Who cares if AMD are faster.
Apple is not going to use AMD anyway.
This is more a PR stunt to promote the product.
I won't buy a non-Mac laptop, but I will buy a Linux workhorse desktop. In fact, I bought a Linux Threadripper system back in 2018. That system is dying now after a lot of heavy use. I think Apple *should* be able to offer a superior alternative to AMD for my next workhorse desktop, but it's not at all clear they're going to live up to their potential. The fact that AMD is getting a 4nm product out there ahead of Apple does not bode well.
Is it really that unstable?
It was good for four years but lately has been turning itself off so often that it’s essentially useless.
The new AMD chip will be released in March, 15 months after M1 Pro became available in a shipping product. It will be a few months more before the new AMD chips appear in a laptop I would expect.
I think the M1 Max in a MBP would the appropriate comparison to an Ryzen 9 laptop chip, at least until Apple releases an updated Mx Max SOC. I would think the M1 Pro should go against new Ryzen 7 chips which would be an equivalent position in the respective ranges.
using the 4nm process while M1 Pro still at 5nm probably explains a lot of the performance gain
using Blender for the comparison. While Blender has recently been converted to Apple silicon and most importantly, Metal, it will still have a ways to go before it is optimised for M series and MacOS compared with x86. So AMD was smart to chose Blender.
Battery life at claimed 30 hours sounds amazing for x86 and appears to give Apple a run for its money. Let’s see how it does in real life.
the question is what does Apple have in the pipeline? It seems M2 Pro, max and ultra have been delayed so much, and with this AMD and Intel competition I would not be surprised if Apple waits a few months and releases something at 3nm, which they might still call M2 or jump to naming it M3 ( for 3nm hahah!).
What resolution was the monitor running? At what hz? At what brightness? Using what capacity battery? In what form factor? With what thermal overhead? At what power usage? Was the video test in a low power mode at minuscule brightness, low resolution and a small monitor? Was the blender render a known size that ram the majority of time at the 5.2 ghz boost in order to pad the numbers? What did the battery life look like during that process?
A ton of questions and little to no answers.
We can only test the veracity of these claims when these chips are in similar enclosures to Apples chips with comparable screen resolution, brightness, battery capacity, etc.
without these known factors, it’s all pie in the sky for now. We will see how thirst hold up against Apple’s freshman effort in a few months. Unfortunately for AMD, these chips will have to face reality in the real world. Beyond that, Apples sophomore lineup will be hitting the market around that time as well. And that’s just making matters more difficult.
So right now it’s a nothing burger. We will see what’s what in due time.
Also, these chips won’t be just competing against M Pro. It has the Max to deal with also. Just because AMD wants to pick on the pro doesn’t mean Apple can’t pick on them back. AMD doing the old MS Surface laptop be old iPad specs again. M series ipad basically pooped on that party. Going to happen here too, but the Pro will likely do the job itself.
It beats the M1 through very selective benchmarks. The raytracing benchmark is testing hardware that the M1 doesn’t even have… The M1 doesn’t boost the clock from 3.2 to 5.2 GHz for short time gains at the expense of efficiency.
the fact that Apple Silicon is among the benchmark comparisons along with Intel and AMD indicates the strength of it.
Came here to say this. In a few short years we've gone from "ARM chips will never be powerful enough to be used inside PCs" to "an ARM chip is now the benchmark for other top-end chips".
FWIW, the article has one fact completely wrong. The A16 is 4nm (N4). Also, the M2 (N5P) does not use the same process as the M1 (N5) series.
And, of course, N4 is TSMC’s third-generation enhancement of 5nm. It isn’t a new node. The difference between N5 and N5P is the similar to the difference between N5P and N4. A reminder:
A14/M1/M1+ = N5 A15/M2 = N5P A16 = N4
Finally, there are two more generations of TSMC’s 5nm process to come (N4P and N4X). It’s possible AMD is using one of these for this, so it could still be a step beyond the A16. Same goes, however, for the M2+ = likely to be N4 at the very least.
I disagree with the idea that Apple will skip M2+ and go straight to M3+. They will still call it M2 Pro/Max/Ultra even if it is 3nm, or N4X (my original guess). It would be bonkers to sacrifice/waste the marketing resonance of these low digits just to momentarily line up with TSMC.
If running cherry picked benchmarks was my primary source of income, I would be interested. Otherwise, if you use Apple products you are locked into Apple Silicon forever. Likewise you’ll never see Apple Silicon in a non-Apple product. So it’s all about bragging rights with no way for end users to pick & choose best of the best anyway. It’s a package deal.
FWIW, the article has one fact completely wrong. The A16 is 4nm (N4). Also, the M2 (N5P) does not use the same process as the M1 (N5) series.
And, of course, N4 is TSMC’s third-generation enhancement of 5nm. It isn’t a new node. The difference between N5 and N5P is the similar to the difference between N5P and N4. A reminder:
A14/M1/M1+ = N5 A15/M2 = N5P A16 = N4
Finally, there are two more generations of TSMC’s 5nm process to come (N4P and N4X). It’s possible AMD is using one of these for this, so it could still be a step beyond the A16. Same goes, however, for the M2+ = likely to be N4 at the very least.
I disagree with the idea that Apple will skip M2+ and go straight to M3+. They will still call it M2 Pro/Max/Ultra even if it is 3nm, or N4X (my original guess). It would be bonkers to sacrifice/waste the marketing resonance of these low digits just to momentarily line up with TSMC.
Anandtech has a table of TSMC process nodes. The real dates are probably 6 to 12 months delayed compared to the ones listed in the table, as this article was written about 12 months ago:
AppleInsider has a weird bug up its butt about calling TSMC N4 or 4nm nodes as 5nm. Wonder if they will say Intel 7 is really just 10nm? Intel is choosing "7" and "4" for their process branding for obvious reasons.
My bet is that the M2 Max derived SoCs (Pro, Max and Ultra) will be TSMC N4 or N4P. I think the only real difference between the two is yield. That's important for AMD as they can squeeze more SKUs out, but for Apple? They get first dibs anyways, and are only interested in 1 SKU. If it was N5P, they would have shipped new MBP, iMac, and Mac mini machines last Fall.
If running cherry picked benchmarks was my primary source of income, I would be interested. Otherwise, if you use Apple products you are locked into Apple Silicon forever. Likewise you’ll never see Apple Silicon in a non-Apple product. So it’s all about bragging rights with no way for end users to pick & choose best of the best anyway. It’s a package deal.
In one aspect, you can choose the best, you can’t unplug anything Intel, and AMD design from a wall in the laptop form factor if that is a need.
In another market, one Apple so far chooses not to compete in, web services and more specifically servers, there may be something else to choose from in that market if Qualcomm and those three ex Apple engineers design a chip or chips that will work in a server in addition to it’s main goal, to have them work in smartphones and tablets, but they can only go so far without a in house OS.
If running cherry picked benchmarks was my primary source of income, I would be interested. Otherwise, if you use Apple products you are locked into Apple Silicon forever. Likewise you’ll never see Apple Silicon in a non-Apple product. So it’s all about bragging rights with no way for end users to pick & choose best of the best anyway. It’s a package deal.
In one aspect, you can choose the best, you can’t unplug anything Intel, and AMD design from a wall in the laptop form factor if that is a need.
In another market, one Apple so far chooses not to compete in, web services and more specifically servers, there may be something else to choose from in that market if Qualcomm and those three ex Apple engineers design a chip or chips that will work in a server in addition to it’s main goal, to have them work in smartphones and tablets, but they can only go so far without a in house OS.
Huawei has done exactly that and for a few years now.
CANN Mindspore Ascend (Max down to nano) TaiShan HarmonyOS / OpenEuler PV ...
Sanctions forced it to re-jig its supply chain almost from top to bottom over the last two or three years but they have just announced 'business as usual' for 2023.
A recent analysis of a 5G base station revealed about 1% of US technology.
Full stack/full scenario solutions for literally everything... except one. Huawei laptops and desktops ship with Windows for consumers in the West. Those systems do however run with some core elements of HarmonyOS to tie in with the HarmonyOS ecosystem. In China business systems can run a Chinese based Linux system. Eventually everything will probably run HarmonyOS for CE and OpenEuler for servers etc.
FWIW, the article has one fact completely wrong. The A16 is 4nm (N4). Also, the M2 (N5P) does not use the same process as the M1 (N5) series.
And, of course, N4 is TSMC’s third-generation enhancement of 5nm. It isn’t a new node. The difference between N5 and N5P is the similar to the difference between N5P and N4. A reminder:
A14/M1/M1+ = N5 A15/M2 = N5P A16 = N4
Finally, there are two more generations of TSMC’s 5nm process to come (N4P and N4X). It’s possible AMD is using one of these for this, so it could still be a step beyond the A16. Same goes, however, for the M2+ = likely to be N4 at the very least.
I disagree with the idea that Apple will skip M2+ and go straight to M3+. They will still call it M2 Pro/Max/Ultra even if it is 3nm, or N4X (my original guess). It would be bonkers to sacrifice/waste the marketing resonance of these low digits just to momentarily line up with TSMC.
Anandtech has a table of TSMC process nodes. The real dates are probably 6 to 12 months delayed compared to the ones listed in the table, as this article was written about 12 months ago:
AppleInsider has a weird bug up its butt about calling TSMC N4 or 4nm nodes as 5nm. Wonder if they will say Intel 7 is really just 10nm? Intel is choosing "7" and "4" for their process branding for obvious reasons.
My bet is that the M2 Max derived SoCs (Pro, Max and Ultra) will be TSMC N4 or N4P. I think the only real difference between the two is yield. That's important for AMD as they can squeeze more SKUs out, but for Apple? They get first dibs anyways, and are only interested in 1 SKU. If it was N5P, they would have shipped new MBP, iMac, and Mac mini machines last Fall.
Their press releases introduce the idea of the “N5 platform” which is comprised of N5, N5P, N4, N4P, and N4X processes. Another way they refer to this is as a “design library”—the “N5 design library” (shared by all of the above processes) and the “N3 design library” … For example: https://n3.tsmc.com/english/dedicatedFoundry/technology/N3.htm
Note the change to “FinFlex” (N3 and N2) “methodology” from FinFET (FET = Field-Effect Transistor) used from 16nm to N5.
FWIW, the article has one fact completely wrong. The A16 is 4nm (N4). Also, the M2 (N5P) does not use the same process as the M1 (N5) series.
And, of course, N4 is TSMC’s third-generation enhancement of 5nm. It isn’t a new node. The difference between N5 and N5P is the similar to the difference between N5P and N4. A reminder:
A14/M1/M1+ = N5 A15/M2 = N5P A16 = N4
Finally, there are two more generations of TSMC’s 5nm process to come (N4P and N4X). It’s possible AMD is using one of these for this, so it could still be a step beyond the A16. Same goes, however, for the M2+ = likely to be N4 at the very least.
I disagree with the idea that Apple will skip M2+ and go straight to M3+. They will still call it M2 Pro/Max/Ultra even if it is 3nm, or N4X (my original guess). It would be bonkers to sacrifice/waste the marketing resonance of these low digits just to momentarily line up with TSMC.
Correct. 3nm doesn’t equal 3rd generation. It’s just a more efficient version of the same architecture due to the smaller process and overhead breathing room it creates.
A 3nm m2 Max for example will be able to be either more efficient or more powerful than a 5nm. It’s still an M2 architecture.
I think the 3nm M2s are what we will see in the new MBPs. And they will easily be the best chips overall in a mobile form factor. Apple will lead in this area for a long time.
What Apple needs to focus on are the desktop (iMac, Mac Pro) versions of its chips. SOCs that are allowed to eat much more and clock much higher. A D or X series chip that doesn’t have the mobile consideration constraints applied.
When I buy a Mac Pro, I don’t give a rip how power efficient it is. I want it to kick EVERYONES butt and take names. Dial up the GHz, increase the performance core count like crazy, limit efficient cores to two, Keep the same case design meant for amazing cooling and let it rip.
Once that is on the table, AMD and Intel won’t be able to play these types of games any longer. It will be take over. The only hope at that point is for them to rely on discrete graphics cards with their own power and heat issues to compare gaming benchmarks with Apple Silicon.
FWIW, the article has one fact completely wrong. The A16 is 4nm (N4). Also, the M2 (N5P) does not use the same process as the M1 (N5) series.
And, of course, N4 is TSMC’s third-generation enhancement of 5nm. It isn’t a new node. The difference between N5 and N5P is the similar to the difference between N5P and N4. A reminder:
A14/M1/M1+ = N5 A15/M2 = N5P A16 = N4
Finally, there are two more generations of TSMC’s 5nm process to come (N4P and N4X). It’s possible AMD is using one of these for this, so it could still be a step beyond the A16. Same goes, however, for the M2+ = likely to be N4 at the very least.
I disagree with the idea that Apple will skip M2+ and go straight to M3+. They will still call it M2 Pro/Max/Ultra even if it is 3nm, or N4X (my original guess). It would be bonkers to sacrifice/waste the marketing resonance of these low digits just to momentarily line up with TSMC.
Anandtech has a table of TSMC process nodes. The real dates are probably 6 to 12 months delayed compared to the ones listed in the table, as this article was written about 12 months ago:
AppleInsider has a weird bug up its butt about calling TSMC N4 or 4nm nodes as 5nm. Wonder if they will say Intel 7 is really just 10nm? Intel is choosing "7" and "4" for their process branding for obvious reasons.
My bet is that the M2 Max derived SoCs (Pro, Max and Ultra) will be TSMC N4 or N4P. I think the only real difference between the two is yield. That's important for AMD as they can squeeze more SKUs out, but for Apple? They get first dibs anyways, and are only interested in 1 SKU. If it was N5P, they would have shipped new MBP, iMac, and Mac mini machines last Fall.
Their press releases introduce the idea of the “N5 platform” which is comprised of N5, N5P, N4, N4P, and N4X processes. Another way they refer to this is as a “design library”—the “N5 design library” (shared by all of the above processes) and the “N3 design library” … For example: https://n3.tsmc.com/english/dedicatedFoundry/technology/N3.htm
Note the change to “FinFlex” (N3 and N2) “methodology” from FinFET (FET = Field-Effect Transistor) used from 16nm to N5.
No idea about the daftness in tech media regarding technical details. Some people have been working in the tech media business for the better part of 2 decades, or more, and should have absorbed technical knowledge through osmosis at least. Perhaps they have to balance being propagandists for OEMs in order to get access versus the technical and factual quality of their content.
Like this AMD statement: "AMD claims up to 30 hours of video playback thanks to the efficiency gains and a powerful on-chip AI." This was right after a graph saying "Apple is also guilty of overplaying its hand in a keynote presentation with selective benchmarks".The former is obviously "overplaying".
AMD is straight up lying with that statement. By saying "claims", AI is just falling to a propaganda technique, or public relations technique, which relies on the media to repeat the lie with a soft disclaimer. The disclaimer doesn't work in readers or viewers minds, and the media just ends up repeating the lie. You could put a desktop processor in a laptop and claim 30 hrs of video or whatnot. As long as there is a big enough battery, it will be true! Except for the part of it not being true for 99.999% of the devices on the market.
(A video playing benchmark is just a display brightness test now as the display will use 80 to 90% of the power in such a test. It will be sensitive to the idle power consumption of the SoC, and hundreds of milliwatts will matter here, but at the limit, it will be battery capacity, display size, power efficiency of the display, and the idle power consumption of the computing bits. TSMC 4nm will get the idle power consumption down, but they aren't getting 30 hrs unless they pair a 13" display with a 100 WHr battery, or with a 50 WHr and having the display at unusable brightness.)
Comments
- The new AMD chip will be released in March, 15 months after M1 Pro became available in a shipping product. It will be a few months more before the new AMD chips appear in a laptop I would expect.
- I think the M1 Max in a MBP would the appropriate comparison to an Ryzen 9 laptop chip, at least until Apple releases an updated Mx Max SOC. I would think the M1 Pro should go against new Ryzen 7 chips which would be an equivalent position in the respective ranges.
- using the 4nm process while M1 Pro still at 5nm probably explains a lot of the performance gain
- using Blender for the comparison. While Blender has recently been converted to Apple silicon and most importantly, Metal, it will still have a ways to go before it is optimised for M series and MacOS compared with x86. So AMD was smart to chose Blender.
- Battery life at claimed 30 hours sounds amazing for x86 and appears to give Apple a run for its money. Let’s see how it does in real life.
the question is what does Apple have in the pipeline? It seems M2 Pro, max and ultra have been delayed so much, and with this AMD and Intel competition I would not be surprised if Apple waits a few months and releases something at 3nm, which they might still call M2 or jump to naming it M3 ( for 3nm hahah!).without these known factors, it’s all pie in the sky for now. We will see how thirst hold up against Apple’s freshman effort in a few months. Unfortunately for AMD, these chips will have to face reality in the real world. Beyond that, Apples sophomore lineup will be hitting the market around that time as well. And that’s just making matters more difficult.
And, of course, N4 is TSMC’s third-generation enhancement of 5nm. It isn’t a new node. The difference between N5 and N5P is the similar to the difference between N5P and N4. A reminder:
A14/M1/M1+ = N5
A15/M2 = N5P
A16 = N4
Finally, there are two more generations of TSMC’s 5nm process to come (N4P and N4X). It’s possible AMD is using one of these for this, so it could still be a step beyond the A16. Same goes, however, for the M2+ = likely to be N4 at the very least.
I disagree with the idea that Apple will skip M2+ and go straight to M3+. They will still call it M2 Pro/Max/Ultra even if it is 3nm, or N4X (my original guess). It would be bonkers to sacrifice/waste the marketing resonance of these low digits just to momentarily line up with TSMC.
AppleInsider has a weird bug up its butt about calling TSMC N4 or 4nm nodes as 5nm. Wonder if they will say Intel 7 is really just 10nm? Intel is choosing "7" and "4" for their process branding for obvious reasons.
My bet is that the M2 Max derived SoCs (Pro, Max and Ultra) will be TSMC N4 or N4P. I think the only real difference between the two is yield. That's important for AMD as they can squeeze more SKUs out, but for Apple? They get first dibs anyways, and are only interested in 1 SKU. If it was N5P, they would have shipped new MBP, iMac, and Mac mini machines last Fall.
In another market, one Apple so far chooses not to compete in, web services and more specifically servers, there may be something else to choose from in that market if Qualcomm and those three ex Apple engineers design a chip or chips that will work in a server in addition to it’s main goal, to have them work in smartphones and tablets, but they can only go so far without a in house OS.
CANN
Mindspore
Ascend (Max down to nano)
TaiShan
HarmonyOS / OpenEuler
PV
...
Sanctions forced it to re-jig its supply chain almost from top to bottom over the last two or three years but they have just announced 'business as usual' for 2023.
A recent analysis of a 5G base station revealed about 1% of US technology.
Full stack/full scenario solutions for literally everything... except one. Huawei laptops and desktops ship with Windows for consumers in the West. Those systems do however run with some core elements of HarmonyOS to tie in with the HarmonyOS ecosystem. In China business systems can run a Chinese based Linux system. Eventually everything will probably run HarmonyOS for CE and OpenEuler for servers etc.
Note the change to “FinFlex” (N3 and N2) “methodology” from FinFET (FET = Field-Effect Transistor) used from 16nm to N5.
Like this AMD statement: "AMD claims up to 30 hours of video playback thanks to the efficiency gains and a powerful on-chip AI." This was right after a graph saying "Apple is also guilty of overplaying its hand in a keynote presentation with selective benchmarks".The former is obviously "overplaying".
AMD is straight up lying with that statement. By saying "claims", AI is just falling to a propaganda technique, or public relations technique, which relies on the media to repeat the lie with a soft disclaimer. The disclaimer doesn't work in readers or viewers minds, and the media just ends up repeating the lie. You could put a desktop processor in a laptop and claim 30 hrs of video or whatnot. As long as there is a big enough battery, it will be true! Except for the part of it not being true for 99.999% of the devices on the market.
(A video playing benchmark is just a display brightness test now as the display will use 80 to 90% of the power in such a test. It will be sensitive to the idle power consumption of the SoC, and hundreds of milliwatts will matter here, but at the limit, it will be battery capacity, display size, power efficiency of the display, and the idle power consumption of the computing bits. TSMC 4nm will get the idle power consumption down, but they aren't getting 30 hrs unless they pair a 13" display with a 100 WHr battery, or with a 50 WHr and having the display at unusable brightness.)