bigadv frame time check

[Mtnduey]

got two of the dual hex's setup with windows now, need someone to help me check the frame times, good/bad/ugly/other

boxen1 = P2686 = 26m 31sec frames
boxen2 = P2684 = 34m 53sec frames

tks in advance

 

[nitrobass24]

looks about right to me

 

[Kendrak]

Spiffy

What OC you got on 'em?

 

[Mtnduey]

these are stock clocked X5560 dual's (HP DL380 G6), per CPUZ it looks like turbo is keeping it right about 3.0ghz

funny that neither core temp nor CPUZ register the voltage the proc's are running at.

Still looking for an i7 OC utility for windows to see how high I can push these bad boys.

I fixed an airflow problem in the data center that was causing the temps to stay pretty high so cooling shouldn't be an issue now.

 

[musky]

Sounds slow actually. From my L5640 thread:

Update - I backed the SR-2 down to stock speed and voltages, which should be close to what you would see with a standard server motherboard and two of these. Power draw is 290W, which actually sounded high to me, but that is what it is showing. The SR-2 may have higher overhead, which would increase power usage over a normal server board. On a P2684, it is turning 29:30 frames. This matches my 970 @ 4.26GHz almost exactly. It would mean that you could expect around 63K ppd with a good work unit, at 217.3 ppd/W worst case. It also would be an extremely stable machine. It is close to 25% faster than my dual X5550 machine at 2.933, at a similar power draw

This would be a dual hex machine at 2.4GHz. That is over 5 minutes/frame faster than your X5650s at 2.933 GHz (I think that's what they run at stock.) Something is up here.

 

[Mtnduey]

Sounds slow actually. From my L5640 thread:



This would be a dual hex machine at 2.4GHz. That is over 5 minutes/frame faster than your X5650s at 2.933 GHz (I think that's what they run at stock.) Something is up here.

hmmm....something is indeed. Wonder if I should switch to Win7 instead of Server 2008 R2

using ddr3 10600 ECC RDIMM's wouldn't slow it down that much would it?

 

[DooKey]

X5650s run at 2.66ghz stock hex cores.

X5560 run at 2.8ghz stock quad cores.

 

[sirmonkey1985]

id say its because of the memory.. server 08 R2 is a stripped down version of windows 7 so it shouldnt make a difference..

 

[may i be worthy]

it might. Bigadv folding is the most sensitive thing I have ever seen to memory speed.

I just went through getting my 24GB (12 sticks) stable, and benchmarking using a captured 2685 unit, restarting it each time to see difference. (same project, same frame)

Loosening memory timings from 7 8 7 20 to 8 9 8 24 etc lost me only 0.2% (and got me stable), but if I went from 1T to 2T timing I lost 3% off frame times So if you can up your RAM frequency do so, feel free to loosen timings, but try to retain 1T command rate.

So now project 2685 takes me 13 mins 5 sec per frame
(dual hexcore x5650 @ 4.20GHz, 191 bclck, mem 2:8/1066ddr, 8 9 8 24 1T)

At 2:6 = 800ddr memory, my frame time was 14mins11sec

Check taskmanager to see if you are using all threads, if so, forget about which windows version.

 

[musky]

X5650s run at 2.66ghz stock hex cores.

X5560 run at 2.8ghz stock quad cores.

That is a good point. I assumed it was a typo and these are X5650s. If they are quads, then your times look about right, compared to my dual X5550 machine.

 

[lehmann]

My i7-920@4.1 does a 2686 in 30:04 im gonna say your times look very slow. turn off any windows service you dont need! and any hardware you dont need sound cards extra raid controlers

 

[musky]

This is my dual X5550 @ stock 2.933 GHz box on these two units. It also is pushing a GTX 460 for all of these:

Project ID: 2684
Core: GRO-A3
Credit: 8955
Frames: 100

Name: believer
Path: \\BELIEVER\fah\
Number of Frames Observed: 224

Min. Time / Frame : 00:33:40 - 31,527.3 PPD
Avg. Time / Frame : 00:34:25 - 30,502.4 PPD
Cur. Time / Frame : 00:34:29 - 30,372.7 PPD
R3F. Time / Frame : 00:34:30 - 30,358.0 PPD
All Time / Frame : 00:34:24 - 30,446.2 PPD
Eff. Time / Frame : 00:34:49 - 30,081.9 PPD


Project ID: 2686
Core: GRO-A3
Credit: 8955
Frames: 100

Name: believer
Path: \\BELIEVER\fah\
Number of Frames Observed: 299

Min. Time / Frame : 00:24:48 - 49,866.6 PPD
Avg. Time / Frame : 00:25:16 - 48,491.4 PPD

 

[Mtnduey]

boy I really need to start getting more sleep. You'd think I would notice the 4 cores per proc in coretemp staring right at me...

guess the memory timings might be causing the percieved slowness, I dont recall seeing anything in bios to tighten things up tho.

blah

would be so much easier to just gut the bloody thing and toss it on an SR-2 board.

 

[may i be worthy]

Now that I am finally up and running with RAM at a decent speed I have got some 2686's and benched a saved 2684:

I will just post best frame time, because I use this machine as a workstation the averages are misleading if I was rendering something else half the day...

Project ID: 2686
Name: SR2 - X5650 +++
Min. Time / Frame : 00:12:21 - 141,901 PPD

Project ID: 2684
Min. Time / Frame : 00:16:51 - 89,000 PPD

 

[Chilly]

it might. Bigadv folding is the most sensitive thing I have ever seen to memory speed.

I just went through getting my 24GB (12 sticks) stable, and benchmarking using a captured 2685 unit, restarting it each time to see difference. (same project, same frame)

Loosening memory timings from 7 8 7 20 to 8 9 8 24 etc lost me only 0.2% (and got me stable), but if I went from 1T to 2T timing I lost 3% off frame times So if you can up your RAM frequency do so, feel free to loosen timings, but try to retain 1T command rate.

So now project 2685 takes me 13 mins 5 sec per frame
(dual hexcore x5650 @ 4.20GHz, 191 bclck, mem 2:8/1066ddr, 8 9 8 24 1T)

At 2:6 = 800ddr memory, my frame time was 14mins11sec

Check taskmanager to see if you are using all threads, if so, forget about which windows version.

Whats your frame time on 2684?

 

[flyz]

3.8GHz L5640, 7-7-7-20 1T 651 ram on a 2864
Min. Time / Frame : 00:32:25 - 33,368.32 PPD


I just bumped up my oc to 3.9GHz and i'll see what's the difference.