Athlon XP power saving

[time]

After recently wrestling with cooling an Athlon XP, I came across this. I've slowly realised that power saving (on idle) may be broken in many Athlon systems.

In particular, I suspect it may be completely broken on certain VIA chipsets. The one in question reported the idle temperature of an Athlon XP 2000 as 65C! I expect variations in monitoring chips, but IMO the heatsink really did feel too warm. The board was a KM400A from Asus.

If you can understand the poor translation, the CoolON Project and other sites have something to add, including problems associated with USB 2.0.

Surely someone here knows all about this?

 

[Mercutio]

I'm not sure if it's a load of crap or not, but I have some candidate machines to try it on.
I'll probably have a chance to try it later this week.

 

[time]

Wow, it really works!

I ran CoolON on an nForce2 board with an undervolted XP2500 CPU. This setup already uses 20-25% less power than normal, so I didn't expect much improvement. The temperature at idle dropped from 41C to 33C!

This is insane. I thought the issue had disappeared with the release of Windows 2000. I'm going to try this on every Athlon I come across. Now I need to find a utility to work with Sempron 2600/2800 ...

 

[Mercutio]

I've got a room full of Barton 2500s on SiS748s that I think are about to melt down. So I'm very interested in this.

 

[timwhit]

Is this OS specific?

I am running an Athlon XP 3000+ on Wink2k3.

When I try to start the program I get an error that says: "This program must be run on AMD Athlon/Duron."

What am I doing wrong here?

 

[Mercutio]

Two PCs sitting side by side in a room with the thermostat set to 23C.
Both Barton 2500s on GA-7S748 motherboards, in Enlight desktop cases.

Base CPU Temp after 20 minutes being on and idle (measured with motherboard monitor 5):

PC1: 71C
PC2: 68C

Installed CoolOn on PC1, rebooted.
(BIOS reported temp of 70C when I looked)

Waited 10 minutes.

PC1: 52C and it's still slowly dropping.
PC2: 68C

I think I'll be adding CoolOn to my standard image for these machines.

 

[P5-133XL]

If you have a Barton 2500 at 71 while idling, I'd hate to think of what it is like during operation. Perhaps, you need to do some thermal surgery with a little Artic Silver.

While this little program may do the job while idling, I seriously doubt it will help when operating and that is where the temps really matter!

Next Q, why not fold: We could use you.

 

[Mercutio]

1. All the machines aren't on very often.
2. Having machines at 100% CPU utilization frightens and confuses the other instructors. I got yelled at last time I had folding set up.
3. The PCs in question were using AMD retail heat sinks with artic silver in place of the standard thermal pad. I talk more about the specifics of how my classrooms are set up here.

Anyway, I went ahead and deployed CoolOn and Motherboard Monitor to a whole classroom. I'm seeing 15 - 20 C drops in CPU temperature on most of those machines, if Motherboard Monitor is to be believed. All the PCs are reporting 50C or less, instead of 65C+.

This may just be a very happy little program.

 

[P5-133XL]

Regardless of the previous Artic Silver installation there is something wrong with temps at that level, even with horrible stock fans. Perhaps it was installed too thick (The stuff acts as an insulator if there is too much). I suggest that some experimentation is justified, if for no other reason than to learn proper technique.

Sorry about the reaction of your instructors. They need some education to disipate their fears. You should not have to risk your job for such though.

 

[Mercutio]

In this instance I believe that the incredibly poor ventilation from desktop cases with 17" monitors sitting on top of them is the main cause of difficulty. In the past I've tried premium copper heat sinks and various different thermal compounds without much luck.

 

[P5-133XL]

In all the desktop cases I have seen, there has been an 80mm fan mounting point in the front left in addition to the PS fan. It may be worth adding a fan there. Another alternative would be to mount a HD fan in a surplus external bay, even if there is no HD mounted there. Another option is a slot-fan especially if you are using onboard video because then there could be direct airflow from the processor.

 

[Mercutio]

None of the desktop cases I have support a fan of any meaningful size in the rear of the case. None of them have air-holes for ventilation in the front of the case (none I can think of anyway), although two of them have a place to mount a front fan.
As I mentioned in the thread I referenced before, they're pretty crappy cases.

 

[P5-133XL]

Where's the PS -- Does it not push air out the back?

As with many front mounted fan locations there may be apparent vetiliation holes. Regardless, there may be air gaps that allow for air flow that are just not very visible. It would be much better if the holes were easy to spot but it is still worth a shot. A big potential problem may be to find a fan mounting bracket for the case: Some cases hare rather non-standard mounting holes.

Assuming a horizontal bay structure, A HD bay fan may actually be an excellent option because it may pull the hot air from the top. You may have to put a cardboard cover over the space right behind the PS to prevent an airflow short-circuit

Even with all that, you still have the option of a slot fan. I wouldn't normally recomend those (or in general extra case fans), but in your case with no air flow at all; Something is much better than nothing. It may also be an excellent option, if there are no cards blocking airflow from the CPU.

Regardless, of what you choose as a solution, something needs to be done because you're gonna burnout your CPU's by doing nothing. I have never really trusted AMD's thermal solution preventing CPU burnout. Even if you don't specificly burn out the CPU, the motherboard itself may be at risk of a much shortened lifespan if the case temperatures are that high. The CPU cooling program is not really a solution because those CPU temps are gonna skyrocket when the machines are being used. All the above solutions are relatively inexpensive and while I can't specificly identify one that is best for your situation, it is easy enough to do a few trials to see what works best.

 

[time]

Mark, I've been through all this a few times. Cooling is tricky - you can never be certain how effective different techniques will be in a given situation. Adding vents and fans doesn't always achieve what you hoped, particularly in a small case.

I wasted countless hours (I need my head examined) on more than one occasion trying to cool the 65C Athlon 2000 I mentioned earlier. It's a tiny tower with some front vents and an additional 60mm exhaust fan near the CPU. I tried folded ducts, reversing fans and changing speeds, removing card slot covers, etc, etc, etc.

I reinstalled the CPU cooler several times, using a magnifying glass to check the thinness of the thermal transfer compound. Surfaces were cleaned each time, prepared with goop, etc. Different compounds were tried, including Arctic Silver and the original AMD stuff (when the cooler was changed). The CPU was changed from an XP2500 to an XP2000 (which as you'd expect from the specs, was worse).

Mercutio may not have been obsessive enough to do all that, but I imagine he's nonetheless a little offended by your suggestions.

 

[P5-133XL]

I'll shut up then: I wouldn't want to offend.

 

[Mercutio]

I guess I need to post a photo of the cases in question. There's just shy of nothing I can do to improve airflow in any of them.

Anyway, 50C is about my comfort level. Obviously, the AMD chips WILL work when they're much hotter than that; these PCs were running at over 60C for two and three years, but at 50C I'm no longer worried; at 50C the case temps are essentially room temperature, so I'm thinking fanless northbridges and video cards are going to last substantially longer in that environment.

time is right that there's a limit to what I'm willing to do to maintain dozens of machines, but over the years I've done quite a bit of experimentation to lower temps in those machines.

 

[time]

Two KT600 PCs:

XP2500 Before = 51C After = 33C

XP1800 Before = 57C After = 33C

Oddly, the temp on the XP1800 seems to drop about 15C instantly when CoolON is enabled; the rest of the fall is more sedate. Could be a measurement error in this case, or something ...

Unless your CPU utilization is always at 100%, you'll see at least some benefit from this technique - the PC doesn't have to be completely idle.

Mark, please don't be offended by the word "offended".

 

[time]

Whoops! The second one should read:

XP1800 Before = 57C After = 23C

 

[Will Rickards]

Thanks for this I think it will help my dad's computer

 

[Mercutio]

Where's the PS -- Does it not push air out the back?

Something else:

Because of the orientation of parts in a desktop case, the PSU is not "above" the CPU (warm air rises etc etc), it's to one side of the CPU. I believe that reduces or possibly minimizes the amount of airflow and/or ventilation provided by the power supply, which could very well be the big problem with ATX desktop cases generally.

 

[paugie]

I wonder how this CoolOn works. Soon as I read this thread, I went to the site, downloaded and installed.

My 2400Sempron which is SocketA and T-Bred based went from 58deg idle (it's summer here now and it's one of the hottest summers in memory and the temps are expected to peak in June, just before the rains come) went to 44deg in less than 5 minutes. Speedfan doing the monitoring. This is going to be recommended on my Filipino Techie Board right away!

 

[CougTek]

As anyone noticed if it affects peformances? Ran before/after benchmarks? It is hard to believe that such a temperature drop comes without any drawback.

 

[time]

The only way it could affect performance is if there was a delay switching out of the halt state - I don't think this is an issue.

Paraphrased from the CoolON Project:

In the ACPI mode of Windows 98 and Windows2000, the idle process issues a HALT instruction.

When the HALT instruction is executed With Athlon/Duron, the phase lock loop continues and the CPU doesn't enter the low power consumption state (because AMD insisted that it return promptly from HALT).

To put the CPU into the low power consumption state, it is necessary to separate the North Bridge from the CPU bus when the HALT instruction is executed.

I tried CPUMark, 3DMark, the benchmark built-in to 7-Zip, and our new Acrobat Reader benchmark and couldn't detect any difference. But then, it won't have any effect with the CPU at 100%.

 

[timwhit]

So is this needed/useful on Windows XP or 2003 Server?

 

[time]

If I understand correctly, it doesn't have anything to do with the OS (provided it issues Halts in the 'idle' process) - it's a motherboard issue. What motherboard/chipset do you have?

 

[paugie]

So is this needed/useful on Windows XP or 2003 Server

Well, on my rig which uses XP, it is quite useful, dropping my temps 10-14deg C when I'm just surfing or word processing.
When the machine is idle, the temp difference is around 16deg C.

 

[Will Rickards]

Finally got around to installing something similar from cpuidle.de on my dad's computer.
When I came over the computers idle temp was 66C.
This is an athlon xp 1800.
The idle temp dropped to 29C.

I was going to reinstall the heatsink but I don't think I'll bother now.