2019 Formula Boinc Sprints

[Orange Kid]

Asteroids@home day one.
Had to stretch it out another place so we would show up. Eleventh place.
Guess we should have kept all our old hardware, then we could be doing better.
I sure we'll inch up a spot or two before it's over. We have maxed out our placement in the marathon for this year most likely as it is.




From FreeDC

 

[crashtech]

Yeah, maybe we should give LAF a chance...

 

[Orange Kid]

Asteroids@home day two.
In tenth now. Moving up slowly. How high can we go?



From FreeDC

 

[Howdy]

Sadly (I guess) I'm/we are crippled because GPU technology has advanced, yet projects do not advance at the same rate.

 

[Ken g6]

Sadly (I guess) I'm/we are crippled because GPU technology has advanced, yet projects do not advance at the same rate.

Asteroids is mostly a CPU project anyway. I set a couple of 3-core DigitalOcean cloud instances on it. Each instance does about as much work as my GTX 1060.

We just have too many CPU cores on WCG.

 

[StefanR5R]

Hehe, I completed¹ 3 tasks so far. One is showing in Orange Kid's stats, another one was validated after FreeDC took its snapshot, and the 3rd is still pending validation.

Edit, looking more carefully, I see that both validated results are in O.K.'s stats already.

________
¹) The technically more correct term is "reported".

 

[ao_ika_red]

Hehe, I completed¹ 3 tasks so far. One is showing in Orange Kid's stats, another one was validated after FreeDC took its snapshot, and the 3rd is still pending validation.

________
¹) The technically more correct term is "reported".

This is like an annoucement for "storm is coming".

 

[TennesseeTony]

I have shut down the 196 threads of nearly 10 year old CPUs. 423K ppd is impressive, but not 2600 watts impressive.

Now that my WCG bunkers are no longer "networking suspended" I have of course replaced the CPU power with 1620 watts of GPU power.....we shall see how that goes as to ppd. Asteroids is still most profitable with CPUs, however, it would appear.

 

[StefanR5R]

storm is coming

Yeah, it'll be a tempest. A tempest in a teapot.

Among written notes of mine from a year or two ago, I found the datum that a 1080Ti completes a job at 1/10th the time of a CPU job. I think this relates to Xeons at low clocks and with Hyperthreading. Which places the Asteroids performance of a GPU of mine between a 4-core and a 6-core low-power Xeon.

 

[TennesseeTony]

Yep, still a CPU project. I do love the science though!

 

[Howdy]

If it's better for CPU then GPU, I'm doing something wrong then. (imagine that) GPU for me is 3:1 vs CPU with the machine I have running it. Is there an app_config I'm supposed to be using?

 

[StefanR5R]

GPU for me is 3:1 vs CPU with the machine I have running it.

If you operate a Quadro GP100 in a Pentium 4, this could explain it. :-)

Are you comparing the GPU with the entire CPU, or with one thread of the CPU?

Is there an app_config I'm supposed to be using?

I am not aware of any.

 

[Howdy]

LOL funny guy funny funny guy!! 1 GPU VS 1 CPU thread...........

1070- 10-17 minutes x2 GPUs
CPU- 1.75-2.5 hours @ 1 for each of 12 threads

 

[StefanR5R]

OK, so the 6c/12t CPU is producing

12 simultaneous_threads * 24 h/d / (2.5 h...1.75 h/task)​

= 110...160 tasks/day,​

while one GPU does

1 simultaneous_thread * 24 h/d * 60 min/h / (17...10 min/task)​

= 80...140 tasks/day,​

right?

 

[StefanR5R]

PS, leading the top hosts is a Ryzen 9 3900X of a SETI.Germany guy who is AWOL from the B'day. This host is taking at the order of 1 hour per task (with large +/- variation, as is apparently common with current job batches).

 

[Howdy]

OK, so the 6c/12t CPU is producing

12 simultaneous_threads * 24 h/d / (2.5 h...1.75 h/task)​

= 110...160 tasks/day,​

while one GPU does

1 simultaneous_thread * 24 h/d * 60 min/h / (17...10 min/task)​

= 80...140 tasks/day,​

right?

Well I guess when you put it that way..........All is good

 

[lane42]

I guess my little 6 core i5 coffee lake is as fast as some 12 thread cpu's.
Can that be. iam seeing times of an hour or less.

 

[TennesseeTony]

Oh YES! I have a Skylake XEON quad and even with single channel RAM it is a beast!

I think the new AMD 3800X is doing tasks in 50 minutes. But Stefan has pointed out there are at least two varieties of task, long and short....not sure which I saw running.

 

[Howdy]

I must be getting all the long tasks and taking it in the short(s)

I might have to back off though, temps are getting kind of high.

 

[TennesseeTony]

Lol you silly goose, less than 40C....

I watched Joker 2019 today, are you trying to be funny?


The scores do not seem to update hourly, for Asteroids. Maybe every two hours at best. Let's hope the last hour gets an update.

Oh Joy and Happiness, 800,000 points waiting for validation! Ah well, it was a good run.

 

[ao_ika_red]

I must be getting all the long tasks and taking it in the short(s)

I might have to back off though, temps are getting kind of high.

View attachment 13316

That 34°C is lower than my afternoon room temperature.

 

[Howdy]

That 34°C is lower than my afternoon room temperature.

Ut oh I had a spike, furnace came on!!



This is only Asteroids though, PG LLR is a whole different story

And yes I watch my temps too much, when they get 55*-60*C + I get "irritated" break out a fan, open the case .................OCD I guess

 

[Orange Kid]

Seventh place for the TeAm.
Just ahead of SG and LAF to keep them at bay in the marathon, nice work all.
This was a tough race due to the WCG challenge going on, but we finished well. Most likely we'll have more points in the coming days as all the tasks finally get validated than we had during the sprint.




From FreeDC



Brazil F1 results

• Max Verstappen 1:33:14.678
• Pierre Gasly +6.077
• Carlos Sainz +8.896

 

[TennesseeTony]

AAAAAAAHHHHHHHHHH. So close to 6th place!!

Good last minute run to all! And thanks for the Stats, Kid!

 

[ao_ika_red]

break out a fan

Yeah, this fan.

Most likely we'll have more points in the coming days as all the tasks finally get validated than we had during the sprint.

Agreed. Not a single chance those xii5ku and lane42 lads only produced less than 1k.

Brazil F1 results
https://www.newegg.com/delta-afb121...yMark=False&IsFeedbackTab=true#scrollFullInfo

• Max Verstappen 1:33:14.678
• Pierre Gasly +6.077
• Carlos Sainz +8.896

I was too tired to watch the race, but from the news I've read that's one hell of a race. A joy for McLaren fans as well, first podium since 2014.