Wattage guesses for Apple Silicon

[mrmister]

I'm very curious about this. I have been around long enough to remember when switching from a PowerBook to a MacBook Pro how the wattage increased—my last PowerBook adapter ran 65 watts, and my first MacBook Pro one was 85w, and enormously large.

In recent times, here are the wattage sizes of modern Apple laptop from their adapters:

12" MacBook — 29 watts
13" MacBook Air — 45 watts
13" MacBook Air Retina — 30 watts
13" MacBook Pro — 60 watts and 61 watts
15" MacBook Pro — 60 watts, 61 watts, and 87 watts
16" MacBook Pro — 96 watts

In all cases the wattages grew throughout the lives of the laptops, except the MacBook Air dropped by a third when it was rearchitected for Retina in 2017.

Both architecturally (RISC instruction set) and reputationally, I'd expect Apple's new chips to result in laptops that drink less power. At the same time, they are going to want to crush the performance of the old laptops. What are people's guesses as to what the lineup will look like?

 

[verticalines]

Power adapter is not quite the power envelope. Of course you want more power to charge but that's not usage. We tend to go off processor TDP for these numbers.

The Airs and single-fan 13-inch Pro have a target of around 15W of dissipation, the other 13-inch Pro is 28W (the one with dual fans and TB at each side), the 15" I think is 90-100W (both CPU and GPU at full load). I haven't used a large Pro in years so don't quote me on that.

Going off Anandtech's iPhnoe 11 review, the A13 ramps up to about 5W maximum in CPU. In graphics the whole looks to have peaked at close to 6W. Let's keep it at 6W for ease. I'd say the Air and 13-inch Pro chassis is fine to dissipate passively two A13s on one board without issue. I don't think it'll need that much performance so one at 6W and the rest is leftover void for a quiet computer with more batteries to boost battery life. Or thinner I guess.

The Pros can have say 60W. I still feel Apple is going to offer something from an outside vendor (AMD or Nvidia) until they can supplement the performance in-house. Imagine two A13s on one board, then a 45W GPU. Still hot but won't get as hot or draining or loud. Or you can add ten A13s but that probably is not ideal or realistic. Wasted silicon is wasted resources.

A12z is on a 7nm process. It's before A13 on a 7nm+. And A14 is said to be shipping on 5nm EUV. There's going to be efficiency gains just progressing through the manufacturing. But of course desktop apps will run a bit differently than iPad apps--modeling in Maya is totally different than pulling shapes in some iOS app. We'd have to see how the processors scale up in clock-speed and whether that affects the voltage which in-turn affects power usage. I'm not concerned about their low-power cores, I'm curious how high their high-performers can go.

 

[JPack]

MacBook Pro with ARM will probably come in around 10-15W, so definitely not passively cooled but relatively small fans.

Keep in mind Apple will have to add interconnect fabric to their current silicon, PCIe controller, and other uncore items. They might even add an L3 cache. And fire up the dual- or quad-channel memory controller. These all consume power.

The power adapter, probably 30W to 40W tops.

 

[glindon]

Power adapter is not quite the power envelope. Of course you want more power to charge but that's not usage. We tend to go off processor TDP for these numbers.

The Airs and single-fan 13-inch Pro have a target of around 15W of dissipation, the other 13-inch Pro is 28W (the one with dual fans and TB at each side), the 15" I think is 90-100W (both CPU and GPU at full load). I haven't used a large Pro in years so don't quote me on that.

Going off Anandtech's iPhnoe 11 review, the A13 ramps up to about 5W maximum in CPU. In graphics the whole looks to have peaked at close to 6W. Let's keep it at 6W for ease. I'd say the Air and 13-inch Pro chassis is fine to dissipate passively two A13s on one board without issue. I don't think it'll need that much performance so one at 6W and the rest is leftover void for a quiet computer with more batteries to boost battery life. Or thinner I guess.

The Pros can have say 60W. I still feel Apple is going to offer something from an outside vendor (AMD or Nvidia) until they can supplement the performance in-house. Imagine two A13s on one board, then a 45W GPU. Still hot but won't get as hot or draining or loud. Or you can add ten A13s but that probably is not ideal or realistic. Wasted silicon is wasted resources.

A12z is on a 7nm process. It's before A13 on a 7nm+. And A14 is said to be shipping on 5nm EUV. There's going to be efficiency gains just progressing through the manufacturing. But of course desktop apps will run a bit differently than iPad apps--modeling in Maya is totally different than pulling shapes in some iOS app. We'd have to see how the processors scale up in clock-speed and whether that affects the voltage which in-turn affects power usage. I'm not concerned about their low-power cores, I'm curious how high their high-performers can go.

Apple already detailed that Apple Silicon is using a unified memory architecture so that completely rules out a GPU from AMD or Nvidia as their current gpus would have to be completely redesigned because they would have to allow application memory to be stored in their video memory. Unified memory means that both gpu and cpu have equal access to one set of memory. This is not to be confused with "shared graphics memory" like integrated gpu's have where your 16gb is partitioned so 4gb goes to video and 12gb is what the system can use. Even their graphic from the presentation showed apple silicon will have a "High-Performance gpu."
[automerge]1593233117[/automerge]

Power adapter is not quite the power envelope. Of course you want more power to charge but that's not usage. We tend to go off processor TDP for these numbers.

The Airs and single-fan 13-inch Pro have a target of around 15W of dissipation, the other 13-inch Pro is 28W (the one with dual fans and TB at each side), the 15" I think is 90-100W (both CPU and GPU at full load). I haven't used a large Pro in years so don't quote me on that.

Going off Anandtech's iPhnoe 11 review, the A13 ramps up to about 5W maximum in CPU. In graphics the whole looks to have peaked at close to 6W. Let's keep it at 6W for ease. I'd say the Air and 13-inch Pro chassis is fine to dissipate passively two A13s on one board without issue. I don't think it'll need that much performance so one at 6W and the rest is leftover void for a quiet computer with more batteries to boost battery life. Or thinner I guess.

The Pros can have say 60W. I still feel Apple is going to offer something from an outside vendor (AMD or Nvidia) until they can supplement the performance in-house. Imagine two A13s on one board, then a 45W GPU. Still hot but won't get as hot or draining or loud. Or you can add ten A13s but that probably is not ideal or realistic. Wasted silicon is wasted resources.

A12z is on a 7nm process. It's before A13 on a 7nm+. And A14 is said to be shipping on 5nm EUV. There's going to be efficiency gains just progressing through the manufacturing. But of course desktop apps will run a bit differently than iPad apps--modeling in Maya is totally different than pulling shapes in some iOS app. We'd have to see how the processors scale up in clock-speed and whether that affects the voltage which in-turn affects power usage. I'm not concerned about their low-power cores, I'm curious how high their high-performers can go.

Their presentation slide showed (what looks like a range of chips) mostly same power consumption as notebooks, so anywhere from 15w to 45w.

 

[verticalines]

Apple already detailed that Apple Silicon is using a unified memory architecture so that completely rules out a GPU from AMD or Nvidia as their current gpus would have to be completely redesigned because they would have to allow application memory to be stored in their video memory. Unified memory means that both gpu and cpu have equal access to one set of memory. This is not to be confused with "shared graphics memory" like integrated gpu's have where your 16gb is partitioned so 4gb goes to video and 12gb is what the system can use. Even their graphic from the presentation showed apple silicon will have a "High-Performance gpu."
[automerge]1593233117[/automerge]


Their presentation slide showed (what looks like a range of chips) mostly same power consumption as notebooks, so anywhere from 15w to 45w.
View attachment 928219

You don't go from 5W SoC graphics to a 5300M overnight. If they're serious about the professional range too, how do you extract the performance of the currently offered Vega Duo? They are licensing more from Imagination but it's going to be an uphill battle against two titans with a lot of brainpower and ambitious roadmaps too. And patents, don't forget about those.

Unified memory is nothing new. AMD has been trying to push that awhile with NUMA. Consoles have been leveraging GDDR. But they're also from pretty experienced players and bandwidth is great for graphics and in a controlled entertainment box but performance computing? Latency comes into play. You'd want to stick to DDR4/DDR5 instead of bandwidth-eager but loose latency GDDR/HBM.

Apple is ambitious and we'll have to see if they stumble or luck out early. A chart from marketing doesn't mean anything.

 

[Moonjumper]

You don't go from 5W SoC graphics to a 5300M overnight. If they're serious about the professional range too, how do you extract the performance of the currently offered Vega Duo? They are licensing more from Imagination but it's going to be an uphill battle against two titans with a lot of brainpower and ambitious roadmaps too. And patents, don't forget about those.

Unified memory is nothing new. AMD has been trying to push that awhile with NUMA. Consoles have been leveraging GDDR. But they're also from pretty experienced players and bandwidth is great for graphics and in a controlled entertainment box but performance computing? Latency comes into play. You'd want to stick to DDR4/DDR5 instead of bandwidth-eager but loose latency GDDR/HBM.

Apple is ambitious and we'll have to see if they stumble or luck out early. A chart from marketing doesn't mean anything.

I doubt Apple are trying to do it overnight. This has been coming for a long time.

Arm have been designing chips for servers for a while, and the chosen option for scaling has been core count, see Ampere’s 80 core chip for example. While the chips will be different in Apple Silicon, I can still see core counts being used to raise performance. To take 80 cores as an example, it wouldn’t surprise me to see a SoC with 16 CPU cores and 64 GPU cores.

Whatever we see in Apple Silicon, it will be far beyond the limits of 5W SoC from day 1 of the consumer launch. I’m expecting an announcement more akin to the XDR monitor launch with a series of great specs than saying here’s an A12Z + 10%.

 

[jinnyman]

SoC for a mobile machine I understand.
But Apple has to consider what they would offer for a heavy lifter?
Whatever unified or shared memory for CPU and GPU is good, but not enough for dGPU of 5600M caliber. What about Vega and Navi caliber? Will Apple abandon that sector?

 

[ondert]

I read that A12Z reaches to 4w tdp while on a single core full load. So, there are 4 high performance core there, it can go up to 16w already with all core loads. With additional cooling, Apple will increase the frequencies much higher than 3GHz while A14 alone is expected to reach 3Ghz in iPhone 12.

 

[aaronage]

So many possibilities!

I wonder if both 13" and 16" MacBook Pro will share the same SoC. There shouldn't be a disparity in terms of performance as both sizes will be able to support the fastest internals (discrete GPUs are gone).

"Advanced Silicon Packaging" was called out on the Apple Silicon reveal slide. What could that mean? It could be something exotic like 3D stacking, it could be HBM memory on package, or it could be something like chiplets.

If I had to guess, I wouldn't expect the first generation SoC to be super exotic but still offer a lot:
- 15-25W TDP
- 5nm >200mm2 die size (monolithic, no chiplets)
- >3GHz peak boost
- 8P/4E CPU cores (as rumoured)
- ~4 TFLOP GPU
- 32MB L2
- Quad-channel LPDDR4X for >120GB/s
- All the other goodness like Neural Engine, custom video encode/decode, custom crypto etc.

Bonus things we might see:
- USB 4
- PCI-E 4 for storage
- HBM instead of LPDDR4X? I'm doubtful since it would be very expensive, especially to offer a 32GB model.

These guesses are based on the existing A12Z as a starting point, and what little information Apple revealed in WWDC sessions.

I don't think these guesses are unreasonable given everything we know. If anything, I would expect Apple to have some surprises for us ?

EDIT: Figured this comment is a little off topic, so started a thread to speculate on SoC specs here https://forums.macrumors.com/threads/mac-silicon-what-do-you-expect-from-the-first-mac-socs.2243510/ ?

 

[BigPotatoLobbyist]

You don't go from 5W SoC graphics to a 5300M overnight. If they're serious about the professional range too, how do you extract the performance of the currently offered Vega Duo? They are licensing more from Imagination but it's going to be an uphill battle against two titans with a lot of brainpower and ambitious roadmaps too. And patents, don't forget about those.

Unified memory is nothing new. AMD has been trying to push that awhile with NUMA. Consoles have been leveraging GDDR. But they're also from pretty experienced players and bandwidth is great for graphics and in a controlled entertainment box but performance computing? Latency comes into play. You'd want to stick to DDR4/DDR5 instead of bandwidth-eager but loose latency GDDR/HBM.

Apple is ambitious and we'll have to see if they stumble or luck out early. A chart from marketing doesn't mean anything.

Lmfao. I love how poorly the usual "scaling" and artificial mountains-of-molehills armchair commentary like this aged.