The Scottish Boinc Team

http://boincstats.com/en/stats/challenge/team/chat/800

Name Compete against Gridcoin: DDM Monthly Race #001
Status Upcoming
Project DistributedDataMining
Issued by gridcoin
Start time 2016-07-17 00:00 UTC
End time 2016-07-24 00:00 UTC
Late entrants allowed? Yes
Number of teams participating » 8
Number of users participating 0

I'm running this for TSBT in case anyone else wants to jump in.

I was going to help SETI.USA but they appear to be easily running away with it.

scole250 wrote:I'm running this for TSBT in case anyone else wants to jump in.

Bowmore said he was going to come along for the ride ... assuming he isn't needed elsewhere

I'd love to see GC finish 3rd in their inaugural challenge! If they continue the way they are that could actually happen. The big boys are asleep.

Just as two days ago at the same time, the challenge stats have been borked...but now those vandals seem to have really wrecked the stats

@Scole, don't back off ... we are closing on them. I just got the 2nd CPU running so that'll add another 100k/day to the good guys.

@Dirk, remember our conversation about being at opposite ends of the spectrum on HW ... well I pulled the trigger

I couldn't get any work for a while. I have work now and everything is wide open.

Bryan wrote:@Dirk, remember our conversation about being at opposite ends of the spectrum on HW ... well I pulled the trigger

I can't exactly pull the trigger, but I managed 16th place so far. I look forward to your performance next -YAFU- challenge, as they have such nasty multi-thread apps nowadays.

scole250 wrote:I couldn't get any work for a while. I have work now and everything is wide open.

I took the time to load up some VMs ... only on the 24 thread and higher machines. So the next time it craps out I can switch over. Right now we are out pulling GC so we might actually be able to pull it off. Wouldn't that be something - squash the borg collective.

@Dirk, I won't be running the Yafu challenge. After hitting 130M on the project I think I'm done for life. While I was waiting for dDM work I have to admit I started up 4 16t tasks at the same time ... just to see what happened. It actually worked

BTW, something I just found out, Windoze only supports 64 threads. It has to do with memory management (NUMA). Linux will run all 72 threads with no problem. So the new machine will spend most of its life on Linux.

Bryan wrote:BTW, something I just found out, Windoze only supports 64 threads. It has to do with memory management (NUMA). Linux will run all 72 threads with no problem. So the new machine will spend most of its life on Linux.

This is golden You of course know that the newest and best Xeons now have 24 cores and 48 threads, while the yet-to-come Zen Opterons even are rumoured to have 32 cores and 64 threads?

The near future will be running Windoze VMs under Linux... Until Windoze 11 comes out (or IBM extends its grip into the x86 server market with a new version of VM, VM/x64 or so).

Dirk Broer wrote: You of course know that the newest and best Xeons now have 24 cores and 48 threads, while the yet-to-come Zen Opterons even are rumoured to have 32 cores and 64 threads?

And you of course know how much those suckers cost I'll stick with my baby 18C/36T E5-2695 V4s They are good enough for "hobby" grade. They are slow but there is a bunches of them!

Bryan wrote:BTW, something I just found out, Windoze only supports 64 threads. It has to do with memory management (NUMA). Linux will run all 72 threads with no problem. So the new machine will spend most of its life on Linux.

What version of Windows and what is the reason?

scole250 wrote:

Bryan wrote:BTW, something I just found out, Windoze only supports 64 threads. It has to do with memory management (NUMA). Linux will run all 72 threads with no problem. So the new machine will spend most of its life on Linux.

What version of Windows and what is the reason?

Any version. NUMA is non-uniform memory access and is the "new" memory architecture. Memory is assigned to processors as "nodes". MS allows for 64 thread support. If you are writing SW you could set each processor as it own NODE and assign memory to each one. The problem would then be that you would have to run 2 versions of BOINC with each one being assigned to a different node.

MS NUMA

It's kinda hard to catch GC when you can't turn in completed work or get new WU. They may hold onto their number 2 spot because of the project. Dang, we would have passed them

I call shenanigans!

Still two days and 6 hours to go...

five hours to go, but you boys have a lead on the Borg since two hours ago

Dirk Broer wrote:five hours to go, but you boys have a lead on the Borg since two hours ago

If the project stays up the Borg may get their collective butts kicked by the small but dedicated TSBT crunchers - all 3 of us

Obviously GC only crunches for money and not for team honor. Where's the fun in that?