2019 Formula Boinc Sprints

[Orange Kid]

Moo!Wrapper day one.
A very respectable 3rd place for us. Crunchers at EVGA are close on our heels though. We seem so be missing a few key players these days. Are they wit us or have they moved on to greener fields? Time will tell. Lettuce hope they are bagg'n.



From FreeDC;

 

[Ken g6]

FYI, aborting CUDA 3.1 work so you can get more of a better project results in you getting...more CUDA 3.1 work because the project thinks an error occurred.

 

[Kiska]

*sigh*
I am finding moo very annoying to work with:

Timeout is 3 mins...

 

[Orange Kid]

Moo! Wrapper day two.
In third and putting some space between us and C@EVGA. Are there bunkers in play on this sprint? Or will it all just flow nicely to the end? Skillz is in, but where are the lost boyz?



From FreeDC

 

[Orange Kid]

FYI, aborting CUDA 3.1 work so you can get more of a better project results in you getting...more CUDA 3.1 work because the project thinks an error occurred.

My one that was on cuda switched to open_cl all by it's self. I have my days set to .1 and .1. Maybe the minimal task setting has something to do with it. I only have one GPU per machine, just another thought.

 

[lane42]

got a bunch of cuda31 work again, o well.

 

[crashtech]

I didn't know that one kind of task was any better than another on this project.

 

[lane42]

I don't know crash. The cuda 31 work use's 4 gpu's on 1 task, so 1 at a time. faster.
The other workunits use 1 gpu per task, so 4 at a time. slower.
That's on a 4 gpu rig. So I don't know what is better.

 

[Ken g6]

Well, my 1060 can do one OpenCL faster than it and my 750 Ti can do a CUDA 3.1 together.

 

[Orange Kid]

The problem with two different cards running one task is that it runs at the lowest cards speed. So effectively you are running two 750ti's.

 

[Ken g6]

The problem with two different cards running one task is that it runs at the lowest cards speed. So effectively you are running two 750ti's.

I rather doubt that's correct. I found an app_config.xml file to run on one card per task, so we'll see.

 

[biodoc]

On one of my GTX 980's , the cuda 3.1 tasks take around 1450 seconds and the cuda 6.7 tasks take around 1000 seconds and opencl tasks are pretty close at around 900 seconds.

 

[Orange Kid]

I rather doubt that's correct. I found an app_config.xml file to run on one card per task, so we'll see.

When running a Cuda task using both cards, yes, that is correct. That is how the dnet client works. If you can run one task per GPU then they will run at card speed.

 

[Ken g6]

Running a CUDA 3.1 task on both my cards took 1250 seconds. Running one on just the 1060 took 1500 seconds. I think both cards work as well as they can for the client when used together. Of course, OpenCL is better.

 

[Ken g6]

Running a CUDA 3.1 task on both my cards took 1250 seconds. Running one on just the 1060 took 1500 seconds. I think both cards work as well as they can for the client when used together. Of course, OpenCL is better.

...Or the app_config didn't actually tell the CUDA code to use one GPU, and it's using both along with an OpenCL task and just muddying the issue.

Still, slowing the slowest card a lot didn't slow the work much.

 

[IEC]

Or you can be like me and run matched dual GPU hosts so you don't worry about "slower" cards

 

[Orange Kid]

Moo! Wrapper Final.
Well, a few MOAB's and as the dust settles, we finish in second!
A fine race all and thanks to all large and small.



From FreeDC



...and le France F1 results

• Lewis Hamilton 1:24:31.198
• Valtteri Bottas +18.056
• Charles Leclerc +18.985

 

[ao_ika_red]

Competition heats up once certain profit-based team wasn't on this race. And some houses are heating up as well, due to the workload. I'm glad having sprint on this time of the year when southern, cooler air, is still breezing (thank you, Kiska, I guess).
And ever since that farce on CanadaGP, I finally ditched F1 for good and having better weekend since. Because now I can spend more time doing my hobbies.

 

[bill1024]

Congratulations on 2nd place in the sprint. By quite a margin too.
Good job!

 

[RFGuy_KCCO]

Nice job in the sprint. Thank you for the assistance in the end. Much appreciated.

 

[zzuupp]

So that I remember to check tomorrow for the announcement of the next one, I'll bookmark this post.

 

[ao_ika_red]

Less than 12 hours to upcoming Sprint project announcement.

 

[Ken g6]

Well, some Einstein decided to run another GPU project.

 

[emoga]

Packed my Tahiti's away just the other day...I guess they're needed once more this summer.
Luckily its raining out and the basement is kind of cool. This will change.

 

[ao_ika_red]

Finally, a sprint without any major hiccup (hopefully).