Frame times on A5 BigAdv WUs

The most demanding Projects are only available to a small percentage of very high-end servers.

Moderators: Site Moderators, PandeGroup

Re: Frame times on A5 BigAdv WUs

Postby muziqaz » Fri Apr 15, 2011 9:31 pm

Used the bug in rev 24 of v7 and downloaded a5 unit for my Phenom x6@3.9ghz. Well I suppose when you specify more cores than you have to the client, performance tanks down a lot. I just finished 1st frame and it took me 59mins per frame :D
It was just an experiment anyway. Will leave it for couple more frames, and then back to beta testing.
Signature is too distracting
User avatar
muziqaz
 
Posts: 588
Joined: Sun Dec 16, 2007 6:22 pm
Location: London

Re: Frame times on A5 BigAdv WUs

Postby orion » Fri Apr 15, 2011 11:42 pm

Muz, what project did you get?

I got a 6900 on v7.1.21 with a TPF 36 minutes. That's 6 minutes slower than running a 6900 on v6.
iustus quia...
User avatar
orion
 
Posts: 285
Joined: Sun Dec 02, 2007 12:45 pm
Location: neither here nor there

Re: Frame times on A5 BigAdv WUs

Postby muziqaz » Sat Apr 16, 2011 12:09 am

6901
User avatar
muziqaz
 
Posts: 588
Joined: Sun Dec 16, 2007 6:22 pm
Location: London

Re: Frame times on A5 BigAdv WUs

Postby Athlonite » Wed Apr 20, 2011 1:40 pm

muziqaz wrote:Used the bug in rev 24 of v7 and downloaded a5 unit for my Phenom x6@3.9ghz. Well I suppose when you specify more cores than you have to the client, performance tanks down a lot. I just finished 1st frame and it took me 59mins per frame :D
It was just an experiment anyway. Will leave it for couple more frames, and then back to beta testing.


so that would be about 4 days 9hrs to complete then roughly
Asus M23A32MVP-Deluxe AMD Phenom II x4 940BE @ 3.41GHz
2xCorsair Dominator twin2x2048 PC8500C5D DDR2-1066MHz 5-5-5-15 CL2
2x HIS ATI Radeon HD5770Fan 1GB @ 940/1275
SilverStone RV02B-W case & ST75F-P 750W full modular PSU
Windows7 X64
Athlonite
 
Posts: 91
Joined: Fri Dec 18, 2009 12:52 am
Location: Napier, NZ

Re: Frame times on A5 BigAdv WUs

Postby tear » Wed Apr 20, 2011 2:01 pm

Muziqaz, if you'd like to experiment, The Kraken has NT override feature for that specific use case.
One man's ceiling is another man's floor.
Image
tear
 
Posts: 857
Joined: Sun Dec 02, 2007 4:08 am
Location: Rocky Mountains

Re: Frame times on A5 BigAdv WUs

Postby muziqaz » Wed Apr 20, 2011 5:10 pm

Tear, thanks, but I use that PC for other stuff, so I might hurt the project. If it was dedicated, I would consider. Waiting for bulldozer now.
User avatar
muziqaz
 
Posts: 588
Joined: Sun Dec 16, 2007 6:22 pm
Location: London

Re: Frame times on A5 BigAdv WUs

Postby jimerickson » Wed Apr 20, 2011 5:55 pm

tear
just started using The Kraken. thanks a bunch!!
jimerickson
 
Posts: 682
Joined: Tue May 27, 2008 11:56 pm
Location: ames, iowa

Re: Frame times on A5 BigAdv WUs

Postby BaBa » Wed Apr 20, 2011 6:56 pm

Hydra's back and she is bad!

Thanks again to tear, sfield and blackshard for making this possible.

Project: 6901
Average time/frame: 00.04.58 Theoretical PPD 550k+
CPU: Opteron 6172 engineering sample @ 2.9Ghz
# of CPU sockets:4
# of Physical cores:48

# of FAH CPU processes:1
# of FAH GPU Clients:0

RAM GB installed:32
RAM Type: DDR3
RAM Speed:1333 6-7-5-20 northbridge 2200
NUMA enable +The Kraken

OS name/kernel version Ubuntu 10.10 server running native


Still more to come from the ram i think and once cooling is better maybe 3Ghz :mrgreen:
BaBa
 
Posts: 13
Joined: Thu Dec 23, 2010 7:40 pm

Re: Frame times on A5 BigAdv WUs

Postby [Ars] For Caitlin » Fri Apr 22, 2011 1:44 pm

Stock HP ProLiant DL580 G7

Project: 6901
Average time/frame: 00:08:50
CPU: Intel Xeon "Beckton" X7542 @ 2.66GHz
# of CPU sockets: 4
# of Physical cores: 24
# of FAH CPU processes: 24
# of FAH GPU Clients: 0

RAM GB installed: 64
RAM Type: DDR3 DIMM
RAM Speed: 1066
NUMA enabled/disabled/not used: Enabled

OS name/kernel version: RHEL5 2.6.18-238.9.1.el5
Client: Linux 6.34
Running in VM: No
[Ars] For Caitlin
 
Posts: 47
Joined: Sun Jan 06, 2008 11:06 pm

Re: Frame times on A5 BigAdv WUs

Postby foo_fight » Sun Apr 24, 2011 8:51 am

Project: P6901
Average time/frame: 00:12:30
CPU: Xeon X5650 @ 3,4 GHz
# of CPU sockets: 2
# of Physical cores: 12

# of FAH CPU processes: 24
# of FAH GPU Clients: 0

RAM GB installed: 12 GB
RAM Type: DDR3 PC12800
RAM Speed: 1780 MHz C8
NUMA enabled/disabled/not used : not used

OS name/kernel version: Ubuntu 10.10 / Kernel 2.6.35 CK
Client: Linux 6.34
Running in VM: No
Alliance francophone member (51)
foo_fight
 
Posts: 17
Joined: Mon Aug 17, 2009 9:29 am
Location: France

Re: Frame times on A5 BigAdv WUs

Postby muziqaz » Sun Apr 24, 2011 8:56 am

BaBa wrote:Hydra's back and she is bad!

Thanks again to tear, sfield and blackshard for making this possible.

Project: 6901
Average time/frame: 00.04.58 Theoretical PPD 550k+
CPU: Opteron 6172 engineering sample @ 2.9Ghz
# of CPU sockets:4
# of Physical cores:48

# of FAH CPU processes:1
# of FAH GPU Clients:0

RAM GB installed:32
RAM Type: DDR3
RAM Speed:1333 6-7-5-20 northbridge 2200
NUMA enable +The Kraken

OS name/kernel version Ubuntu 10.10 server running native


Still more to come from the ram i think and once cooling is better maybe 3Ghz :mrgreen:


I think you said you had it folding at 3.1ghz in other forum. Or was it unstable?
Imagine 16c Bulldozer with turbo core at similar clocks :D
User avatar
muziqaz
 
Posts: 588
Joined: Sun Dec 16, 2007 6:22 pm
Location: London

Re: Frame times on A5 BigAdv WUs

Postby BaBa » Mon Apr 25, 2011 5:09 pm

Not had her upto 3.1 (yet) i think that was Patriot over at [H]ard with
his dual setup.
BaBa
 
Posts: 13
Joined: Thu Dec 23, 2010 7:40 pm

Re: Frame times on A5 BigAdv WUs

Postby Alpineb1 » Wed Apr 27, 2011 12:29 am

Project: P6901
Average time/frame: 00:22:32 57 577,0 PPD
CPU: I7 2600K @ 4.9 GHz
# of CPU sockets: 1
# of Physical cores: 4 + 4HT

# of FAH CPU processes: 8
# of FAH GPU Clients: 0

RAM GB installed: 8 GB
RAM Type: DDR3 PC12800
RAM Speed: 1600 MHz 9/9/9/24
NUMA enabled/disabled/not used : not used

OS name/kernel version: 7 Ultimate 64bits
Client: 6.34
Running in VM: Yes


and the second one:


Project: P6901
Average time/frame: 00:23:14 - 54 994,5 PPD
CPU: I7 2600K @ 4.6 GHz
# of CPU sockets: 1
# of Physical cores: 4 + 4HT

# of FAH CPU processes: 8
# of FAH GPU Clients: 0

RAM GB installed: 4 GB
RAM Type: DDR3 PC12800
RAM Speed: 1600 MHz 9/9/9/24
NUMA enabled/disabled/not used : not used

OS name/kernel version: 7 Ultimate 64bits
Client: 6.34
Running in VM: Yes
Image
I7 2600k|P67A-UD4-b3|12G|Corsair AX1200|Asus GTX560 DCU TOP|
I7 2600k|P67A-UD4-b3|8G|Corsair AX850|Asus 8600GT|
2X AMD Opteron 6128 8 cores|8G|ASUS KGPE-D16|Noctua NH-U9D0
Alpineb1
 
Posts: 45
Joined: Mon Jun 08, 2009 1:11 pm
Location: Montréal

Re: Frame times on A5 BigAdv WUs

Postby [Ars] For Caitlin » Wed Apr 27, 2011 11:45 pm

One third more physical cores running at two thirds the clock (compared to above) results in equivalent performance. HT adds nothing. Top shows %sy roughly 4X on this box. Thread management/locking?

Stock HP ProLiant DL580 G7

Project: 6901
Average time/frame: 00:08:50
CPU: Intel Xeon "Beckton" X7550 @ 2.00GHz
# of CPU sockets: 4
# of Physical cores: 32
# of FAH CPU processes: 64
# of FAH GPU Clients: 0

RAM GB installed: 384G
RAM Type: DDR3 DIMM
RAM Speed: 1333
NUMA enabled/disabled/not used: Enabled

OS name/kernel version: RHEL5 2.6.18-238.9.1.el5
Client: Linux 6.34
Running in VM: No
[Ars] For Caitlin
 
Posts: 47
Joined: Sun Jan 06, 2008 11:06 pm

Re: Frame times on A5 BigAdv WUs

Postby bruce » Thu Apr 28, 2011 2:35 am

[Ars] For Caitlin wrote:One third more physical cores running at two thirds the clock (compared to above) results in equivalent performance. HT adds nothing.



On a scale of 0 to 10, where 0 is false and 10 is true, you're about at an 8 with those statements.

In an ideal world, one third more physical cores running at two thirds the clock can never be quite as fast as the original, though the difference may be negligible. If any of the cores gets interrupted, the fast ones still have to wait for the slow ones.

HT does add something every time there is a pipeline stall. Again, that may be small in comparison to the total, but "nothing" is an exaggeration. Typical SMP or Uniprocessor Gromacs results are often as much as 10%-20% but nowhere near the 100% that you'd get from doubling the number of independent CPUs.
bruce
 
Posts: 22437
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

PreviousNext

Return to SMP with bigadv

Who is online

Users browsing this forum: No registered users and 3 guests

cron