Welcome to the MacNN Forums.

BTBlomberg · Jun 21, 2006, 10:49 PM

BOINC's estimates seam worthless so my estimates based on it may be crap so nevermind. So much for trying to guesstimate on a slow machine with BOINC's output. Real data when it's done.

bobpalmer · Jun 22, 2006, 12:38 AM

The wisdom files in the slots/{0,1,2,3} directories don't seem to match the wisdom.sah file generated by fft_test3. The file sizes and content are different.
projects/setiathome.berkeley.edu/wisdom.sah = 16976 bytes (reference)
slots/0/wisdom.sah = 6902 bytes
slots/1/wisdom.sah = 6903 bytes
slots/2/wisdom.sah = 6612 bytes
slots/3/wisdom.sah = 7383 bytes

I am running boinc 5.4.9 CLI version, and Alex's enhanced v5 worker. I killed boinc before regenerating the reference wisdom.sah file. I copied the new wisdom.sah file to the projects/setiathome.berkeley.edu directory, and also into each of the slots/{0,1,2,3} directories. So I know the wisdom files in the slot directories are not hold overs from an earlier source. Any ideas from the experts on why the wisdom files do not match? Since the slot wisdom files do not match the reference, it appears the worker is not using the results from fft_test3.

Putting all the wisdom files here overflows the post size limit.

halimedia · Jun 22, 2006, 12:41 AM

Thanks for v6, Alex! Great stuff!!

Results from G4s and G5s of all walks of life indicate speed gains between 5 and 13%. Excellent!

ChillyWilly5280 · Jun 22, 2006, 11:10 AM

Wow, thanks to the great work done by Alex optimizing the Mac worker, I have now set SETI's resource share back up where it was before enhanced was introduced.

Rock 'n roll baby, lets find us some aliens!

chboss · Jun 22, 2006, 03:23 PM

Alex's new creation shows around 12% speed improvment over V5.

A big thank you!

Thanar · Jun 22, 2006, 03:45 PM

I would like to share Bob's query about whether putting the custom wisdom file in the project's directory adds something to the calculation speed or not. I would like to note that there is NO reference to the wisdom file in the app_info XML file.

zombie67 · Jun 22, 2006, 10:40 PM

My question is: When I generate the wisdom file, does it matter if I am using my computer for other things at the same time? Is it somehow measuring the performance of my machine? Sorta like making sure you aren't using your internet connection while doing speed tests. Or do I want to have everything else shut down while it runs? Does it matter at all?

TIA

Gecko_r7 · Jun 22, 2006, 10:59 PM

Originally Posted by zombie67

My question is: When I generate the wisdom file, does it matter if I am using my computer for other things at the same time? Is it somehow measuring the performance of my machine? Sorta like making sure you aren't using your internet connection while doing speed tests. Or do I want to have everything else shut down while it runs? Does it matter at all?

TIA

It's better to have as few processes running as possible.

ChillyWilly5280 · 11:42 PM

Originally Posted by zombie67

My question is: When I generate the wisdom file, does it matter if I am using my computer for other things at the same time? Is it somehow measuring the performance of my machine? Sorta like making sure you aren't using your internet connection while doing speed tests. Or do I want to have everything else shut down while it runs? Does it matter at all?

TIA

Probably best to run it in single user mode, or console. Barring that, fire up Activity Monitor and quit all the other apps & background process' that you can without logging yourself out to do so (and then quit Activity Monitor too).

Here's a link to Gecko_r7's instructions on running fft_test3 in single user mode that I used successfully on three Macs:

fft_test3 in SU mode

zombie67 · Jun 23, 2006, 01:39 AM

Originally Posted by ChillyWilly5280

Probably best to run it in single user mode, or console. Barring that, fire up Activity Monitor and quit all the other apps & background process' that you can without logging yourself out to do so (and then quit Activity Monitor too).

Here's a link to Gecko_r7's instructions on running fft_test3 in single user mode that I used successfully on three Macs:

fft_test3 in SU mode

I suspected as much. BTW, *great* instructions.

Question: Is it normal that when one boots singe user, you then need to mount the drive to write files?

It's been about 15 years since I've had the need to do that kind of stuff. But I don't recall having to mount drives in single user.

ChillyWilly5280 · Jun 23, 2006, 02:32 AM

Originally Posted by zombie67

I suspected as much. BTW, *great* instructions.

Question: Is it normal that when one boots singe user, you then need to mount the drive to write files?

It's been about 15 years since I've had the need to do that kind of stuff. But I don't recall having to mount drives in single user.

I am by no stretch of the imagination even remotely knowledgeable about matters of the command line. But yes, you do. I imagine because when finished it writes the wisdom file to the drive. But I could be wrong.

zombie67 · Jun 23, 2006, 08:40 AM

Originally Posted by ChillyWilly5280

I am by no stretch of the imagination even remotely knowledgeable about matters of the command line. But yes, you do. I imagine because when finished it writes the wisdom file to the drive. But I could be wrong.

Yes, I understand that you need to do that *now*. My question is has it always been that way for SunOS (bsd) or Solaris (svr4)? I don't recall the need to have to mount the drive in the past.

Jun 23, 2006, 10:15 AM

I guess it is sort of a safety issue. Since MacOS X i is created to be a consumer friendly product.

If someone, knowingly or unknowingly enters single user mode at startup, the unmounted drive (read only) set-up is propably configured to be the default. In that way, you prevent people who don't know what they're doing from messing up the computer.

A Sun system however, is not targeted at the consumer level, and therefore you propably don't have to mount the drive before performing tasks that requires the need to write to the drive, when in single user mode...

Just a thought :-)

BTBlomberg · Jun 23, 2006, 11:00 AM

Now I have to say I am seeing a about a 12% improvement overall vs. V5. My first one that was only about 8% but it went 26.667% with V5 and then ran V6 so no suprize. I guess that teaches mne to trust Boinc Completion times for estimating. I like those numbers a lot though.

Knightrider · Jun 23, 2006, 06:47 PM

Originally Posted by bobpalmer

on why the wisdom files do not match? Since the slot wisdom files do not match the reference, it appears the worker is not using the results from fft_test3.

The wisdom files are, as I understand it, generated 'on the fly' by the regular sah version of the worker, and the self generated ones were used more in hope than in certain knowlege that they would work. That said, Alex has now linked V6 to the fft3 wisdom files so they should be effective when running V6.

Originally Posted by Alexkan

v6:

G4 version
G5 version

Oh, and this is linked with the (slightly) faster FFTW compile I rolled for fft_test3, so that doesn't hurt, either.

K.

bobpalmer · Jun 24, 2006, 12:36 AM

Originally Posted by Knightrider

The wisdom files are, as I understand it, generated 'on the fly' by the regular sah version of the worker, and the self generated ones were used more in hope than in certain knowlege that they would work. That said, Alex has now linked V6 to the fft3 wisdom files so they should be effective when running V6.

K.

I'm now running v6, and it is still not using the wisdom generated by fft_test3. So the question is which program creates the slots directories, the worker (Alex's v6 in this case), or boinc? I'm running the command line version of boinc 5.4.9. There is no reference to the wisdom.sah file in the app_info.xml file.

-- bob

bobpalmer · Jun 24, 2006, 12:47 AM

Anybody else seen their Quad shutting down due to overheating while running SAH?

Running with CPU performance set to highest. Both cores in CPU A are running significatly higher than CPU B when crunching. From Temperature Monitor I see the CPU A temperature fluctuate from 96C down to 74C, and back to 96C every 5-10 seconds. When the machine was new, and it was colder in the house, CPU A run hotter at about 85C, but did not fluctuate. This is the 2nd time it shutdown this week. It looks like the shutdown occurs when you hit 100C. I guess I'll have to switch CPU performance to Automatic. Here's the system.log entry from today's shutdown.

Jun 23 12:30:50 Bashful kernel[0]: SMU_Neo2_PlatformPlugin core dump:
Jun 23 12:30:50 Bashful kernel[0]: IOHWControls:
Jun 23 12:30:50 Bashful kernel[0]: [0] "BACKSIDE" Type:"fan-rpm" Id:1 TGT:1100 CUR:1097
Jun 23 12:30:50 Bashful kernel[0]: [1] "CPU B PUMP" Type:"fan-rpm" Id:7 TGT:3600 CUR:3625
Jun 23 12:30:50 Bashful kernel[0]: [2] "DRIVE BAY A INTAKE" Type:"fan-rpm" Id:0 TGT:1000 CUR:995
Jun 23 12:30:51 Bashful kernel[0]: [3] "CPU A INTAKE" Type:"fan-rpm" Id:2 TGT:3104 CUR:3103
Jun 23 12:30:51 Bashful kernel[0]: [4] "EXPANSION SLOTS INTAKE" Type:"fan-rpm" Id:8 TGT:1560 CUR:1561
Jun 23 12:30:51 Bashful kernel[0]: [5] "CPU B INTAKE" Type:"fan-rpm" Id:3 TGT:3104 CUR:3103
Jun 23 12:30:51 Bashful kernel[0]: [6] "CPU B EXHAUST" Type:"fan-rpm" Id:5 TGT:3200 CUR:3197
Jun 23 12:30:51 Bashful kernel[0]: [7] "CPU A PUMP" Type:"fan-rpm" Id:6 TGT:3600 CUR:3571
Jun 23 12:30:51 Bashful kernel[0]: [8] "CPU A EXHAUST" Type:"fan-rpm" Id:4 TGT:3200 CUR:3199
Jun 23 12:30:51 Bashful kernel[0]: IOHWSensors:
Jun 23 12:30:51 Bashful kernel[0]: [0] "CPU A0 POWER" Type:"power" Id:13 CUR:51.47424 W
Jun 23 12:30:51 Bashful kernel[0]: [1] "CPU A1 POWER" Type:"power" Id:23 CUR:53.43072 W
Jun 23 12:30:51 Bashful kernel[0]: [2] "CPU B0 POWER" Type:"power" Id:33 CUR:45.17403 W
Jun 23 12:30:51 Bashful kernel[0]: [3] "CPU B1 POWER" Type:"power" Id:43 CUR:50.638 W
Jun 23 12:30:51 Bashful kernel[0]: [4] "CPU B0 VCORE" Type:"voltage" Id:30 CUR:1.14144 V
Jun 23 12:30:51 Bashful kernel[0]: [5] "CPU B0 DIODE TEMP" Type:"temp" Id:31 CUR:54.12288 C
Jun 23 12:30:51 Bashful kernel[0]: [6] "CPU B1 DIODE TEMP" Type:"temp" Id:41 CUR:57.14336 C
Jun 23 12:30:51 Bashful kernel[0]: [7] "CPU B0 CORE CURRENT" Type:"current" Id:39 CUR:37.15104 A
Jun 23 12:30:51 Bashful kernel[0]: [8] "CPU B1 CORE CURRENT" Type:"current" Id:46 CUR:40.26368 A
Jun 23 12:30:51 Bashful kernel[0]: [9] "MLB INLET AMB" Type:"temp" Id:9 CUR:29.36864 C
Jun 23 12:30:51 Bashful kernel[0]: [10] "CPU A0 VCORE" Type:"voltage" Id:10 CUR:1.16384 V
Jun 23 12:30:51 Bashful kernel[0]: [11] "BACKSIDE" Type:"temperature" Id:6 CUR:40.16384 C
Jun 23 12:30:51 Bashful kernel[0]: [12] "CPU A1 VCORE" Type:"voltage" Id:20 CUR:1.18944 V
Jun 23 12:30:51 Bashful kernel[0]: [13] "CPU A0 CORE CURRENT" Type:"current" Id:19 CUR:41.24832 A
Jun 23 12:30:51 Bashful kernel[0]: [14] "CPU A1 DIODE TEMP" Type:"temp" Id:21 CUR:99.6144 C
Jun 23 12:30:51 Bashful kernel[0]: [15] "CPU A1 CORE CURRENT" Type:"current" Id:26 CUR:41.40960 A
Jun 23 12:30:51 Bashful kernel[0]: [16] "DRIVE BAY" Type:"temperature" Id:4 CUR:39.32768 C
Jun 23 12:30:51 Bashful kernel[0]: [17] "TUNNEL" Type:"temperature" Id:1 CUR:35.57344 C
Jun 23 12:30:51 Bashful kernel[0]: [18] "TUNNEL HEATSINK" Type:"temperature" Id:2 CUR:32.16384 C
Jun 23 12:30:51 Bashful kernel[0]: [19] "CPU B1 VCORE" Type:"voltage" Id:40 CUR:1.15584 V
Jun 23 12:30:51 Bashful kernel[0]: [20] "KODIAK DIODE" Type:"temperature" Id:7 CUR:62.0 C
Jun 23 12:30:51 Bashful kernel[0]: [21] "CPU A0 DIODE TEMP" Type:"temp" Id:11 CUR:95.15360 C
Jun 23 12:30:51 Bashful kernel[0]: [22] "Slots Power" Type:"power" Id:3 CUR:25.14720 W
Jun 23 12:30:51 Bashful kernel[0]: [23] "gpu-diode" Type:"temp" Id:-1558188032 CUR:47.0 C
Jun 23 12:30:51 Bashful kernel[0]: IOHWCtrlLoops:
Jun 23 12:30:51 Bashful kernel[0]: [0] "CPU A0 Liquid Cooling" Id:3 MetaState:0
Jun 23 12:30:51 Bashful kernel[0]: [1] "CPU A1 Liquid Cooling" Id:4 MetaState:0
Jun 23 12:30:51 Bashful kernel[0]: [2] "CPU B0 Liquid Cooling" Id:5 MetaState:0
Jun 23 12:30:51 Bashful kernel[0]: [3] "CPU B1 Liquid Cooling" Id:6 MetaState:0
Jun 23 12:30:51 Bashful kernel[0]: [4] "Clock Slew" Id:0 MetaState:0 "Dynamic Power Step"
Jun 23 12:30:51 Bashful kernel[0]: [5] "Drive Bay Fan" Id:1 MetaState:0 "Normal"
Jun 23 12:30:51 Bashful kernel[0]: [6] "U3/Backside Zone" Id:2 MetaState:0 "Normal"
Jun 23 12:30:51 Bashful kernel[0]: [7] "PCI Slot Fan" Id:7 MetaState:0 "Normal"
Jun 23 12:30:51 Bashful kernel[0]: ---------------------------------

alexkan · Jun 24, 2006, 12:49 AM

Whoops...yeah, the crunchers aren't using the wisdom generated by fft_test3. However, at this point, I'm inclined to say that it doesn't matter. The improvement in FFT speed is negligible, especially since the FFTW_MEASURE planner setting (the one that SETI uses by default) is better than I previously thought.

To make things worse, I also realized that because of the way that SETI uses FFTs, the numbers that fft_test3 puts up don't reflect actual processing at all. I'll post a revised app if you want to see what performance is actually like when running SETI, but only if you're curious--there won't be any speed gains from running it.

Knightrider · 01:44 AM

Originally Posted by bobpalmer

Anybody else seen their Quad shutting down due to overheating while running SAH?

Yes: my quad is at the garage being repaired at the moment. The logic board and cpu's are being replaced. Still under warrenty thank goodness.

We must be carefull here, not to stray to far off topic. Start a new thread if this developes further.

They tell me that the quad is not built to run 24/7, which I was doing with seti and other dc's and it was going to sleep/shut down often. There are a few reasons why and running sah 24/7 with cpu set to max is one suspect, which may have damaged one of the cpu's by overheating. When I reduced the cpu level it ran longer before shutting down than it did at the high level.

Having two drives, which I also have, can apparently also cause overheating and is a ventilation design fault. I shall be getting an enclosure for the second drive.

Dust clogging up the 'radiator' is another problem, although I had cleaned that up about six weeks ago.

Static damage to a component is another possibility.

There are sleep/shut down osx/quad issues as well. Search the issue at the Apple support discussion groups.

Keep an eye on the red lights that may show up inside the front of the box.

Run the hardware test, if it will not run you may have a problem.

K.

Edit: It might be a good idea to continue this discussion at THIS thread.

K.