPD go?   Well it took a major dip when you started watching full screen 4K you tube videos and you can see that on the chart that you're way below the average line... etc.

bu-bu-bu, muh pixels!

 

I'll definitely be curious to know how my 1070ti's do against the average considering my more aggressive power limits

[marknd59]

@LAR_Systems If you have a moving plot isn't that going to start 3d usage ? I thought the idea of disable animations was to get away from that ?

[GOTSpectrum]

1 minute ago, marknd59 said:

@LAR_Systems If you have a moving plot isn't that going to start 3d usage ? I thought the idea of disable animations was to get away from that ?

Would you not just be able to generate an image and to have it refresh that element of the page every X number of seconds? 

 

 

[LAR_Systems]

Just now, marknd59 said:

@LAR_Systems If you have a moving plot isn't that going to start 3d usage ? I thought the idea of disable animations was to get away from that ?

That's the testing part.   Charts would be optional and you can enable / disable them.

The 3D usage issue seen before was with animated GIFs etc. which engage a different part of the browser rendering system from "text" updates.    This is a pure JS chart component which should not engage the exact same systems and ties in to the existing text update loop.

That all said, there may be a performance hit... and what I want to test really well is if it's GPU (PPD) or CPU based.

The alternative option will be charts disabled, but the +/- percent info displayed in tab for the slot for context and still showing the text values updating for all slots.
 

If there is a minor performance hit from the chart, it might be worth it to users regardless to understand how the folds change throughout a work unit or when changing power limits etc. just not running it all the time.

Also toying with the idea of logging the chart data to memory, but then only displaying when a user wants to see it... but I only have soo much time to dedicate to this.

[TVwazhere]

4 hours ago, LAR_Systems said:

but I only have soo much time to dedicate to this.

What are you telling us you have a real job or something? You dont just live on the Internet? Pfft

[LAR_Systems]

1 hour ago, TVwazhere said:

What are you telling us you have a real job or something? You dont just live on the Internet? Pfft

My real job requires me to live on, build and maintain small chunks of the internet for money.     

Then my free time is... living on and maintaining small chunks of the internet that actually interest me.

[Sequence211]

The RTX 3080 just got added to the lists

https://folding.lar.systems/folding_data/gpu_ppd_datasheet?gpu_model=ga102_geforce_rtx_3080

seems promising, #1 PPD estimate, but it does not say how many WU samples is used for the calculation.

[LAR_Systems]

Just now, Struck said:

The RTX 3080 just got added to the lists

https://folding.lar.systems/folding_data/gpu_ppd_datasheet?gpu_model=ga102_geforce_rtx_3080

seems promising, #1 PPD estimate, but it does not say how many WU samples is used for the calculation.

Bottom of the profile shows it's all from one project right now, it will smooth out once a range of projects are run.

[Sequence211]

1 minute ago, LAR_Systems said:

Bottom of the profile shows it's all from one project right now, it will smooth out once a range of projects are run.

Yes i saw that. but there might be several different submissions for that specific project.

[LAR_Systems]

7 minutes ago, Struck said:

Yes i saw that. but there might be several different submissions for that specific project.

23 PPD samples all on the same project thus far (it's fun querying SQL DBs)

So to be expected not a lot of variance of the samples all on the same GPU and Project,  once there is a wider field of project and samples we will see the average likely settle down below the 3090.

That said... the 3080 could perform better than the 3090 if the projects are better optimized for the 3080 or able to just use it more efficiently.  

This has been seen in the past where Titans and 2080Tis would be in close races with 2080 Supers in some cases because the projects did not have need or optimization to use all of the more powerful cards to their advantage leaving processing power unused and making the lower end cards better value.

Hope the people running them keep the plugin running so we can find out.   At least until I can get ahold of one to run in my lab. 

[Sequence211]

52 minutes ago, LAR_Systems said:

23 PPD samples all on the same project thus far (it's fun querying SQL DBs)

So to be expected not a lot of variance of the samples all on the same GPU and Project,  once there is a wider field of project and samples we will see the average likely settle down below the 3090.

That said... the 3080 could perform better than the 3090 if the projects are better optimized for the 3080 or able to just use it more efficiently.  

This has been seen in the past where Titans and 2080Tis would be in close races with 2080 Supers in some cases because the projects did not have need or optimization to use all of the more powerful cards to their advantage leaving processing power unused and making the lower end cards better value.

Hope the people running them keep the plugin running so we can find out.   At least until I can get ahold of one to run in my lab. 

Yeah, i am also just waiting to order a 3080 FE. but it seems impossible.

[LAR_Systems]

1 hour ago, Struck said:

Yeah, i am also just waiting to order a 3080 FE. but it seems impossible.

I have multiple sites on "notify me" have not seen a single email.

Think they are holding back next wave of shipments until the entire crashing thing is confirmed driver vs hardware.

[DirectingGnu2]

well looks like the 3080 went back to where it belongs (behind my 3090s :P) lol

[Baha]

On 6/1/2020 at 11:02 PM, cbigfoot said:

yes, use chrome-extension://alpjkkbjnbkddolgnicglknicbgfahoe/fah_custom.html?nofade=true

it will prevent the extehsion from sleeping.

Just wanted to say thanks for the tool, and comment that LAR_Systems might want to stick this on the front page.

I was troubleshooting getting my client to stay active to help populate the DB with results for the new cards, and sticking ?nofade=true at the end of the standard client URL as the first post implies (the setup instructions tell you to use the regular folding client URL) doesn't seem to work.

[cbigfoot]

1 minute ago, Baha said:

Just wanted to say thanks for the tool, and comment that you might want to stick this on the front page.

I was troubleshooting getting my client to stay active to help populate the DB with results for the new cards, and sticking ?nofade=true at the end of the standard client URL as the first post implies (the setup instructions tell you to use the regular folding client URL) doesn't seem to work.

it has been updated there is an option at the bottom to disable sleep or something like that...