LTT Folding Team's Emergency Response to Covid-19

[DESTROYER17]

20 minutes ago, Stealther said:

Thanks for organizing this, its has been a while since I folded like this!  I can't find the link to the result spreadsheet and I'm curious how I did.  Would someone kindly send me the link?  TIA!

9 minutes ago, Tad Bittoomuch said:

Contained within are links to all the blog posts from the event, in which are links to each day's spreadsheets:

Day Zero: Users List

Day One Stats

Day Two Stats

Day Three Stats

Day Four Stats

Day Five Stats

Day Six Stats

Day Seven Stats

Day Eight Stats

Day Nine Stats

Day Ten Stats

Day Eleven Stats

Day Twelve Stats

Day Fourteen Stats

 

I can't find Day Thirteen, but I don't think there where stats day 13, because @GOTSpectrum was quite sick and almost had to go to the Emergency room (correct me if I'm wrong).

 

[Psedog]

Just now, ObeliskAG said:

Looks like some people are contemplating getting some more folding power. As ongoing folding is going to use a lot of electricity, I've updated my table. Assuming the P106-100 as the reference - the higher Relative Efficiency the better:

 

It looks like the RX 5600 XT is excellent value.

 

Attached the spreadsheet for your enjoyment.

GPU Specs and Timings.xlsx 23.92 kB · 0 downloads

That's crazy how powerful the Vegas are at folding, but not so much in gaming.

[mIllIs]

44 minutes ago, rkv_2401 said:

If you leave it on -1, it should automatically set it for 10 threads in runtime.

-1 was my initial setting, that however left me with "11" on the log.

forcing it to 10 solved the issue though

[leadeater]

18 minutes ago, Psedog said:

That's crazy how powerful the Vegas are at folding, but not so much in gaming.

Vega kinda isn't really that good, GFLOPs is one thing but actual PPD is another.

 

Vega 56 has the same GFLOPs as a 5700 XT but half the PPD. Even Radeon VII isn't actually all that great, Vega architecture doesn't seem to do all that well efficiency wise for PPD. Polaris isn't much better either, not for PPD to GFLOPs ratio anyway.

 

Scale an RX 580 to 12 TFLOPs and it'll only be doing 660k PPD, about half what would be consider good for that raw compute power. 

[Psedog]

Just now, leadeater said:

Vega kinda isn't really that good, GFLOPs is one thing but actual PPD is another.

 

Vega 56 has the same GFLOPs as a 5700 XT but half the PPD. Even Radeon VII isn't actually all that great, Vega architecture doesn't seem to do all that well efficiency wise for PPD. Polaris is much better either, not for PPD to GFLOPs ratio anyway.

 

Scale a RX 580 to 12 TFLOPs and it'll only be doing 660k PPD, about half what would be consider good for that raw compute power. 

Considering the age and I paid less than 400 a year and a half ago, my stats are pretty good.

 

 

 

 

 

[Stormseeker9]

1 hour ago, marknd59 said:

Are you running a stock cooler and if so what temp is it running at ? I just built a new system yesterday with 2700X in it and it's at 70C with at 70% CPU load using 8 threads for F@H. I'm trying to figure out if that's OK or if I need to be looking to buy a AIO water cooler for it.

Check my sig. cpu has no issue; it’s the GPU which gave up after a solid 6 days of folding; no more WU’s getting assigned. 
 

already using the advanced flag 

[Federated]

For the last few days i've struggled for WU's for my 1080 but my old RX480 gets them quite regularly which is a bit annoying

[marknd59]

2 minutes ago, Stormseeker9 said:

Check my sig. cpu has no issue; it’s the GPU which gave up after a solid 6 days of folding; no more WU’s getting assigned. 
 

already using the advanced flag 

No you missunderstood me I was not saying your CPU was a problem I was asking what temp. it was running at.

 

I've just built a new system with a 2700X and and I'm trying to figure out if the temp. is OK or if I need to get a water AIO for it. At the moment it's running at around 70C at 70% load. I was woundering what  temp. yours was at so I could compare it to mine, my R7 1700 runs at around 60C at 70% load.

 

The way you have to mess about getting that stupid cooler latch on I was wondering if I had a bad paste job.

[marknd59]

Finding WU should be getting easier. There was a big drop in the assign rate but it's now back over 100K an hour so things should get better.

[Valentyn]

1 hour ago, leadeater said:

Vega kinda isn't really that good, GFLOPs is one thing but actual PPD is another.

 

Vega 56 has the same GFLOPs as a 5700 XT but half the PPD. Even Radeon VII isn't actually all that great, Vega architecture doesn't seem to do all that well efficiency wise for PPD. Polaris isn't much better either, not for PPD to GFLOPs ratio anyway.

 

Scale an RX 580 to 12 TFLOPs and it'll only be doing 660k PPD, about half what would be consider good for that raw compute power. 

I'm surprised the 2080Ti is beating the Titan V there; massive difference in TFLOPs; and memory bandwidth.

I wonder how well whatever new 7nm+ 3000 series Titan will be able to do. I hope NVIDIA deliver a substantial improvement given the jump from 16-7nm, and that it's been more than 2 years since Volta.

[leadeater]

19 minutes ago, Valentyn said:

I'm surprised the 2080Ti is beating the Titan V there; massive difference in TFLOPs; and memory bandwidth.

I wonder how well whatever new 7nm+ 3000 series Titan will be able to do. I hope NVIDIA deliver a substantial improvement given the jump from 16-7nm, and that it's been more than 2 years since Volta.

You could probably get a fair bit more out of a Titan V & V100 if you optimized more for it but I don't think that kind of hardware is much of a focus for FAH. For what ever reaons Turing does exceptionally well at FAH, Navi projected out to 16.9 TFLOPs is 2.3m PPD so that really does just show how good the 2080 Ti is for this, price ignored of course.

[Spotty]

17 minutes ago, Valentyn said:

I'm surprised the 2080Ti is beating the Titan V there; massive difference in TFLOPs; and memory bandwidth.

I think that's more to do with how the chart collects its data. It's user submitted. More people have 2080Tis than Titan Vs, so there's a larger pool of data to pull from. There's only 1 user submission for the Titan V while there are 9 for the RTX2080Ti so you get a more accurate representation for the 2080Ti.

Not only that but the Titan V is a year older and its score there is likely based off of Core 21 work units while the 2080Ti is newer and more likely based off of Core 22 data which can potentially give a 10-20% boost in points.

 

The chart is here if you're interested. https://docs.google.com/spreadsheets/d/1vcVoSVtamcoGj5sFfvKF_XlvuviWWveJIg_iZ8U2bf0/pub?output=html

Which comes from the overclock.net forums https://www.overclock.net/forum/55-overclock-net-folding-home-team/475163-gpu-projects-ppd-database.html

 

That said, the Titan cards don't necessarily see much of a boost over the mainstream GTX/RTX cards.

[Xsilent(X)]

Guys my rig is now averaging around 2.5M PPD....

[sese317j]

47 minutes ago, marknd59 said:

Finding WU should be getting easier. There was a big drop in the assign rate but it's now back over 100K an hour so things should get better.

I dont think so atm, because the "biggest" work server (fah1), which had an assign rate of over 28000/hr, just ran out of disk space at some point in the last 24 hours. But maybe there is hope, two new work servers appeared in the list, which are hosted by AWS. They are the first servers in the list with a new software version (9.6.4) but i think they aren't  ready to start assigning and collecting WU just yet.

[Valentyn]

49 minutes ago, Spotty said:

I think that's more to do with how the chart collects its data. It's user submitted. More people have 2080Tis than Titan Vs, so there's a larger pool of data to pull from. There's only 1 user submission for the Titan V while there are 9 for the RTX2080Ti so you get a more accurate representation for the 2080Ti.

Not only that but the Titan V is a year older and its score there is likely based off of Core 21 work units while the 2080Ti is newer and more likely based off of Core 22 data which can potentially give a 10-20% boost in points.

 

The chart is here if you're interested. https://docs.google.com/spreadsheets/d/1vcVoSVtamcoGj5sFfvKF_XlvuviWWveJIg_iZ8U2bf0/pub?output=html

Which comes from the overclock.net forums https://www.overclock.net/forum/55-overclock-net-folding-home-team/475163-gpu-projects-ppd-database.html

 

That said, the Titan cards don't necessarily see much of a boost over the mainstream GTX/RTX cards.

Seems about right. I see larger PPD for my stock/reference Titan V at CUDA clocks ( 1335Mhz ), than I see some folks here with 2080Tis

So it could be Core21 results, mixed with lack of Titans. Since NVIDIA more than doubled the price on them they're scarcer. Athlough getting "professional" drivers now is wonderful.

 

[Alex Atkin UK]

I'm guessing that table is not from Core 22?  As my 2080 constantly reports between 1800k to 2000k.  T2 gets 930K.

[bafo_ah]

3 hours ago, leadeater said:

Vega kinda isn't really that good, GFLOPs is one thing but actual PPD is another.

 

Vega 56 has the same GFLOPs as a 5700 XT but half the PPD. Even Radeon VII isn't actually all that great, Vega architecture doesn't seem to do all that well efficiency wise for PPD. Polaris isn't much better either, not for PPD to GFLOPs ratio anyway.

 

Scale an RX 580 to 12 TFLOPs and it'll only be doing 660k PPD, about half what would be consider good for that raw compute power. 

It's rate on base clock? because my RTX 2080 (1950Mhz) produce at least 1.8-2.1 ppd and my 2060 (1920Mhz) have around 1.1-1.3 ppd (depend on wu, I use beta flag, with assumption I get steady wu, next-percentage-flag is 95)

[Xsilent(X)]

Picture of the update of my folding rig now with over 24 hours of folding...

[FloRolf]

On 4/10/2020 at 10:39 AM, FloRolf said:

Well, as someone who actually did buy it (Ultra version tho) and ran it the last 2 weeks, maybe I should put some results in the database

 

But yes, the cooler is garbage but it produces an average of ~1.1m so that's pretty good for a 330€ card imo. 

 

@ffbr @leadeater @GOTSpectrum

 

Here are some results of the 2060KO Ultra. Average is 1085k

 

and compared to a GTX1080. Average is 865k

 

i'll try to see if i can get some more common WU the directly compare the two even better. 

 

[Zberg]

12 hours ago, cbigfoot said:

they do fold quite well make sure to ramp the fan curve if you know how makes me feel better seeing it chigging away at 62C 

 

<rant> DAM you @Zberg For some reason yet beyond my comprehension i have orders the Meshify S2 IN WHITE.  mostly because 1 i want to move my computer and 2 testing Flexo he doesnt like the confines of the XPS 7100 so he will be getting the Corsair Chassis... and 3 I want GLASS i might have scratched the plexy the Corsair chassis has on it...and im too anal about that... and 4 your computer looked good in white hence why im blaming you...i know its the meshify but i want airflow lol...but do like the look of the define...</rant>

I too will be meshifying "the box" which currently is too packed with no airflow!

 

Probably the black meshify for me though!

 

[leadeater]

46 minutes ago, FloRolf said:

Here are some results of the 2060KO Ultra. Average is 1085k

Anyone got current stats for the reg 2060?

 

Edit:

But basically no difference but I would find it interesting if just one project did run faster, just being hopeful lol

[GOTSpectrum]

13 hours ago, justpoet said:

@GOTSpectrum, when you do the awards ceremony.  Don't forget to include the amount of alcohol you consumed as a winning stat!  

 

Feel better before worrying about writing it though.  An extra day or two to get a free game key isn't going to kill anybody.

Appreciate your understanding on the matter, I've now done all the final works on the sheet, and assigned who is winning what through a RNG.

 

 

[abrackers]

4 hours ago, leadeater said:

Vega kinda isn't really that good, GFLOPs is one thing but actual PPD is another.

 

Vega 56 has the same GFLOPs as a 5700 XT but half the PPD. Even Radeon VII isn't actually all that great, Vega architecture doesn't seem to do all that well efficiency wise for PPD. Polaris isn't much better either, not for PPD to GFLOPs ratio anyway.

 

Scale an RX 580 to 12 TFLOPs and it'll only be doing 660k PPD, about half what would be consider good for that raw compute power. 

Those Numbers seem very low. My vega 64 gets 1.0 - 1.3m ppd on standard wu's.

[GOTSpectrum]

4 hours ago, DESTROYER17 said:

Day Twelve Stats

Day Fourteen Stats

 

I can't find Day Thirteen, but I don't think there where stats day 13, because @GOTSpectrum was quite sick and almost had to go to the Emergency room (correct me if I'm wrong).

 

Unfortunately that was the case. But just to be clear it wasn't illness, just stupid stockpilers meaning I couldn't get my meds. I have the meds now but the downside is I have to deal with the side effects again while I start taking them again. Not fun!

[leadeater]

3 minutes ago, abrackers said:

Those Numbers seem very low. My vega 64 gets 1.0 - 1.3m ppd on standard wu's.

They will be, all the older existing GPUs will have Core21 stats submissions where Core22 is now the common and produces higher PPD. It's actually a little dangerous to use that spreadsheet right now but oh well.

 

Edit:

All the Turing GPUs will be Core22 in that list, same with Navi.