M4 Mac Announcements

[KingOfPain]

Gotta remind you here that chip design is a slow process.

And here I thought it was as easy as ordering a pizza...

 

[Cmaier]

And here I thought it was as easy as ordering a pizza...

making a pizza is slow. First you need a water buffalo. I don’t remember the rest of the steps.

 

[mr_roboto]

Oh and because I call it out when other companies do it: Apple seriously comparing the M4 Max (and even the M4 Pro) to the Lunar Lake 258V is silly. Of course the M4 Max and Pro are faster, the Lunar Lake chip isn't in the damn weight class, not even close (especially to the Max GPU and to the CPU for either of them). I suppose the Arrow Lake laptops aren't available for testing yet and Apple wants to compare themselves against Intel and Apple's clearly got a thing for these new "AI PC" monikers, but still.

I can see some justification in selecting the 258V. I looked at Intel's Lunar Lake processor list and all mobile segment LL CPUs are 4P+4LPE core configurations. (LPE = low power E core, Intel's got three kinds of cores now.)

Intel® Core™ Ultra Processors (Series 2)

ark.intel.com

There's nothing higher than 4 P cores in any currently available Intel mobile segment CPU built on TSMC N3. Everything with more is a 125W TDP desktop part. Furthermore, while there's other higher end members of the 4P+4LPE mobile family, Intel's binning is so flat that I'm unsure why they bothered (seems to just be reflex for them). The 258V seems like a decent representative of the whole range.

Apple probably should have tossed one of those 125W desktop parts into the mix just for fun, because I bet M4 Max would still do well, just not 2.5x. In a development which I'm sure will shock everyone reading this, the PPA (performance-power-area) of Apple and Intel cores on N3 process nodes seems to indicate that chip design is not, in fact, just like ordering pizza.

 

[dada_dave]

I can see some justification in selecting the 258V. I looked at Intel's Lunar Lake processor list and all mobile segment LL CPUs are 4P+4LPE core configurations. (LPE = low power E core, Intel's got three kinds of cores now.)

Intel® Core™ Ultra Processors (Series 2)

ark.intel.com

There's nothing higher than 4 P cores in any currently available Intel mobile segment CPU built on TSMC N3. Everything with more is a 125W TDP desktop part.

Aye I admitted that Arrow Lake mobile isn’t out yet. For the current generation chips it would have to be AMD and Qualcomm and both of them would get curb stomped by the the Pro, never mind Max - the Strix Halo might do better but like Arrow Lake mobile, it isn't out yet. But I suppose as Cliff said AMD represents supposedly a minority of laptop market share, I’ve seen estimates of about 20% of the client PC market for both desktop and laptop, while Qualcomm’s laptop share is unknown but presumably marginal. But the Intel LL CPU/GPU losing 2.5x/4x is just not a fair argument for anything meaningful. In terms of the overall devices, well, the M4 Max starts at $3200. Meanwhile, the MSI laptop Apple compared themselves to costs $1400. I'm not sure they can count as competitors. So I admit Apple is a bit stuck in good comparisons if it has to be Intel ... unless ...

Furthermore, while there's other higher end members of the 4P+4LPE mobile family, Intel's binning is so flat that I'm unsure why they bothered (seems to just be reflex for them). The 258V seems like a decent representative of the whole range.

Apple probably should have tossed one of those 125W desktop parts into the mix just for fun, because I bet M4 Max would still do well, just not 2.5x.

Oh it absolutely would have. Very well in fact. I would've loved to see them make that comparison in lieu of a mobile chip. They couldn't do it versus the highest end desktop chip, but the middle ones with still comparatively high power draws? Oh yeah. That would've been funny (albeit maybe too petty for a professional press conference) to say we couldn't find a relevant mobile chip that could actually compete with the Max in CPU power so here is an Intel desktop chip ...

However, for GPU, they could've chosen an appropriate dGPU to compare against the Max since there isn't anything like it in the iGPU space until Strix Halo comes out next year.

In a development which I'm sure will shock everyone reading this, the PPA (performance-power-area) of Apple and Intel cores on N3 process nodes seems to indicate that chip design is not, in fact, just like ordering pizza.

 

[Yoused]

making a pizza is slow. First you need a water buffalo. I don’t remember the rest of the steps.

Next is the surgery. Then, once they are properly attached and healed up, the arduous task of teaching it to fly begins. I think that part involves squirting tabasco sauce on its hooves.

 

[Altaic]

Next is the surgery. Then, once they are properly attached and healed up, the arduous task of teaching it to fly begins. I think that part involves squirting tabasco sauce on its hooves.

Turning a water buffalo into a water + air = lightning buffalo, now we’re talking. Always wondered how pizza gets cooked.

 

[theorist9]

Oh and because I call it out when other companies do it: Apple seriously comparing the M4 Max (and even the M4 Pro) to the Lunar Lake 258V is silly. Of course the M4 Max and Pro are faster, the Lunar Lake chip isn't in the damn weight class, not even close (especially to the Max GPU and to the CPU for either of them). I suppose the Arrow Lake laptops aren't available for testing yet and Apple wants to compare themselves against Intel and Apple's clearly got a thing for these new "AI PC" monikers, but still.

what laptop would you like Apple to compare to? What’s comparable to an MBP nowadays?

I agree with @dada_dave that comparing to Lunar Lake 258V because it's Intel's most recently-released laptop processor (Q3 2024), while not accounting for the class of that processor, doesn't make sense. At the same time, if Apple chose to compare the M4 Max MBP against the most powerful current laptop processor (i9-14900HX, released Q1 2024), there would be complaints that it wasn't the most recent. So the correct and sensible solution (and when do marketing departments act thusly?) would be to bracket the MPB by including both in the comparison. That would give a much more complete accurate picture about how Apple fares against today's best.


More broadly, I think the best comparison would be to show how the M4 Max MBP does against the most powerful current x86 laptop configuration (i9-14900HX with NVIDIA 4090 laptop GPU; see example below*), as well as the most efficient laptop that consumes about the same power as the M4 Max. Then you can use former to compare max performance (both on and off battery) and the relative wattage and fan noise needed to attain that performance, and the latter to compare max performance at iso power.

*Example:
Razer Blade 16" Gaming Laptop
Specs: i9-14900HX, RTX 4090, 64 GB 5600 MHz DDR 5 RAM, 4 TB SSD, 240 Hz 2560 × 1440 OLED display, 22 mm thick, 5.4 lbs, $4,400

https://www.amazon.com/Razer-Blade-Gaming-Laptop-i9-14900HX/dp/B0D21RBD4C/ref=pd_ci_mcx_pspc_dp_2_t_4?pd_rd_w=9Fo9f&content-id=amzn1.sym.08bf8351-b34a-4ca2-ab23-21462ed87c92&pf_rd_p=08bf8351-b34a-4ca2-ab23-21462ed87c92&pf_rd_r=NFG57XZ7HQWGKWW257ZN&pd_rd_wg=hSdxZ&pd_rd_r=18da41cf-b3ba-4cfc-9696-646a661c8bb9&pd_rd_i=B0CPP726HK&th=1

Comparable 16" M4 Max MBP:
Specs: 16-core CPU, 40-core GPU, 64 GB 8533 MHz LPDDR5x RAM, 4 TB SSD, 120 Hz 3456 x 2234 LED display, 25 mm thick, 4.7 lbs, $5,200

 

[dada_dave]

Turning a water buffalo into a water + air = lightning buffalo, now we’re talking. Always wondered how pizza gets cooked.

I'm too tired to choose between a joke involving buffalo mozzarella or buffalo wings.

I agree with @dada_dave that comparing to Lunar Lake 258V because it's Intel's most recently-released laptop processor (Q3 2024), while not accounting for the class of that processor, doesn't make sense. At the same time, if Apple chose to compare the M4 Max MBP against the most powerful current laptop processor (i9-14900HX, released Q1 2024), there would be complaints that it wasn't the most recent. So the correct and sensible solution (and when do marketing departments act thusly?) would be to bracket the MPB by including both in the comparison. That would give a much more complete accurate picture about how Apple fares against today's best.


More broadly, I think the best comparison would be to show how the M4 Max MBP does against the most powerful current x86 laptop configuration (i9-14900HX with NVIDIA 4090 laptop GPU; see example below*), as well as the most efficient laptop that consumes about the same power as the M4 Max. Then you can use former to compare max performance (both on and off battery) and the relative wattage and fan noise needed to attain that performance, and the latter to compare max performance at iso power.

*Example:
Razer Blade 16" Gaming Laptop
Specs: i9-14900HX, RTX 4090, 64 GB 5600 MHz DDR 5 RAM, 4 TB SSD, 240Hz 2560 × 1440 OLED display, 22 mm thick, 5.4 lbs, $4,400

https://www.amazon.com/Razer-Blade-Gaming-Laptop-i9-14900HX/dp/B0D21RBD4C/ref=pd_ci_mcx_pspc_dp_2_t_4?pd_rd_w=9Fo9f&content-id=amzn1.sym.08bf8351-b34a-4ca2-ab23-21462ed87c92&pf_rd_p=08bf8351-b34a-4ca2-ab23-21462ed87c92&pf_rd_r=NFG57XZ7HQWGKWW257ZN&pd_rd_wg=hSdxZ&pd_rd_r=18da41cf-b3ba-4cfc-9696-646a661c8bb9&pd_rd_i=B0CPP726HK&th=1

Comparable M4 MBP:
16" Max, 40-core GPU, 64 GB 8533 MHz LPDDR5x RAM, 4 TB SSD, 120 Hz 3456 x 2234 LED display, 25 mm thick, 4.7 lbs, $5,200

Although ... doing everything you suggested is generally more involved than Apple typically likes to get in its slides recently (certainly this presentation - they did more of it at the start of the M-series), sadly. Indeed presenting all that info, especially fan noise*, would atypical for anyone as I know and is typically done by review sites. But yes I agree that I'd have like to have seen a better comparison - although the 4090 (especially on wall power/performance mode) would be way beyond the M4 Max the other way IMO. Truly "fair" comparisons are hard, but LL is definitely not a fair comparison for the M4 Max.

*not saying it isn't important but I don't think I've ever seen a promotional bit report fan noise. On the other hand, given Apple's penchant for both cool and quiet operations ... maybe they should.

 

[KingOfPain]

A few more thoughts on the comparison between the chips...

I forgot (suppressed?) that the M3 Pro went down to 150GB/s bandwidth from 200GB/s in the previous two Pro versions. One more factor for the M3 Pro being primarily a laptop chip.
This is the first time a Max chip breaks the 400GB/s bandwidth threshold. Probably just comes down to the RAM timing, but still interesting.

I'm not sure if a Mac Studio with an 36GB M4 Max could be slightly cheaper than a Mac mini with 48GB M4 Pro or not.
If that were the case, the smallest Mac Studio would be the better choice as long as one doesn't need the extra RAM.

I'll still wait to see what the Mac Studio brings to the table.
Please correct my numbers, but my current estimations from MacBook Air M1 to Mac mini M4 Pro are:

• About 30% ST performance increase.
• Almost 3x MT performance.
• Over 3x graphics performance.

This definitely would be a beefy upgrade...

ArsTechnica posted a comparison article:

Apple’s M4, M4 Pro, and M4 Max compared to past generations, and to each other

M4 and M4 Pro get core count bumps; M4 Max relies on architectural improvements.

arstechnica.com

Two takes from it:

• the “E” in “E-core” does not stand for “exciting”
• The M4 Pro is the most interesting year-over-year upgrade, though this says more about the M3 Pro than anything else.

I would agree with the last one, but I think one should not underestimate the E-cores.

 

[theorist9]

I'm too tired to choose between a joke involving buffalo mozzarella or buffalo wings.


Although ... doing everything you suggested is generally more involved than Apple typically likes to get in its slides recently (certainly this presentation - they did more of it at the start of the M-series), sadly. Indeed presenting all that info, especially fan noise*, would atypical for anyone as I know and is typically done by review sites. But yes I agree that I'd have like to have seen a better comparison - although the 4090 (especially on wall power/performance mode) would be way beyond the M4 Max the other way IMO. Truly "fair" comparisons are hard, but LL is definitely not a fair comparison for the M4 Max.

*not saying it isn't important but I don't think I've ever seen a promotional bit report fan noise. On the other hand, given Apple's penchant for both cool and quiet operations ... maybe they should.

Yeah, maybe a 4080 would be a better comparison, since those are more common, and their on-wall power consumption is not so off-the-wall .

 

[mr_roboto]

Truly "fair" comparisons are hard, but LL is definitely not a fair comparison for the M4 Max.

To be clear, despite my post, I agree. Just thought that it was possible to see how Apple's benchmark marketing folks talked themselves into going with that comparison.

Also, it would've been 100% fine (and far more meaningful) if they'd compared single thread speed instead, just as a way to say "hey Intel, you might be on almost the same node now, but you're still eating our dust".

 

[theorist9]

I'm not sure if a Mac Studio with an 36GB M4 Max could be slightly cheaper than a Mac mini with 48GB M4 Pro or not.
If that were the case, the smallest Mac Studio would be the better choice as long as one doesn't need the extra RAM.

I'm guessing an M4 Max Studio with 36 GB RAM should be the same as the M2 with 32 GB. With the lower-end Max chip and 512 GB storage, that's $2,000.

And the upper-end M4 Pro Mini with 48 GB/512 GB is also $2,000. So yeah, for the same price it seems you could replace a high-end Pro chip with a low-end Max chip, and also gain more ports, if you were willing to sacrifice 12 GB RAM.

This of course assumes there will be an M4 Max Studio, rather than an M4+ or M5.

 

[theorist9]

Welp, I went ahead and ordered my upgrade for the next few years. Interestingly, it says I should be able to grab it day one from the Apple Store near my office. Usually have to wait a bit for these CTO builds.

14" MBP with M4 Pro, 48GB, 2TB

You shouldn't have told us. Now we're going to expect detailed graphs and charts!

 

[casperes1996]

And here I thought it was as easy as ordering a pizza...

Isn’t it? A very high IQ person told me it is

 

[KingOfPain]

Isn’t it? A very high IQ person told me it is

IQ is similar to latency, the lower the better, right?

EDIT: To add more than a joke. At times like these it's painful that Anandtech is dead. I have no idea where else I could find deep-dives for these chips.

 

[The Flame]

EDIT: To add more than a joke. At times like these it's painful that Anandtech is dead. I have no idea where else I could find deep-dives for these chips.

Chips&Cheese will do a deep dive if you send them a device.

 

[Nycturne]

I'm not sure if a Mac Studio with an 36GB M4 Max could be slightly cheaper than a Mac mini with 48GB M4 Pro or not.
If that were the case, the smallest Mac Studio would be the better choice as long as one doesn't need the extra RAM.

In this case, you’d likely be trading RAM for GPU cores and a larger box. Unless you need more than 36GB, the Studio would likely be the better value IMO.

I'll still wait to see what the Mac Studio brings to the table.
Please correct my numbers, but my current estimations from MacBook Air M1 to Mac mini M4 Pro are:

• About 30% ST performance increase.
• Almost 3x MT performance.
• Over 3x graphics performance.

This definitely would be a beefy upgrade...

I don’t know about the other numbers, but seems more like ~58% more ST perf, based on Geekbench. The M1/M2 iPad Pro Geekbench benchmarks aren’t that far off from the Mac benchmarks of the same chip, so unless the M4 numbers somehow are bad, 30% seems pretty low.

 

[Nycturne]

You shouldn't have told us. Now we're going to expect detailed graphs and charts!

The bad news is that my measurements will be in “M33s per hour".

 

[theorist9]

The bad news is that my measurements will be in “M33s per hour".

You'll have to explain that one to me. Cortex M-33's? Or would that be how many photos of Messier M33 you can process per hour? [You did mention getting into astrophotography....]

 

[ArgoDuck]

You'll have to explain that one to me. Cortex M-33's? Or would that be how many photos of Messier M33 you can process per hour? [You did mention getting into astrophotography....]

It’s a pretty nice galaxy…