Brony@Home would like to formally challenge Team AnandTech

[TennesseeTony]

The dates are approaching quickly, thanks for the reminder that we should probably gets some details hammered out.

Absolutely certain we start our TeAm (internal) race on December 1st.

 

[iwajabitw]

Non-stop action, blow them out whilst they run!

Haha, I did that once and burned up a fan. 90psi is too much when there running, from my 30gal air compressor. Plus I like the dust to go flying out in the driveway.

 

[TennesseeTony]

The dates are approaching quickly....we should probably gets some details hammered out.

So....let's start hammering. @petrusbroder is our Folding Race king, so I defer to him for the final answer(s)

Start date/time (for both our internal and external [Brony] race ) is December the 1st at.........???? [UTC] @tiwake can we do 1800UTC? I know you wanted 0000UTC, but 1800 works better for our stats keeper, as he does daily stats around that same time (when he is not on call and saving lives at the hospital).

And the end date/time for the internal competition is ??????

And the end date/time for the Brony competition is ?????? I think we tentatively agreed to two months here. I plan to FOLD for the majority of the winter, plus Mark will, and many of the other regular folders too, so if not everyone can join for the second month, that's fine. Maybe just weekly updates on the scoreboard, petrusbroder, for the second month (2nd), to reduce the workload of doing the stats?

@tiwake be warned, last year we gained 1 Billion points in December.

Here is to wishing both teams the best of luck, and higher points than ever before (for science! of course).

 

[TennesseeTony]

I've messaged Titillating and Ryan Smith to see if they plan to pit us against Tom's Hardware (for charity) again this year. That would be very much fun, again.

Last year's front page announcement(s) added some highly valued new members to our ranks, in addition to some DC 'testers' that hopefully would come back for another few weeks at least.

 

[Pokey]

I expect petrusbroder will start an "official" Folding thread to start things off. Looking forward to seeing all the horsepower that shows up this year.

 

[TennesseeTony]

"The New Boss" replied. A race with Tom's is not off the table, but.....clearly wasn't planned before now. Not likely to happen, although Ryan did express his admiration of our efforts.

 

[crashtech]

I hope the bosses will consider it!

 

[TennesseeTony]

Brony is going ahead with a 2 month race, and with 1800 UTC as the start and end time, December 1st through Feb. 1st.

They even have a cool graph, updated hourly (which will reset once the race begins).

 

[Markfw]

They even have prizes for the winner ! But their head guy has 11 million ppd to my 10 million ppd, so our chances of winning a prize are low (or are we even considered ? Internal prizes ?)

Regardless, let the race begin ! Cancer is the loser ! And I hope for me too ! Chemo is hell.....

 

[crashtech]

Well I'm going to run with what I have, all out of money for more hardware this year. What we really could use are a few more new participants!

 

[Markfw]

Well I'm going to run with what I have, all out of money for more hardware this year. What we really could use are a few more new participants!

Yes, I spent over $8k in the last 2 months for 2 threadrippers and 5 1080TI cards, and now I have over $5000 in medical bills due to my cancer. I don't regret the computer spending due to my condition, but, I am done upgrading this year due to finances.

 

[StefanR5R]

The dates are approaching quickly, thanks for the reminder that we should probably gets some details hammered out.

So....let's start hammering.

Thank you Tony.

Brony is going ahead with a 2 month race, and with 1800 UTC as the start and end time, December 1st through Feb. 1st.

@tiwake, thanks to Brony@Home for being flexible about the start date and time.

They even have prizes for the winner ! [...] Internal prizes ?

Yes, they are for their own internal race.

No bunkers in F@H

That'll inconvenience me personally for sure. But joking aside, what really inconveniences me about Folding@Home is that of all DC projects which I tried so far, F@H is the one which causes the most coil whine on my GPUs.

Well I'm going to run with what I have, all out of money for more hardware this year.

I'm down to 3 GPUs which I can use for Folding. (A mix of GTX 1080 and 1080Ti in a single watercooled host.) Parts are supposed to arrive soon to bring me back up to 5 GPUs, but due to constrained spare time the second host may not be ready on day 1 already.

What we really could use are a few more new participants!

Even getting as many participants on board as last year won't be easy. A spirited front page announcement like last year would be great.

 

[iwajabitw]

I only had 2 cards last year. Have 6 now, but it's mostly a DP fleet, so we will see what it can do. $1200 set of tires, some medical bills has put a crunch on the Christmas funds much less any new hardware. I'm bringing Enigma tasks to a stop so that I will have a few days to test my other cards with F@H.

 

[Markfw]

I tried to register on their site, but could not answer the security question "who is the antagonist that sought to bring the eternal night" (or something like that.) Even after googling, I have no idea.

 

[TennesseeTony]

Try Princess Luna

 

[TennesseeTony]

I only had 2 cards last year. Have 6 now, but it's mostly a DP fleet, so we will see what it can do. .......

Just finishing up a set of tasks on my 280X's, and getting 130,000 ppd per card. I didn't save the math, however some quick calculating (triple checked due to disbelief) showed my 10-series cards give 4.7x more points-per-watt, than the R9-280X fleet. Ouch.

edit: (both test tasks are 9415, I'll need to try and get some different tasks for better comparison)

 

[Markfw]

Try Princess Luna

That worked, I joined.

 

[TennesseeTony]

Thanks for not asking how I know the answer.

Back to my R9-280X's, the next set of tasks were also 94xx series (9413 and 9431), and also are paying about 130,000 ppd. So it would take 4 R9-280X's to get the ppd of a single stock GTX-1070.

 

[crashtech]

Stop it, Tony, most of my GPUs are 7970/280x.

 

[TennesseeTony]

Still, not a bad way to take the chill out of the air!

 

[iwajabitw]

Well that blows about the 280x's ppd for F@H. Guess the GTX 980's will carry the load and I may have to leave the others on MW or Enigma, unless it gets into a tight race.

 

[TennesseeTony]

For direct comparison, a 9415 task running on one of my 1080's is estimating 660,000 ppd. Ouch ouch ouch, and for 30 fewer watts at least.

 

[TennesseeTony]

For clarity: This post is regarding ppd of a GPU, based on the IPC of the CPU (instructions per clock).

The 660,000ppd for the 1080 above on a 9415 task, was with a Xeon e5v3 @2.8GHz. On a consumer chip of the same family, but at 3.3GHz, the ppd was close to 700,000 (40,000 more). I pulled one of ThunderStrike's cards, put it in an AMD FX8320E system at 3.5GHz (but pcie 2.0 slots) and am only seeing 650,000ppd (bus utilization only at 44-48%, so pcie 2.0 vs 3.0 not a factor).

GTX1080 @1848GHz (stock), vs CPU IPC

• e5-2695-v3@2.8GHz..........660k ppd edit: CPU seems to be peaking at 2.7GHz currently
  
• Core i7-5820K@3.3GHz....700k ppd
• AMD FX8320E@3.5GHz....650k ppd edit: as low as 630k ppd, varies as it progresses

Everything except system memory (DDR4 vs 3) was pretty much equal, OS, 1848MHz non-OC'd core clock all around, number of threads free to assist the GPU, etc.

I was really hoping that 3.5GHz older AMD would at least perform as well as the 2.8GHz....because I was thinking of:

• turning ThunderStrike into a Linux machine, CPU tasks only, and
  
• having those 4 (usually idle) GPU's on a dedicated GPU machine, only running when needed/wanted, and
  
• resizing the power supplies to better match the 50-75% load sweet-spot for efficiency's sake.
  
• (8 months out of the year I'm pulling less than 200 watts out of 650/850/850/1200/1200/1375 watt power supplies, lol. )

 

[Markfw]

I don't see CPU making a difference. My 1800X@3.7 is currently giving less ppd than my 2583v3@2.5 on 1070's. Its all over the place. And with boxes that have 2 of the same cards in them, the ppd is quite different between the 2 cards ! Study the belowexample, Threadripper 1 932 vs 983, same box, same unit)

 

[TennesseeTony]

Looks like the 1800X and 2583 are running different tasks though, Mark. ThreadRipper 1, however is on the same task per GPU, and at about the same percentage of completion, but indeed shows 10,000 points difference for just those single tasks.
And that particular task is likely...90 minutes on a Ti?, so that would certainly cause the 50k ppd difference. Interesting.

Once the Sun sets, I'll try getting the GPU to work again on Linux, for my AMD setup, see if that makes a difference, then go back to Windows and toss in another 2 1080s to see if that has any negative effect on that system, strains the BUS, etc.