Mac Mini DTK A12Z faster than MBP 16" i9 in native benchmarks.

[deconstruct60]

Actually Apple, after announcing the transiton to Intel in June 2005, introduced the first Intel iMacs in Jan 2006, only three months after they released new iMac G5s in Oct 2005. They won't wait a year to release AS iMac. They discontinued the last iMac G5s after three months. These newly released Intel iMacs can meet the same fate this time too.

In 2005 , Intel had a complete line up of CPU processor products. (in 2004 same thing. 2003 again same thing). Intel was going to have a complete line up in 2006 whether Apple moved over or not. Intel had reference boards for al the major system type variations also ( working laptop, desktop , workstation reference implementation boards )

2020 is not 2005. Apple does not have a complete line up of CPU processor products right now at all. The newest iPhones get a new processor each year, but that is about it. The majority of the iPhone product line up consists of "hand me down' processors. The majority of the iPad line up consists of "hand me down" processors. (only the iPad Pro regularly get "new" ones. Un early 2020, it didn't really even get that. A12Z was a A12X die with a core turned on rather than off. ). The iPad Pro's have only gotten new updates on process shrink ( so on a 2-3 year cycle).

There is no track record of Apple highly concurrently doing a wide variety of new SoC all at the same time at all. So the notion that "this is almost exactly like 2005" is relatively bankrupt. There is lots of wishful thinking that it might be like 2005, but truth is that Apple is jumping to an unproven vendor at this point for this task of covering the whole Mac line up.

Apple just released a 2020 iMac 27". There is zero pressing need to obsolete that in less than 3-6 months at all. This system just jumped to the T2.

The other issue at play here is that Apple is jumping to 5nm later this year. Doing much larger dies on a brand new process is asking for trouble. There is no good reason to put Macs at substantively higher risks.

2005-2006 also didn't have a global pandemic either. The 24" iMac intel could have quite easily been decoupled from the 26" iMac Intel in the release process. The LG Ultrafine moved to the 24" panel last year (off of the 21.5"). Again the current iMac getting off a panel that "nobody" buys anymore. (not that there every were lots of 21.5" buyers).

Apple gave themselves two years. They can use it and have not broken any "promise". Last time they could lean on Intel to do major work for them. Major rumbling at the time is that the Mac Pro logicboard was largely intel with minorApple tweaks sprinkled on top to justify the "designed in Cupertino" badge also slapped on. There is lots of dotting i's and crossing t's that Intel did for Apple on last transition that Apple will have to do themselves this time. A "Big Bang" transition where it is all done in a handful of months is unlikely. It isn't impossible ( this is usually where folks hand waving at the Scrooge McDuck money pit and saying Apple is just going to throw buckets of cash at it to make it go as fast as possible ... no expense spared. ), but it isn't likely.

Neither is Mac SoC's being on a bleeding edge iPhone Pro like cadence going forward either. If Mac SoCs are going to have 18-24 month iteration cycles then it doesn't make any sense to tightly couple the whole line up together. laptops and desktops playing leapfrog would be a better stream of newer products. ( not like the iPhone market where can just trot out 3 year SoC to hit lower price points. )

 

[deconstruct60]

I think the first silicon Macs will be the air, 13/14” pro, and the Mac mini. I don’t think the 16” pro and iMac will be on silicon until the A15 variant is ready. No Mac Pro until the A16 is ready.

Probably not going to work that way. The 'X' variation of the A-series was only picking up update every process shrink ( 18-24 months). The Macs SoC in the desktop range are likely even bigger SoC dies. Apple probably isn't going to push the biggest dies onto the most bleeding edge process that TSMC has available. That is just asking for yield issues (costs ).

The A_X series could probably better get away with the long cycle because nothing was really touching the iPad Pro in performance anyway. That series has been skipping the odds. ( no A13X , no A11X , is an A9X but that isn't where there in the sub 10nm range shrinks, design, and fab cost issues. It is costing more to go dramatically smaller. )

Desktop/Workstation wise that probably isn't going to be as easy a reach for Apple. AMD isn't screwed up their fab process ( using the same subcontractor Apple is. ). Two years Intel will be closer to being back on track also.

I’d be all over a mac mini with A14x/z, 16GB of memory, and SSD for $799.

A Mac Mini with a A14X probably woulnd't have Thunderbolt. Which apple has said the new Macs would. It is probably going to be a different SoC than what the iPad Pro uses.

IF Apple did a one-two port wonder laptops with just USB-C again then perhaps that would be a more minor variation of the A14X but probably not a mini.

The DTK is actually problematical development wise for those who need to work on I/O drivers. ( no Thunderbolt ). The DTK isn't a "Mac" product.

The Mac Mini had finally got to the stage that it was using "desktop" processors ( although in a BGA package, but a mainstream desktop die and settings. ). If Apple wanted to backslide the Mini back into using MBP processors (and MBP ports. Dropping down to perhaps 2-4 Thunderbolt ports ). then that could come out shorter term Mini transition.

IMHO, that is probably not a good place for the Mini long term. But it is path Apple could take.

 

[Malus120]

8-core A12Z 17% faster than 8-core 2.3 - 4.8GHz i9 in native benchmarks. Rosetta means a 52% performance hit.



View attachment 948606

Thanks for sharing

Those numbers are... good. Really good. Obviously I can't say how representative this benchmark is but assuming it's representative... If Apple can deliver this level of performance in a 12" MacBook form factor this fall with an A13/A14 derivative that would be a powerful statement that would convince a lot of the people who have been hesitant or downright negative about the switch to Apple Silicon (myself included) to give it a go.
Hell, if performance really is that good, I'll probably buy one, maybe even two depending on pricing/software.availability.

(That said, the Ryzen 4800U, a 15W part is faster than the 16MBP in Cinebench R20 so again, while cautiously optimistic I still wanna see what an A14 series chip can do before I say Apple made the right call)

 

[Homy]

In 2005 , Intel had a complete line up of CPU processor products. (in 2004 same thing. 2003 again same thing). Intel was going to have a complete line up in 2006 whether Apple moved over or not. Intel had reference boards for al the major system type variations also ( working laptop, desktop , workstation reference implementation boards )

2020 is not 2005. Apple does not have a complete line up of CPU processor products right now at all. The newest iPhones get a new processor each year, but that is about it. The majority of the iPhone product line up consists of "hand me down' processors. The majority of the iPad line up consists of "hand me down" processors. (only the iPad Pro regularly get "new" ones. Un early 2020, it didn't really even get that. A12Z was a A12X die with a core turned on rather than off. ). The iPad Pro's have only gotten new updates on process shrink ( so on a 2-3 year cycle).

There is no track record of Apple highly concurrently doing a wide variety of new SoC all at the same time at all. So the notion that "this is almost exactly like 2005" is relatively bankrupt. There is lots of wishful thinking that it might be like 2005, but truth is that Apple is jumping to an unproven vendor at this point for this task of covering the whole Mac line up.

Apple just released a 2020 iMac 27". There is zero pressing need to obsolete that in less than 3-6 months at all. This system just jumped to the T2.

The other issue at play here is that Apple is jumping to 5nm later this year. Doing much larger dies on a brand new process is asking for trouble. There is no good reason to put Macs at substantively higher risks.

2005-2006 also didn't have a global pandemic either. The 24" iMac intel could have quite easily been decoupled from the 26" iMac Intel in the release process. The LG Ultrafine moved to the 24" panel last year (off of the 21.5"). Again the current iMac getting off a panel that "nobody" buys anymore. (not that there every were lots of 21.5" buyers).

Apple gave themselves two years. They can use it and have not broken any "promise". Last time they could lean on Intel to do major work for them. Major rumbling at the time is that the Mac Pro logicboard was largely intel with minorApple tweaks sprinkled on top to justify the "designed in Cupertino" badge also slapped on. There is lots of dotting i's and crossing t's that Intel did for Apple on last transition that Apple will have to do themselves this time. A "Big Bang" transition where it is all done in a handful of months is unlikely. It isn't impossible ( this is usually where folks hand waving at the Scrooge McDuck money pit and saying Apple is just going to throw buckets of cash at it to make it go as fast as possible ... no expense spared. ), but it isn't likely.

Neither is Mac SoC's being on a bleeding edge iPhone Pro like cadence going forward either. If Mac SoCs are going to have 18-24 month iteration cycles then it doesn't make any sense to tightly couple the whole line up together. laptops and desktops playing leapfrog would be a better stream of newer products. ( not like the iPhone market where can just trot out 3 year SoC to hit lower price points. )

You have some valid points but let's not forget that these kind of developments start several years before the actual launch. I don't think Apple started to develop their Mac SoCs just after WWDC 2020. So just because they don't have any current Mac CPUs it doesn't have to mean it will take a long time before we'll see some. We may not see a new iMac in three months but early next year after the introduction of A14 in Oct is very possible. Leakers are even talking about 24" iMac in Oct. Personally I also think maybe June 2021 but my point was that it doesn't have to take that long since developers want to know what hardware they will develop for.

 

[dmccloud]

While I haven't had a chance to run any benchmarks on a DTK, a friend of mine did let me test his unit out last week. What I can say is that even on the A12 SoC, the DTK was just as responsive (if not slightly more so) than the i3 MBA I'm typing on at this time. It's definitely still preview hardware/software though, so I'd expect performance to get even better as more apps are ported from x86 and the finalized hardware is released.

 

[aednichols]

Last time they could lean on Intel to do major work for them. Major rumbling at the time is that the Mac Pro logicboard was largely intel with minorApple tweaks sprinkled on top to justify the "designed in Cupertino" badge also slapped on. There is lots of dotting i's and crossing t's that Intel did for Apple on last transition that Apple will have to do themselves this time. A "Big Bang" transition where it is all done in a handful of months is unlikely.

Apple's market cap is ten times that of Intel. I don't think they will have any trouble hiring engineers to perform work that Intel would have done for them in years past.

 

[Shivetya]

Anything that Apple ships this year that does not have its own separate GPU (theirs or some other) isn't going to mean much from benchmarks. What will be interesting in the bench mark wars is how people dismiss benchmarks Apple cannot run (namely some games)

 

[MalcolmH]

Apple's market cap is ten times that of Intel. I don't think they will have any trouble hiring engineers to perform work that Intel would have done for them in years past.

And they started this process .. 3? Years ago .. not in 2020

 

[dmccloud]

Apple's market cap is ten times that of Intel. I don't think they will have any trouble hiring engineers to perform work that Intel would have done for them in years past.

They already have the engineers on hand - after all, they've been building iPhones and iPads built on Apple SoCs for quite a while now.

 

[Kostask]

And they started this process .. 3? Years ago .. not in 2020

if you are talking about the AS Mac SoCs, I think it probably happened 5+ (from the days of the A6/A7) years ago. Originally, it was just to port over MacOS (OSX at the time) over, just to see what it would look like. When they started to see decent performance, the gears started to turn, combined with the documented failures on Intel's part, and here we are.

 

[the8thark]

https://www.macworld.co.uk/news/mac/apple-silicon-beats-macbook-pro-in-speed-comparison-3794690/

Here's the original text of the article via the wayback machine, before it was all removed

Spoiler: Click this spoiler tag to read it

Our colleagues at Macwelt in Germany have published an interesting insight into how the Mac mini Apple is sending out to developers testing how their apps work on Apple Silicon compares to the 16in MacBook Pro. The encouraging conclusion: "Apple Silicon is already competitive and comparable to current Intel chips. It can be assumed that Apple will continue to drastically improve the performance of its own chips. The first commercial Macs with Apple Silicon should be in no way inferior to the current Intel Macs, at least in terms of CPU performance."

Macwelt published speed comparisons using a native, self-developed benchmark program - and they claim to have seen "sensational results". A few month ago benchmarks for the developer Mac mini showed up on Geekbench, but at the time it was noted that these used benchmarking software that was running via Rosetta 2 emulation, and therefore wouldn’t demonstrate how the machine would perform when running natively.

"That costs a lot of performance and does not show the real possibilities of the Apple chip," notes Macwelt, adding: "So we sat down, installed the current beta version of Xcode 12 and wrote our own benchmark tool called 'Apple Silicon Benchmark' in Swift and SwiftUI. It performs three extremely CPU-intensive functions and measures the time in milliseconds."

Macwelt’s tool generates a million random numbers each in different number formats (integers in 8 to 64 bits, floating point, floating point for graphic output). It performs a million Fast Fourier Transforms (FFT) to transform a data packet (sine wave) from time to frequency domain (often used in audio applications). To do this it uses the “Accelerate” framework from Apple, which accelerates FFT operations.


It then starts the open source ray tracer "SwiftRay" (thanks to Renaud Pradenc) and calculates a virtual scene by ray tracing.

The first two tests only use one CPU core (single core), while SwiftRay tries to use all cores at the same time (multi-core), explains Macwelt. "The overall result should result in a good cross-section of CPU-heavy applications," they claim. Macwelt is not measuring graphics performance.


The site ran the Apple Silicon Benchmark tool three times in total. First using the DTK in native mode, then under the Intel emulation Rosetta. They also ran the benchmarking tool on a 16-inch MacBook Pro with an Intel Core i9 processor (8 cores). In both cases the Macs were running the current developer beta of macOS 11 Big Sur.

The results show that running through Rosetta actually absorbs 50 percent of the power. Macwelt claims that the DTK is "about as fast in existing Intel applications as a 15in MacBook Pro from 2012."

Macwelt adds that when it came to the native results the DTK achieved the same performance as the current 16in MacBook Pro with an eight-core Intel chip.

The Apple’s Developer Transition Kit (DTK) costs $500 - although it’s a loan that you have to return to Apple and uses the Apple A12Z Bionic (as seen in the iPad Pro).

We have this article comparing Intel to Silicon. And here's our guide to Apple's Silicon plan

http://www.macwelt.de/news/Apple-Silicon-schlaegt-Macbook-Pro-16-im-Geschwindigkeitsvergleich-10873144.html

This link doesn't exist but it's the benchmark comparison article that was referenced in the original post here. I found it also via the wayback machine. Though the video contained within was taken down.

Spoiler: Click this spoiler tag for the German original

Apple Silicon schlägt Macbook Pro 16" im Geschwindigkeitsvergleich

Der erste Mac mit Apple-CPU ist eingetroffen, Macwelt macht den Geschwindigkeitsvergleich. Ein natives, selbst entwickeltes Benchmark-Programm liefert sensationelle Ergebnisse.

Den allerersten Mac mit Apple-Silicon-Prozessor gibt es nicht zu kaufen. Nur Entwickler bekommen das sogenannte Apple „Developer Transition Kit“ kurz :“DTK“ und müssen dafür mehr als 500 Euro berappen. Für eine Leihgabe wohlgemerkt, denn man muss das DTK wieder an Apple zurückschicken.

Wir haben uns um ein DTK beworben und Ende August 2020 tatsächlich eines von Applebekommen. Das erste was uns interessiert, ist natürlich die Performance. Ausgerüstet ist der Mac Mini mit einem Apple A12Z Bionic, der bereits im iPad Pro zum Einsatz kommt. Das Problem: Es gibt noch keine nativen Benchmark-Programme, die die echte ungebremste Geschwindigkeit der CPU messen können. Zwar laufen aktuelle Benchmarks wie Geekbench 5 und die Ergebnisse können Sie im Internet auch sehen. Doch sie werden auf dem DTK im Intel-Emulations-Modus unter Rosetta ausgeführt. Das kostet jede Menge Performance und zeigt nicht die wahren Möglichkeiten des Apple-Chips.

Also setzten wir uns hin, installierten die aktuelle Beta-Version von Xcode 12 und schrieben unser eigenes Benchmark-Tool namens „Apple Silicon Benchmark“ in Swift und SwiftUI. Es führt drei extrem CPU-lastige Funktionen aus und misst dabei die Zeit in Millisekunden:

• Es erzeugt eine Million zufällige Zahlen jeweils in verschiedenen Zahlenformaten (Ganzzahlen in 8 bis 64 Bit, Fließkomma, Fließkomma für Grafikausgaben)
• Es führt eine Million Fast Fourier Transformationen (FFT) durch, um ein Datenpaket (Sinuswelle) vom Zeit in den Frequenzbereich zu transformieren (wird oft in Audio-Anwendungen verwendet). Dazu benutzt es das „Accelerate“-Framework von Apple, das FFT-Operationen beschleunigt.
• Es startet den Open-Source-Raytracer „SwiftRay“ (Dank an Renaud Pradenc) nach und berechnet eine virtuelle Szene per Raytracing.

Die ersten beiden Tests verwenden nur einen CPU-Kern (Single-Core), SwiftRay hingegen versucht alle Kerne gleichzeitig einzusetzen (Multi-Core). Das Gesamtresultat dürfte einen guten Querschnitt CPU-lastiger Anwendungen ergeben. Die Grafikperformance wird hierbei nicht gemessen.

Wir lassen Apple Silicon Benchmark insgesamt dreimal laufen. Auf dem DTK im nativen Modus, dann unter der Intel-Emulation Rosetta und auf einem aktuellen 16-Zoll Macbook Pro mit Intel Core i9 Prozessor (8 Kerne). Immer jeweils unter der aktuellen Entwickler-Beta von macOS 11 Big Sur.

Die Ergebnisse zeigen, dass Rosetta ziemlich genau 50 Prozent der Leistung schluckt. Das DTK ist damit in vorhandenen Intel-Anwendungen etwa so schnell wie ein 15-Zoll-Retina-Macbook Pro aus dem Jahre 2012.

Die nativen Ergebnisse sind jedoch fast schon eine Sensation, denn das DTK erreicht hier aus dem Stand die gleiche Performance des aktuellen 16-Zoll-Macbook Pro mit Achtkern-Intel-Chip. Das haben wir so nicht erwartet!

Im folgenden Video haben wir den Vorgang aller drei Tests festgehalten.

Fazit

Apple Silicon ist schon jetzt konkurrenzfähig und vergleichbar mit aktuellen Intel-Chips. Man kann davon ausgehen, dass Apple die Performance seiner eigenen Chips weiterhin drastisch steigern wird. Die ersten kommerziellen Macs mit Apple Silicon dürften, zumindest was die CPU-Leistung angeht, den momentanen Intel-Macs in nichts nachstehen.

Spoiler: Click this spoiler tag for the English google translated version of the article

Apple Silicon beats Macbook Pro 16 "in the speed comparison

The first Mac with Apple CPU has arrived, Macwelt is doing the speed comparison. A native, self-developed benchmark program delivers sensational results.

The very first Mac with an Apple Silicon processor is not available for sale. Only developers get the so-called Apple “Developer Transition Kit” for short: “DTK” and have to shell out more than 500 euros for it. For a loan, mind you, because you have to send the DTK back to Apple.

We applied for a DTK and actually got one from Apple at the end of August 2020. The first thing that interests us is of course the performance. The Mac Mini is equipped with an Apple A12Z Bionic, which is already used in the iPad Pro. The problem: There are still no native benchmark programs that can measure the real unrestrained speed of the CPU. Current benchmarks such as Geekbench 5 are running and you can also see the results on the Internet. But they are run on the DTK in Intel emulation mode under Rosetta. That costs a lot of performance and does not show the real possibilities of the Apple chip.

So we sat down, installed the current beta version of Xcode 12 and wrote our own benchmark tool called "Apple Silicon Benchmark" in Swift and SwiftUI. It performs three extremely CPU-intensive functions and measures the time in milliseconds:

It generates a million random numbers each in different number formats (integers in 8 to 64 bits, floating point, floating point for graphic output)
It performs a million Fast Fourier Transforms (FFT) to transform a data packet (sine wave) from time to frequency domain (often used in audio applications). To do this, it uses the “Accelerate” framework from Apple, which accelerates FFT operations.
It starts the open source ray tracer "SwiftRay" (thanks to Renaud Pradenc) and calculates a virtual scene by ray tracing.
The first two tests only use one CPU core (single core), while SwiftRay tries to use all cores at the same time (multi-core). The overall result should result in a good cross-section of CPU-heavy applications. The graphics performance is not measured here.

We run Apple Silicon Benchmark three times in total. On the DTK in native mode, then under the Intel emulation Rosetta and on a current 16-inch Macbook Pro with an Intel Core i9 processor (8 cores). Always under the current developer beta of macOS 11 Big Sur.

The results show that Rosetta pretty much swallows 50 percent of the power. The DTK is therefore about as fast in existing Intel applications as a 15-inch Retina Macbook Pro from 2012.

The native results are almost a sensation, however, because the DTK achieves the same performance as the current 16-inch Macbook Pro with an eight-core Intel chip right from the start. We didn't expect that!

In the following video we recorded the process of all three tests.


Conclusion

Apple Silicon is already competitive and comparable to current Intel chips. It can be assumed that Apple will continue to drastically improve the performance of its own chips. The first commercial Macs with Apple Silicon should be in no way inferior to the current Intel Macs, at least in terms of CPU performance.

 

[cmaier]

The original article is available on the Web Archive: https://web.archive.org/web/2020082...eats-macbook-pro-in-speed-comparison-3794690/

Here is the full article in German:https://web.archive.org/web/2020082...16-im-Geschwindigkeitsvergleich-10873144.html

My favorite german word is either Geschwindigkeitsvergleich or Fremdenverkehrsburo. (Sorry about the missing umlauts. Keyboard is in english mode at the moment).

 

[MisterMe]

My favorite german word is either Geschwindigkeitsvergleich or Fremdenverkehrsburo. (Sorry about the missing umlauts. Keyboard is in english mode at the moment).

Macs have always had the ability to type diacriticals from the English keyboard. In the case of the umlaut, type [ opt ]+[ u ] followed by the letter over which you want the umlaut to appear. For ü, type [ opt ]+[ u ] [ u ] For ë, type [ opt ]+[ u ] [ e ]. Similar keystrokes are available for most other popular diacriticals. The Keyboard Viewer menu app is a handy guide and has been for decades. Enable it in Keyboard Preferences. Within Keyboard Viewer, press the modifier keys and combinations of modifier keys to see the available glyphs.

 

[Diskutant]

In the case of the umlaut, type [ opt ]+[ u ] followed by the letter over which you want the umlaut to appear. For ü, type [ opt ]+[ u ] [ u ]

Or just press and hold the u and a iOS-like menu appears where you can choose umlauts and other variants.

 

[MisterMe]

Or just press and hold the u and a iOS-like menu appears where you can choose umlauts and other variants.

Amazing. I have used Macs for decades. I was not aware of this feature. I am not sure that it is faster, but it is certain.

Thank you.

 

[MacPoulet]

Or just press and hold the u and a iOS-like menu appears where you can choose umlauts and other variants.

I had no idea you could do that! Fantastic!

 

[Homy]

I think this was interesting, Apple trying to make the best Mac, not the best CPU/GPU. Hope that 3A gaming is a part of that "best Mac".

 

[Homy]

Here is my unscientific speculation based on the presentation of A14:

A12 6.9 billion transistors, 6 CPU cores, 4 GPU cores, Geekbench 5 single core 1112, multi-core 2866, Metal 4641

A12X 10 billion transistors, 8 CPU cores, 7 GPU cores

A12Z 10 billion transistors, 8 CPU cores, 8 GPU cores, Geekbench 5 single core 1118, multi-core 4631, Metal 10337

A13 8.5 billion transistors, 6 CPU cores, 4 GPU cores

A14 11.8 billion transistors, 6 CPU cores, 4 GPU cores, 40% faster CPU than A12, 30% faster GPU than A12, Geekbench 5 single core 1557, multi-core 4012, Metal 6033

A14Z 17.1 billion transistors, 8 CPU cores, 8 GPU cores, 40% faster CPU than A12Z, 30% faster GPU than A12Z, Geekbench 5 single core 1565, multi-core 6483, Metal 13438

A14Z Mac SoC 25.8 billion transistors, 12 CPU cores, 32 GPU cores, 50% faster multi-core than A14Z, 300% faster GPU than A14Z, Geekbench 5 single core 1565, multi-core 9725, Metal 53752

That would put A14Z Mac on par with 10-core i9 in iMac 2020 and the GPU between RX 580 and RX Vega 56.

 

[iPadified]

Here is my unscientific speculation based on the presentation of A14:

A12 6.9 billion transistors, 6 CPU cores, 4 GPU cores, Geekbench 5 single core 1112, multi-core 2866, Metal 4641

A12X 6.9 billion transistors, 8 CPU cores, 7 GPU cores

A12Z 6.9 billion transistors, 8 CPU cores, 8 GPU cores, Geekbench 5 single core 1118, multi-core 4631, Metal 10337

A13 8.5 billion transistors, 6 CPU cores, 4 GPU cores

A14 11.8 billion transistors, 6 CPU cores, 4 GPU cores, 40% faster CPU than A12, 30% faster GPU than A12, Geekbench 5 single core 1557, multi-core 4012, Metal 6033

A14Z 11.8 billion transistors, 8 CPU cores, 8 GPU cores, 40% faster CPU than A12Z, 30% faster GPU than A12Z, Geekbench 5 single core 1565, multi-core 6483, Metal 13438

A14Z Mac SoC 11.8 billion transistors, 12 CPU cores, 32 GPU cores, 150% faster multi-core than A14Z, 400% faster GPU than A14Z, Geekbench 5 single core 1565, multi-core 9725, Metal 53752

That would put A14Z Mac on par with 10-core i9 in iMac 2020 and the GPU between RX 580 and RX Vega 56.

Prediction in performance looks reasonable but the transistor count is the same for A14 and A14Z. That is impossible. That predicted A14Z will be 2-3 as large as the A14.

 

[dmccloud]

Prediction in performance looks reasonable but the transistor count is the same for A14 and A14Z. That is impossible. That predicted A14Z will be 2-3 as large as the A14.

Not necessarily. We know that the A12 in the DTK has additional cores enabled, but it is the same size as the A12 in the iPad Pro. A common practice among all CPU manufacturers is to "bin" individual CPUs based on how many working cores there are, the max speed it can run at, etc. Intel can disable cores to make an 8-core with HT a 6-core with HT, or just disable HT entirely, depending on how the individual processor handles the testing process. This is how Intel has been creating their various models of the core series for years. It wouldn't be difficult for Apple to do the same with these new SoCs.

 

[iPadified]

Not necessarily. We know that the A12 in the DTK has additional cores enabled, but it is the same size as the A12 in the iPad Pro. A common practice among all CPU manufacturers is to "bin" individual CPUs based on how many working cores there are, the max speed it can run at, etc. Intel can disable cores to make an 8-core with HT a 6-core with HT, or just disable HT entirely, depending on how the individual processor handles the testing process. This is how Intel has been creating their various models of the core series for years. It wouldn't be difficult for Apple to do the same with these new SoCs.

Possible but in my opinion unlikely. In that case an iPhone chip has the same chip layout as a Mac chip fit for an iMac. Binning to some extent is fine (6 to 8 cores) but going from 2P cores to 8P or from 4 GPU to 32 GPU cores using binning seem like a waste of chip real estate.

 

[dmccloud]

Possible but in my opinion unlikely. In that case an iPhone chip has the same chip layout as a Mac chip fit for an iMac. Binning to some extent is fine (6 to 8 cores) but going from 2P cores to 8P or from 4 GPU to 32 GPU cores using binning seem like a waste of chip real estate.

Apple could also do binning in each processor subgroup. Assuming that the new Macs use a processor in the A14 family, it is possible that the A14, A14X, and A14Z could all use different die sizes. There may even be an A14M for the Macs, we just don't know what Apple has planned right now. For example, the distinction between an A14 and A14 Bionic could be the number of active cores. On the Mac side, the number of working cores could determine whether that specific chip goes in a Mac Pro, MBP, or MacBook.

 

[Homy]

Prediction in performance looks reasonable but the transistor count is the same for A14 and A14Z. That is impossible. That predicted A14Z will be 2-3 as large as the A14.

I have corrected the numbers. Strange! The Wiki page I used as source said 6.9 billion in A12Z but it has 10 billion. In that case A14Z should have 17.1 billion transistors and Mac SoC many more. Also the Metal score for A13 is already 6513, higher than my calculated A14, but I used Apple's statement about A14 having 30% better GPU performance than A12. If A14 GPU performs more than 30% than A12 in Metal it's great news.

 

[thunng8]

I have corrected the numbers. Strange! The Wiki page I used as source said 6.9 billion in A12Z but it has 10 billion. In that case A14Z should have 17.1 billion transistors and Mac SoC many more. Also the Metal score for A13 is already 6513, higher than my calculated A14, but I used Apple's statement about A14 having 30% better GPU performance than A12. If A14 GPU performs more than 30% than A12 in Metal it's great news.

Apple wildly under quoted the GPU performance gains for the a12 -> a13. The number they specified at the keynote was 20%, but in reality for compute is was closer to 40% and in games and gaming benchmarks it was way more than 20% (varies from 20%-60%)

The Apple iPhone 11, 11 Pro & 11 Pro Max Review: Performance, Battery, & Camera Elevated

www.anandtech.com

It is definite that they are underquoting the a12->a14 performance gains as well.

 

[thenewperson]

It is definite that they are underquoting the a12->a14 performance gains as well.

Fingers crossed ?