G5: Automatic vs. highest performance mode

Does the machine actually run faster in Highest setting, or is this just an artifact of the testing method?
dyehANTISPAM wrote on :

I got a G5 iMac, and in the energy saver control panel there is an option for Automatic vs. Highest Performance setting. As has been noted on the web, XBench scores are significantly higher in Highest Performance setting (for me, CPU score of 90 vs. 170). So what does this mean? Does the machine actually run faster in Highest setting, or is this just an artifact of the testing method? If so, why would I want to set it on Automatic (which is the default, and thus presumedly, Apple's suggested setting).

Thanks, ddave

Tacit replied on :

I got a G5 iMac, and in the energy saver control panel there is an option for Automatic vs. Highest Performance setting. As has been noted on the web, XBench scores are significantly higher in Highest Performance setting (for me, CPU score of 90 vs. 170). So what does this mean? Does the machine actually run faster in Highest setting, or is this just an artifact of the testing method?

The machine runs faster in the Highest Performance setting. In Automatic, the processor clock speed changes with processor load--the processors actually slow down when there is a lighter load.

If so, why would I want to set it on Automatic (which is the default, and thus presumedly, Apple's suggested setting).

On Automatic, the processors consume a lot less power, and generate far less heat. For most applications--word processing, surfing the Web, and so on--it makes no difference in performance; these are not processor-intensive tasks.

For processor-intensive tasks, running the processors on their highest performance setting results in better CPU performance, at a cost of more electricity consumed and more heat generated.

Morten replied on :

On 2004-09-25 05:40:05 +0200, dyehANTISPAM@redacted.invalid said:

I got a G5 iMac, and in the energy saver control panel there is an option for Automatic vs. Highest Performance setting. As has been noted on the web, XBench scores are significantly higher in Highest Performance setting (for me, CPU score of 90 vs. 170). So what does this mean? Does the machine actually run faster in Highest setting, or is this just an artifact of the testing method? If so, why would I want to set it on Automatic (which is the default, and thus presumedly, Apple's suggested setting).

In automatic mode the CPU clocks down to arround 1.4GHz. It has been a feature of the PPC's since the iBook 366MHz (with firewire) and PB/PM G4 +667MHz

Frank Malczewski replied on :

dyehANTISPAM@redacted.invalid wrote:

I got a G5 iMac, and in the energy saver control panel there is an option for Automatic vs. Highest Performance setting. As has been noted on the web, XBench scores are significantly higher in Highest Performance setting (for me, CPU score of 90 vs. 170). So what does this mean? Does the machine actually run faster in Highest setting, or is this just an artifact of the testing method? If so, why would I want to set it on Automatic (which is the default, and thus presumedly, Apple's suggested setting).

Thanks, ddave

Running it at slower speed for most activities and highest speed for time critical activities is apparently a way to increase the lifetime of your system. (See the Apple Support Discussions boards (G5); there's been a few discussions about this.) Apparently running it full-bore all the time is not so much of a good thing.

3662 replied on :

On 25/9/04 1:40 PM, in article cj2pal$v0$1@redacted.invalid, "dyehANTISPAM@redacted.invalid" dyehANTISPAM@redacted.invalid wrote:

I got a G5 iMac, and in the energy saver control panel there is an option for Automatic vs. Highest Performance setting. As has been noted on the web, XBench scores are significantly higher in Highest Performance setting (for me, CPU score of 90 vs. 170). So what does this mean? Does the machine actually run faster in Highest setting, or is this just an artifact of the testing method? If so, why would I want to set it on Automatic (which is the default, and thus presumedly, Apple's suggested setting).

Thanks, ddave HI,

Yea, I have it in my G5 Dual 1.8 and I have it on Auto.

I have noticed that I will be surfing the web and maybe one time it will start to "rev up"

The noise from it will increase, I spoze the fans kick in (9?) and that makes the noise etc.

The thing is I am just using the web, burn a few discs etc no really intensive tasks on this machine.....

Cheers.

I LOVE MY MAC :)

James Glidewell replied on :

Tacit wrote:

On Automatic, the processors consume a lot less power, and generate far less heat. For most applications--word processing, surfing the Web, and so on--it makes no difference in performance; these are not processor-intensive tasks.

For processor-intensive tasks, running the processors on their highest performance setting results in better CPU performance, at a cost of more electricity consumed and more heat generated.

I had always assumed that running my G5 in "Automatic" would result in it "gearing up to full speed" when a processor intensive load was presented to it for an extended period of time (seconds to minutes, tops).

But apparently this isn't the case:

http://docs.info.apple.com/article.html?artnum=86527

states:

"When using professional applications, games with intensive video processing, or other applications that use the microprocessor intensively, the Automatic option in Energy Saver may not provide the best microprocessor performance. If you know you are going to be placing a significant demand on the microprocessor, change your Energy Saver setting to "Highest", as seen in Figure 1."

This seems a bit silly to me - why should I have to turn the knob up to "11"... :-)

Has anybody implemented a "click the box for me" mechanism that monitors system load and switches gears appropriately?

Will setting my dual G5 to "Highest" result in more heat/fan noise?

Jim Glidewell replied on :

In article I4q3ux.83q@redacted.invalid, James Glidewell jimglidewell@redacted.invalid wrote:

Tacit wrote:

For processor-intensive tasks, running the processors on their highest performance setting results in better CPU performance, at a cost of more electricity consumed and more heat generated.

I had always assumed that running my G5 in "Automatic" would result in it "gearing up to full speed" when a processor intensive load was presented to it for an extended period of time (seconds to minutes, tops).

But apparently this isn't the case:

http://docs.info.apple.com/article.html?artnum=86527

states:

"When using professional applications, games with intensive video processing, or other applications that use the microprocessor intensively, the Automatic option in Energy Saver may not provide the best microprocessor performance. If you know you are going to be placing a significant demand on the microprocessor, change your Energy Saver setting to "Highest", as seen in Figure 1."

This seems a bit silly to me - why should I have to turn the knob up to "11"... :-)

Has anybody implemented a "click the box for me" mechanism that monitors system load and switches gears appropriately?

Looks like Apple did this - see below.

Will setting my dual G5 to "Highest" result in more heat/fan noise?

Doh! You had already answered that question...

I did a small bit of testing on my G5 (dual 2Ghz, rev A) and found a couple interesting things...

First, Apple's "Help" for the "Automatic" setting is somewhat at odds with the technote quoted above (and more in line with my initial impression):

"Some models support the Automatic setting, which allows your computer to switch rapidly back and forth between the Highest and Reduced settings in order to optimize energy use, depending on how much work the processor is doing."

I tried running the UT2004 benchmark (highest settings, 1024X768, Flythru, Antalus):

58.636288 / 91.627533 / 234.244781 fps -- Score = 79.148766

58.037689 / 92.658813 / 230.954803 fps -- Score = 79.559006

42.529785 / 87.168091 / 227.123886 fps -- Score = 77.035271

(results for Auto, Highest, Reduced, respectively)

So "auto" ~= "highest", with "reduced" being 6-7% slower on average.

But UT2004 is likely graphics-card limited on my system, so I went back to a Java benchmark that had been posted in April.

<http://groups.google.com/groups?selm=jimglidewell-230032.20433608042004% 40netnews.comcast.net>

(numbers are in milliseconds, lower is better...)

Auto: 354 Highest: 247 Reduced: 379

Hmmmm....

When I ran this before, I noticed something odd - if I ran 4 executions, stacked on a single line like:

G5:~ jim$ java -server Test;java -server Test;java -server Test;java -server Test; 352 247 248 247

I got one "slow" number, followed by three (or more) fast numbers

Running them manually sequentially always yielded a "slower" number:

G5:~ jim$ java -server Test 333 G5:~ jim$ java -server Test 333 G5:~ jim$ java -server Test 346 G5:~ jim$ java -server Test 360

I really couldn't figure out why this would be...

... but now I can.

It looks to me like "Automatic" mode kicks things into high gear pretty quickly (in a fraction of a second) when a heavy CPU load is present:

Reduced:

G5:~ jim$ java -server Test;java -server Test;java -server Test;java -server Test 381 381 381 381

Highest:

G5:~ jim$ java -server Test;java -server Test;java -server Test;java -server Test 248 250 249 248

Automatic:

G5:~ jim$ java -server Test;java -server Test;java -server Test;java -server Test 360 248 248 247

Note that after the first 1/3 of a second, Automatic mode ran just as fast as "Highest" mode. The fact that the benchmark test executes in almost the same time that it takes for the G5 to "kick into high gear" was just a "lucky" coincidence.

Just to be sure, I changed the inner loop by a factor of 4X:

G5:~ jim$ #Auto G5:~ jim$ java -server Test4;java -server Test4;java -server Test4;java -server Test4 1049 974 971 972 G5:~ jim$ #Highest G5:~ jim$ java -server Test4;java -server Test4;java -server Test4;java -server Test4 973 974 976 972 G5:~ jim$ #Reduced G5:~ jim$ java -server Test4;java -server Test4;java -server Test4;java -server Test4 1492 1493 1496 1497

Based on these numbers, I'd say that it takes somewhere between 0.2 and 0.35 seconds to get up to speed when running in automatic mode. After that, your CPU is running just as fast as if you had set it for "Fastest".

The good news is - automatic should be fine for most all purposes on a G5. There is no performance penalty when doing long-running CPU intensive things like DVD rendering, MP3 encoding, etc.

The bad news is - I don't get a speed boost by simply clicking a pop-up menu... :-)

If anyone has any evidence of "Automatic" seriously compromising a G5's performance, please follow-up. But for now, I'm leaving my G5 on "Auto"...