Times (Elapsed / CPU) for BRP5/6/6-Beta on various CPU/GPU combos - DISCUSSION Thread

MarkHNC
MarkHNC
Joined: 31 Aug 12
Posts: 37
Credit: 170,965,842
RAC: 0

RE: RE: RE: My 1st beta

Quote:
Quote:
Quote:
My 1st beta 1.52 is running on my Intel HD4000

Actually we are not sure yet whether we want to keep serving the rather big BRP6 work units to the Intel iGPUs. At least the less powerful among them like the HD 2500 will take longer to crunch than we usually like tasks to take to complete. It is quite possible that we'll stop BRP6 beta on the Intel iGPUs after some initial tests.

HB

I have an i3 3220/HD 2500 that runs at 100% CPU and 100% GPU nearly 24/7 and that picked up two of these big tasks. I have it setup to run two Einstein tasks at the time, because it was peaking around 70% GPU utilization when running the single-unit Arecibo tasks. I also run three WCG units alongside, mostly FAAH-Vina with the occasional CEP2. So it is now running two of these biggies.

First is 50%, elapsed 23h30m, remaining 11h52m
Second is 48%, elapsed 24h49m, remaining 12h58m

These two finally completed:
First ended up run time 165,685.50, CPU time 1,104.69
Second ended up run time 166,069.17, CPU time 1,105.88

I am now back to (so far) receiving single-unit Arecibo tasks on this machine.

Richard Haselgrove
Richard Haselgrove
Joined: 10 Dec 05
Posts: 2,143
Credit: 2,924,560,825
RAC: 882,603

RE: HD2500 has just 6 EUs,

Quote:
HD2500 has just 6 EUs, compared to 16 in HD4000 and 20 in HD4600.


Doesn't appear to make as big a difference as you might expect. The fastest HD 2500 I've been able to find in the database (I haven't checked them all) is computer 6828436 (Intel(R) Core(TM) i5-3570 CPU @ 3.40GHz), regularly getting below 1Ksec.

Stef
Stef
Joined: 8 Mar 05
Posts: 206
Credit: 110,568,193
RAC: 0

So these are my

So these are my results.
Looking forward for the new cuda version.

ExtraTerrestrial Apes
ExtraTerrestria...
Joined: 10 Nov 04
Posts: 770
Credit: 569,010,339
RAC: 149,837

RE: Doesn't appear to make

Quote:
Doesn't appear to make as big a difference as you might expect. The fastest HD 2500 I've been able to find in the database (I haven't checked them all) is computer 6828436 (Intel(R) Core(TM) i5-3570 CPU @ 3.40GHz), regularly getting below 1Ksec.


Impressive performance of the small HD2500 and actually confirms my suspicion that 1.34(*) was/is limited elsewhere. Main memory bandwidth is a prime candidate, especially with all those copy operations taking place in 1.34. The CPU L3$ size may also play a significant role, as the fastest memory access is always the one avoided. The i5 has 6 MB, whereas Mark's i3 has just 3 MB.

@MarkHNC: are you running dual channel memory?

@HBM: I can has Intel 1.52?

MrS

(*) Just replace 1.39 in my previous post with 1.34 - thanks for spotting!

Scanning for our furry friends since Jan 2002

Harri Liljeroos
Harri Liljeroos
Joined: 10 Dec 05
Posts: 4,196
Credit: 3,115,086,992
RAC: 1,840,389

Is the application name for

Is the application name for version 1.52 identical for Intel, Nvidia and AMD GPU? I haven't so far received any 1.52 BRP6 for my Intel but I am doing 3 at a time on my GTX970. If the application name is same for Intel my app_config makes Intel (HD4000) also try 3 at a time. This will surely fail and may crash a whole lot more WUs on other projects too (like CPDN).

So please make Intel applications have a different application name so we can easily experiment separate settings for them.

floyd
floyd
Joined: 12 Sep 11
Posts: 133
Credit: 186,610,495
RAC: 0

RE: So please make Intel

Quote:
So please make Intel applications have a different application name so we can easily experiment separate settings for them.

You don't need different application names, you can distinguish the versions by plan class.

Holmis
Joined: 4 Jan 05
Posts: 1,118
Credit: 1,055,935,564
RAC: 0

RE: So please make Intel

Quote:
So please make Intel applications have a different application name so we can easily experiment separate settings for them.


The plan classes are different and that's enough to separate them in app_config.xml.
I use the following to run x2 on my Nvidia card and only one task on my Intel iGPU:

[pre]

einsteinbinary_BRP6
BRP6-Beta-cuda32-nv301
0.5
0.5


einsteinbinary_BRP6
BRP6-Beta-opencl-intel_gpu
1.0
1.0

[/pre]
The difference is that you change the -tag to and insert a -tag.
Then for some reason the Boinc developers choose to rename the and tags to and .
It's all documented on this page.

MarkHNC
MarkHNC
Joined: 31 Aug 12
Posts: 37
Credit: 170,965,842
RAC: 0

RE: RE: Doesn't appear to

Quote:
Quote:
Doesn't appear to make as big a difference as you might expect. The fastest HD 2500 I've been able to find in the database (I haven't checked them all) is computer 6828436 (Intel(R) Core(TM) i5-3570 CPU @ 3.40GHz), regularly getting below 1Ksec.

Impressive performance of the small HD2500 and actually confirms my suspicion that 1.34(*) was/is limited elsewhere. Main memory bandwidth is a prime candidate, especially with all those copy operations taking place in 1.34. The CPU L3$ size may also play a significant role, as the fastest memory access is always the one avoided. The i5 has 6 MB, whereas Mark's i3 has just 3 MB.

@MarkHNC: are you running dual channel memory?

@HBM: I can has Intel 1.52?

MrS

(*) Just replace 1.39 in my previous post with 1.34 - thanks for spotting!

Yes: 2 x 4GB PC3-12800

Harri Liljeroos
Harri Liljeroos
Joined: 10 Dec 05
Posts: 4,196
Credit: 3,115,086,992
RAC: 1,840,389

Thank you for the plan_class

Thank you for the plan_class information. I'll give it a try when I get home.

Edit: Where you can find the valid plan_classes?

Richard Haselgrove
Richard Haselgrove
Joined: 10 Dec 05
Posts: 2,143
Credit: 2,924,560,825
RAC: 882,603

RE: Thank you for the

Quote:

Thank you for the plan_class information. I'll give it a try when I get home.

Edit: Where you can find the valid plan_classes?


If you look on the project's applications page, the plan class names are shown in brackets after the version number.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.