Real-world PowerPC Energy Consumption

[AphoticD]

I was writing up a response in another thread and thought this might deserve a thread of it's own.

There aren't a lot of power consumption details on older PowerPC hardware, I think this is mostly because the architecture demanded a lot of energy to perform. Apple love to boast about their 6W/hour mac mini, but we didn't see any marketing material about how much POWER the G4s and G5s consumed.

So, here are my own non-scientific and slightly rounded-out results. I have plugged everything into a digital watt/voltage/amperage meter at the wall and noted the steady idle consumption and the variances under load (running GeekBench 2.x).

My Power Usage Results:

• G4 Sawtooth 350Mhz with Radeon 9700 GPU (no display)
  • Idle: 55w. 95w under load.

• G4 Mac mini 1.42Ghz (no display)
  • Idle: 30w. 60w under load.

• G5 Dual 1.8 (2003) 970 with FX 5200 GPU + ADC 17" Display
  • Idle: 170w. 275w under load.

• G5 Dual 2.0 (2003) 970 with 6800 GT GPU (no display)
  • Idle: 170w. 400w under load.

• G5 Dual 2.3 (Late ‘05) 970MP with Quadro FX 4500 GPU (no display)
  • Idle: 150w. 250w under load.

• Non-PPC for comparison: Mac Pro 8 Core 3.2Ghz (2008) Xeon with GTX 680 GPU (no display)
  • Idle: 220w. 270w under load.

Notes:

• The G5s were tested at idle consumption with the CPU speed set to Reduced in Energy Saver, then set to Highest Performance and run through Geekbench for load testing.
• Note that the 6800 GT, FX 4500 (and GTX 680) GPUs are all power hungry. Stock-standard graphics cards such as the FX 5200 or ATI Radeon 9600 consume much less power.
• Some of these units have more internal devices (expansion cards, hard drives, optical drives, etc) than others, so this is just to get a rough idea of how the different machines consume in the real world.
  
• It's interesting that the Dual 2.0 maxed out so much higher than the dual 1.8 when they share the same period architecture, (logic board and PSU) and have interchangeable CPUs.
• For reference, the Mac Pro's original GT120 graphics card would draw approx 60 watts less while idle than the GTX 680.

It would be interesting to see results from the 970fx model G5s (including the iMacs) as well as the Quad and more G4s, like the Quicksilver, MDDs and lampshade iMacs.

Feel free to discuss and/or post your results!

 

[eyoungren]

I find these kinds of energy posts interesting, although not really relevant to me. Not to put them down or anything, but I have yet to have a computer in my home that out consumes my air conditioner for power in August.

My energy bill has gone up maybe $30 since having two G5s always on at home. But my biggest issue in the summer is the A/C. Summer means a difference of $200 in our electric bill by August over say January or February and that's always because of the A/C and not any computers I've left on.

 

[AphoticD]

I find these kinds of energy posts interesting, although not really relevant to me. Not to put them down or anything, but I have yet to have a computer in my home that out consumes my air conditioner for power in August.

My energy bill has gone up maybe $30 since having two G5s always on at home. But my biggest issue in the summer is the A/C. Summer means a difference of $200 in our electric bill by August over say January or February and that's always because of the A/C and not any computers I've left on.

Living costs vary from country to country, but for the most part I get the impression that electricity, fuel, food and water are all very affordable in the USA. It can be tough to get a perspective on consumption priorities when everything is readily available at a reasonable cost.

It’s not uncommon for the average quarterly bill for a 4 person household in Australia to be up around $800 - $1000, especially during summer. We are getting hammered by our gov on living costs.

I am not complaining though, the fact is that we have readily available electricity, unlike some parts of the world. The key is to balance the usage.

Knowing that energy can only deplete, I always make a conscious effort to switch everything off at night. The kids still need their night lights, but all electronics are switched off, except for the modem.

A/C comes on only when it hits 40C+. No electric heating in winter. We’re in a mostly warm climate and this works for my family (because I tell them so). I wouldn’t expect the same for anyone else, especially where it actually gets cold or unbearably hot.

Getting back to the point, the computers have a sleep option, which people often decline to use because of things like file sharing access. It’s a matter of training yourself to consider the consumption as you operate.

I know that running all my Macs at the same time for an 8 hour stint is going to cost me about $2 - $3. If I put this into perspective of the cost of buying all new equipment just for the sake of lower power consumption, it just doesn’t add up. The real cost of mass production and throw-away, landfill mentality is too high for me.

So i’ll use what I’ve got to get the job done and I put them to sleep or switch off if I’m focusing on something else.

Some days, I’ll turn off everything and work the pencil and paper. It can be a grounding experience! (And something I almost forgot I could do)

 

[eyoungren]

Living costs vary from country to country, but for the most part I get the impression that electricity, fuel, food and water are all very affordable in the USA. It can be tough to get a perspective on consumption priorities when everything is readily available at a reasonable cost.

It’s not uncommon for the average quarterly bill for a 4 person household in Australia to be up around $800 - $1000, especially during summer. We are getting hammered by our gov on living costs.

I am not complaining though, the fact is that we have readily available electricity, unlike some parts of the world. The key is to balance the usage.

Knowing that energy can only deplete, I always make a conscious effort to switch everything off at night. The kids still need their night lights, but all electronics are switched off, except for the modem.

A/C comes on only when it hits 40C+. No electric heating in winter. We’re in a mostly warm climate and this works for my family (because I tell them so). I wouldn’t expect the same for anyone else, especially where it actually gets cold or unbearably hot.

Getting back to the point, the computers have a sleep option, which people often decline to use because of things like file sharing access. It’s a matter of training yourself to consider the consumption as you operate.

I know that running all my Macs at the same time for an 8 hour stint is going to cost me about $2 - $3. If I put this into perspective of the cost of buying all new equipment just for the sake of lower power consumption, it just doesn’t add up. The real cost of mass production and throw-away, landfill mentality is too high for me.

So i’ll use what I’ve got to get the job done and I put them to sleep or switch off if I’m focusing on something else.

Some days, I’ll turn off everything and work the pencil and paper. It can be a grounding experience! (And something I almost forgot I could do)

I think it comes down to location. My wife and I live in Phoenix, but we are not natives (both of us born and lived in California). We moved here in 2000 because things were getting too expensive where we lived and earnings/salary was not enough. At the time it was an employers market and they were able to demand literally anything legal and pay almost nothing for it.

Electricity is our singular largest utility bill every month. If I turned everything off, got the sliding glass door fixed (broke the outside pane over 10 years ago) and had the house fixed for drafts with the vents cleaned we'd still have a high bill because of the damn A/C. Part of it is because of the mickey mouse job for the ventilation system done in the attic that the landlord denies. Essentially I pay a small amount every month to air condition our attic when it's 110º outside. Short of paying to have that fixed myself, not going to get fixed.

All that said, it's a much easier bill during the fall, winter and early spring. Probably higher than most though, but a price I am willing to pay because I keep the network devices and computers on 24/7. The laptops are all that are allowed to sleep and usually only when the lids are closed.

I just prefer to be able to walk up to a computer, move the mouse and get going. I used to make them sleep, but with the amount of trouble I've had in the past of trying to get sleeping Macs/PCs to wake up without a forced reboot it's just easier to have the displays turn off while the Mac remains on at full power.

All of that said, I suppose it might be comical to know that we make sure lights are shut off when we aren't in a room.

 

[mp2017]

Living costs vary from country to country, but for the most part I get the impression that electricity, fuel, food and water are all very affordable in the USA. It can be tough to get a perspective on consumption priorities when everything is readily available at a reasonable cost.

It’s not uncommon for the average quarterly bill for a 4 person household in Australia to be up around $800 - $1000, especially during summer. We are getting hammered by our gov on living costs.

[ snip ]

I know that running all my Macs at the same time for an 8 hour stint is going to cost me about $2 - $3. If I put this into perspective of the cost of buying all new equipment just for the sake of lower power consumption, it just doesn’t add up. The real cost of mass production and throw-away, landfill mentality is too high for me.

These two ideas seem to be at odds with one another. One doesn't have to buy the latest systems in order to achieve lower power consumption. Used systems can be had for low cost, maybe even free depending on the model, and may be able to outperform many of the systems in your OP. At least they can here in the US

While I am a fan of the PPC based systems if power consumption were a concern of mine I definitely would not use one for anything other than playing around (which is exactly what I do with them but I know many here use them for real work).

 

[AphoticD]

It’s a balance between making use of equipment which others consider only good for landfill and keeping things running in good economy.

There is also the enthusiast's angle of wanting to collect quality, old Macs here which may be going against possible better judgement.

 

[mp2017]

It’s a balance between making use of equipment which others consider only good for landfill and keeping things running in good economy.

Barring a special circumstance it's unlikely any PPC system is going to have an advantage over an older, but more modern Intel based system when it comes to operational or possibly productivity costs.

There is also the enthusiasts angle of wanting to collect quality, old Macs here which may be going against possible better judgement.

I do not deny the enthusiast angle of the platform. However that's an entirely different argument than attempting to justify using a power hungry PPC based system as more economical than an older, but newer, based system. They're great systems and I have a few of them. But if it comes to operational costs or productivity, barring special circumstances, a newer system is likely to beat the PPC based system in both.

 

[AphoticD]

I have no doubt I could get cheap, old Intel Macs and/or PCs which will run more efficiently than a G5 or G4 if lowest power consumption was the aim of the game.

Considering we’re posting on the PowerPC section, there is an understanding, sometimes regardless of a “best value” option, of this being a platform/architecture used by choice.

 

[mp2017]

I have no doubt I could get cheap, old Intel Macs and/or PCs which will run more efficiently than a G5 or G4 if lowest power consumption was the aim of the game.

Wasn't it? Your response to eyoungren left me with the impression electric costs in Australia were high but the cost of acquiring new equipment was greater than using the current equipment. My comment was merely to say faster, more power efficient, newer (but not current) equipment can be picked up at little to no cost.

Considering we’re posting on the PowerPC section, there is an understanding, sometimes regardless of a “best value” option, of this being a platform/architecture used by choice.

I understand the use of PPC based systems is one of choice. My response was never to argue against such a choice.

 

[AphoticD]

Wasn't it? Your response to eyoungren left me with the impression electric costs in Australia were high but the cost of acquiring new equipment was greater than using the current equipment. My comment was merely to say faster, more power efficient, newer (but not current) equipment can be picked up at little to no cost.

I understand the use of PPC based systems is one of choice. My response was never to argue against such a choice.

True. I must have gotten off track. Running costs are high yes as are new acquisitions. Your advice is sensible for anyone looking to get into a low-cost, Intel Mac entry-point.

 

[Slix]

Any idea how these compare to a MacBook Air (mine is 2012) running directly off of power? I imagine it's quite low, but I'm still curious.

 

[AphoticD]

Any idea how these compare to a MacBook Air (mine is 2012) running directly off of power? I imagine it's quite low, but I'm still curious.

There's a series of archived power consumption reports at:
https://images.apple.com/environment/reports/

This looks like it began for '08/'09 products.

The report for the 2012 13" MBA states that it idles at about 9w, but it doesn't state maximum under load. The battery is rated at 50Whr and the magsafe charger is 45w? This would be the maximum draw while charging, but if the charge is full and the machine is under load, I imagine it would be around the 30w mark.

 

[eyoungren]

@AphoticD

Here is the temp and rpm stats for my Quad at this particular moment. You will note that the GPU is 149º. That's running 2º cooler than the temp it was running during the height of summer., although the week the A/C was out I was hitting 159º or so.

Fans within the range I mentioned either here or the other thread.

Edit: Obviously the backside temp is entirely off!

 

[AphoticD]

@AphoticD

Here is the temp and rpm stats for my Quad at this particular moment. You will note that the GPU is 149º. That's running 2º cooler than the temp it was running during the height of summer., although the week the A/C was out I was hitting 159º or so.

Fans within the range I mentioned either here or the other thread.

View attachment 725014


Edit: Obviously the backside temp is entirely off!

I was going to say, that's a hot backside!

While we're comparing apples with apples. Here's the DC 2.3 G5:

Ambient temps are a mild 28°C today.

When I'm not running the '03 G5s, my workstation is silent. But if the outside temps go up, the G5s will be the first to alert me with fan noise and hot air

 

[eyoungren]

I was going to say, that's a hot backside!

While we're comparing apples with apples. Here's the DC 2.3 G5:

View attachment 725022

Ambient temps are a mild 28°C today.

When I'm not running the '03 G5s, my workstation is silent. But if the outside temps go up, the G5s will be the first to alert me with fan noise and hot air

Screen grab through VNC…my DC 2.3 running about the same as yours.

 

[mp2017]

Decided to add some of my own numbers. I don't have many PPC based systems so I used my sole G5 and two older, Intel based systems along with my nMP. Here is the summary with comments to follow:

• PowerMac 7,3: Dual 2.3GHz G5 processor, 3GB RAM, ATI Radeon 9600 w/128MB, Leopard 10.5.8
  
  • Idle: 142 watts, Geekbench load: 255 watts, Handbrake load: 259 watts
  • Handbrake encode time: 20:33 (mm:ss)

• MacBook 2,1: Dual core 2.16GHz Core 2 Duo, 2GB RAM, GMA950 w/64MB, Snow Leopard 10.6.8
  • Idle: 17 watts, Geekbench load: 30 watts, Handbrake load: 41 watts
  • Handbrake encode time: 26:09 (mm:ss)

• Mac Pro 1,1: Dual, dual core 2.66GHz Xeon processor, 5GB RAM, GeForce 7300 GT w/256MB, Lion 10.7.5
  • Idle: 158 watts, Geekbench load: 251 watts, Handbrake load: 248 watts
  • Handbrake encode time: 09:55 (mm:ss)

• Mac Pro 6,1: Hexa core 3.5GHz Xeon E5 processor, 16GB RAM, dual D300 w/2GB each, High Sierra 10.13.1
  • Idle: 42 watts, Geekbench load: 133 watts (spike), Handbrake load: 138 watts
  • Handbrake encode time: 02:41 (mm:ss)

Notes:

• Unlike AphoticD I did not change any of the power settings. They were left on the OS defaults.
• Power consumption is the highest observed and not average.
• Highest power consumption highlighted in red.
• The MacBook's display was at the highest intensity level (it's old so needs higher intensity).
• I felt there was a piece missing in AphoticD's initial post. Specifically a time element. If one compares the PowerMac 7,3 results with the Mac Pro 1,1 results you see the power consumption is similar. However the Mac Pro 1,1 completes the Handbrake encode in half the time as the PowerMac 7,3. Thus for the same workload the Mac Pro uses half as much power (for this specific test).
• The Mac Pro 6,1 133 watt Geekbench load is a little misleading as it spend most of its time under triple digit power consumption. The 133 watt measurement was a single observence during the entire run. The Handbrake load is consistent.

And finally this is not intended to try and convince people to avoid using PPC based systems. I have only a single G5 system for which to bench. My only other PPC based systems are iBooks (one a G3, the other a G4). I didn't feel like running the benchmarks on the iBooks.

Comments / questions / thoughts welcome.

 

[AphoticD]

Great post!

Your results confirm the substantial power efficiency improvements made with the 970fx when compared to my ‘03 G5’s 970 CPUs. It would be worth noting a comparison when the G5’s Performance option is set to Highest/Fastest.

What is surprising is the faster encode time on the G5 when compared to your MacBook C2D. I have only ever witnessed the C2Ds outperform the G5s (at any speed) when it comes to video encoding.

The MP1,1 is a great, mostly under appreciated Mac. They sell for quite cheap (often same or lower than a dual G5) due to their lack of post-Lion support. Power consumption is relatively low - or about the same as the 970fx but with better efficiency and storage capacity is huge. Are they quiet or noisy operators? How does the running noise compare to the G5?

The 6,1 truly shines and for the price you would expect so!

Just to clarify, I watched activity monitor and noted the constant draw while the CPU pushed ~100% then did a little rounding up/down to the nearest 5. There are always going to be fluctuations and in the case of your 6,1, spikes in power.

 

[Orizence]

What is surprising is the faster encode time on the G5 when compared to your MacBook C2D. I have only ever witnessed the C2Ds outperform the G5s (at any speed) when it comes to video encoding.

I think this one might be a outlier and would be proved wrong when tried again. I have never seen a G5 faster than a C2D in video encoding either.

 

[mp2017]

Great post!

Your results confirm the substantial power efficiency improvements made with the 970fx when compared to my ‘03 G5’s 970 CPUs. It would be worth noting a comparison when the G5’s Performance option is set to Highest/Fastest.

There does appear to be a significant decrease in power consumption with the 970fx. Here are the results for the high performance setting:

• PowerMac 7,3: Dual 2.3GHz G5 processor, 3GB RAM, ATI Radeon 9600 w/128MB, Leopard 10.5.8
  • Idle: 151 watts, Geekbench load: 245 watts, Handbrake load: 256 watts
  • Handbrake encode time: 20:56 (mm:ss)

I also decided to try the reduced power setting mode too:

• PowerMac 7,3: Dual 2.3GHz G5 processor, 3GB RAM, ATI Radeon 9600 w/128MB, Leopard 10.5.8
  • Idle: 136 watts, Geekbench load: 199 watts (spiked, 188 average), Handbrake load: 199 watts
  • Handbrake encode time: 26:0 (mm:ss)

Notes:

• The high performance setting didn't really change much from the automatic setting. Test variances could explain the differences.
• The reduced power setting reduced power consumption by approximately 22% while performance was reduced approximately 19%.
• Interestingly the power consumption in reduced power mode never reached above 199 watts. An interesting delineation point.

What is surprising is the faster encode time on the G5 when compared to your MacBook C2D. I have only ever witnessed the C2Ds outperform the G5s (at any speed) when it comes to video encoding.

The MacBook did have a 6% clock deficit right out of the gate. Thermal throttling may have reduced it even more. I'd say the MacBook did very well especially when you consider power consumption.

The MP1,1 is a great, mostly under appreciated Mac. They sell for quite cheap (often same or lower than a dual G5) due to their lack of post-Lion support. Power consumption is relatively low - or about the same as the 970fx but with better efficiency and storage capacity is huge. Are they quiet or noisy operators? How does the running noise compare to the G5?

The interesting thing to note about the Mac Pro 1,1 is that while it is more power efficient I am surprised it only outperformed the PowerMac 7,3 by 50%. It has a 15% higher clock speed (2.66GHz versus 2.3GHz) and has twice the cores. It's too bad I don't have a quad as that would be fairly comparable. It's possible the quad would be faster even if just slightly so.

Mac Pro 1,1 is very quiet. So are my Mac Pro 3,1 and 5,1's. Of course the 6,1 is very quiet).

The 6,1 truly shines and for the price you would expect so!

The Mac Pro 6,1 is, IMO, overpriced given the age of the technology inside. But it's the top Mac right now so nothing to drive prices down.

Just to clarify, I watched activity monitor and noted the constant draw while the CPU pushed ~100% then did a little rounding up/down to the nearest 5. There are always going to be fluctuations and in the case of your 6,1, spikes in power.

I was really surprised in the fluctuations in the Geekbench readings for the Mac Pro 6,1. They were always changing. With the other systems I could obtain a feel for what an "average" was. Not so with the 6,1. It was all over the place. But I have to say it spent significantly more time under 90 watts than it ever did above. If I were to guess at a more stable number I'd say 85 watts.
[doublepost=1507915180][/doublepost]

I think this one might be a outlier and would be proved wrong when tried again. I have never seen a G5 faster than a C2D in video encoding either.

I'll run it again. I even have two MacBook 2,1's I can test with (one a 2.0GHz and the other the 2.16GHz)

 

[mp2017]

There was concern the MacBook number was an outlier so I ran the benchmark again. Following are the results for two MacBook 2,1 systems:

• MacBook 2,1: Dual core 2.16GHz Core 2 Duo, 2GB RAM, GMA950 w/64MB, Snow Leopard 10.6.8
  • Handbrake encode time: 26:08 (mm:ss)

• MacBook 2,1: Dual core 2.0GHz Core 2 Duo, 2GB RAM, GMA950 w/64MB, Snow Leopard 10.6.8
  • Handbrake encode time: 27:43 (mm:ss)

I did not include power consumption as the second run of the MacBook 2,1 2.16GHz came out almost identical to the first (being faster by one second) and the power consumed was not in question. The 2.16GHz system is clocked 8% faster than the 2.0GHz system and the results show it to have been 6% faster.

 

[mectojic]

Sorry to revive an old thread, but has anyone run any tests about power consumption on any G3 PPC Macs?

As far as I've read, the G3 was a sort of perfection state in terms of its energy efficiency to performance, which was why the Clamshells and slot-load iMacs were able to run fanless (and not suffer any problems, unlike the G4 Cube).

Especially with these new M1 Macs, it would be interesting to get some stats about the G3s for comparison (the M1 Macbook Air was Apple's first fanless computer since the 2000 Clamshell /Cube).