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

[marknd59]

[Baha]

6 minutes ago, cbigfoot said:

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

Ah! You're right, thank you. I somehow managed to completely miss the text down there. Whoops.

[Sequence211]

Do i understand this right?.

If the web UI sleeps it will not upload the results to the PPD db?

[cbigfoot]

1 hour ago, Struck said:

Do i understand this right?.

If the web UI sleeps it will not upload the results to the PPD db?

that is correct last i knew unless his skills have worked around it @LAR_Systems 

[marknd59]

I just have it open in all the time it's own window.

[LAR_Systems]

3 hours ago, DirectingGnu2 said:

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

Annnnd the 3080 is back in 1st place.   Realtime data is a real mofo!

I suspect F@H may have optimized work units for the more "common" card LOL that they have had an extra week to mess with data.

[DirectingGnu2]

1 hour ago, LAR_Systems said:

Annnnd the 3080 is back in 1st place.   Realtime data is a real mofo!

I suspect F@H may have optimized work units for the more "common" card LOL that they have had an extra week to mess with data.

O cmon... I was at disadvantage, I did some WUs before CUDA support. I guess I am running really hot, it ain't easy cooling 2 of them with air. Time to bring in some OC now

[cbigfoot]

3 hours ago, DirectingGnu2 said:

O cmon... I was at disadvantage, I did some WUs before CUDA support. I guess I am running really hot, it ain't easy cooling 2 of them with air. Time to bring in some OC now

 Cry me a river im running 2x 2060 and a 5700xt lol... 

But good news is that the 3090 is on top again trailed by a 2080TI!!!! then the 3080...WTH Nvidia!!!

[DirectingGnu2]

44 minutes ago, cbigfoot said:

 Cry me a river im running 2x 2060 and a 5700xt lol... 

But good news is that the 3090 is on top again trailed by a 2080TI!!!! then the 3080...WTH Nvidia!!!

yea I was gonna say, why is there a 2080ti up there

[LAR_Systems]

1 hour ago, DirectingGnu2 said:

yea I was gonna say, why is there a 2080ti up there

I don't know what the Rev A 2080 Ti is.. but it's not your normal 2080 Ti it's always been trading blows with the top of the leaderboard.

Will this owner of this unicorn please stand up and identify the make and model.

[tep]

2 hours ago, LAR_Systems said:

I don't know what the Rev A 2080 Ti is.. but it's not your normal 2080 Ti it's always been trading blows with the top of the leaderboard.

Will this owner of this unicorn please stand up and identify the make and model.

I'm contributing with a strix 2080ti. It's been consistent at 5-5.5m PPD since cuda was enabled.

 

[Yabdat]

 

I've apparently got a Rev.A card, but its just a 'ASUS GeForce RTX 2080 Ti DUAL Advanced' card I bought right after launch.

Mine is also running in the ~5.5m PPD range after cuda was enabled.

[LAR_Systems]

3 hours ago, Yabdat said:

 

I've apparently got a Rev.A card, but its just a 'ASUS GeForce RTX 2080 Ti DUAL Advanced' card I bought right after launch.

Mine is also running in the ~5.5m PPD range after cuda was enabled.

Thanks for letting me know, and congrats on the unicorn cards.   There seems to be a little magic in the Rev. A.

[Favebook]

@LAR_Systems If I exit chrome or close tab of your extension before WU is finished or do not even run it until it's close to end, but I run it at like 90% of WU. Will the WU be submitted normally or does your extension have to be active for all 100% of WU being folded for submissions to be valid?

[LAR_Systems]

1 minute ago, Favebook said:

@LAR_Systems If I exit chrome or close tab of your extension before WU is finished or do not even run it until it's close to end, but I run it at like 90% of WU. Will the WU be submitted normally or does your extension have to be active for all 100% of WU being folded for submissions to be valid?

It’s ok to close before done it takes multiple ppd samples during a WU to see ppd fluctuations. The last X percent is not needed as long as it got to track most of the Wu.

[Favebook]

4 hours ago, LAR_Systems said:

It’s ok to close before done it takes multiple ppd samples during a WU to see ppd fluctuations. The last X percent is not needed as long as it got to track most of the Wu.

That means, I SHOULD let it run during the folding of a WU.

 

May I know at what % does it take samples? Or is it time based?

[LAR_Systems]

7 hours ago, Favebook said:

That means, I SHOULD let it run during the folding of a WU.

 

May I know at what % does it take samples? Or is it time based?

It's time based approximately every 15-20 mins to account for fast/slow folding cards and catch the change.

This timeframe will be optimized to be algorithmic in upcoming versions of the extension to be more focused on capturing swings in ppd vs random time samples. 

[LAR_Systems]

7 hours ago, yaboistar said:

turns out that the best 3080 is a 2080Ti

As long as it's the REV A

That said, there are not a lot of 3080s in the wild right now (that are also using the extension) so I doubt it's getting a wide sampling of projects consistently right now.

This is why I'm on multiple notification lists for when they are in stock as I really want a 3080 for 24/7 folding in my lab to keep the numbers right as I assume the majority want them for gaming at least some of the time.
 

[LAR_Systems]

31 minutes ago, yaboistar said:

I’m going after the 3090 next

It's funny... I have a 2080 Super pulling 3.5 Million right now... it's like they have optimized, the optimized CUDA projects. 

[DirectingGnu2]

From a quick glance, the COVID projects seem to give a more steady and higher PPD than the non covid projects

[LAR_Systems]

6 hours ago, DirectingGnu2 said:

From a quick glance, the COVID projects seem to give a more steady and higher PPD than the non covid projects

Can confirm.   The covid projects more less rule the GPU/PPD leaderboard until you get a good way down the list.

https://folding.lar.systems/folding_data/gpu_ppd_by_wu

[Baha]

On 10/5/2020 at 9:03 AM, LAR_Systems said:

As long as it's the REV A

That said, there are not a lot of 3080s in the wild right now (that are also using the extension) so I doubt it's getting a wide sampling of projects consistently right now.

This is why I'm on multiple notification lists for when they are in stock as I really want a 3080 for 24/7 folding in my lab to keep the numbers right as I assume the majority want them for gaming at least some of the time.
 

I'd just like to add to this and mention that I've been leaving a 3080 running with the extension for a few days now, and even when limiting the power to 70%, I'm still seeing constant reminders of supposedly outperforming the database's expected results by significant margin:

[LAR_Systems]

2 hours ago, Baha said:

I'd just like to add to this and mention that I've been leaving a 3080 running with the extension for a few days now, and even when limiting the power to 70%, I'm still seeing constant reminders of supposedly outperforming the database's expected results by significant margin:

And you will keep seeing that until the new core is a few weeks old and the average PPD per project has adapted to the new Nvidia CUDA normal.

[Baha]

2 hours ago, LAR_Systems said:

And you will keep seeing that until the new core is a few weeks old and the average PPD per project has adapted to the new Nvidia CUDA normal.

Makes sense! I'm getting back into folding after a long break, so I'd forgotten the CUDA patch was so recent. Thanks for the info!

[A-J-M]

Just installed another 2080ti Rev. A (second hand): ROG-STRIX-RTX2080TI-011G-GAMING 

 

 

EDIT: But I guess that the true unicorn that raises the average to such heights is a special model, with binned chips, like this one:

https://www.techpowerup.com/gpu-specs/gigabyte-aorus-rtx-2080-ti-xtreme-waterforce-wb.b6363

 

nvidia-smi (out of the box - look at the wattage): 

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 440.100      Driver Version: 440.100      CUDA Version: 10.2     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce RTX 208...  On   | 00000000:01:00.0 Off |                  N/A |
| 38%   52C    P2   316W / 366W |    541MiB / 11018MiB |     93%      Default |
+-------------------------------+----------------------+----------------------+

UPDATE: I also have those binned chips (300A-K1-A1) and with just Afterburner (now under water) I can reach 

Microsoft Windows [Version 10.0.19041.508]
(c) 2020 Microsoft Corporation. All rights reserved.
C:\Users\ajmch>nvidia-smi
Mon Oct 12 18:53:47 2020
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 456.38       Driver Version: 456.38       CUDA Version: 11.1     |
|-------------------------------+----------------------+----------------------+
| GPU  Name            TCC/WDDM | Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce RTX 208... WDDM  | 00000000:17:00.0  On |                  N/A |
|  0%   55C    P2   312W / 325W |    757MiB / 11264MiB |     94%      Default |
+-------------------------------+----------------------+----------------------+

To get to 366 W, the Waterforce (factory OC) uses a special Bios: https://www.techpowerup.com/vgabios/211063/gigabyte-rtx2080ti-11264-181121-3