M1Pro 32GB vs M1Max 64GB: battery life and performance test suggestions?

[yurkennis]

Still wondering how exactly M1Max 64GB compares to M1Pro 32GB in terms of battery life, real-life performance or anything else?

I have both of them for side-by-side tests for limited time. Any suggestions on tests to make are welcome: reproducible, scientific, reliable, actionable, you name it?

Personally, I'm most interested in:

• comparing real-life time on battery charge; "work done per battery Watt spent"
• whether I actually need 64GB memory for my workload
• whether Max vs Pro makes any difference for me as a non-creator (occasional gaming? transcoding movies for travel devices? else?). Entry-level 25-cores Max assumed (which is only +$200), not the 32-cores one I've got for tests

I've got:

Max: 32 cores GPU / 10 cores CPU, 64GB, 1TB.

Pro: 16 cores GPU / 10 cores CPU, 32GB, 1TB.

 

[yurkennis]

My first test: video playback, two side-by-side Safari windows each playing the same 8K video (New York), time per single battery charge:

M1 Pro 32GB (16 GPUs): 8h 57m
M1 Max 64GB (32 GPUs): 7h 18m

MAX is 81% of PRO time, PRO is 122% of MAX time

Temperature (average CPU, per iStat):
- after 1st period (5h 50m): M1Max 39ºC, M1Pro 36ºC
- after 2nd period (1h 05m): M1Max 45ºC, M1Pro 46ºC

Both ran from 100% charge to 1% battery left.
Both Max and Pro dropped about 81% of video frames, as per YouTube "Stats for nerds".
Surprisingly, MAX played AND dropped +1% frames during first period.
However, during second period MAX played +5% while dropped +7% frames.

There is some chance that I didn't ensure they operated on the same brightness, so maybe I will run the test again.

Any suggestions welcome.

 

[blaraka]

Wonderful tests - just what I was looking for.

 

[MrGunnyPT]

This is such a great information! Thanks for sharing.

 

[throAU]

Given fast charge I’d say just work out if you need the horsepower and power consumption be damned.

Don’t forget these have fast charge!

 

[LinkRS]

My first test: video playback, two side-by-side Safari windows each playing the same 8K video (New York), time per single battery charge:

M1 Pro 32GB (16 GPUs): 8h 57m
M1 Max 64GB (32 GPUs): 7h 18m

MAX is 81% of PRO time, PRO is 122% of MAX time

Temperature (average CPU, per iStat):
- after 1st period (5h 50m): M1Max 39ºC, M1Pro 36ºC
- after 2nd period (1h 05m): M1Max 45ºC, M1Pro 46ºC

Both ran from 100% charge to 1% battery left.
Both Max and Pro dropped about 81% of video frames, as per YouTube "Stats for nerds".
Surprisingly, MAX played AND dropped +1% frames during first period.
However, during second period MAX played +5% while dropped +7% frames.

There is some chance that I didn't ensure they operated on the same brightness, so maybe I will run the test again.

Any suggestions welcome.

Hi yurkennis,

What is "1st period" and "2nd Period"? Also, are you saying that both systems dropped 81% of the frames while playing? That sounds horriblel?

Thanks!

Rich S.

 

[yurkennis]

Hi yurkennis,

What is "1st period" and "2nd Period"?

The post was poorly edited, now it's a bit better.

Anyway, the test was run with a break: first period (5h 50m) late in the evening, second one in the morning (1h 05m on both, plus some more time on the Pro alone).

Also, are you saying that both systems dropped 81% of the frames while playing?

Exactly. Keep in mind it's 8K video, though--and there are two of them playing side by side, in a browser.

 

[Fomalhaut]

These tests are interesting, but unfortunately, they don't allow us to know what proportion of the difference is a resultof having double the RAM, or double the GPU cores. It would have been nice to see the test with a 32GB M1 Max so that we can see the impact of the GPU cores on battery life. I suspect that the increased RAM may be responsible for a significant part of the reduced battery life.

 

[yurkennis]

These tests are interesting, but unfortunately, they don't allow us to know what proportion of the difference is a resultof having double the RAM, or double the GPU cores. It would have been nice to see the test with a 32GB M1 Max so that we can see the impact of the GPU cores on battery life. I suspect that the increased RAM may be responsible for a significant part of the reduced battery life.

See my next reply below. Looks like the entire premise "Pro is cooler system with better battery life" is only marginally true.

 

[yurkennis]

Another run for side-by-side video playback, this time:

• brightness was definitely synced at 4 ticks down from the maximum
• Low Power Mode enabled
• video is now 4K (as 8K is not available for this video under Safari any more)
• as before, the entire playback was recorded with Cmd-Shift-5

Time per single battery charge:
M1 Pro 32GB (16 GPUs): 8h 14m
M1 Max 64GB (32 GPUs): 7h 49m

So the difference now is only 5%(!).

Comparing these to my original test, some takeaways:

• +16 GPU cores, +32GB RAM, Max vs Pro together result in only 5% difference in Low Power Mode
• at low workloads, brightness level alone is much bigger factor than both Low Power Mode, Pro vs Max, number of GPU cores or RAM size

---

Small-print disclaimers on the test (which unlikely affect the bottom line):

• there was a ~half-hour pause in playback, 8m longer on Pro (Pro 34m, Max 26m)
• playback was shortly switched to lower resolution: 720p on Pro (8m in single window plus 2m in both windows); 1440p on Max (5m in both windows)

Both were spontaneous, possibly a result of short network failure.

 

[yurkennis]

These tests ... don't allow us to know what proportion of the difference is a result of having double the RAM, or double the GPU cores.

If you have any suggestions on how I can measure this with the 32gbPro vs 64gbMax I have in hand, please let me know.

 

[yurkennis]

My next test: offline side-by-side video playback, this time by VLC.

Time on full battery charge (100% charge to shutdown on battery depleted):

• difference is only 6%(!)
• M1 Pro 32GB (16 GPUs): 12h 08m
• M1 Max 64GB (32 GPUs): 11h 24m

Thermals, sensors, power draw (all per iStat Menus 6.61 / 1185; all averages are mine, "non-scientific"):

• fans: both Max and Pro had running both fans all the time, at constant 1499 rpm
• CPU core average: 40ºC on both Max and Pro (non-scientific average)
• Total Power draw: Max 8.7W, Pro 8.03W (non-scientific average)

Test conditions:

• brightness: 4 ticks from max
• Normal Power Mode
• video: No Time To Die, 3840x2160 H265 MKV (75GB)
• video looped with "Repeat One" playback mode
• full-screen playback with menu bar made visible by a hovering mouse pointer
• the entire playback was screen-recorded with Cmd-Shift-F5

 

[kardus]

I assume this was on regular/high power mode for both for your last test? Does the power saving mode throttle the cpu at all or is it just the gpu?

I'm really interested in what the battery life differences between pro and max will be for my use case. My idea was hopefully getting both the pro and max to test out myself, but it seems like due to shortages the return window for the pro will expire before I will even get to test out the max. So I will basically have to decide if I am keeping the pro or not before being able to test myself.

My typical max usage would be something like:
- 11-16 screen-on hours daily
- Firefox running multiple tabs, web browsing and reading pages for most of the day
- mixed video playback usage, largely 1080p but some 4k/higher reses as well
- UTM/qemu running an ARM linux and/or windows vm (use in the guest OS will mostly be idle/web browsing a few tabs in firefox, maybe some photoshop/office software)
- light photoshop/video editing/productivity software use
- light dev work/software compilation
- light video rendering, 5-10 mins daily

I don't mind giving up maybe 5-10% of battery life for the option of better performance (saving time in other aspects when I don't want to wait), but losing much more battery life than that might be too much of a tradeoff. I think for my use case pro seems more of the logical choice that will result in greater battery life, but I am getting half the memory in that case. I'm not too sure how GPU acceleration in utm/qemu would affect battery life (for example if i were to be using photoshop in windows in qemu).

Are there any other ways to manually control/throttle the gpu usage? The inability to control these things was my major reason for distancing from mac os/apple over 5 years ago in the first place.

 

[yurkennis]

I assume this was on regular/high power mode for both for your last test?

As follows:

Regular power mode:

My first test: video playback, two side-by-side Safari windows each playing the same 8K video (New York), time per single battery charge

Low power mode:

Another run for side-by-side video playback, this time:

• Low Power Mode enabled
• video is now 4K (as 8K is not available for this video under Safari any more)

Regular power mode:

My next test: offline side-by-side video playback, this time by VLC.
...
Test conditions:

• ...
• Normal Power Mode

---

---

Does the power saving mode throttle the cpu at all or is it just the gpu?

I never tried to research that.

---

---

I'm really interested in what the battery life differences between pro and max will be for my use case.

I do not have much experience with photo editing, video rendering, virtual machines--so take the following with the grain of salt. Basing on my limited knowledge, your workload will unlikely saturate even 24 GPUs and, quite likely, won't even need 64GB. Eg Windows under Parallels takes nearly zero energy / CPU, and is quite friendly to memory pressure. The only serious reason for going 64GB could be: regularly working with datasets / projects of nearly that size--this is not only a consensus among all reviews, but also my own takeaway from 2 months of testing 64GB Max vs 32GB Pro.

Are there any other ways to manually control/throttle the gpu usage?

Subjectively, I would expect that: idle GPUs take virtually zero energy; moderate load results in exactly as much energy required to get job done; under heavy GPU load you would prefer speed over battery spending rate.

 

[vivid.gear71]

Thanks for doing this and happy since I have M1Max 64

 

[brosenz]

More RAM and more GPU cores require more energy, It will run warmer, therefore with higher fan usage, all of that will have a battery toll

 

[yurkennis]

More RAM and more GPU cores require more energy, It will run warmer, therefore with higher fan usage, all of that will have a battery toll

The battery life difference is only 5% under moderate workload, per my experiments earlier on the thread.