In commenting on my previous post about the extravagant cost of the new Mac Studio (as compared to the 27″ iMac SKU it has seemingly replaced), Thomas Speck quite reasonably pointed out that the reduced power consumption of the M1 silicon would result in savings over the lifetime of the device. I was sceptical, and have now done some rough and ready calculations to see just how much those savings would add up to.
As with any such exercise, the methodology is key. So this is the approach I used.
The inputs were:
- The current daily cost of electricity to derive a cost per kw/h
- The power consumption of an M1 Ultra at maximum load: 100W
- The combined power consumption of an Intel i9 and AMD Radeon 5700 GPU (as fitted in my Intel iMac), again at maximum, standard (not over-clocked or exotically configured) load: 450W
I then assumed that I would be leaving both systems on 24×7.
Finally, since any calculation of actual workload power consumption would be synthetic, rather than a complicated recipe to derive some approximation of actual workloads, I settled on simply assuming that both systems were running constantly flat-out.
In theory, this would give the M1 silicon the best possible chance of demonstrating savings as the difference is greatest at those peak loads:
For simplicity I ignored the power consumption of ancillary equipment including the displays since peripherals would be the same and, when it comes to the displays:
- Apple don’t provide power consumption figures for the Studio Display
- Even if they did, spec-wise it is more or less identical to the integrated display in the iMac and since I don’t run that at max brightness (the 100 nits brighter Studio Display being the only significant difference between the two), it would be safe to assume that if running at the same brightness levels, power consumption should be within a gnat’s spit of each other.
i.e. I am considering only the difference in power consumption of the CPU & GPU package.
So here are the numbers I came up with:
| Typical household electricity consumption per Day | 56 kwh |
| Daily Cost | $ 11.96 |
| $/kwh | $ 0.21357 |
| M1 Ultra power consumption per Day @ 100W | 2.4 kwh |
| Intel i9 + Radeo Pro 5700 power consumption per Day @ 200W+250W | 10.8 kwh |
| M1 Ultra power saving per day | 8.4 kwh |
| $1.794 | |
| M1 Ultra Mac Studio price premium over equiv. Intel i9 iMac | $ 4,500 |
| Time to recoup cost in power savings | 6 yrs 319 days |
$’s are NZ $’s.
Electricity cost does vary over the course of the year, but the current price here in NZ sits nicely at the median cost at this time of year, so is usefully indicative. I had to calculate our kwh unit cost since our electricity provider does not provide that information directly, but does provide daily cost and consumption stats, so it can be derived. NOTE: This does include a component of “fixed line cost”.
The actual time to recoup the additional cost would be greater if you don’t leave your system on 24×7.
The actual time to recoup the additional cost will be greater in any event since the idea that these systems would run at sustained, maximum load is wholly unrealistic; the power consumption advantage of the M1 vs Intel, whilst still significant, is reduced at lower loads.
Conclusion
Whilst the performance per watt advantage of Apple Silicon is undeniable, unless you intend holding on to your device for a very long time indeed, it does not justify nor offset the Apple Tax being levied on the Mac Studio.
Arguably, if you are someone that leaves your system on 24×7, there are likely to be greater savings to be had by employing judicious power saving on the Intel equipment than by switching to M1 silicon.
If “long term energy cost savings” are a factor for you in considering the additional outlay on one, you should do some careful calculations first (more careful than the rough and ready approach here, that’s for sure! :)).
And finally, if you are enthusiastic about energy saving in and of itself and less concerned with recouping the costs of the more expensive equipment to achieve it, then the arithmetic is obviously very different.
Footnote: Caveats Galore!
After compiling the above numbers, I then found this CPU comparison:
Which suggests a different differential [sic] for the CPU power efficiency than for the GPU, above. It isn’t clear to me whether the 60W of power drawn by the Ultra as a CPU on this chart is included in the 100W referenced on the GPU comparison or in addition to it.
If it is included, then there’s no real impact on the figures I’ve come up with here as I had already assumed a single 100W draw for the combined CPU+GPU M1 package.
But if not, then the reduced differential for the CPU (160W/450W vs 100W/450W) and increase in overall consumption, would result in a dramatically reduced cost/benefit case for the Mac Studio, at least in M1 Ultra configurations.
Apple’s way of comparing their silicon with the competition has been frustratingly oblique from the outset, making it very difficult to draw solid conclusions.
If anyone can shed any light into this murk and clarify things (with actual numbers, not interpretations of Apple’s graphs or footnotes), I’d be very interested.