M4+ Chip Generation - Speculation Megathread [MERGED]

[smithdr]

I am considering updating my 16” MBP M1Max to a 16” MBP M4Max but only if there is an update to the media engine/encoders. It seems I read something about this somewhere, but do not recall. Does anyone have any insight if the media engine/encoders are going to be updated in the M4Max?

Don

 

[ondioline]

They added AV1 decoding in the M3, so “yes,” but the rest is the same afaik.

 

[EugW]

I am considering updating my 16” MBP M1Max to a 16” MBP M4Max but only if there is an update to the media engine/encoders. It seems I read something about this somewhere, but do not recall. Does anyone have any insight if the media engine/encoders are going to be updated in the M4Max?

M3 series got a new media engine with AV1 support, and M3 Max has faster Final Cut Pro export speeds vs M1 Max (but it would depend upon the settings and the content).

 

[Boil]

View attachment 2436073

This picture provides more context. The top of the picture includes the plate with all the ports, indicating just how wide the Mac mini chassis is, meaning the logic board only takes up about 3/4 of the width of the chassis.

That "empty space" to the right of the logic board is where the PSU goes...

 

[EugW]

That "empty space" to the right of the logic board is where the PSU goes...

Yes we are aware. It was already discussed in our posts.

To summarize my prior discussion on this: We don't know for sure if the new Mac mini will have an internal PSU, but if it does, it does not have to go beside the logic board. While the Mac mini would be built narrower, it could also be built taller to provide space vertically for the PSU.

 

[PaulD-UK]

Or flat below. The ASDisplay shows how thin Apple can make the PSU. It’s behind the 11mm thick screen panel.

 

[deconstruct60]

I am actually more concerned, or let's say more interested in how Apple tackles the height. The current mini height has a perhaps coincidental ideal height of less than 1U. Many of the 1 Litre small form factor PC kind of followed this dimension blue print since people found out this is nice size for half rack deployment. If you are just putting one or two in your cabinet you won't turn it sideways like the data center would. Being less than 1U means you can easily slide it in a 1U shelf, or a drawer even.

Just simply dropping it on a plain 1U shelf isn't really secure. ( Ground does shake in some places. )

If it is 2U and get to put more on a shelf that is the primary trade-off. I doubt one-sy , two-sy only rack deployments are going to push the overall design. At the 'one' level the footprint as a literal desktop likely will dominate the design process. The smaller desktop footprint likely will win (there is other stuff on peoples desk competing for space). Two is weak also since Apple has stackable on desktop as a factor also ( some use cases doesn't even need a rack). Rack customers that buy 1-2 aren't going to look any different to Apple in aggregate numbers than desktop customers that buy 1-2 .

The 'rack thing' likely gets more traction in the 20, 50, 100+ range.

As a result I think less than 2U height is going to be its limit. Apple may look at the Studio's ratio between its dimensions, and just down scale it for the new mini, so to maintain a consistent look.

Less than 2U, but greater than 1U doesn't really 'buy' much. Pragmatically it is a 2U.
3U is viable. Some back of the envelope numbers.


18.3 wide ( 19" rack )
1.75 tall. ( 1U )

Mini 1.41 (pragmatically 1.5 ) tall and 7.75 (pragmatically 8" ) wide (and deep)
[ Note; Apple TV on it side is 3.66" which is incrementally over 2U. ]

flipped onto side 4.42U so 5U .

On its side on a 5U tray can fit 12 Minis.

5 1U tray can each have 2 Minis. 5 x 2 = 10 .
( horizontal is less dense which is why few scaling deployments use that).



If Apple went to a smaller Studio (let height approximately double and shave 2 inches off other dimensions).
3" tall and 5.75" (and deep )


flipped onto side 3.28U so 4U

On its side a 4U tray can fit 6 'taller' Minis.

2 2U trays can each have 3 Minis 2* 3 = 6 [ so doesn't matter]


If Apple squeeze a bit more to an even smaller Studio 2.75" tall and 5.25"

On its side a 3U exactly

flipped onto side 3U tray can fit 6 Minis ( if keep the 3" tall still 6 ).

2 2U trays can each have 3 Minis 2 * 3 = 6


42U high rack there are

8 5Us 8 * 12 = 96 classic Mini
10 4Us 10 * 6 = 60 smaller Studio
14 3Us 14 * 6 = 84 even smaller Studio
21 2Us 21 * 3 = 63 smaller Studio (horizontal)
42 1Us 42 * 2 = 84 classic Mini (horizontal)

Packing the width of the rack is the more effective scaling/density factor for deploy Mac systems. (given macOS
restrictions on multitenant hosting, that is primarily scale issue. ). 4U tall when flipped, but 8 on a tray would be miss of the 3U count. Even 9 would miss the classic 5U mini tray count. The 5U classic mini is about a 14% uplift from the 3U count (or -13.5% backslide).

The depth of the rack reduced width-depth Mini's take up is getting smaller also. So with
multiple rows of racks there could be footprint savings there. The customer racks for
MacStadium/Coloc have mini's on both sizes (rows of Mini's only 8" deep doesn't really make sense).
That means the USB ports on back not open access on the backside of the rack. There is another mini with no
front ports there either. If there were maintenance ports on both side of rack that would be an
improvement.

A standard depth 1U server is substantially deeper than that. And will need to consolidate
network traffic into backhaul switches coming off the rack anyway (the direct Ethernet connection from
Mini isn't going to leave the rack. )

Squeezing the power supplies out into a bigger 'brick' is just a 'balloon squeeze'. Mini gets smaller but
the tray now has to host a bigger brick. Likely isn't going to save tray space. Replacing failed supplies would
be easier though. (desktop wise though it is another brick soaking up space somewhere else. )


To put some SoC performance context to this, I suspect that Apple would talk to the Scaling folks that the plain M4 is so much better than the old Intel stuff that customers that might have rented 2 (or more) Minis might be collapsing down to just one. The rack density of the systems went down, but the compute per rack went up. So there will be a need fewer systems impact. ( which brings power and datacenter space savings . ). Big picture, it is deploying enough systems to get the work done as opposed to deploying the most systems possible. Mini's with a Mn Pro are going to help get more work done with fewer systems (than old days). Folks running a website and some odd-bits of server hosting don't have skyrocketing compute requirements every year.


P.S. A Mac Pro pro with 6 PCIe cards that each had two plain M4 would be 13 Macs in a 5U space. That would be more Macs than the classic Mini in 5U. [ Not really set up internally to power that currently. but 'thinner' (horizontally) and 5U can pack density. ] Even on M4 class per PCI-e slot and the plus much bigger one isn't too bad if actually need a small cluster ( 6 clients and server. ) if can route the intra-cluster network traffic over the PCI-e layer.

 

[Chuckeee]

flipped onto side 3U tray can fit 6 Minis ( if keep the 3" tall still 6 ).

But that is just a single row, the advantage of the smaller mini is that it’s feasible to install multiple rows onto a single 3U shelf. This applies to your other calculations too.

 

[deconstruct60]

underwhelmed users that compared it to the Nvidia 4090 (roughly twice as performant as the highest spec Apple).

If a hypothetical 600W Apple M2 machine was required to compete with the Nvidia 4090, and next year, power users will be comparing the Mac to the 5090 (which is being estimated as 50%-70% faster than the 4090) - then what Apple needs to build will need to be something equivalent to a 960W M2 machine.

It is highly unlikely that Apple is going to get into a power consumption pissing match with Nvidia. Apple consistently gets up and preaches sermons on Perf/Watt. Doubtful they are changing religion any time soon. The 5090 is going to consume more power and extremely likely going to be even more expensive. The Mac Studio really can't do down either one of those paths in a monomaniacal fashion. Cranking the costs even higher draws even more fratricide with the Mac Pro. The Mac Pro already is in that higher cost zone so has taking the unit volume 'hit'.

The Mac Pro is also sitting there already with a substantially larger power supply (than the Mac Studio). [ May loose the 8-pin aux power if the primary SoC soaks up far more power. However, the chassis and power supply to play the game is already in place. Likely one reason why it wasn't cancelled in the transition. ]

The 4090 - 5090 - 6090 path is increasingly shrinking in scope ( higher and higher prices and fewer and fewer customers). Apple using that as a primary benchmark would be looking at a single tree instead of the whole forest. Nvidia will drift into more and more of a bragging rights card rather than a viable return on investment market card. As long as they can print money from the AI-Datacenter side to cover up the costs they'll play that game.

....- its seems more likely to me than not that the Studio and the Pro will both offer same highest spec SoC.

Mac Mini M4 and M4 Pro SoC.
Mac Studio M4 Max and Ultra SoC.
Mac Pro ... just one.

One of these things is not like the others. Yes the iMac is sitting there was just one , but that is primarily because Apple painted themselves into a corner making the iMac thinner than the iPad (and constraining the SoC placement to under the 'chin'). The iMac isn't the strategic 'king of desktops' that it once was in the Intel era. It is at the fringe of the desktop campaign now. The Mac Pro chassis has no such 'corner painting' constraints slapped on it. It is distinctly missing something that hasn't shown up yet. 8-pin power that is pragmatically a 'bridge to no where' doesn't make much sense.

'Doubling' everything for M-series means pragmatically doubling RAM ( the increased memory I/O comes with increased minimal threshold capacity). It is already the case that Apple delivers more "VRAM" than 4090 ( or 5090) will deliver. Pushing the compute units way up also drives the RAM costs up. Which drives pricing way up. Pragmatically priced even higher than 5090/6090 really won't help much in the aggregate. It will be increasingly more a fringe subgroup buy.

Apple primarily just needs something incrementally bigger than the Ultra for the Mac Pro. "Doubling" isn't necessarily a requirement. It they had something that could use 1, 2, or 3 CPU/GPU blocks that would work.
The path of biggest die coupled to biggest package possible has substantive issues coming down the road in terms of affordability.

 

[deconstruct60]

But that is just a single row, the advantage of the smaller mini is that it’s feasible to install multiple rows onto a single 3U shelf. This applies to your other calculations too.

You only get a single row if the device putting on a 3U shelf is 3U tall. The parenthetic comment is about going from 3" tall when nominal (horizontal ) to 2.75 tall nominal isn't buying more density on the single shelf in the flipped orientation. Making the shelving gap smaller is what is helping across the full rack ( turning a bit over 3U which effectively turns into 4U into 3U. Don't have to 'round up'. ).


Multiple rows effective means multiple shelves. I did the 1U versus 3U calculations. (when you flip the Mini horizontal the count on a row goes substantively down. Even at 5" square footprint. ) If hand waving at stacking minis 3 high on a single shelf all that saves sit the shelf height in a rack which isn't making a huge difference in a 42U rack. Stacking isn't as secure placement as bracketing into place.

 

[Chuckeee]

5 1U tray can each have 2 Minis. 5 x 2 = 10 .

That is off by a factor of 2

 

[wojtek.traczyk]

I am totally buying Apple sermons of Perf/Watt (nice pun @deconstruct60, btw!) — because it’s really lone voice of taking computing-as-we-know-it environmental costs seriously.

The digital industry is spending energy as if every corner of the globe were sunny California with limitless renewable energy sources — and doing so at accelerating rate which seems… nuts if you really think about it (global warming, anyone?).

Sooooo… every time I am considering Apple product strategy I got their 2030 pledge as a background. Thus, making — in a perspective — computing bit less power/material hungry makes a’lotta more sense. Even physically shrinking them (manufacturing plus logistics, anyone?).

 

[Antony Newman]

“It is highly unlikely that Apple is going to get into a power consumption pissing match with Nvidia.…”

If the M4 is twice the performance per watt of the M2 - and Apple are willing to increase the height of the Mac Studio and resort to PGS heat spreaders to dissipate at 4 x times the rate - I am suggesting that Apple would be able to push the 295W continuous spec of the M2 Ultra to 500W continuous in a (quiet) M4 Studio … for those moments when the machine is running flat out (eg video rendering - or real time playback of colour graded footage).

If a 500W M4 SoC is in the same ball park of performance as an Nvidia 5090, and could be put inside a Mac Studio body - I doubt many users would need the 1,280 watts continuous rating of the Mac Pro.

If Apple were to release a 500w package - and moved from the M2 ultra with SoC connections on one side to two cx adjacent sides - they could release:
+) 4 x 125w (500W) M4 SoC <— the “Hidra”
+) 3 x 125w (375W) M4 SoC <— 3 x headed
+) 2 x 125w (250W) M4 SoC <— 2 x headed ‘Ultra’
+) 1 x 125w (125W) M4 SoC <— Base configuration - 25% faster than the M4 Max

Dedicated silicon for the Studio and Mac Pro that could keep power users happy for 3 years.
Higher yields than using 2 x monstrous 250W SoC

 

[leman]

If the M4 is twice the performance per watt of the M2 - and Apple are willing to increase the height of the Mac Studio and resort to PGS heat spreaders to dissipate at 4 x times the rate - I am suggesting that Apple would be able to push the 295W continuous spec of the M2 Ultra to 500W continuous in a (quiet) M4 Studio … for those moments when the machine is running flat out (eg video rendering - or real time playback of colour graded footage).

If a 500W M4 SoC is in the same ball park of performance as an Nvidia 5090, and could be put inside a Mac Studio body - I doubt many users would need the 1,280 watts continuous rating of the Mac Pro.

If Apple were to release a 500w package - and moved from the M2 ultra with SoC connections on one side to two cx adjacent sides - they could release:
+) 4 x 125w (500W) M4 SoC <— the “Hidra”
+) 3 x 125w (375W) M4 SoC <— 3 x headed
+) 2 x 125w (250W) M4 SoC <— 2 x headed ‘Ultra’
+) 1 x 125w (125W) M4 SoC <— Base configuration - 25% faster than the M4 Max

Dedicated silicon for the Studio and Mac Pro that could keep power users happy for 3 years.
Higher yields than using 2 x monstrous 250W SoC

All this assumes that M4 can be clocked substantially higher on demand. That does not seem to be a property of any Apple Silicon design until now.

 

[smalm]

All this assumes that M4 can be clocked substantially higher on demand. That does not seem to be a property of any Apple Silicon design until now.

What a nice way to say "pah, humbug!"

 

[Antony Newman]

All this assumes that M4 can be clocked substantially higher on demand. That does not seem to be a property of any Apple Silicon design until now.

If the M1 Studio Ultra was designed to expel 370w of heat continuously and the M4 enclosure is being made larger (Gurman rumour), it would suggest that Apple intend for the top end M4 Mac studio to have an SoC that uses more power (wattage).

A 30% taller M4 enclosure could mean a Mac Studio that peaks 480w - potentially more if they upgrade the heatsink technology (enough in my speculation to compete head to head with the Nvidia 5090).

Apple Plans to Overhaul Entire Mac Line With AI-Focused M4 Chips

Apple Inc., aiming to boost sluggish computer sales, is preparing to overhaul its entire Mac line with a new family of in-house processors designed to highlight artificial intelligence.

www.bloomberg.com

If Gurman thinks the Hidra is only coming to the Mac Pro - the only explanation I find rational would be if the new M4 Mac Studio enclosure inadequate to dissipate the required heat.

This could mean the M4 Mac Pro gets a >500w four way Hidra SoC and >125w for each Hidra SoC.

The M2 Ultra peaked at 89W : Perhaps a 125W M4 Hidra SoC is physically larger (for additional GPU cores) + has uprated memory bandwidth (20%) + has additional silicon inter-Hidra connectivity (eg 4TB/s) + the same CPU core frequency 4.41GHz.

If Apple are not pushing the thermal envelope required for the M4 Hidra - then Gurman's speculation that it would only be available in the Mac Pro seems unlikely (as Apple will loose $ales from those only considering the Mac Studio form factor).

 

[tenthousandthings]

If the M1 Studio Ultra was designed to expel 370w of heat continuously and the M4 enclosure is being made larger (Gurman rumour), it would suggest that Apple intend for the top end M4 Mac studio to have an SoC that uses more power (wattage).

A 30% taller M4 enclosure could mean a Mac Studio that peaks 480w - potentially more if they upgrade the heatsink technology (enough in my speculation to compete head to head with the Nvidia 5090).

Apple Plans to Overhaul Entire Mac Line With AI-Focused M4 Chips

Apple Inc., aiming to boost sluggish computer sales, is preparing to overhaul its entire Mac line with a new family of in-house processors designed to highlight artificial intelligence.

www.bloomberg.com

If Gurman thinks the Hidra is only coming to the Mac Pro - the only explanation I find rational would be if the new M4 Mac Studio enclosure inadequate to dissipate the required heat.

This could mean the M4 Mac Pro gets a >500w four way Hidra SoC and >125w for each Hidra SoC.

The M2 Ultra peaked at 89W : Perhaps a 125W M4 Hidra SoC is physically larger (for additional GPU cores) + has uprated memory bandwidth (20%) + has additional silicon inter-Hidra connectivity (eg 4TB/s) + the same CPU core frequency 4.41GHz.

If Apple are not pushing the thermal envelope required for the M4 Hidra - then Gurman's speculation that it would only be available in the Mac Pro seems unlikely (as Apple will loose $ales from those only considering the Mac Studio form factor).

In general/layman terms, there are two ways to improve both performance and efficiency, [1] make changes to the “chassis” and/or [2] make changes to the silicon.

The second of these now appears to be on a regular cadence, and that will unfold nicely going forward, M3 was just the start of a long run of steady, stable improvements to Apple silicon, in line with the company’s fifteen years of experience in the iPad/iPhone environment (starting with A4 in 2010). But I think it’s the first of these, irregular adjustments to the chassis, where you’ll see the big, disruptive leaps forward.

We’ve already seen the great age of Apple-silicon chassis changes, from the M1 iMac, to both MacBooks, to the Studio. Not to mention iPad/iPhone. Next, apparently, we’ll see that for the Mini.

I see speculation in this thread above for further moves with regard to the Studio/Pro universe on both sides of the equation: silicon and chassis.

I think we won’t see this with M4, but rather M5, or maybe more like M6 (2026) or M7 (2027). The “Hidra” (why the “i” ?) rumor, plus the unveiling of whatever Apple is doing inside the Private Cloud Compute servers (Mserve, anyone?), it’s all quite fun to watch, as a lifelong Mac user.

 

[altaic]

Good grief. I sometimes wonder if some users are alts of MR staff. Anyway, the thermal dissipation of the M2 Ultra SoC is not the maximum rating of its power supply. Get a clue.

 

[MRMSFC]

To be honest, I'm still not convinced Apple really cares much about this. BTW, this same rack mounting argument was why other people in these forums said they wouldn't change the Mac mini form factor, yet here we are.

As a homelabber with an unhealthy obsession with rackmouting all of his equipment, I really hope that Apple at least keeps the new chassis to 1u.

Also this begs the question, was anyone dissatisfied with the size of the Mac Mini? I’m sure there are plenty of people who just want a chassis change for the sake of “newness” but I don’t think there’s anyone who said “I like the Mac Mini, but it takes up too much space”

 

[EugW]

As a homelabber with an unhealthy obsession with rackmouting all of his equipment, I really hope that Apple at least keeps the new chassis to 1u.

Also this begs the question, was anyone dissatisfied with the size of the Mac Mini? I’m sure there are plenty of people who just want a chassis change for the sake of “newness” but I don’t think there’s anyone who said “I like the Mac Mini, but it takes up too much space”

I thought the current Mac mini was too big. Not unmanageably big but chunky nonetheless for a 2024 (or even 2020) SFF model of its class. Plus, people still complain about Bluetooth issues so something had to change anyway.

That said, it’s unfortunate that going to a smaller size may come with a reduction in the number of USB ports.

 

[Antony Newman]

Good grief. I sometimes wonder if some users are alts of MR staff. Anyway, the thermal dissipation of the M2 Ultra SoC is not the maximum rating of its power supply. Get a clue.

Guilty as charged.

Perhaps a better way to present this should have been:

+) We know the existing Studio Enclosure can dissipate >25% more heat than is used by the M2 Ultra (Rated 370w for the M1 Ultra vs 295w for the M2 Ultra)

+) The M4 Architecture is roughly 2 x the performance per watt of the M2

+) The Nvidia 4090 is performing roughly 2 x as fast (App performance) as the M2 Ultra

+) The lower estimate of the Nvidia 5090 is that it is 50% faster than Nvidia 4090.

+) If Apple do increase the physical size of the M4 Studio enclosure - I speculate this is not because the PCB / RAM / CPU / Sockets / Power supply take up more space - but to increase the air volume to dissipate more heat than the M1 Ultra, whilst still remaining quiet at full load.

If the M1 enclosure can expel 25% more heat than the M2 enclosure it seems reasonable to predict the highest spec M4 Studio could if needed consume >25% more electrical power than the existing highest spec M2 Ultra.

+) If Apple were on the cusp of needing to increase the size of the M4 Studio enclosure, the maximum top spec M4 Studio performance would be :

125% (M4 enclosure uprated to run at M1 Ultra continuous power) x 200% (increase in perf/watt of M4 vs M2) = 2.5 x times the performance of the current top spec M2 Ultra.

This would be about 25% faster than an Nvidia 4090

+) If Apple increase the volume of the M4 Studio enough to increase the heat dissipation by an additional 20% then the maximum performance would be :

125% (M4 enclosure uprated to run at M1 Ultra continuous power) x 200% (increase in perf/watt of M4 vs M2) x 120% (larger M4 studio enclosure) = 3 x times the performance of the current M2 Ultra.

This would be roughly 1.5 x the speed of an Nvidia 4090 - and the lower estimate of the performance of an Nvidia 5090.

+) If a 4 'headed' Hidra (Gurman believed to be destined for the M4 Mac Pro) is found to be roughly 2 x times are fast an M4 Ultra (based on 2 x M4 Max) - and there are 3 x times less Mac Studios sold than Mac Pro's:

Loading…

www.cultofmac.com

Is it more likely that Apple would utilise the Scalable SoC used in the Hidra inside the M4 Mac Studio as well (Rather than 2 x M4 Maxs); And so an M4 Ultra would based on 2 x headed Hidra?

+) If Apple have the space in the M4 Enclosure for a 4 headed Hidra but needed an extra (speculation) 33% of heat to be dissipated by the enclosure, Apple could offer:
- a 3 x headed Hidra in the existing M2 enclosure (133% x 3/4 = 100%)
- introduce thermal throttling (when the 4-headed Hidra is running at full tilt)
- increase the heat dissipation by 33% (eg by increasing the M4 enclosure size).

Speculative Conclusion

If Apple build a top spec M4 Studio using a configuration of Hidra (3 x at full tilt, or 4 x with more aggressive thermal throttling) - they could offer Mac Studio that is up to 25% more performant than a Nvidia 4090.

If Apple increase the internal volume of the Mac Studio enclosure to allow for 20% more continuous heat output - a configuration of 4 x Hidra would allow for a performance this about 50% faster than an Nvidia 4090 (ie the lower estimates of a Nvidia 5090).

 

[DrWojtek]

Guilty as charged.

Perhaps a better way to present this should have been:

+) We know the existing Studio Enclosure can dissipate >25% more heat than is used by the M2 Ultra (Rated 370w for the M1 Ultra vs 295w for the M2 Ultra)

+) The M4 Architecture is roughly 2 x the performance per watt of the M2

+) The Nvidia 4090 is performing roughly 2 x as fast (App performance) as the M2 Ultra

+) The lower estimate of the Nvidia 5090 is that it is 50% faster than Nvidia 4090.

+) If Apple do increase the physical size of the M4 Studio enclosure - I speculate this is not because the PCB / RAM / CPU / Sockets / Power supply take up more space - but to increase the air volume to dissipate more heat than the M1 Ultra, whilst still remaining quiet at full load.

If the M1 enclosure can expel 25% more heat than the M2 enclosure it seems reasonable to predict the highest spec M4 Studio could if needed consume >25% more electrical power than the existing highest spec M2 Ultra.

+) If Apple were on the cusp of needing to increase the size of the M4 Studio enclosure, the maximum top spec M4 Studio performance would be :

125% (M4 enclosure uprated to run at M1 Ultra continuous power) x 200% (increase in perf/watt of M4 vs M2) = 2.5 x times the performance of the current top spec M2 Ultra.

This would be about 25% faster than an Nvidia 4090

+) If Apple increase the volume of the M4 Studio enough to increase the heat dissipation by an additional 20% then the maximum performance would be :

125% (M4 enclosure uprated to run at M1 Ultra continuous power) x 200% (increase in perf/watt of M4 vs M2) x 120% (larger M4 studio enclosure) = 3 x times the performance of the current M2 Ultra.

This would be roughly 1.5 x the speed of an Nvidia 4090 - and the lower estimate of the performance of an Nvidia 5090.

+) If a 4 'headed' Hidra is found to be roughly 2 x times are fast an M4 Ultra (based on 2 x M4 Max) - and there are 3 x times less Mac Studios sold than Mac Pro's:

Loading…

www.cultofmac.com

Is it more likely that Apple would utilise the Scalable SoC used in the Hidra inside the M4 Mac Studio as well (Rather than 2 x M4 Maxs); An so an M4 Ultra would based on 2 x headed Hidra?

+) If Apple have the space in the M4 Enclosure for a 4 headed Hidra but needed an extra (speculation) 33% of heat to be dissipated by the enclosure, Apple could offer:
- a 3 x headed Hidra in the existing M2 enclosure (133% x 3/4 = 100%)
- introduce thermal throttling (when the 4-headed Hidra is running at full tilt)
- increase the heat dissipation by 33% (eg by increasing the M4 enclosure size).

Speculative Conclusion

If Apple build a top spec M4 Studio using a configuration of Hidra (3 x at full tilt, or 4 x with more aggressive thermal throttling) - they could offer Mac Studio that is up to 25% more performant than a Nvidia 4090.

If Apple increase the internal volume of the Mac Studio enclosure to allow for 20% more continuous heat output - a configuration of 4 x Hidra would allow for a performance this about 50% faster than an Nvidia 4090 (ie the lower estimates of a Nvidia 5090).

Wielding that level of power in that small of enclosure is too great and terrible to imagine!

 

[mr_roboto]

Guilty as charged.

Perhaps a better way to present this should have been:

I'm afraid your posts are a maze of bad information and poor reasoning based on it.

The bad info: you really need to look at Apple's "power consumption and thermal output" support document for a Mac model. When you do you'll find numbers far different from what you were using, both M1 and M2.

Mac Studio power consumption and thermal output (BTU) information - Apple Support

Learn about the power consumption and thermal output of Mac Studio computers.

support.apple.com

Note #2 at the bottom of this document is very important to read and understand:

“Max” is defined as running a compute-intensive test application that maximizes processor usage and therefore power consumption. No external peripherals are attached during testing.

"No external periperhals" is the key phrase. You can't look up the specs of the M1 or M2 Studio PSU and decide that 100% of the electrical power output must be getting dumped into the enclosure. These computers have a bunch of USB/Thunderbolt ports that can provide power to external peripherals, so the PSU's rated for significantly more output power than Apple expects to actually be dissipated inside the Studio's enclosure.

Next, you appear to think that increasing the volume of the Studio enclosure by 25% would also increase maximum heat dissipation by 25%. First, has there been even a hint that Apple plans to do this? Second, why do you think volume maps to heat dissipation in linear fashion?

Another one - you bought into the "Consumer Intelligence Research Partners" junk data which has iMac at 4% of Apple's Mac sales, Mac Pro at 3%, Mac Mini at 1%, and Mac Studio at 1%. If you google "Consumer Intelligence Research Paftners" the first thing you see is "CIRP surveys consumers on recent purchase activity to generate timely, accurate, and focused insights into consumer behavior." So these guys are not reporting actual sales data, instead they're a nonpolitical polling firm. If there's selection bias in the population of people they survey, and/or their sample size isn't large enough, their numbers can be off.

And these are extremely off! They are trying to claim the Mac Pro outsells the Mac Mini 3 to 1! I'm sure Apple would love it if that were true, but it obviously is not. Same thing applies to Pro vs Studio - no way that's 3:1 either. The Studio should be far more units than the Pro, because the only reason you buy a M2 Pro over a M2 Studio is if you need PCIe slots. Otherwise, you spend $3000 less on the same computer in a much friendlier form factor.

 

[Boil]

If Apple have the space in the M4 Enclosure for a 4 headed Hidra but needed an extra (speculation) 33% of heat to be dissipated by the enclosure, Apple could offer...

Apple could offer the Mac Pro Cube...! ;^p

 

[jdb8167]

Another one - you bought into the "Consumer Intelligence Research Partners" junk data which has iMac at 4% of Apple's Mac sales, Mac Pro at 3%, Mac Mini at 1%, and Mac Studio at 1%. If you google "Consumer Intelligence Research Paftners" the first thing you see is "CIRP surveys consumers on recent purchase activity to generate timely, accurate, and focused insights into consumer behavior." So these guys are not reporting actual sales data, instead they're a nonpolitical polling firm. If there's selection bias in the population of people they survey, and/or their sample size isn't large enough, their numbers can be off.

And these are extremely off! They are trying to claim the Mac Pro outsells the Mac Mini 3 to 1! I'm sure Apple would love it if that were true, but it obviously is not. Same thing applies to Pro vs Studio - no way that's 3:1 either. The Studio should be far more units than the Pro, because the only reason you buy a M2 Pro over a M2 Studio is if you need PCIe slots. Otherwise, you spend $3000 less on the same computer in a much friendlier form factor.

Since the CIRP survey appears to be self reported I’d imagine the highly inflated Mac Pro numbers are coming from average consumers who call their MacBook Pro a Mac Pro on the survey response.