Posts by Toby

21) Message boards : Number crunching : good *quiet* cooling fan (Message 739709)
Posted 16 Apr 2008 by Profile Toby
I say you can't have too many 120mm fans. All the new hardware I'm building has them. They are just so much quieter while still moving enough air to keep things cool. I just got a new Lian Li case for my fileserver that has two 120 fans in the front and one in the back. The front ones have air filters on them so hopefully there is a little bit of positive air pressure in the case to keep the dust out.
22) Message boards : Number crunching : MYSQL question (Message 739674)
Posted 16 Apr 2008 by Profile Toby
Well you can still put my queries inside of a php/perl/java/shell script. I just prefer doing things in one place. So instead of spreading the logic between the php script and the SQL I just put it all in the SQL. Either way will work fine.
23) Message boards : Number crunching : good *quiet* cooling fan (Message 739593)
Posted 16 Apr 2008 by Profile Toby
Yeah I have to say I'm not a huge Thermaltake "fan." I do have a power supply of theirs that I have no complaint with but the first model of heatsink I tried was horribly loud and the second one's fan seized up much quicker than others I have used. I am currently very impressed with my Zalman CNPS9700 (runs 20C cooler than AMD's stock heatsink) although I will admit that it was no simple task to install. They do have a slightly smaller version (9500) as well as other designs that may be more suited for your setup.
24) Message boards : Number crunching : MYSQL question (Message 739587)
Posted 15 Apr 2008 by Profile Toby
No need to involve php in the date computations...

This will work if your column is a date, timestamp or datetime type:

For an int column with a unix timestamp in it:

More about dates in mysql: Mysql manual page on date and time functions
25) Message boards : Number crunching : AMD Phenom (Message 736737)
Posted 10 Apr 2008 by Profile Toby
I think I heard that the black edition comes with a decent heatsink like my x2 4400+ came with. I could even see that one holding up to some overclocking. It has a copper base, 4 heatpipes and aluminum fins.

However the "normal" Phenom that I have comes with a chunk of aluminum that can't even stand up to stock speeds. I have positively confirmed that this was my problem. With my new Zalman 9700 the CPU runs a full 20C cooler (!!) under load and basically idles right at ambient. It even looked like it was idling below ambient (15C) but I'm guessing there was a cold draft coming in from the window at the time. Everything is running at full (stock) speed and I haven't had a reboot since the new heatsink was installed.

Note: installing that beast of a heatsink was no simple matter! I just uploaded a picture of it after installation to my profile. It really dominates the case now.
26) Message boards : Number crunching : What's up with LHC@home? (Message 735678)
Posted 7 Apr 2008 by Profile Toby
I had wondered if they started updating their boinc server software to a non-archaic version and ran into trouble... but that seems pretty unlikely at this point. They would have at least put up a static index.html page to indicate what was going on. But the fact that work is being generated is rather odd. If the server software were broken then the scheduler would probably not be working... if it is legal action you would think that any court order would include submission of new work...
27) Questions and Answers : Unix/Linux : Problem with download ( not installing (Message 735218)
Posted 6 Apr 2008 by Profile Toby
I would suggest installing BOINC from the Ubuntu repositories. That way it is automatically integrated into your system. It will automatically start running during boot and as new versions come out, it will be updated through Ubuntu's update manager.

Use the Synaptic package manager (System -> Administration). Check your repositories (Settings -> Repositories) and make sure the "universe" box is checked. Once the list of packages have been updated there should be several boinc related packages available. You need to install "boinc-client" and "boinc-manager." You do not need "boinc-app-seti."

Once it is installed, BOINC should be in your Applications menu under Accessories.
28) Message boards : Number crunching : Refresh My Memory, Why can't we detect CPU to use optimized (Message 735180)
Posted 6 Apr 2008 by Profile Toby
Its not that it can't be done... It just hasn't been done yet. David Anderson ceated Trac ticket 562 one month ago that describes how he would like to see this handled. Of course it would only work with reasonably new clients that actually report the CPU capabilities to the server.
29) Message boards : Number crunching : AMD Phenom (Message 735174)
Posted 6 Apr 2008 by Profile Toby
Well I think I have reached the end of my investigation. With "supercooling" in place the machine remained stable for over 12 hours. I have ordered a Zalman CNPS9700 heatsink which seems to score in the top few spots in most heatsink reviews.

So to summarize for those thinking about buying a Phenom: The stock AMD heatsink provides insufficient cooling for running SETI on the Phenom. You're going to want some copper heatpipe action or if you're truly crazy, maybe some water cooling :)
30) Message boards : Number crunching : AMD Phenom (Message 733929)
Posted 3 Apr 2008 by Profile Toby
Well I got my replacement CPU yesterday. They sent me a whole new retail package so now I have two AMD heatsinks. Maybe I can get rich selling them on ebay! Oh wait... Anyway, I installed it and started up slowly. I ran the AMD "stability test" for an hour which it passed with flying colors. But seti was still too much for it. It took a whole 20 minutes but eventually it rebooted again. I did notice that my office was a little bit cooler than usual so I ran a test over night. I opened my window all the way and put a fan in front of it to blow cold air directly at the front of my computer. 5 hours later it was still happily crunching away. I'm doing the same thing right now. If it survives for the rest of the evening then I think I can pretty definitively say that the Phenom has some kind of overheating problem.

Note that I do not keep my apartment overly warm. It is usually around 70F (21C). The temperature sensors in the cores top out at ~50C which is warm but nowhere near critical and certainly reasonable considering I'm using the stock AMD heatsink (straight aluminum block). However the temperature sensor labeled "TMPIN2" does go up to the mid 60s which is kind of worrysome. I suspect that somewhere in the middle of the CPU (where the memory controller is) the temperature is spiking up into the 70s or higher, which is what is probably causing the reboots. My next step will be to get a better heatsink and see if that "fixes" it although I really think a CPU should be able to run without overheating with a stock heatsink unless you are overclocking.

My old 4400+ came with a good stock heatsink. It has a copper base and 4 heatpipes. That chip doesn't get above about 42C under full load. I was hoping that they were still shipping these heatsinks with the Phenoms which is why I didn't buy a better one from the start. Live and learn I guess.
31) Message boards : Number crunching : AMD Phenom (Message 732847)
Posted 31 Mar 2008 by Profile Toby
Well my opinion of AMD customer support remains high! I 2 day aired my CPU to California. It was delivered to AMD at 9 AM on Friday and John called me at 3 PM the same day and said "It took an hour but it eventually locked up in testing. We'll be sending you a new one on Monday morning."

32) Message boards : Number crunching : AMD Phenom (Message 730687)
Posted 26 Mar 2008 by Profile Toby
And if it does not fail?

What are the chances your sample is just a little slow? Did you ever try these same things at a lower clock rate?

If it does not fail they will 2 day air me my original CPU back :/
Yes, underclockig the memory controller to 1.4 yielded relative stability. It still rebooted once a week or so. Of course every time it did I was in the middle of a game with some friends. :( But since the memory controller is on the die that still means a bad CPU. Underclocking the actual cores had no impact.

Also beware the motherboards that can't supply the core supply current required and cause one of the cores to appear to fail when stressed... There's quite a flurry of web comments about the third core failing in that way.

I have seen some of this... I thought it was mostly a problem with slightly older AM2 motherboards - am I mistaken? I have a brand new AM2+. Also if this was a problem with the 3rd core, I would expect prime95 to cause problems in the "Small FFT" mode where the cores all heat up to higher than normal temps which of course means that they are drawing more power than usual.

Guess we'll see what they find. I just dropped it off at a FedEx location over lunch. Should be in California by Friday.
33) Message boards : Number crunching : AMD Phenom (Message 730645)
Posted 26 Mar 2008 by Profile Toby
Well I finally got fed up with my Phenom and called AMD to do an RMA. I'm almost certain mine has a bad memory controller. Prime95 runs ok if I use the "Small FFTs" option. The cores even heat up a good 8 degrees more than when they run seti. But when I use the slightly larger "In-place FFTs" which start to hit the memory, the box reboots within 2 minutes, just like it does with seti. This would seem to indicate that the cores themselves are ok but as soon as I start stressing the memory controller it wiggs out.

Anyway, I do have to give AMD an "Excellent" rating on their customer service! I had to navigate a menu system but once I said "connect me to a human" I was talking to "John" within 5 seconds. About 7 minutes later I had an RMA number and an email with the address to send the CPU to. He will test it on their test bench and If it fails, they will 2 day air me a new one.

*crosses fingers*
34) Message boards : Number crunching : Windows vs. Linux debate (Message 730629)
Posted 26 Mar 2008 by Profile Toby
I hope I came to the right forum. If not please excuse me.

I just bought an AMD-62 X2 TL-64, with 2 gigs of ram, and have installed BOINC under Vista (Home Premium 32 bit version) and Edubuntu 7.10 (64 bit version)

I have noticed that running in Vista BOINC will resolve each chunk of data in about 4 hours, while Edubuntu takes roughly 16 hours. I am just learning about the difference in 32 vs 64 bit processing, but it seems that a 64 processor/64 OS/64 program would kick the tar out of a 64 processor/32 OS/32 program.

What am I missing? Could someome please educate me on this?

Thanks in advance.


Please see the "Windows vs Linux" thread that is currently on the 2nd page of the forum. Specifically, post 725163 by me concerning frequency scaling. I suspect you are seeing the same thing.
35) Message boards : Number crunching : Stats Sites (Message 728064)
Posted 20 Mar 2008 by Profile Toby
My site does exactly what richard suggested. I query the tables.xml file every hour and compare that to the timestamp in my database. If it is newer then I download the big XML files and update the project. Once per day (1ish AM Central US) I combine all the individual projects into the cross project stats. Seti only publishes new XML once per day. Some projects do it much more often (I believe BURP does it every 2 hours) so update times appear somewhat random.
36) Message boards : Number crunching : Windows vs. Linux (Message 726677)
Posted 15 Mar 2008 by Profile Toby
Searching around a little it looks like Mandriva might use a program called "kpowersave" (part of KDE) to manage frequency scaling and other power control systems. I haven't used KDE in a while so I'm not sure how it works. But it would seem that your CPU(s) are now running at full speed so I guess you don't really need to worry about it :)
37) Questions and Answers : Unix/Linux : lost work units on restart (Message 725845)
Posted 14 Mar 2008 by Profile Toby
No, it should save its state every few minutes at most. How did you install and how are you running BOINC? could there be a permission problem or something like that?
38) Message boards : Number crunching : Windows vs. Linux (Message 725707)
Posted 13 Mar 2008 by Profile Toby
I would argue that the way linux handles idle processes as it relates to increasing CPU speed is probably better than windows. If a process is set to such a low priority then by definition it doesn't really NEED any CPU time - but it will take whatever the OS decides it can spare. For every application except distributed computing I would say leaving the CPU in a low power state would be preferable, especially considering all the energy conservation efforts being promoted these days. There are utilities you can get for Windows (like AMD's Power Monitor utility) that allow you to change the default behavior kind of like how cpufreq-selector works.
39) Message boards : Number crunching : Best case for best cooling (Message 725439)
Posted 13 Mar 2008 by Profile Toby
I really like my Lian-Li case. They tend to be pretty clean and simple looking on the outside but all the edges are rolled or rounded so you don't cut yourself while trying to plug in a cable that goes in some cramped corner somewhere. Mine also has a slide-out motherboard tray that makes working on it very easy. Just about everything is fastened down with thumb screws for easy access.

The front 120mm fan has a filter over it that is easy to clean. Other than that it has one 80mm fan in the back and one in the top plus the 120mm fan in the power supply (bought separately). My phenom runs right around 50C with the stock AMD all aluminum heatsink.
40) Message boards : Number crunching : Windows vs. Linux (Message 725163)
Posted 12 Mar 2008 by Profile Toby
I see the same behavior on my laptop running Ubuntu. Idle processes do not cause the kernel to initiate CPU power level change. You can change this behavior with the cpufreq-selector command. To make the CPU run at full speed all the time, just type this command into a terminal:
sudo cpufreq-selector -g performance

This causes the kernel to use the "performance" governor to determine when speedstep/powernow features are enabled. The other governors that are available are powersave, conservative, ondemand and userspace. I believe the default one is "ondemand" which, as the name suggests, keeps the CPU at low speed unless more power is needed. You can also lock the CPU in at a specific frequency with this command:
sudo cpufreq-selector -f 1670000

The 1670000 means 1.67 GHz (my CPU can run at 1, 1.67 or 2 GHz).

I'm sure there is some GNOME/KDE interface to this feature that doesn't require breaking out a terminal but thats the quick and dirty way :)

You can view the speed your CPU is currently running at with this command:
cat /proc/cpuinfo

and looking for the line that starts with "cpu MHz." There is also a GNOME panel widget (or whatever they're called) under the "System & Hardware" section entitled "CPU Frequency Scaling Monitor" that will graphically tell you at what speed your CPU is running.

Previous 20 · Next 20

©2023 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.