Join the MicroCenterOfficial Folding@Home Team: #257944

cine_chris

MightyMayfield said:

I'd say the additions are working 

Good to see, as that brings the team back to the ~100M PPD mark with @MightyMayfield as the Foundation & back into the elite ranks of the Top50 donors too.
Currently, 100M PPD puts the team near the Top30 threshold. 
Cooler weather might bring a few people back too.
Anyone else in a holding pattern waiting for the new gear to arrive?
The arrival of the CUDA client (only date I can get is soon) should pump new folding-power into the Nvidia gear.
AMD news:

AMD Announced two significant dates in October; the first one at the 8th addresses Zen3 and thus Ryzen 4000 series processors (Vermeer). This event is scheduled for October 8th. The name of the event is ‘Next Generation Ryzen Desktop Processors.’

In what now can be called Red October on the 28th will be about RDNA2 Graphics cards. The Radeon RX 6000 for which the name now has been confirmed in a tweet. This event will be called ‘Next Generation Radeon Graphics. With announcements in October, there is the possibility you may expect the availability of the new graphics cards and processors in November.

kmiller922

Oh lord AMD is having a star trek moment mixed with the Hunt for Red October. LOL

cine_chris

I'm taking a different approach at gauging or looking at participation.
http://bit.ly/MCOstats   
Are you aware that 40.2% of the Top1K folders have abandoned & producing Zed-Zero-Nada PPD?  It's like Home Run Stats, Babe Ruth's name is always near the top.  The EOC page link provides the box scores to what our team is really doing today.
What's missing is a view of what I call "the Power of Many", an aggregate view of what contribution every group makes to the team.
Something like this:

Yes, @0legravy_leg & I are the slackers w/11%, he needs to step it up a bit.
Let me know what you think. 
**NOTE: the 100Ks sum is higher, just what I could extract in one page.  MightyMayfield is the heart of the team, but as a group, the 100ks are the body. 

 

cine_chris

The EOC MCO team page with some annotated highlights. Showing activity that you never see when looking at cumulative stats.

MightyMayfield

cine_chris said:

**NOTE: the 100Ks sum is higher, just what I could extract in one page.  MightyMayfield is the heart of the team, but as a group, the 100ks are the body. 

 

Indeed! The strength of distributed computing is that everyone can take part. One strong computing body will rarely be as powerful as many nodes combined, even if latencies are a bit wacky sometimes
Edit almost a day later:
I think that 35.7Mppd is a daily output record for our store. More power and tweaking needed 

kmiller922

@MightyMayfield congratz on cracking top 200 way to go

MightyMayfield

kmiller922 said:

@MightyMayfield congratz on cracking top 200 way to go

200 more to go

cine_chris

MightyMayfield said:

kmiller922 said:

@MightyMayfield congratz on cracking top 200 way to go

200 more to go

4B!4MM

MightyMayfield

cine_chris said:

MightyMayfield said:

kmiller922 said:

@MightyMayfield congratz on cracking top 200 way to go

200 more to go

4B!4MM

I notice secretlabs turned off production...I wonder if there are any possible upgrades they could be making to surpass us

cine_chris

Even being a long time MC customer (decades, folder too), I've not rcv'd any news on the new Nvidia gear.
Saw mention that the initial MC store allotment was ~20 units.  I suspect it won't improve until the boat shipments start to arrive, air shipping is severely impacted due to travel restrictions.  A shrewd mfg would be loading 3070s onto ships now & flood the market when released.
Sea freight shipping takes 24-28 days.

cine_chris

I notice secretlabs turned off production...I wonder if there are any possible upgrades they could be making to surpass us

That'd take some serious rackspace. If surpass means catch.
As #471 NetAppAI NetApp @75,408,293PPD  won't catch you until 11.09.20, 5am / 1.7 Month
An MMclone would put MCO near Top20 PPD rank! ~150M PPD
Beyond Top20 it gets tough, Top15 jumps to 200M PPD.

Yellowbillycat

cine_chris said:

...
Cooler weather might bring a few people back too.
...

Yup.  It was 39 degrees when I woke up this AM.  Turned 3 FaH machines back on.

JADO

cine_chris..where do you get the data you report.  I don't see it on the FAH web statistics page?  Thanks

cine_chris

FAH released some #'s on the CUDA core development.

core22 0.0.13 in BETA testing : CUDA support at last!

TPF 73s - GTX 1080Ti running OpenCL/ 1.554 M PPD
TPF 57s - GTX 1080Ti running CUDA / 2.253 M PPD
TPF 49s - RTX 2080Ti running OpenCL/ 2.826 M PPD
TPF 39s - RTX 2080Ti running CUDA / 3.981 M PPD
TPF 36s - RTX 3080 running OpenCL / 4.489 M PPD
TPF 31s - RTX 3080 running CUDA / 5.618 M PPD
So if Windows-OpenCL is our baseline, then on this project, the speed improvements are:
Windows-CUDA: +53%
Debian-OpenCL: +24%
Debian-CUDA: +98%
CUDA average: +75%
==================================
This might've been #'s specific to Moonshot Free Energy WUs?
Good indication that CUDA will mean a significant boost for Nvidia gear in the near future. Much higher than I expected.  The exceptional boost for the Debian-CUDA combo, has me planning to re-install a Mint system.  I now have HTM running and can make meaningful comparisons, having logged data for nearly 1K WU this month.   The AMD core update is was scheduled to arrive 9/18, so hang on.

cine_chris

JADO said:

cine_chris..where do you get the data you report.  I don't see it on the FAH web statistics page?  Thanks

I plucked this link from the Discord server:
https://foldingforum.org/viewtopic.php?p=343068#p343068
I'm still trying to get a feel for how they function, Chodera is always very open, but others can be hostile.
CUDA is an excellent example, I couldn't get any info last week, other than "very soon", same answer for many weeks.  Turns out that soon, was beta testing.   Perhaps, he listened?  As this is the exact type of info I was seeking, 'what can we expect to see'.  As boring as Folding is, some good news helps to keep people attached.

cine_chris

JADO said:

cine_chris..where do you get the data you report.  I don't see it on the FAH web statistics page?  Thanks

Perhaps you meant the tabular data?
Extreme Overclocking, I made a short link
https://bit.ly/MCOstats
The EOC data is updated every 3hrs CentralST & 15-20min to complete.

cine_chris

RDNA2 is still an unknown, but could be a piece of the puzzle too.  

0legravy_leg

How many PPD does a thread ripper generate (which one)?  I have been only running GPU folding, the power draw and heat output into my loops would actually slow the GPUs down and the added PPD of the CPUs would just barely make up for that loss.   

cine_chris

0legravy_leg said:

How many PPD does a thread ripper generate (which one)?  I have been only running GPU folding, the power draw and heat output into my loops would actually slow the GPUs down and the added PPD of the CPUs would just barely make up for that loss.   

GPUs are more efficient.  The 3070/80's, when we can get them will pack some punch. Based on the data above, I'm looking at migrating to Linux folders in Oct.  I'll be ready  for CUDA's arrival, that'll be a nice addition.  I'm struggling to make my Sept goal of Top1K, then 1B.  There are many small clusters passing me that keep moving the Top1K spot higher, it's currently moving at ~1.7M PPD on average.

cine_chris

Latest Client Beta test #s
OpenCL is what the current FAHclient is running, I calculated % gain w/ CUDA
Updated with PPDs - project 11765 :

TPF 73s - GTX 1080Ti running OpenCL/ 1.554 M PPD

TPF 57s - GTX 1080Ti running CUDA / 2.253 M PPD   45%

TPF 49s - RTX 2080Ti running OpenCL/ 2.826 M PPD

TPF 39s - RTX 2080Ti running CUDA / 3.981 M PPD  41%

TPF 36s - RTX 3080 running OpenCL / 4.489 M PPD

TPF 31s - RTX 3080 running CUDA / 5.618 M PPD  25%

TPF 30s - RTX 3090 running OpenCL / 5.901 M PPD

TPF 26s - RTX 3090 running CUDA / 7.314 M PPD  24%

teamtempest

Should we congratulate ourselves for contributing over 1M work units? I personally account for 0.1% of that!

But I see former team members with more points than I have with only a fraction of the number of work units - one in particular, 8M points but only 82 work units. A lot of non-active team members have disparities like that, though not all so extreme. Is there something frustrating about so many points and so few work units completed? Would more people stick around if more effort was spent on matching jobs to capabilities?

ryau

Today F@H just announced a speed increase for those running NVIDIA cards. 15%-30% increase for typical WUs and up to 400% increase for Moonshot:

"To see these speed boosts, you won’t have to do anything—the new 0.0.13 release of core22 will automatically roll out over the next few days on many projects, automatically downloading the CUDA-enabled version of the core and CUDA runtime compiler libraries needed to accelerate our code. If you have an NVIDIA GPU, your client logs will show that the 0.0.13 core will attempt to launch the faster CUDA version."

https://foldingathome.org/2020/09/28/foldingathome-gets-cuda-support/

cine_chris

@teamtempest
Any milestone is worth a celebration, some bigger than others, every day Folding is a mini-milestone. 
A 'smaller' GPU or CPU can churn away on a WU for many hours and at some point, people feel they aren't making an impact.  After researching this, it actually starts to occur at 1MPPD.  At  <500K PPD attachment to the project fades quickly.  
It's a complex topic.
Currently, there are so many powerful, huge clusters supporting FAH that people shouldn't feel frustration, angst or guilt about any contribution they are making.  During the surge, Anonymous went from 13MPPD to 10BPPD (mostly clusters).  Any contribution is enough.
I'm hoping to see MCO grow organically and looking at ways to give visibility to everyone.  Like a suggestion to the EOC site to show total group contribution at all levels.  As the view we see, Cumulative Points, Top Ranks, etc is misleading.  Adding a completed WU column is a good idea too.

cine_chris

Got my first CUDA core WUs: 2060s@2MPPD  1660Ti@1.16MPP
UPDATE: Going back through HFM logs, I've been getting core13 CUDA WU since 9/25.  That explains the jumps I was seeing, DOH!  Shows 31 WUs all 1448# projects.  So, apparently, the project needs to be CUDA aware.   I'm showing the 2070s as high as 2.87MPPD with a range of 2.5-2.8M PPD.  All data is Win10.
UPDATE: currently seeing roughly a 50-60% boost with CUDA

Yellowbillycat

cine_chris said:

Got my first CUDA core WUs: 2060s@2MPPD  1660Ti@1.16MPP
UPDATE: Going back through HFM logs, I've been getting core13 CUDA WU since 9/25.  That explains the jumps I was seeing, DOH!  Shows 31 WUs all 1448# projects.  So, apparently, the project needs to be CUDA aware.   I'm showing the 2070s as high as 2.87MPPD with a range of 2.5-2.8M PPD.  All data is Win10.

This post made me look and while I watched...

"21:52:59:WU01:FS01:0x22:  Configuring platform CUDA

21:53:05:WU01:FS01:0x22:  Using CUDA and gpu 0"

2070 running at 2.4M PPD.  I'm no expert, but that's a lot more than without CUDA!

MightyMayfield

I've got a few 2080ti's showing 3.9mPPD so far but most machines aren't affected yet.

0legravy_leg

I have been super busy with work so I have not really been watching too closely but at some point today i saw my 2080ti at around 4.5m ppd.. it was in the 3.9m range most of September which was disappointing after being in the 4.2 range for most of July.  

kmiller922

@cine_chris congratz on cracking the 1k markaand @MightyMayfield 186 to go i'm sure you two will get there way before me

ryau

MightyMayfield said:

I've got a few 2080ti's showing 3.9mPPD so far but most machines aren't affected yet.

Did you update the NVIDIA drivers?
"To get the most performance out of the new CUDA-enabled core, be sure to update your NVIDIA drivers! There’s no need to install the CUDA Toolkit."

cine_chris

kmiller922 said:

@cine_chris congratz on cracking the 1k markaand @MightyMayfield 186 to go i'm sure you two will get there way before me

Thanks for the mention!
It was a BIG Folding weekend for me.  Top1K, then 1B points and then the CUDA arrival.
You can't go to the pub and celebrate with like-minded folders, there aren't any. LOL
It's also 20 years for Folding too!