It is finally time to kick this project off, so lets go! I will get some preliminary information out of the way first.
The temperature in my room is 80°F/27°C.
My Radeon 9800 Pro has a fresh thermal paste job using Arctic Silver 5.
My Motorola G4 1.33GHz 7455B has a fresh thermal paste job using Cooler Master High Performance Thermal Compound (HTK-002).
My memory sticks have heat spreaders on them.
Before I get creative and start moving fans around and swapping them out I need to get a set of baseline temperatures. I will be using the stock fans in their original locations and orientations. I will use the resulting baseline as a metric to determine the success of all the future cooling modifications during this project. Dan (PPCLuddite) suggested using CHUD in OS X for nap mode and powernowd in Debian for CPU frequency scaling. These should obviously help bring the temperatures down so I would like to test them out. Unfortunately I have hit a snag. I am unable to get powernowd working on my Quicksilver. I fear that my CPU is not compatible with powernowd. When I run
powernowd as root I get the following results...
powernowd: PowerNow Daemon v1.00, (c) 2003-2008 John Clemens
powernowd: Found 1 scalable unit: -- 1 'CPU' per scalable unit
/sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_max_freq: No such file or directory
powernowd: Found 1 scalable unit: -- 1 'CPU' per scalable unit
/sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_max_freq: No such file or directory
For grins I ran the same command on my iBook G3 as its CPU does scale with powernowd. Here are it's results...
powernowd: PowerNow Daemon v1.00, (c) 2003-2008 John Clemens
powernowd: Found 1 scalable unit: -- 1 'CPU' per scalable unit
powernowd: cpu0: 400Mhz - 700Mhz (2 steps)
powernowd: Found 1 scalable unit: -- 1 'CPU' per scalable unit
powernowd: cpu0: 400Mhz - 700Mhz (2 steps)
I do not have an original Quicksilver CPU so I am currently unable to test against that fear. After running
dmesg I found the following encouraging line.
[ 0.000000] Processor NAP mode on idle enabled.
Even though I cannot currently get my CPU to scale with powernowd on my Quicksilver, at least I have nap mode in both Debian and OSX. I feel like I have talked enough, lets get onto the tests and results, shall we?
Ten minutes of idle. OS X 10.4.11. CHUD nap mode disabled:
CPU: 57°C
GPU: 43°C
RAM: 43°C
Intake air: 38°C
Exhaust air: 33°C
Ten minutes of idle. OS X 10.4.11. CHUD nap mode enabled:
CPU: 42°C
GPU: 42°C
RAM: 39°C
Intake air: 34°C
Exhaust air: 33°C
Ten minutes of idle. Debian Wheezy:
CPU: 42°C
GPU: 42°C
RAM: 42°C
Intake air: 34°C
Exhaust air: 34°C
My first stress test will be web browsing. I will have three tabs open, PowerPC Liberation, FrozenCPU.com and Newegg. During the test I will flip between FrozenCPU and Newegg, browsing both sites for fans.
Surfing under TenFourFox 31. OS X 10.4.11:
CPU: 51°C
GPU: 43°C
RAM: 43°C
Intake air: 38°C
Exhaust air: 34°C
Surfing under Ice Weasel 24.7.0. Debian Wheezy:
CPU: 51°C
GPU: 43°C
RAM: 43°C
Intake air: 38°C
Exhaust air: 34°C
Now for my second test. I will rip a portion of a DVD using handbrake under OS X. I will be using Handbrake 0.9.1. My codec choice will be XviD. My average bitrate choice will be 2000kbps. I will be using 2-pass encoding and the AVI container. These settings came from Zen's excellent XviD guide. I recommend reading it if you haven't already.
The DVD rip results are...
CPU: 61°C
GPU: 46°C
RAM: 48°C
Intake air: 42°C
Exhaust air: 36°C
The third test will be the playback of the resulting AVI file. I will use VLC under OS X and Debian. In addition to VLC, I will test mplayer in Debian.
OS X 10.4.11 VLC 0.9.10 full screen playback:
CPU: 43°C
GPU: 39°C
RAM: 39°C
Intake air: 34°C
Exhaust air: 33°C
OS X 10.4.11 VLC 0.9.10 windowed playback:
CPU: 43°C
GPU: 42°C
RAM: 41°C
Intake air: 34°C
Exhaust air: 33°C
Debian Wheezy VLC 2.0.3 full screen playback:
CPU: 46°C
GPU: 43°C
RAM: 42°C
Intake air: 36°C
Exhaust air: 34°C
Debian Wheezy VLC 2.0.3 windowed playback:
CPU: 46°C
GPU: 43°C
RAM: 42°C
Intake air: 38°C
Exhaust air: 34°C
Debian Wheezy mplayer svn34540 full screen playback:
CPU: 46°C
GPU: 43°C
RAM: 43°C
Intake air: 36°C
Exhaust air: 34°C
Debian Wheezy mplayer svn34540 windowed playback:
CPU: 46°C
GPU: 43°C
RAM: 42°C
Intake air: 38°C
Exhaust air: 34°C
This experiment needs more first person shooters. Time for some Quake III Arena!
Quake III Arena OS X 10.4.11:
CPU: 57°C
GPU: 43°C
RAM: 43°C
Intake air: 38°C
Exhaust air: 34°C
This has been fun. I could have just cheese balled my way through this and used something like geek bench. I opted to use real world tasks instead as they are more important to me than benchmarking software magic numbers.
I did notice something interesting about CHUD and OS X nap mode. I belive that OS X, for my Quicksilver, enables nap mode by default. I can install CHUD, turn off nap mode and watch my temperatures climb. I can then turn it back on and watch them go back down. This is where it gets interesting. If I turn nap mode off using CHUD, then remove CHUD and reboot, my temperatures will go back down to nap mode levels. In the article Dan brought to my attention this is not the case. The MDD owner in the article can monitor his temperatures using software and watch them drop after enabling nap mode with CHUD. Based on my findings and the MDD owner's findings, I believe CHUD's benefits will vary model to model.
Given the period of time the Quicksilver came from, I do not believe these temperatures are bad. The highest temperature I got was 61C during a handbrake rip. I have seen plenty of PCs of similar vintage push out higher temperatures. However, I want the absolute best for my Power Mac. Despite these positive results, I still feel like I can make improvements to the cooling. I will continue to keep pushing onward.
The next phase will be experimenting with the fan orientation. Stay tuned and thank you for reading!
A well done and detailed followup.
ReplyDeleteOne thing to keep in mind about most thermal paste is that it can take up to 2-3 weeks to take full effect. That is the one downfall of quality paste; it takes a while to fully settle in its gap because it's more durable.
The real upside to quality paste is that it can last a good 8-10 years if the sink isn't removed afterward. The paste on my main 7448 CPU is at least 3 years old, and going strong. Never used coolermaster paste, but I normally use arctic silver. I have also gotten good use out of a few different OCZ pastes.
I look forward to the progression of this project. When I have time I will share my nap experiences, but that was with MDD's, not Quicksilvers. MDD's are a whole different beast compared to all other G4's.
Interesting project! I've done some similar testing with cubes and powerbooks. If you want to load the CPU the best things I've found to stress test the cooling setup are Power Fractal and dnetc from distributed.net ...looking forward to your results!
ReplyDeleteRipping video, as he did, is as effective as anything else to peg the CPU at 100%. You would have to loop that fractal app to keep it going, and it would give the CPU brief lulls, which would contribute to a slightly lower overall temp.
DeleteDon't follow this advice, Mark.
Thanks for reading. These programs look interesting, but I am going to continue to use the same set of tests. People obviously use Power Fractal and dnetc; however, I believe the current set of tests applies to a much wider range of users.
DeleteThere are a lot of ways to "peg" the CPU, but if you're trying to evaluate the cooling setup (which it seems like you're going to do), then the programs that use the vector unit and aren't bound by the throughput of the bus consume the greatest power and thus yield the highest CPU temperatures than other methods. PowerFractal and dnetc both do this. You can change the settings in Power Fractal so that it takes a LONG time to complete.
ReplyDeleteYou're missing his point that he wants to use real world tests. What you suggest is great if the goal is to do the exact opposite of what Mark (fiftysixk) is trying to do.
DeleteAlso, the fact that you're visiting here with Safari 3 doesn't bode well for you. No offense. I appreciate that you obviously trust this site, but Safari 3??? Come on man/woman!
What you suggest would be worthwhile to look into for his own machines personal safety, in a worst case scenario, but that is up to him. Also, the 7455 chips (at least the stock Apple ones) all have an auto system shut down once it gets around 70 C, so there may be no need to worry in the first place. I have never tested this extreme temp on a 3rd party daughtercard, which is what he has, so I can't comment on that. Maybe I should test that myself. I have a Sonnet/Motorola 1.0GHz 7455B.
A client of mine once accidentally turned his MDD on with no heatsink, and it simply just shut itself down within a few seconds, then booted fine (once the sink was reattached) and runs to this day. That was about 4 years ago.
I would be curious to see if my third party CPU did have a shutoff point like the Apple 7455 Zen encountered. I acknowledge that your recommendations, anonymous reader, appear to be a good way to do that. I will not risk my current CPU while this project is in progress.
DeleteMy goal is to examine the impact of cooling upgrades and modifications has on various points in my Power Mac. The CPU is one of those, but it is not the only one I am interested in. Using these programs or something like geek bench would make it much easier; however, I want to stress my system like users actually would. I believe that my battery of real world tests, while they may be more time consuming, will accomplish the project's goal.
What model of Radeon 9800 Pro do you have?
ReplyDeleteI've got an Hercules 3D Prophet flashed Radeon 9800, and it gets to about 60°C after 5 minutes of MacTubes Quicktime video playback in 480p.
Is it normal?
Sorry for the late reply! My card is the plane old 128mb 9800 pro. The part number is "pn 109-a07500-00" I had very little faith in the stock heat sink and fan so I opted for a Zalman VF700-CU cooler. A friend of mine gave me some ram sinks so I am using those as well.
DeleteI looked at the Hercules 3D Prophet on the Mac Elite wiki and the cooling looks better than the default 9800 pro cooling. Have you re-applied the thermal paste? If it still has the original thermal paste that could be a cause for higher temperatures. You could also upgrade the cooler as I did. The Zalman VF700-CU I use is discontinued but Newegg still stocks a Zalman VF900-CU LED. It boasts Radeon 9 Series compatibility with the exception of the 9550 and 9600.
I would start with the thermal paste if it hasn't already been changed. If you do re-apply the thermal paste take special care not to over apply it. You don't want to short out any transistors or diodes.
What kind of Power Mac do you have? How are your PCI ports populated? Is there a PCI card right next to your Radeon?
Hi! Don't worry :)
DeleteYes, it has a nice cooling system (heatsinks on every VRAM chip, heatsink on the back of the card, heatsink on the front with a VERY NOISY and rattling fan. I may need to replace it. It spins though, so it's still fine). I'm not monitoring the temperatures via software, I'm using a thermal probe attached to the heatsink on the back of the card, so 60°C is the temperature on the back of the card.
I don't think thermal paste is a problem because the heatsink actually gets warm before the Mac completes the bootup process. It's a 2001 QuickSilver with upgraded Dual 800MHz CPU (from another QuickSilver). What a machine.
Unfortunately I live in Italy, so I can't buy anything from Newegg. I may look for that cooler on other websites.
The only other PCI card I have in the Mac is an USB 2.0 card, but it's in the 3rd slot (far from the graphics card).
I have another question for you: do you still use the stock 344W PSU in your QuickSilver? I'd like to add a 2nd hard drive to run Debian, but I have fear that the PSU can't cope with it (haven't tried yet).
It has a Dual 800MHz G4, 1GB of RAM, Radeon 9800 Pro, USB 2.0 PCI card and an IDE 80GB Maxtor drive with OS X Leopard on it. What do you think?
Thank you!
I would replace that heat sink and fan, it sounds like its bearings are failing. My Italian is bad (and by bad I mean non existent) but I believe I found a VF700-CU on amazon.it. This would perform better and it will be much quieter. Feel free to email me and I will forward you the link.
DeleteYour system sounds very nice! I am still using the stock power supply. I have two hard drives, one of which, is connected to a SATA PCI controller. You can see the rest of my hardware on the author & hardware information page to the right. I would imagine that I am taxing the stock power supply but I have yet to get into trouble.
Based on my experience with my Quicksilver, I would say you can go ahead and follow through with your plan for a second drive. Would the second drive be an IDE drive?
I can assure both of you that you're not overexerting a Quicksilver PSU. They simply can't hold enough hard drives to overpower a 344 watt PSU, especially when you consider a typical magnetic drive only consumes 10-15 watts.
DeleteYou both have Radeon 9800's, which need a 300+ watt PSU (ATI's recommendation), but when they say 300 watts they don't mean the 9800 uses that much. More like 120~ watts.
There isn't a G4-capable (even one that is flashed) video card out there that can overpower a Quicksilvers PSU; nor can they hold enough hard drives to do that. Even 10 drives would only be about 100~ watts.
The only AGP-capable G4 towers that can't handle a 9800 and a few drives on the stock PSU are the Sawtooth and the Gigabit models, because they only have 200 watts.
I was using a modified ATX PSU in my main Sawtooth when I was using my 9800, but that PSU died, so I'm back on the stock one again, along with my Geforce 6200.
Thank you zen :)
DeleteI actually added the second drive today and installed Debian 7.7 on it. What an amazing OS! Flies on this machine, so pleased!