Apple Silicon M1 Die-Shots

https://www.techinsights.com/blog/two-new-apple-socs-two-market-events-apple-a14-and-m1

by stefanhoelzl • 6 years ago

185 107 6 years ago

How does one identify which regions of a die-shot is the CPU or GPU? Are the structures on silicon easily recognizable from such pictures?

How far can someone with time and money go towards reverse engineering a chip from such pictures?

wmf3• 6 years ago

For very old chips reverse engineering is relatively easy; here's one example: http://www.righto.com/2020/04/inside-am2901-amds-1970s-bit-s... For modern chips you need an electron microscope and there are a lot of metal layers to deal with.

wtallis2• 6 years ago

You don't need an electron microscope to discern large-scale repeated structures like CPU and GPU cores, or to tell the difference between random logic and highly regular memory arrays (usually cache, but sometimes you can see microcode or register files if you look at a small enough area). An electron microscope is only useful if you want to fully reverse-engineer the entire circuit schematic, but that's overkill for basically any purpose except manufacturing your own clone.

duskwuff• 6 years ago

In short: educated guesswork. The lighter blocky regions in the die shot are recognizable as memory; a lot of the rest is based on recognizing duplicated or mirrored elements (like the two CPU1 cores) and making guesses based on their number, relative size, and what they're close to.

Identifying the GPU, for example, is based on recognizing that there are eight repeated elements within it, and matching that with the known fact that there are eight GPU cores.

mhh__• 6 years ago

I assume you could at least make an educated guess based on the layout and ratios of the area of the caches (assuming you know their topology and sizes) and therefore their proximity to non-cache stuff.

sydney6• 6 years ago

This [1] is a Talk from John McMaster about "Low-Cost IC Emission Reverse Engineering" using, iirc, commodity cameras.

[1]https://www.youtube.com/watch?v=iAuLXg2fVz0

saagarjha• 6 years ago

In general, the other comments give an idea of how to do it. But Apple shared their own die shots in their presentation ;)

ogre_codes• 6 years ago

And they were actually accurate which is unusual for this sort of thing.

baybal2• 6 years ago

How does one identify which regions of a die-shot is the CPU or GPU? Are the structures on silicon easily recognizable from such pictures?

Infrared camera

systemvoltage3• 6 years ago

Kind of sad to think about this - Why does marketing dept not actually show these and instead put up a cartoonized version of Apple's chips [1]? That's in every keynote. This is a rhetorical question, the real mystery is why does an average consumer want to see unreal things to convince them its cooler/better? You'd think seeing actual die shots would be more convincing to average consumer, but they love marketing that's untruthful, fake and exaggerated. Not the fault of the marketing department. This is what consumers want and this is what sells. Sad to think how shiny things sell and the whole luxury industry in general.

[1] https://www.apple.com/v/mac/m1/a/images/overview/chip_cpu__d...

madeofpalk• 6 years ago

As a technical consumer I don’t know why I would want to see the actual die shots - it really doesn’t tell me anything at all.

The graphics they put on slides or marketing sites aren’t supposed to be technical reference material, usually it’s just about conveying the idea of a SoC or performance improvements, or relative sizes, or whatever.

FartyMcFarter2• 6 years ago

Because it's more real, and more inspiring to look at real technical material.

systemvoltage3• 6 years ago

The inquiry I am making is more general - infantilization of consumers and the idea of dumbing things down for consumers like they're blobs of meat with cash to spare.

You don't need to see the cartoonized version of it either. May be its easier as a schematic to explain? Not sure. A list of M1's features would work equally well. I think its just marketing pizzaz.

asadlionpk• 6 years ago

Does it make a difference? The ones from the marketing department convey the message better and are pretty to look at. These are not as cool to average consumer.

sergiomattei• 6 years ago

Get real - which consumer ever really cares about die shots?

I'm a rather technical folk and even I don't care.

amelius• 6 years ago

I think if Apple wants to be a tech leader they should open up more. The secrecy is totally unnecessary.

manigandham• 6 years ago

Are you claiming that they are not a tech leader right now?

eyelidlessness• 6 years ago

They've actually done this for some web pages, you likely can find them on Wayback.

soneil• 6 years ago

They do have a shot on https://www.apple.com/mac/m1/ (the first non-logo image) which looks like it matches up (you'll need to rotate one by 180°). So it's not like they're hiding it or actively avoiding it. It apparently just doesn't suit their needs.

userbinator2• 6 years ago

The degree to which they hype UMA (a term of art within the PC industry for decades) on that page is amusing --- if Intel did the same sort of marketing for its integrated GPUs, they would almost certainly be laughed at.

userbinator2• 6 years ago

The other discussion here recently about how Apple is very silo'd internally is relevant: the marketing department very likely doesn't have access to real die photos.

dannyw2• 6 years ago

They do, die shots are literally on Apple's marketing page: https://www.apple.com/mac/m1/

A quick visual comparison shows this to be pretty accurate, just inverted and rotated.

FartyMcFarter2• 6 years ago

What is a "system cache" ? Is that an L3 cache?

chrisseaton4• 6 years ago

It's L3, but I think it's called 'system cache' because it is also used for the bus that loads memory from RAM into the GPU.

bullen3• 6 years ago

I would really like to know where the L1 caches are for proportions!

my123• 6 years ago

The system-level cache isn't an L3 cache specific to the CPU only, it's common across all the blocks in the SoC.

It is used by the GPU, neural engines,... in addition to the CPU itself.

segfaultbuserr2• 6 years ago

I know nothing about chips, but I guess it's probably the last-level CPU cache but the GPU shares it as well, similar to Intel iGPU, so they decided to call it "system cache".

segfaultbuserr2• 6 years ago

Only a die-shot, it's disappointing that there's no comment or explanation. But the website advertises:

A DFR report will be available to TechInsights subscribers on Apple's M1 disruptive chip architecture in January 2021.

So I assume the actual write-up is still a work-in-progress. Perhaps we just need to be patient and come back a month later.

dannyw2• 6 years ago

The actual report will be thousands of dollars because it's for industry professionals trying to compete with Apple's M1.

dismalpedigree3• 6 years ago

The “reports” are targeted at competitors who want to know how and what Apple did. They are quite expensive for a consumer or casual user.

The are more than just a document. A DFR includes in depth analysis, hundreds of supporting images, and usually access to the Circuit Vision for it. CV is a viewer that allows drilling into the chip down to viewing individual transistors along with the chip schematics. Think map tiles but for a chip.

Disclosure: I work at TechInsights.

jsjohnst• 6 years ago

Only a die-shot, it's disappointing that there's no comment or explanation.

A die shot that Apple themselves showed at the event. Which has had the same level of annotations previously by Anantech.

Sure, the DFR in early 2021 will be much more detailed, but also expensive to the point 0.001% of HN viewers will actually be able to see it. As such, I think your skepticism of this being worthy of home page billing is not out of line.

For the record, this is not meant to be a dig at TechInsights. I’ve seen a DFR before and they really are good for what they are, it’s just not something most HNers would gain value from, even if it were free.

kbumsik2• 6 years ago

Interesting to see the CPU isn't actually that big in the chip. What are the uncommented parts? Maybe peripheral interfaces?

als03• 6 years ago

Cache, GPU, Secure Enclave, Neural processors, peripheral controllers. GPU takes the lion’s share.

coder543• 6 years ago

The Cache, GPU, and NPU are all already labeled in the image.

The Secure Enclave is a good one to point out that is in the unlabeled space.

Certainly some non-DDR peripherals are in the unlabeled space, but it would be neat to have a more comprehensively labeled diagram. I wouldn't expect most peripherals to take up so many transistors compared to the processor cores.

texse• 6 years ago

It may be interesting to look at a more completely labeled die image (eg. a recent AMD laptop chip [1]).

I/O controllers, various PHYs, the display pipeline and the media engine account for a significant fraction of chip.

We also know the M1 has an integrated SSD controller, ISP and Thunderbolt which probably take a good chunk of room too.

[1] https://potatopc.net/amd-renoir-die-annotation-raises-hopes-...

kbumsik2• 6 years ago

We also know the M1 has an integrated SSD controller, ISP and Thunderbolt which probably take a good chunk of room too.

SSD controller and even Thunderbolt. That's the most integrated SoC I've heard of.

Waterluvian• 6 years ago

What's all the stuff that isn't annotated with a yellow rectangle?

deaddodo• 6 years ago

This is a SoC, not a CPU. The majority of the non-annotated is glue logic, I/O, etc.

bullen3• 6 years ago

That's alot of glue! I would say ~50%! What is the system cache? nm, see question below: https://news.ycombinator.com/item?id=25547796

xet7• 6 years ago

Techinsights page returns error.

Archive.org seems to work:

https://web.archive.org/web/20201227010504/https://www.techi...

commandpaul• 6 years ago

Does anyone know if its possible/ tried to do ram upgrades with higher capacity chips by replacing the dram like one does in the newer chip stacked iphones?

JohnJamesRambo3• 6 years ago

Can we step back and marvel at the absurdity that one of the most powerful CPUs ever created was for a phone?

chrisseaton4• 6 years ago

But these devices aren’t ‘phones’ any more are they? That’s disingenuous.

They’re hugely powerful general purpose computing devices - many people’s primary computing device if you go outside the tech bubble. They do compute intensive tasks like photo manipulation, on-line video editing, grading, and compression, they do face detection, depth estimation, etc etc.

I don’t know why people still say ‘wow all this for a telephone’ like all it does is make analog phone calls.

JohnJamesRambo3• 6 years ago

Because it fits in your hand and runs on a battery and is called a phone.

iPhone

As an aside, I was wondering what things does the iPhone 12 do that can even begin to tax all this horespower?

murukesh_s3• 6 years ago

its a shame we can't 'dock' em to a larger monitor to do desktop work yet. There are attempts by few players, but if Apple can do that it would be another revolution..

vnorilo• 6 years ago

From a device class where the newly all-important perf per watt was always the north star, in a market with the largest chip volume, by a company with ample margins and mountains of cash? Does not seem all that absurd :)

murukesh_s3• 6 years ago

yea it's *the world largest company (by market capitalisation), whose main product is their flagship mobile and whose main differentiator is the CPU (almost everything else is beaten or available in competing products like Camera, number of apps etc). The moment the CPU performance falls below Android devices, people will look at iPhones differently.. iOS is a differentiator but doesn't have the x factor to make iPhones justify the premium tag.

2muchcoffeeman• 6 years ago

I always assumed that they would be inferior in terms of absolute performance if you were going for performance per watt. I never imagined just how efficient the silicon could be.

quattrofan• 6 years ago

It's an ultra portable computer that for most people less than a majority of it's time will be used as a phone.

hombre_fatal• 6 years ago

It's not absurd at all, nor surprising that the most power efficient chip was compelled by the most portable general computing device. Plugged-in desktop computers certainly weren't going to bring us here.

This just sounds like "DAE think phone=bad XD" anti-intellectualism.

JohnJamesRambo3• 6 years ago

I think people are misconstruing what I am saying here. I said marvel at it. It's crazy that a handheld battery powered device has one of the greatest processors in the history of human existence in it.

faitswulff• 6 years ago

If you were used to using mainframes, you could have said much the same thing about desktop CPUs in the 90s. "That tiny thing doesn't even fill up a room!"

joakleaf2• 6 years ago

I wonder what they'll do for "M1X" and pro-machines?

Seems like heat isn't an issue, so can they increase the clock-speed of the M1?

And will they create a larger die with 2xGPU and 2xCPUs, or will they "stitch" multiple M1 dies together on one package to have a higher yield and reduce supply chain concerns?

An M1X with 2xGPU will bring it to Radeon 5500M level which is what I have in my current 16", but still not quite to Nvidia 1080 level.

So they'll definitely need a more powerful GPU for the pro-users. I wonder if they can make a 4xGPU, or if the yield will become poor as the die size increases. Maybe they'll have to do a discrete GPU, but that doesn't seem to match their current trajectory.

And then there have been "leaks" (rumors, really) about an 32xcore M1-family CPU, but I imagine the yield of such a beast would be horrible, with all the other SOC components on the same die. Especially, if they put a 32xcore GPU on the same die.

Hamuko• 6 years ago

Well, the prevailing rumor currently is a processor with 8 high-performance cores (as opposed to the 4 on the M1). I don't know if the amount of high-efficiency cores would also be doubled so that it's like 2xCPU when compared to the M1.

joakleaf2• 6 years ago

Yes, that's the M1X rumor, but that isn't going to be enough to compete with high-end CPUs in iMac and Mac Pros. For those, we have the 32-core rumor (but that is a massive die).

And doubling the M1-GPU will not be enough to even surpass current Macbook Pro offerings. So they'll also need to do something more on that front.

trollied• 6 years ago

I'd like to see a pro machine with daughterboard slots, allowing you to add extra CPUs. A bit like the Acorn RISC PC. I guess Apple have already done this to seom extent with the FPGA cards that you can buy for the current Mac Pro. I do wonder if that was a taste of things to come?

ksec2• 6 years ago

Seems like heat isn't an issue, so can they increase the clock-speed of the M1?

They are running into their current node limits, so dont expect a 4Ghz+ coming unless they decide to remake the whole thing in High Performance Node. ( Which means it will cost a lot more for a relatively niche volume.)

I wonder if they can make a 4xGPU

GPU increase should be given, and die size / yield would not be a concern / problem. The question is how will they handle the additional memory bandwidth requirement? As you can see in the Die-Shots those DDR I/O side are already take up much of the side space. Going with HBM2? Or LPDDR5?

32xcore M1-family CPU, but I imagine the yield of such a beast would be horrible,

A 32x CPU / GPU Core M1 will only be 4x increase from its current form. If you exclude I/O part, the CPU + GPU + Cache is only a little more than 50% of die size, or ~60mm2. 4x That is only 240mm2+. The final product would be a sub 350mm2 if you include all the additional. Which is roughly inline with the current estimate of AWS Graviton 2 die size ( ~400mm2).

But yes, M1 ( A14X ) isn't really that interesting. The most exciting part is how they will go about doing high performance chip. Especially on iMac and Mac Pro.

dooferlad• 6 years ago

(microelectronics graduate in 2001, turned software engineer, I have worked for ARM and Imagination Technologies on simulations of CPUs, GPUs, networking products and other bits of SoCs)

Yield wouldn't be that bad - ARM CPUs are small in comparison to their x86 cousins. I would expect that you would get to 32 CPUs the same way everyone else does - build a 32 core design and blow enable fuses on bad CPUs, then badge as appropriate.

To get a desktop class GPU they will need to address memory bandwidth. I am assuming they are using a tile based rendering process so memory bandwidth is less of an issue in comparison to scan line rendering (do desktop GPUs still do that?) I would assume that they are enjoying the benefits of the system level cache with the CPU and GPU sharing. I would expect there to be some careful benchmarking of increasing GPU cores, memory speed, memory size and system cache size going on at Apple.

There isn't anything stopping Apple from supporting external GPUs, but it would require a new SoC with more PCIe lanes. External buses are much more power hungry and take up space on the die. I don't have a mental model of how much space plonking a PCIe x16 interconnect would cost in terms of area or power (taking into account you need to get that data across the chip too), but my gut reaction is that it isn't likely unless there is a customer use case they can't solve on chip.

dismalpedigree3• 6 years ago

I work at TechInsights.

Depending on what the community is looking to know, there may be some things we can share publicly. I’m in the software group and not intimately involved with the RE parts, but I will act as an ambassador and see if I can get the right people involved.

Throw some comments on this thread and we’ll see what we can do to help. Might be next week due to holidays.

ksec2• 6 years ago

I am actually interested in the Non-Labeled part which no one on the Internet have even guessed what they are. Image Signal Processors ( ISP ), Video Decoder and Encoder, there should be a FPGA somewhere, Secure Enclave, NAND Controller, PCI-Express Controller / USB3 / Thunderbolt etc.

Mostly because all the other part are already well known.

kharak• 6 years ago

Can someone explain to me, what the term "bionic" means in this context? How is the chip derived from nature / natural systems?

sradman• 6 years ago

Pure speculation: the word bionic means using an artificial body part. The A14 Bionic includes a new Neural Engine for Artificial Intelligence.

EDIT: the name Bionic has been used since the A11 when the Neural Engine was introduced. Fusion introduced big.LITTLE.

studius• 6 years ago

Could a genetic algorithm have been partially used in the chip design also?

For example, in the layout to the right of the DDR on the M1, it looks like some significant optimization was done, but only in certain areas.

Apple Silicon M1 Die-Shots

Related Stories