GPU Projects PPD Database
2261 - 2280 of 2329 Posts
Joined
·
28,262 Posts
yeah.. then an hour later a 14xxx shows up and cuts productivity in half and takes twice as long to complete. Hard to understand the PPD-math there.
Joined
·
6,960 Posts
That PPD output that you are submitting, is that based on HFM? Becasue I would love to see a sanity check from HFM to see that same PPD.
Joined
·
28,262 Posts
Not using HFM on this rig. Why is that considered a sanity check?
Anyway, here's another similar PPD with 11738 I posted here some time ago (a week maybe?). I didn't enter it into the database... only 11% complete. 11738's are running that high on these 2080Tis which are at 2040/6800 for [email protected]
I'll see if I can catch one further along. But no worries, no need to enter either into the database as far as I'm concerned.
Attachments
Joined
·
5,120 Posts
Actual throughput and PPD will vary depending on upload/download speed and next-unit percentage setting, but the PPD reported by the client is right in line with what is expected of an 11738 running at that TPF. Without knowing the overall elapsed time for the WU there is no way to calculate the actual final points for the WU, but we can estimate the PPD.
To get an exact estimate, you would just need need to parse the logs to get the average of the completed frame-times. Going off the TPF reported by the client (59 seconds) this is what I come up with:
14.4*29389*(sqrt(14.4*.75*8.2/.98))/.98 = 4,105,130 PPD
*If you are a registered user on the Folding Forum, you can get the k-factor and the deadline in the OP of the beta project's announcement thread. You can also get the deadline from the link that @mmonnin posted for active projects. k-factor is generally set at .75, but I think I've seen that vary on a few projects.
https://foldingathome.org/support/faq/points/
https://apps.foldingathome.org/psummary?visibility=ALL
And here's a link that automates the above math for you:
http://www.linuxforge.net/bonuscalc2.php
To get an exact estimate, you would just need need to parse the logs to get the average of the completed frame-times. Going off the TPF reported by the client (59 seconds) this is what I come up with:
14.4*29389*(sqrt(14.4*.75*8.2/.98))/.98 = 4,105,130 PPD
*If you are a registered user on the Folding Forum, you can get the k-factor and the deadline in the OP of the beta project's announcement thread. You can also get the deadline from the link that @mmonnin posted for active projects. k-factor is generally set at .75, but I think I've seen that vary on a few projects.
https://foldingathome.org/support/faq/points/
https://apps.foldingathome.org/psummary?visibility=ALL
And here's a link that automates the above math for you:
http://www.linuxforge.net/bonuscalc2.php
Joined
·
1,762 Posts
Just added a submission for Project 11738 on my 2080ti.
TPF is 53 sec.
Estimated PPD is 4,797,101.
The estimate is pretty spot on since I have gotten a few of these and they ran about the same consistently.
2080ti is running 2100 core and +800 mem.
CPU is 9900k @5.1 with ram running at 4000 cl17
Edit: Forgot to say what driver I am on. 441.66
TPF is 53 sec.
Estimated PPD is 4,797,101.
The estimate is pretty spot on since I have gotten a few of these and they ran about the same consistently.
2080ti is running 2100 core and +800 mem.
CPU is 9900k @5.1 with ram running at 4000 cl17
Edit: Forgot to say what driver I am on. 441.66
Joined
·
28,262 Posts
Yeah, every once in a while these 11738 WUs come in and they run very efficiently on the 2080Ti. I'm trying ot see if it is a 2080Ti thing or they are just more efficiently coded or something. They pull the highest power of any I've looked at.
Joined
·
5,120 Posts
The same is seen on AMD gear. 11738 is a core_22 WU, and this new core is built on a much newer version of OpenMM, which has many optimizations over the older version that was used to build the core_21 WUs.
It is pretty great to see the increase in GPU usage, power usage, and heat, from a pure performance perspective. That is almost always a sign that the compute app is better leveraging the full capabilities of the GPU.
It is pretty great to see the increase in GPU usage, power usage, and heat, from a pure performance perspective. That is almost always a sign that the compute app is better leveraging the full capabilities of the GPU.
Joined
·
8,223 Posts
Really enjoy the PPD the core 22 gets (3.8mil on a stock 2080Ti).
In combination with the 7700k folding at 6 cores I see 550W from my wall powermonitor.
GPU-Z shows a little over 300W on the 2080Ti alone.
I'm not seeing the same PPD/power draw on all the 22 units though, there is some difference.
Joined
·
7 Posts
Just getting into [email protected] with my son over the past week and I've been looking for a PPD CPU database, haven't found much luck other than the newly created one at foldingforum.org at https://foldingforum.org/viewtopic.php?f=38&t=34240 and you can see the database here: https://docs.google.com/spreadsheet...PI_bqH3CeZVU7Wc9CGPL0CdLkg/edit#gid=323814829
It seems to have a few more submissions of the 5700 XT than we have here =)
It seems to have a few more submissions of the 5700 XT than we have here =)
Joined
·
2,545 Posts
Joined
·
6,960 Posts
Good find!
We did attempt a CPU database at one point, but we ran into a bit of a data management issue. Firstly, not everyone folds with the same number of cores. Usually, if people have GPU folding rigs, the remove an even number of threads from the folding process. Also, we noticed that different OS's had a significant effect on PPD output, primarily due to scheduler issues.
I don't think we will have one setup on OCN, so that seems like a great reference point.
In regards to WU submissions, we are a bit restricted based on what users submit. If you have data that you would like added to the database it, please give it an upload!
And if you are trying to submit lets say 15+ WUs, and you are using HFM, please PM me. I can upload mass HFM submissions to the database, in order to streamline the process a bit. Last time we tested this, we managed to get 500 entries added to the database at once.
It's a bit clunky, but it works.
Joined
·
6,960 Posts
Okay sorry for DP, but the Database is now up to date with current submissions! Keep em coming folks!
Joined
·
2,545 Posts
something seems to have went wonky, when sorted by PPD... vega 56 is above and below the radeon VII in the chart and the 5700XT PPD is way down on what it was before and even below what I usually see with my "efficient" profile with lower power limits.
Attachments
Joined
·
5,120 Posts
That's a Google sheet, so Vega 56 and RX Vega 56 are different GPUs.
The 5700XT are probably down in average ppd because there were a fair number of early WUs that had ppd that was under 800k. It has only been with some of the more recent WUs that ppd has gone up in the 1.2M+ average.
@Simmons572 I have a bunch of WUs that I can submit for Radeon VII, but I will probably submit them individually because there are 4 rigs and like 14 slots in my HFM log. :thumb:
The 5700XT are probably down in average ppd because there were a fair number of early WUs that had ppd that was under 800k. It has only been with some of the more recent WUs that ppd has gone up in the 1.2M+ average.
@Simmons572 I have a bunch of WUs that I can submit for Radeon VII, but I will probably submit them individually because there are 4 rigs and like 14 slots in my HFM log. :thumb:
2261 - 2280 of 2329 Posts
