Alleged Apple M1X Processor Specifications Surface

erek

[H]F Junkie
Joined
Dec 19, 2005
Messages
10,785
"Enter the world of the Apple M1X processor. Designed for high-end laptops and the most demanding workloads, the new processor aims to create a new performance level. Featuring a 12-core CPU with eight big and four small cores, the M1X processor is going to deliver much better performance than M1. The graphics and memory configuration are currently unknown, so we have to wait and see how it will look like. The M1X is set to arrive sometime in Q1 of 2021, according to the source of the leak, so be patient and remember to take this information with a grain of salt."

https://www.techpowerup.com/275102/alleged-apple-m1x-processor-specifications-surface
 
So far that just lines up with what was expected. Doubling the big cores will translate to big gains in media work and other things that multi-thread well.

I think the real question for people interested in the higher end chip is going to be neural processors... tensor cores? Whatever we’re calling them now. And GPU size.

I’d also have normally said it would be nice to have confirmation of higher amounts of RAM, and I’m sure we’ll get that at some point, but reports so far are showing that the traditional view of amount of memory needed by a “PC” doesn’t seem to apply to the M1-based systems available so far.
 
So far that just lines up with what was expected. Doubling the big cores will translate to big gains in media work and other things that multi-thread well.

All of the "Media Work" is done by custom ASIC encoders and decoders. Seriously doubt "big gains in media".
 
lol @ everyone that was like "Full power desktop x86 IS STILL FASTER THAN 5 Watt m1!!!! ARM WILL NEVER BE FASTER!!!!"

But that is the kind of people has been ignoring the ARM server hardware released in the last years, and that the fastest computer in the world runs on ARM.

From a source that leaked the M1X specs two days before techpowerup did: "if you think M1 is fast, you haven’t seen M1X"
 
All of the "Media Work" is done by custom ASIC encoders and decoders. Seriously doubt "big gains in media".
That isn't really completely true however. I just posted a video of someone using the M1 mini for Logic in the other M1 thread. You can clearly watch the task manager and see the ARM cores usage going up the more processing being thrown at it. It wasn't all being offloaded, and the ARM cores clearly chewed through 3-5x the workload the previous mini (which yes is long in the tooth at this point) was able to do. The ARM cores where doing the lifting not some ASIC. What impressed me most watching that was just how responsive the machine remained even when the cores where being nailed at near 100%. The interface remained completely responsive and even zippy right up to the point where he pushed it just a bit to far at over 1000 logic plugins running at once.
 
But that is the kind of people has been ignoring the ARM server hardware released in the last years, and that the fastest computer in the world runs on ARM.

From a source that leaked the M1X specs two days before techpowerup did: "if you think M1 is fast, you haven’t seen M1X"
It’s because it’s Apple. If this were anybody else they would be cheering and jumping up and down.
 
All of the "Media Work" is done by custom ASIC encoders and decoders. Seriously doubt "big gains in media".
Not hardly, and this has been proven again and again.
I'm not saying that hardware encoders aren't being used, but Intel x86-64 CPUs have these as well, so that isn't totally inclusive.

What is so shocking is how powerful the Apple M1 ARM64 CPU is for operating in a sub-25 watt TDP, and that includes the GPU.
The GPU is capable of 2.6 TFLOPS, which doesn't sound like a ton, but to put this in perspective, the NVIDIA Series 16 GTX 1650 also nets roughly 2.6 TFLOPS... at a 75 watt TDP, and that is the GPU by itself.

So Apple has managed a powerful SoC (CPU, GPU, etc.) in under a 25 watt TDP.
No other SoC on the market can manage to perform how the M1 does at that TDP in both CPU and GPU, at least to my knowledge.
 
That isn't really completely true however. I just posted a video of someone using the M1 mini for Logic in the other M1 thread. You can clearly watch the task manager and see the ARM cores usage going up the more processing being thrown at it. It wasn't all being offloaded, and the ARM cores clearly chewed through 3-5x the workload the previous mini (which yes is long in the tooth at this point) was able to do. The ARM cores where doing the lifting not some ASIC. What impressed me most watching that was just how responsive the machine remained even when the cores where being nailed at near 100%. The interface remained completely responsive and even zippy right up to the point where he pushed it just a bit to far at over 1000 logic plugins running at once.
Previous mini to M1 is not M1 to M1X. Also the M1 performance is very dependent on the app used. The Logic app is Sound only;. Pretty low bar to set. Most people thing of video and sound (ie. x265 encoding) as media, since that is by far the typical usage.
 
I’d also have normally said it would be nice to have confirmation of higher amounts of RAM, and I’m sure we’ll get that at some point, but reports so far are showing that the traditional view of amount of memory needed by a “PC” doesn’t seem to apply to the M1-based systems available so far.
If the solid state storage is fast enough, then it might be suitable for "normal" tasks as virtual memory. If you can read/write 1-2GB/s then you can swap out memory as needed. For memory intensive tasks this would be a different story, but I can't say I have ever noticed 8GB not being enough memory on my modern Macintosh computers.

If you think about that memory's original purpose was to overcome the bandwidth/latency limitation of traditional storage, we are reaching a point where that is no longer as much a concern in real world usage.
 
Not hardly, and this has been proven again and again.
I'm not saying that hardware encoders aren't being used, but Intel x86-64 CPUs have these as well, so that isn't totally inclusive.

What is so shocking is how powerful the Apple M1 ARM64 CPU is for operating in a sub-25 watt TDP, and that includes the GPU.
The GPU is capable of 2.6 TFLOPS, which doesn't sound like a ton, but to put this in perspective, the NVIDIA Series 16 GTX 1650 also nets roughly 2.6 TFLOPS... at a 75 watt TDP, and that is the GPU by itself.

So Apple has managed a powerful SoC (CPU, GPU, etc.) in under a 25 watt TDP.
No other SoC on the market can manage to perform how the M1 does at that TDP in both CPU and GPU, at least to my knowledge.
Wow, I have 25 TDP, but there's a plug 2 feet away. A lot of people don't care about TDP, and/or doesn't technically come into play in a normal usage scenario. The TDP play on ARM just seems overused and overhyped over and over again.
 
Wow, I have 25 TDP, but there's a plug 2 feet away. A lot of people don't care about TDP, and/or doesn't technically come into play in a normal usage scenario. The TDP play on ARM just seems overused and overhyped over and over again.
You aren't seeing the forest through the trees on this.
Again, having that much CPU and GPU within a sub-25 watt TDP is virtually unheard of, and nothing in the x86-64 camp even comes close.

Sure, x86-64 and a discrete GPU can do this... at a 65 watt TDP for the CPU and a 75 watt TDP for the GPU.
So, when a sub-25 watt TDP SoC can beat a combined 140 watt TDP of an x86-64 build, that is quite the large paradigm shift.

To have that much processing power in such a small thermal envelope is a giant leap forward in computing.
To see the even bigger scope, the main excitement of all of this is that a megacorp like Apple is pushing ARM forward, and will do so in such a way that it will start to propel ARM forward in much quicker and more mainstream ways for both hardware and software in general for other vendors and software developers.

I'm not saying this to stroke Apple's ego, I'm saying this because it is going to help ARM progress in ways that other smaller corporations could not do in the past because of either niche product placement (mobile-only or niche-servers) or lack of widespread/mainstream usage.


I am really starting to see the majority of individuals who comment in these threads are missing that point entirely.
It isn't just "Apple sux, who cares about a low TDP???", it's about how ARM is going to be propelled forward in ways we have never seen before, and especially so on the laptop and desktop segments.

The AMD Jaguar SoC in my sig has a 25 watt TDP, and all 4 CPU cores wouldn't match a single big core on the M1 in either integer or floating point operations.
The Radeon HD 8400E GPU is also part of that SoC, and it might net 120 GFLOPS at best.

That's the point I am trying to make, and the paradigm shift that will be the start of the migration away from x86-64.
This is the only the beginning.
 
Last edited:
You aren't seeing the forest through the trees on this.
Again, having that much CPU and GPU within less than a 25 watt TDP is virtually unheard of, and nothing in the x86-64 ISA even comes close.

Sure, x86-64 and a discrete GPU can do this... at a 65 watt TDP for the CPU and a 75 watt TDP for the GPU.
So, when a sub-25 watt TDP SoC can beat a combined 140 watt TDP of an x86-64 build, that is quite the large paradigm shift.

To have that much processing power in such a small thermal envelope is a giant leap forward in computing.
To see the even bigger scope, the main excitement of all of this is that a megacorp like Apple is pushing ARM forward, and will do so in such a way that it will start to propel ARM forward in much quicker and more mainstream ways for both hardware and software in general for other vendors and software developers.

I'm not saying this to stroke Apple's ego, I'm saying this because it is going to help ARM progress in ways that other corporations could not do in the past because of either niche product placement (mobile-only or niche-servers) or lack of widespread/mainstream usage.
ARM -> "paradigm shift" "Giant leap forward" "forest through the trees" "giant leap in computing"

Been hearing this for over 20 years now. Wake me up in another 10 or 20. Then we will see.
 
ARM -> "paradigm shift" "Giant leap forward" "forest through the trees" "giant leap in computing"

Been hearing this for over 20 years now. Wake me up in another 10 or 20. Then we will see.
Well it’s actually being brought to consumer devices now and Apple has a pretty decent track record of supporting their devices even though there will be growing pains for early adopters.
 
ARM -> "paradigm shift" "Giant leap forward" "forest through the trees" "giant leap in computing"

Been hearing this for over 20 years now. Wake me up in another 10 or 20. Then we will see.
I don't know what you've been hearing for the last 20 years, but you might want to wake up sooner than later...

72aa11cdea4ae9cf710a2fa672411de5c15d88b5.jpg
 
I know a lot of people here couldn’t care less about TDP but I very much do. I have buildings with hundreds of devices each and a 100w TDP savings spread over that many units works out to a substantial energy savings. As it currently stands I am having a hard time not justifying the purchase of a bunch of Mac Mini’s for many of our labs. My next labs on the chopping block are Intel 6700’s running 1080’s. These look to not only out perform them, but would do so in a way that would let me turn off the AC to those rooms.
 
I am really starting to see the majority of individuals who comment in these threads are missing that point entirely.
It isn't just "Apple sux, who cares about a low TDP???", it's about how ARM is going to be propelled forward in ways we have never seen before, and especially so on the laptop and desktop segments.

It'll be nice when we start seeing that power in non-Apple OSes. I don't have anything against MacOS but the software I use at my day job isn't available on the Mac at all, so the great performance envelope isn't useful to me. (Not to mention the whole desktop gaming issue, either.)
 
To me this is more convergence that divergence. As much as a 25w SoC is ahead of the curve, it isn't that far ahead. Being 5nm certainly has some advantages. Having a long linage of low power siblings also helps.

To be sure the floor dropped quite a bit for the other players. We fer sure know where all the wafers went.
 
I'm a die hard PC guy but I've been thinking of picking one of these up as a toy / email / writing appliance.

From the reviews, not only is the performance on par or better than PC counterparts (especially for optimized apps / not Rosetta 2 reliant), but the power drain is incredibly efficient.

I'm always going to be keeping a mega monster PC to play games on, but for carrying around or doing work, I can see adopting one of these, especially if the apps are Big Sur / ARM native and can take advantage of the new hardware. It doesn't take a rocket ship to run Microsoft Office or Excel / Word - which I do 99% of the time.
 
A lot of people don't care about TDP, and/or doesn't technically come into play in a normal usage scenario.
A lot of people doesn't care about hardware details, but those with technical knowledge know how important TDP is.
 
Wow, I have 25 TDP, but there's a plug 2 feet away. A lot of people don't care about TDP, and/or doesn't technically come into play in a normal usage scenario. The TDP play on ARM just seems overused and overhyped over and over again.
You need to imagine, there is a bit of projection going on, if it can do this with just 15 watt in 2020, what will it do in 2021 with 50 watt, 200 watt and just still easy to run from a regular office outlet/regular AC regular 2000 watt.
 
You need to imagine, there is a bit of projection going on, if it can do this with just 15 watt in 2020, what will it do in 2021 with 50 watt, 200 watt and just still easy to run from a regular office outlet/regular AC regular 2000 watt.
You can extrapolate to infinity and you are ignoring how AMD and Intel have also vastly decreased their mobile platform efficiency.
 
You can extrapolate to infinity and you are ignoring how AMD and Intel have also vastly decreased their mobile platform efficiency.
I am not sure how relevant AMD and Intel decrease in mobile platform efficiency is when we try to extrapolate if the M1 can do all that with only 15w, try to imagine with 50-200-2000 watt worth of them.

If a platform use little power, that make it way easier and cheaper to scale (all else being equal).
 
I am not sure how relevant AMD and Intel decrease in mobile platform efficiency is when we try to extrapolate if the M1 can do all that with only 15w, try to imagine with 50-200-2000 watt worth of them.

If a platform use little power, that make it way easier and cheaper to scale (all else being equal).
If power was the only concern, x86 would be gone. ARM fans ignore everything but efficiency for the sake of argument. This is why I stated earlier that this single point of view was / is / and always will be overplayed by a wide margin.
 
Does Cinebench run on the Bigcores only or also the little ones? I would be interested in knowing. Because 8c/16t vs 4c puts one of those at a pretty large disadvantage.
From what I have seen when they want to compare threads, they limit them to either 4 core or even 4 thread for that comparison, so I think only the 4c are running on the m1 on those.
 
If power was the only concern, x86 would be gone. ARM fans ignore everything but efficiency for the sake of argument. This is why I stated earlier that this single point of view was / is / and always will be overplayed by a wide margin.
You going from different extreme, power is not the only concern obviously and obviously it is a relevant concern.
 
You going from different extreme, power is not the only concern obviously and obviously it is a relevant concern.
Obviously. Code portability to an x86 desktop platform comes to mind as well. As well as not having to port to ARM from X86.
 
Does Cinebench run on the Bigcores only or also the little ones? I would be interested in knowing. Because 8c/16t vs 4c puts one of those at a pretty large disadvantage.
It has to run on all cores huge and small. If perfect scaling 4 * 1498 = 5992 < 7508 all core score.
 
From what I have seen when they want to compare threads, they limit them to either 4 core or even 4 thread for that comparison, so I think only the 4c are running on the m1 on those.
https://www.cpu-monkey.com/en/cpu-amd_ryzen_7_4800u-1142

Well, I don't think they did limit them because they have those same scores in their benchmark run for the 4800U against the 3975wx and the 1068G7 for both single and multi-core runs. So it looks like they did a balls-out to balls-out comparison which isn't bad, it's good to know where the CPU's stand, just lets us know that while the M1's may get relatively good multi CPU scores if you know you are needing a very thread heavy load the M1 may not be the chip for you.
 
Back
Top