Discussion Apple Silicon SoC thread

senttoschool said:

It's the same thing. I didn't say anything different.

M1 CPU is the fastest, best performing, best overall, speediest laptop CPU out there. Period. Not controversial.

M1 chip (entire SoC) is the fastest, best performing, best overall, speediest laptop chip out there. Period. Not controversial.

Is it more clear to you now?

Edit: Added "period" and "not controversial" since you seem to get caught up when I either write it or not write it.

Click to expand...

I mean... this is just not true. Also, I'm not the one who got caught up on the period. But I'd not expect you to realize that, since you seem to completely not address all the points I brought up.

Meh, it's not worth talking with someone who's entering this in bad faith, someone who's not open to actually discussing the topic. You're just not worth it.

I feel like a lot of the arguments going on here atm are pretty pointless. So here's a little totally unrelated analysis and some handwavey speculation. Big wall of text, my apologies.

I feel like much of the focus on the M1 has been on its ST performance, probably because of splashy Geekbench and SPEC numbers (and less splashy but still impressive showings in Cinebench etc.). I think the focus on raw ST performance is a poor way to analyze Apple Silicon's strengths and weaknesses, for a few reasons...

1) M1 has a real but marginal ST lead over TGL in best showings, and comes out close in others, 2) M1 matches Desktop Zen 3 in some places and lags it in others, but 3) M1 does all this with a process node advantage. In the real-world I expect Apple will maintain a node or half-node advantage over its rivals for the foreseeable future due to its arrangement with TSMC, and thus may maintain cutting edge ST performance at any point in time. But in a world where everyone was on the same node, as a thought experiment, I expect Intel's mobile chips would match or slightly outperform M1 in ST, and the Zen 3 mobile parts would likely match it as well. It is possible (though we don't know), that M1's ST performance in a Macbook Pro and in an iMac will be necessarily identical due to voltage, and lag Rocket Lake / Zen 3 desktop due to an inability to clock up like the competition.

Apple Silicon's real edge is in efficiency. I don't mean that in the "Arm is for low power uses and x86 is for performance" way that gets tossed around. But Apple's wide and slow approach is achieving their competitive ST performance at unbelievably low power draw, and when high-performance workflows are increasingly massively parallel, that means that Apple probably has a multi-threaded advantage over its competitors even when accounting for process node.

Let's use Cinebench as a test case, explicitly because it's a workflow where the M1 appears at a disadvantage relative to other benchmarks.

- The M1 pulls a sustained 3.8w in ST and 15w in MT for its scores of ~ 1520 / 7800. That MT load engages both the 4 performance cores and the 4 efficiency cores; I don't know what the breakdown is there, but I expect the p-cores are around 13w. There doesn't, as far as I can tell, appear to be any big power spikes during Cinebench, unlike in the x86 world.

- The 4750u pulls a sustained 28w MT, and scores around 8080. It also hit a high of 30w, so no huge spikes.

- The 4800u is frankly hard to find both in real-life and in reliable reporting, but it seems that it pulls a sustained 34w MT for its scores ranging from 9500 to 10150 MT. That slight sustained increase in power draw over the other Renoir-u chips makes sense. The 4800u appears to be aggressively binned (and thus so rare) but also slightly more aggressively clocked to achieve its impressive performance uplift.

- Tiger Lake is almost not worth comparing, and I can't find Cinebench specific numbers, but it appears to pull 28w sustained but well upwards of 50w in spikes under load, and has an r23 MT score of 6260.

We'll know soon enough, but based on these sorts of numbers an 8+4 variant of the M1 should be right on target for a sustained "TDP" of around 28w (accounting for CPU load only, of course), which is not coincidentally the TDP of the Intel chips Apple has traditionally put in the upper-end 13" MBP. Of course, Intel has been running their chips a good bit hotter for extended periods before throttling back to that, whereas I'd anticipate an M1X running consistently at 28w on something like Cinebench, based on the behavior of the M1.

I think Apple was smart to start with the M1, and in chasses where their strengths are balanced between battery life and performance. An M1X in the 13" model would probably have better battery life than an Intel variant because of the e-cores, but much less definitively so. What it would have instead would be a significant performance advantage.

A fairly conservative guess at performance scaling from 4+4 to 8+4 is around a 60% MT improvement for most workloads. Which is to say, I would guess that an M1X at 28w sustained versus a 4800u at 34w sustained should deliver around 12500 MT Cinebench r23 v 10000ish. I anticipate Cezanne will even that up a bit (but again, only for the most aggressively binned parts, and at a slightly higher power draw).

TLDR: As an architecture, I expect that on the same node Apple Silicon can't match the ST performance of Intel and AMD's best examples. But they have a significant efficiency advantage, such that Apple's high-end Macs will almost certainly be the fastest laptop chips for MT by a wide mile, and therein will lie the architectural advantage of Apple Silicon going forward.

Thank you for coming to my TED talk.

I think there are a few things to consider when discussing this chip and comparing it to traditional x86 options available.

While it's certainly fair game to use dedicated hardware to speed up some workload (at the end of the day I don't care how you get the performance if it's faster) it does mean that any future chips with more cores aren't going to see a massive improvement in performance unless they include more of that dedicated hardware and problems can be parallelized to run more of it. Performance improvements are limited to the extent to which additional (or faster) cores can speed up the overall task, which may be a small percentage in some cases.

Apple will only target markets or categories that apple cares about. Even though their GPU tech is rather impressive, particularly given the power envelope for it, I doubt we'll see Apple release anything that is designed to compete against even the upper mid-range parts like GA-104 and Navi 22. Performance will be great for an ultrabook or small form-factor desktop machine, but it won't be scaled to the high-end. The install base may not be good enough for developers to be willing to optimize titles for Apple's GPUs either.

Ability to pick and choose may be more limited due to the way Apple designs the hardware. If you really just want a system with more RAM, you might also need to buy one with the faster and more expensive chip as well. To some degree Apple already does this with their hardware using x86 CPUs where you need to buy a more expensive CPU in order to pair it with the best GPU offerings. Apple tends to value simplicity through limitation of choices so there may be fewer, more arbitrary product categories in their future products.

Also the excellent iGPU etc. A lot of Apple's designs were ending up rather hamstrung by weak integrated graphics performance.

Putting a very solid baseline underneath the entire range is likely to be very useful.

Like with phones - a top device should really be at least very good at nearly everything.

senttoschool said:

List them out.

Click to expand...

No thank you. Watching this thread unfolding around you taught me not to engage you with facts because that eats up precious time of my life for nothing.

People here are getting off track from what matters. The mere fact that people are arguing over how M1 compares to the fastest x86 CPUs shows that Apple achieved their goal: no one buying an M1 Mac is going to feel like they are giving up any performance versus if that same model Mac had the latest x86 CPU instead. Rosetta2 is fast enough that no one is going to worry they're making big sacrifices in the performance of their old x86 applications that haven't been ported to ARM yet.

The only possible objections for Apple customers buying an M1 Mac (and since I'm talking about "Apple customers" the Apple hater arguments about "too expensive" etc. are irrelevant here) are those who need to run a lot of Windows software, or those who need one of the higher end models Apple has yet to release.

When rumors about this move were discussed in the past a lot of people said it would destroy Apple because they'd have to give up too much performance. Even when after years of rumors it finally became official in June there were still many doubters. Now no one doubts the ability of Apple's ARM designs in a Mac to go toe to toe with the best x86 out there, even if they believe Apple is behind they are forced to acknowledge they aren't that far behind.

Erasing doubt was Apple's goal, to make their customers feel comfortable that the migration to ARM will be a win and not require some sacrifice or compromise on their part to stick with macOS. I think pretty much everyone would agree now that Apple's ARM transition will be a success. The only question left in my mind is will they eventually be able to win over enough Windows customers to get the Mac marketshare into double digits, or will the performance (especially performance/watt) advantages of Apple's ARM Macs not be seen as enough to overcome the "but it isn't Windows" resistance to change.

IvanKaramazov said:

1) M1 has a real but marginal ST lead over TGL in best showings, and comes out close in others, 2) M1 matches Desktop Zen 3 in some places and lags it in others, but 3) M1 does all this with a process node advantage.

...delted wall of text.

TLDR: As an architecture, I expect that on the same node Apple Silicon can't match the ST performance of Intel and AMD's best examples. But they have a significant efficiency advantage, such that Apple's high-end Macs will almost certainly be the fastest laptop chips for MT by a wide mile, and therein will lie the architectural advantage of Apple Silicon going forward.

Click to expand...

I think your giant wall of text missed a critical element that invalidates what you set up, and your final conclusion.

Sure, M1 scores a small single thread win with a "node advantage", But the only alternatives that came close, were running comparatively unconstrained power, more than double M1 power.

That process node advantage will mainly give a small power savings in the 20%-30% range at best.

When the competition also gets its 20-30% node improvement in power usage, they will still be behind, because they are behind by much more than the 20-30% node advantage on the element that node improvement impacts.

IvanKaramazov said:

TLDR: As an architecture, I expect that on the same node Apple Silicon can't match the ST performance of Intel and AMD's best examples.

Click to expand...

We can probably confirm that once AMD comes out with Zen4, though AMD will have one more generation of development work under their hat so I'm not sure it's a fair comparison. If Apple had ever released a MacOS product with A13, we could probably confirm/deny your hypothesis.

Also observe what AMD was able to accomplish staying on the same node (7nm). Node advancements don't necessarily account for everything.

Doug S said:

Erasing doubt was Apple's goal, to make their customers feel comfortable that the migration to ARM will be a win and not require some sacrifice or compromise on their part to stick with macOS.

Click to expand...

For the select bits of software that people will probably use pretty often on their Macbooks, yeah, I think they've done pretty well. Edge case users may reach some level of frustration. I don't think we can really accurately guess how many will fall into which categories, but rest assured that Apple will continue to provide support for key developers to move their ARM-based MacOS platform forward. In the long run, developers who get little to no direct support will have to follow suit to continue making money.

The only question left in my mind is will they eventually be able to win over enough Windows customers to get the Mac marketshare into double digits

Click to expand...

M1 will have to be followed up by something better than what's available. It isn't just Windows, its the lack of configurability and the prices. Woof. There's a reason many people just don't like Apple.

DrMrLordX said:

Edge case users may reach some level of frustration.

Click to expand...

Name one "edge case" where you can demonstrate this "level of frustration" you are claiming. I have yet to see anything where the M1 Mac is slower the x86 Mac it is replacing other than something using hand optimized x86 SIMD code. Even when running under Rosetta2 few things are slower on the M1 than on the x86 Mac being replaced.

The hand optimized SIMD code case won't matter in six months, because once the developer ports and optimizes their application for the M1 it will easily beat the x86 SIMD version - perhaps by several hundred percent (as per Apple's claims) since they will not only have the M1's ARM cores but also have a GPU, NPU and IPU at their disposal.

Sure, you can point to benchmarks for recently released Windows laptops using top of the line CPUs not available in Mac laptops (and cost over half of what entire M1 Mac costs) that may beat it in some things, but like AMD performance that's not a valid comparison for a Mac user looking to replace their Mac with another Mac. Which is all that Apple is concerned about at this time.

guidryp said:

Sure, M1 scores a small single thread win with a "node advantage", But the only alternatives that came close, were running comparatively unconstrained power, more than double M1 power.

That process node advantage will mainly give a small power savings in the 20%-30% range at best.

When the competition also gets its 20-30% node improvement in power usage, they will still be behind, because they are behind by much more than the 20-30% node advantage on the element that node improvement impacts.

Click to expand...

It's not just a node advantage - the Zen2 core is also a 1.5 year old core. Zen2->3 was a 20-25% performance improvement (CB20 ST being 20-25% depending on specific Zen2 vs Zen3 comparison), and N7->N5 is a 30% efficiency improvement or 15% speed improvement at iso-power. That calculates to be 20% performance * 15% performance = 38% total performance gain at same TDP.

So we take a theoretical 5800U-XT, a 15W TDP Cezanne (Zen3 core) on 5nm. 4800U (15W) CB ST was 1199. +20% for Zen3, +15% for N7->N5, that's a 38% total improvement, which makes the CB20 ST score estimate 1654. M1 native was 1522. I mean, I think that's completely in the realm of possibility that a Zen3-powered 15W APU on 5nm could actually have faster ST performance than M1. (Heat be damned, anyway.)

The chief issue, of course, is that M1 is actually available. Cezanne has no release date, and Cezanne is on 7nm anyway, so it's all just pie-in-the-sky dreaming. Granted, it's do-able, as the Zen3 core exists and 5nm exists. But this whole conversation could go down a very long rabbit-hole of what-ifs that are pretty irrelevant right now because I can't buy a 5nm Zen3 laptop chip, and may never be able to.

As of now, we only have Renoir, TGL, and M1. And M1 clearly has better efficiency and better single/lightly threaded performance and competes well enough in MT stuff with the 4800U and 4900HS that we can reasonable conclude that an 8+4 M1X or whatever would actually be damn stellar.

On the node advantage for M1 vs. the competition (N5 vs N7), realize that the node advantage also caters more to Apple's approach to chip design. For the M1, Apple has thrown a ton of circuit resources at their cores to enable them to extract as much IPC as possible from each of their fewer clocks. While the node advantage only offers 10-20% better power draw characteristics on two equivalent designs on the new and old node, the extra capacity to pack logic into their cores by making them wider, by having deeper buffers, by having larger internal low level caches (iCache, dCache), etc. It also helps them with secondary effects, by allowing them to maintain their use of denser libraries because of their lower clock targets. At lower clocks, N5 is allowing them to be very sense, enabling a 4+4 design on a very small piece of silicon.

To put that in to perspective: if M1 was implemented on N7, it would either be a much larger chip, physically, reducing the number of die per wafer greatly. This would adversely affect the per chip profit, and make the die plus ram package larger and more expensive. Alternately, they could have maintained the same die size, and per wafer chips, but it would have likely cost them two of the Firestorm cores and at least some of the L2 and SLC. Power draw in single core scenarios would be notably higher, and multi threaded benches would be much slower.

It is my opinion that the node advantage for Apple in this situation is a very big deal. I'm not criticising Apple here, as they paid the money and took the risk to be on that node, and it has paid off. Zen3 on N5, as it currently exists, wouldn't gain a whole lot. It would consume a few less watts. It could sustain higher MC boost clocks. It would be a smaller chip. That's about it. But Zen3 wasn't built with N5 in mind. It was built with the constraints of N7 in mind. Zen4 will be built for N5. It will have extra circuit budget for more CPU resources. It will be power optimized for it's targeted node.

But, it isn't here today. And that's the reality of the market.

amrnuke said:

It's not just a node advantage - the Zen2 core is also a 1.5 year old core. Zen2->3 was a 20-25% performance improvement (CB20 ST being 20-25% depending on specific Zen2 vs Zen3 comparison), and N7->N5 is a 30% efficiency improvement or 15% speed improvement at iso-power. That calculates to be 20% performance * 15% performance = 38% total performance gain at same TDP.

Click to expand...

Zen 3 is already here. It's matching ST performance of M1, using what? 3x-4x the power??

All zen three gains is power improvement which won't even put it in the ballpark. Naturally there will be design improvements as well, but Apple is probably not done improving their chips either.

People seem to be missing the scale of the power usage differences.

Anandtech is measuring the Whole iPad Mini at the Wall. The whole computer maxing out ST uses 15 Watts, including PSU losses.

AMD/Intel power numbers are package power being through sofware. Intel will use up to 50 watts in a single thread, package power, to match the ST performance of the Mini that uses 15 watts for the whole computer.

Doug S said:

Name one "edge case" where you can demonstrate this "level of frustration" you are claiming.

Click to expand...

I'm not DrMrLordX, but I'm a Linux user. My level of frustration is not being able to replace MacOS with Linux even though I do like Apple's approach to hardware design most of the time.

guidryp said:

Zen 3 is already here. It's matching ST performance of M1, using what? 3x-4x the power??

All zen three gains is power improvement which won't even put it in the ballpark. Naturally there will be design improvements as well, but Apple is probably not done improving their chips either.

People seem to be missing the scale of the power usage differences.

Anandtech is measuring the Whole iPad Mini at the Wall. The whole computer maxing out ST uses 15 Watts, including PSU losses.

AMD/Intel power numbers are package power being through sofware. Intel will use up to 50 watts in a single thread, package power, to match the ST performance of the Mini that uses 15 watts for the whole computer.

Click to expand...

Well, we don't know what the Zen3 power usage will be in Cezanne. It looked like Zen2 in Matisse was a big power sucker based on the per core power draw I believe you're referencing, right? But the Renoir chips running full-bore Furmark and Prime95 draw peak power draw (at the wall) of 37.2W - powering a screen as well (reference). The Mac mini pulled 31W in a compute MT load and the power of the monitor is not included. Did Andrei mention what his "compute MT workload" was actually measuring? I didn't see that. Granted, Furmark + Prime95 might not be as taxing as Andrei's compute MT workload. But it sure seems pretty taxing.

Zen3 doesn't only gain power improvement, we already know ST scores are 20-25% higher, IPC is 15-20% higher. To say it only gains power improvement seems wrong. Where did you get that information?

So, sure, single-thread workload uses 15W. But multithread workload uses 31W. And MT+GPU workload on Renoir draws 37.2W (yes, that's at the wall, using a Metrahit - not using the software-provided numbers).

amrnuke said:

Well, we don't know what the Zen3 power usage will be in Cezanne. It looked like Zen2 in Matisse was a big power sucker based on the per core power draw I believe you're referencing, right? But the Renoir chips running full-bore Furmark and Prime95 draw peak power draw (at the wall) of 37.2W - powering a screen as well (reference). The Mac mini pulled 31W in a compute MT load and the power of the monitor is not included. Did Andrei mention what his "compute MT workload" was actually measuring? I didn't see that. Granted, Furmark + Prime95 might not be as taxing as Andrei's compute MT workload. But it sure seems pretty taxing.

Zen3 doesn't only gain power improvement, we already know ST scores are 20-25% higher, IPC is 15-20% higher. To say it only gains power improvement seems wrong. Where did you get that information?

So, sure, single-thread workload uses 15W. But multithread workload uses 31W. And MT+GPU workload on Renoir draws 37.2W (yes, that's at the wall, using a Metrahit - not using the software-provided numbers).

Click to expand...

Note that the 37.2W is using the 4750U, 4800U laptops can be much higher. eg. 57.6W here:

guidryp said:

I think your giant wall of text missed a critical element that invalidates what you set up, and your final conclusion.

Sure, M1 scores a small single thread win with a "node advantage", But the only alternatives that came close, were running comparatively unconstrained power, more than double M1 power.

Click to expand...

I actually agree with you here, my suggestion was that while the M1 might not match the others for raw ST performance (at hugely different power levels, obviously), the efficiency of the M1 is exactly what should set it up to scale so well to more cores and higher raw MT performance. In other words, my expectation is that if one ignores efficiency and ppw entirely, the Apple Silicon chips may not be the clear best in ST, but likely will be in MT once larger chips come out. I obviously could be wrong.

They asked Linus Torvalds if he was planning on buying an M1 Macbook Air, and he said no (as expected) because he cannot run Linux on it.

Something tells me that someone will get Linux working on this thing eventually. If you can run Linux on Nintendo Switch, getting it working on an ARM Macbook should be easy!

moinmoin said:

I'm not DrMrLordX, but I'm a Linux user. My level of frustration is not being able to replace MacOS with Linux even though I do like Apple's approach to hardware design most of the time.

Click to expand...

If I got one of these M1 Macbook systems for job, it would royally piss me off. It can't run Docker (yet), nor can it run VMWare virtualization software. I also have a ton of oddball IT administration software that probably aren't compatible with Big Sur yet.

Of course, this laptop wasn't made for my workload. The 16" Macbook Pro was, though. Which is why I bought one

I found this site, which will (hopefully) tell me when M1 native versions of my applications are available:

ultimatebob said:

Something tells me that someone will get Linux working on this thing eventually. If you can run Linux on Nintendo Switch, getting it working on an ARM Macbook should be easy!

Click to expand...

Switch runs on a standard NVidia Tegra chip, and NVidia actually provides "Linux for Tegra". It's not intended for switch but it would probably be a great starting place for all important drivers.

But for M1 Macs. Apple isn't providing Linux drivers, and M1 isn't based on HW that has Linux drivers, there isn't even much of stating place for drivers.

No doubt someday futher away, it will happen, but likely not in a very usable form.

thunng8 said:

Note that the 37.2W is using the 4750U, 4800U laptops can be much higher. eg. 57.6W here:

The Ryzen 7 4800U is an Absolute Monster: Lenovo Yoga Slim 7 14 Laptop Review

After facing a few delays, AMD's premier Ryzen 7 4800U is finally starting to hit store shelves. Results are insane for a 14-inch subnotebook to the point where it'd be tough to recommend an Intel counterpart when given the same price.

www.notebookcheck.net

Click to expand...

The example I gave is one of an entire laptop, screen included, under full load - including GPU - using less power (37.2W) than a 3700X under a 2-thread load (39.63W - reference). Which sufficiently makes my point that using the power consumption of a desktop chip to criticize the potential power consumption of its laptop companion is a fool's errand.

Non-sequitor, and perhaps straying off-topic from what I intended:

CB20 MT score of the HP EliteBook 4750U I referenced is 3264 which is actually better than Lenovo's Yoga Slim you linked that draws 57.6W, which only got 3213 points. (There is much to be said about manufacturer implementations.) [Begin speculation] Assuming that translates to CB23 MT, the 4750U would get roughly the same score as the 4800U, which is 9286. (Some 4750U implementations carry a TDP of 45W and can hit 9716, but I don't think it's fair to include that.)

If we assume that both the Mac Mini and the EliteBook in question both draw their full peak loads while running CB23 (again, speculation):

4750U CB23MT score of 9286 @ 37.2W = 249.6 CB23/W
M1 CB23MT score of 7833 @ 31W = 252.7 CB23/W

amrnuke said:

The example I gave is one of an entire laptop, screen included, under full load - including GPU - using less power (37.2W) than a 3700X under a 2-thread load (39.63W - reference). Which sufficiently makes my point that using the power consumption of a desktop chip to criticize the potential power consumption of its laptop companion is a fool's errand.

Non-sequitor, and perhaps straying off-topic from what I intended:

CB20 MT score of the HP EliteBook 4750U I referenced is 3264 which is actually better than Lenovo's Yoga Slim you linked that draws 57.6W, which only got 3213 points. (There is much to be said about manufacturer implementations.) Assuming that translates to CB23 MT, the 4750U would get roughly the same score as the 4800U, which is 9286. (Some 4750U implementations carry a TDP of 45W and can hit 9716, but I don't think it's fair to include that.

If we assume that both the Mac Mini and the EliteBook in question both draw their full peak loads while running CB23 (which may be a bit unfair to the EliteBook since that peak power draw includes the GPU under Furmark, and powering the screen too):

4750U CB23MT score of 9286 @ 37.2W = 249.6 CB23/W
M1 CB23MT score of 7833 @ 31W = 252.7 CB23/W

Click to expand...

M1 in a mini draws 15W for cinebench.


I can't find the video source since I watched so many videos, but the macbook air when thermally throttled (at the end of the 30min run) draws 7W for a score of ~5900.

amrnuke said:

The example I gave is one of an entire laptop, screen included, under full load - including GPU - using less power (37.2W) than a 3700X under a 2-thread load (39.63W - reference). Which sufficiently makes my point that using the power consumption of a desktop chip to criticize the potential power consumption of its laptop companion is a fool's errand.

Non-sequitor, and perhaps straying off-topic from what I intended:

CB20 MT score of the HP EliteBook 4750U I referenced is 3264 which is actually better than Lenovo's Yoga Slim you linked that draws 57.6W, which only got 3213 points. (There is much to be said about manufacturer implementations.) Assuming that translates to CB23 MT, the 4750U would get roughly the same score as the 4800U, which is 9286. (Some 4750U implementations carry a TDP of 45W and can hit 9716, but I don't think it's fair to include that.

If we assume that both the Mac Mini and the EliteBook in question both draw their full peak loads while running CB23 (which may be a bit unfair to the EliteBook since that peak power draw includes the GPU under Furmark, and powering the screen too):

4750U CB23MT score of 9286 @ 37.2W = 249.6 CB23/W
M1 CB23MT score of 7833 @ 31W = 252.7 CB23/W

Click to expand...

This is so wrong on so many levels. I can't even...

A lot of AMD fanboys in denial here.



Use of the word fanboys is not allowed.


esquared
Anandtech Forum Director

thunng8 said:

M1 in a mini draws 15W for cinebench.

https://twitter.com/x/status/1328777333512278020

I can't find the video source since I watched so many videos, but the macbook air when thermally throttled (at the end of the 30min run) draws 7W for a score of ~5900.

Click to expand...

So now we're ok with reported power draw? In any case, that entire segment was a complete non-sequitor, as I mentioned, since we don't have actual power draws under CB for both chips, we aren't apparently using the same measurement system, and so on. Continuing down that path is silly, but I thought interesting enough to consider! I really would love to have more data to make a better comparison, but all we have right now is speculation when comparing with contemporary laptop chips.

Anyway, do you care to address my primary response?

"The example I gave is one of an entire laptop, screen included, under full load - including GPU - using less power (37.2W) than a 3700X under a 2-thread load (39.63W - reference). Which sufficiently makes my point that using the power consumption of a desktop chip to criticize the potential power consumption of its laptop companion is a fool's errand."