Frame Times on BigAdv A3 projects.

[Alpineb1]

here another one

Average time/frame: 46mins 37sec{in hh:mm:ss} (R6,C3,G2)
CPU: _I7 860 @ 3.8 GHz
# of CPU sockets: 1
# of Physical cores: 4
# of Physical cores + HT :8

# of FAH CPU processes: 8
# of FAH GPU Clients: 1

RAM GB installed: 8
RAM Type: DDR3
RAM Speed: 1600
OS / Linux kernel: Win Seven 64 bits Ultimate
Running in VM: Yes (linuxrouter image 1.2 in VM player)

[Grandpa_01]

some posters need to recheck what they have posted (like last one 3.1Ghz it should be 4.1 and the 980x at 1.3) as long as Grandpa spots that be funny if they ended up in there as well (none of my systems are running a3 at the moment so i guess they pushing A2 so they run out of work units)

I try to catch the mistakes as I go hopefuly I won't make too many myself.

[leexgx]

Average time/frame: 00:16:07 {in hh:mm:ss}
CPU: 8439 @ 2.8 GHz
# of CPU sockets: 4
# of Physical cores: 24

# of FAH CPU processes: 23
# of FAH GPU Clients: 0

RAM GB installed: 16
RAM Type: {DDR/DDR2/DDR3} DDR2
RAM Speed: 667
OS / Linux kernel: 2.6.31-20
Running in VM: {Yes/No} No

got to give that something thats just mental time frames

[Zagen30]

Average time/frame: 00:46:57 {in hh:mm:ss}
CPU: i7 930 @ 3.8 GHz
# of CPU sockets: 1
# of Physical cores: 4

# of FAH CPU processes: 8
# of FAH GPU Clients: 1

RAM GB installed: 6
RAM Type: DDR3
RAM Speed: 1384 MHz (rated for 1600, running at 1384 due to OC)
OS / Linux kernel: Ubuntu 10.04 / 2.6.32.22
Running in VM: Yes, 4.4 GB RAM available to the VM, though 2684 only seems to use around 1.4

[bishoptf]

Average time/frame: 00:35:00
CPU: amd8356 @ 2.3 GHz
# of CPU sockets: 4
# of Physical cores:16

# of FAH CPU processes: 15
# of FAH GPU Clients: 0

RAM GB installed:32Gig
RAM Type: DDR2
RAM Speed:667
OS / Linux kernel: centos5 - 2.6.18-194.3.1.el5
Running in VM: No

[leexgx]

Average time/frame: 00:48:35
CPU: i7-920 @ 3.8 GHz
# of CPU sockets: 1
# of Physical cores:4

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

RAM GB installed:6Gig
RAM Type: DDR3
RAM Speed:1600 (9-9-9-29)
OS / Linux kernel: VMware LinuxFAH 1.2
Running in VM: yes

ram use 1.7GB

------------------------------------------------

to find out how much ram is been used type TOP in the command line (its in KB the first two numbers will be in GB if FAH is working on an project (1400000kb) ignore the 000, so norm 1.4 GB or higher for an 8 threaded system)

Grandpa_01 add that to your new post as well

little things like above need to be fixed under Linux TOP should auto group the kb into MB or GB, far more readable

[radekboktor]

Average time/frame: 00:52:01
CPU: X5365 @ 3.0GHz
# of CPU sockets: 2
# of Physical cores: 8

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

RAM GB installed: 6GB
RAM Type: DDR2 FB-DIMM
RAM Speed: 667
OS / Linux kernel: OSX 10.5.8
Running in VM: no

Not sure about RAM use, I didn't check while it was running. Sorry.

[leexgx]

this is my Native LinuxFAH Native 1.1 box disappointing 48:23

Average time/frame: 00:48:21
CPU: i7-920 @ 3.8 GHz
# of CPU sockets: 1
# of Physical cores:4

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

RAM GB installed:6Gig
RAM Type: DDR3
RAM Speed: not sure 1600 or 1052 (9-9-9-29)
OS / Linux kernel: native LinuxFAH 1.1
Running in VM: no

1.7gb ram use

[Alpineb1]

Average time/frame: 46mins 47sec{in hh:mm:ss} (R1,C15,G3)
CPU: _I7 860 @ 3.8 GHz
# of CPU sockets: 1
# of Physical cores: 4
# of Physical cores + HT :8

# of FAH CPU processes: 8
# of FAH GPU Clients: 1

RAM GB installed: 8
RAM Type: DDR3
RAM Speed: 1600 (9-9-9-24)
OS / Linux kernel: Win Seven 64 bits Ultimate
Running in VM: Yes (linuxrouter image 1.2 in VM player)

[brentpresley]

Average time/frame: 00:16:51
CPU: 2 x X5670 @ 4.0GHz
# of CPU sockets: 2
# of Physical cores: 12

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

RAM GB installed: 12GB
RAM Type: DDR3 PC3-1600
RAM Speed: 1465MHz
OS / Linux kernel: LinuxFAH 1.1
Running in VM: no

[Kougar]

Somewhat unusual setup here. 980X on Windows 7 64 folding inside LinuxFAH's VMWare image, so capped at 8 threads.

Project 2684 (Run 8, Clone 0, Gen 2)
Average time/frame: 00:34:03
Min frame time: 00:33:34
CPU: Core i7 980X @ 4.40GHz
# of CPU sockets: 1
# of Physical cores: 6

# of FAH CPU processes: 8
# of FAH GPU Clients: 1

RAM GB installed: 12GB (4900MB assigned to VM)
RAM Type: DDR3
RAM Speed: 1600MHz CAS 8
OS / Linux kernel: Windows 7 64bit (LinuxFAH 1.2v VMware)
Running in VM: Yes

[ahu]

Somewhat unusual setup here. 980X on Windows 7 64 folding inside LinuxFAH's VMWare image, so capped at 8 threads.

Project 2684 (Run 8, Clone 0, Gen 2)
Average time/frame: 00:34:03
Min frame time: 00:33:34
CPU: Core i7 980X @ 4.40GHz
# of CPU sockets: 1
# of Physical cores: 6

# of FAH CPU processes: 8
# of FAH GPU Clients: 1

RAM GB installed: 12GB (4900MB assigned to VM)
RAM Type: DDR3
RAM Speed: 1600MHz CAS 8
OS / Linux kernel: Windows 7 64bit (LinuxFAH 1.2v VMware)
Running in VM: Yes

How are you setting the core affinities? Or are they at default? You have 8 cores used and only 6 physical cores, so two cores are logical cores. I wonder how this unbalance would affect the thread synchronization.

I have a dual Westmere Xeon system with 24 cores available and I found the optimal setting for me would be running two VMWare instances. One would be running on the first 8 even numbered cores (the physical ones), and the second one on the last 8 cores (4 physical and 4 logical). The affinities are set in the VMWare config file.

Yes, I'm also very much looking forward to Windows bigadv launch

[Kougar]

They are at default. Windows 7 seems to have fairly intelligent thread handling and there is no ping-ponging of the VMware load across cores, even with other stuff running. (This is my personal work+gaming system). CPU load is consistent 66-67% without any spikes or dips. Unfortunately I can't say the same for the latest version of VirtualBox, the performance was still worse even when running 12 threads, and the CPU usage was erratic to put it politely. I'd see as much as a 80%-100% variation in usage, and the system would "lag". I experienced the same load balancing issue six months ago when testing VirtualBox on a Core i7 920. I've run systems at 100% loads regularly, and there should be zero lag with the load processes configured for low priority.

As you mention it, I did note that inside the VM the new A3 2684 projects generate a load balancing message inside the log. Something about enabling automatic load balancing, this usually occurs around the 15% mark. The log has already been overwritten so I can't just paste it here, but I did think it was interesting to note. TPF's improved only by a measure of seconds after it kicked in though.

With your setup, my first guess would be assigning one VM to each socket would yield the best results? That way each VM would only need to access local memory and not have to call the 2nd socket for memory read/writes off its local memory? This may not be as big an issue with 2684 though given it uses so much less RAM than the A2 cores...

[ahu]

Yes, the overall performance would be slightly better by running one VM per socket. But because of the way the bonus is calclulated, I'm PPD-wise better now with one VM maxed out. As a side note: when the VM spans different sockets, the cores can actually run at slightly different clockspeeds (!) as I have turbo enabled.

[leexgx]

windows 7 thread handling should correctly put the load on all the real cores any thing left would go onto the HT threads evenly (win7 does it tad better then Vista they improved Thread handling due to HT and i7 for the most part as Vista did not always correctly put CPU load onto an real core that was free where as windows 7 does)

if the program on the other hand decided to lock onto first 4 threads then windows 7 would be unable to use the other 4-8 threads that are free (thats program dev issue) and would be missing out on 2-4 real cores

whats interesting your getting the same time Frame i get when i was doing A2 bigadv cores (45-48 now with A3 bigadv), you be getting some good time frames once you get native winSMP A3
due to How Slow A3 Bigadv is compared to A2 Bigadv, 12 threads WinSMP A3 Norm may be faster or same speed as what your doing in VMware linuxfah with 8 threads as all HT threads would be fully loaded as well (should test it)