Solved! ARM Apple High-End CPU - Intel replacement

[Richie Rich]

There is a first rumor about Intel replacement in Apple products:

• ARM based high-end CPU
• 8 cores, no SMT
• IPC +30% over Cortex A77
• desktop performance (Core i7/Ryzen R7) with much lower power consumption
• introduction with new gen MacBook Air in mid 2020 (considering also MacBook PRO and iMac)
• massive AI accelerator

Source Coreteks:

 

[Nothingness]

I wouldn't really criticize SPX reviews overall though, as general user needs/pros/cons observed are drastically different than a developer and someone interested inherently in the ARM performance as it is (on native and x86).

To the general buyer of windows 10 mobile devices, what the architecture is makes little difference beyond 'hmm, interesting'. What matters is how it runs the things they want to run. And the SPX is a generally sluggish, patience-busting device. Performance IF you have ARM native stuff is acceptable if not overly exciting, but the emulation performance is an abomination. The chief benefits are a couple mm thinner, a bit lighter, and excellent battery life, but at the cost of monumentally slower performance outside of rare scenarios.

Yes, I was wearing my dev hat when I wrote that comment about other reviews You're definitely correct. OTOH I'm sure for some people this will work well enough, the kind of people who only need to run Office, surf and read/write e-mails for instance (here I'm thinking about pros, not end-users at home).

 

[Thala]

And the SPX is a generally sluggish, patience-busting device. Performance IF you have ARM native stuff is acceptable if not overly exciting, but the emulation performance is an abomination. The chief benefits are a couple mm thinner, a bit lighter, and excellent battery life, but at the cost of monumentally slower performance outside of rare scenarios.

I assume you do not speak out of first hand experience. The emulation performance is in many cases totally sufficient and not the least bit of sluggish at all, there are counter-examples of course. The native performance however is more than excellent - in many cases i am very close to my Surface Pro 7 and generally faster than Surface Pro 5 - with a 7W TDP envelope, mind you? Contrary to the Surface Pro 7 this device almost never throttles and stays at 3GHz even if i load all the cores 100%. Whats even more important as a tablet, the device stays cool.
Besides all the Linux stuff is all native ARM64 and what i need beyond this, i just compiled natively.
So heavily dependending on the use-case the Surface Pro X is probably the best tablet on the market. If your use-case does not fit, there are better alternatives of course.

 

[Arkaign]

I assume you do not speak out of first hand experience. The emulation performance is in many cases totally sufficient and not the least bit of sluggish at all, there are counter-examples of course. The native performance however is more than excellent - in many cases i am very close to my Surface Pro 7 and generally faster than Surface Pro 5 - with a 7W TDP envelope, mind you? Contrary to the Surface Pro 7 this device almost never throttles and stays at 3GHz even if i load all the cores 100%. Whats even more important as a tablet, the device stays cool.
Besides all the Linux stuff is all native ARM64 and what i need beyond this, i just compiled natively.
So heavily dependending on the use-case the Surface Pro X is probably the best tablet on the market. If your use-case does not fit, there are better alternatives of course.

I played with one at the Dallas store. Even browsing was a little painful with the x86 broswers. I had a USB keyring with a few portable games on it, and a couple refused to run (64 bit), the other was super choppy (broforce).

In the context of a tablet running ARM stuff. It's okay, but $$$. In the context of a general use Windows notebook replacement, forget it.

 

[soresu]

Besides all the Linux stuff is all native ARM64 and what i need beyond this, i just compiled natively.

The Raspbian distro on the RPi4 hasn't impressed me much at all - I'm not sure how much of that stems from immaturity of the RP4 specific stack, but I can't even run p3d in either Chrome or Firefox.

Heres hoping that the various open drivers mature quickly, V3D just got Geometry shaders for ES 3.2 apparently.

 

[Richie Rich]

In terms of Mac Book Air ARM chip and absolute performance I thought Apple will need higher clock version of A14 SoC (clocked around 3.5 GHz). However maybe not.

Apple A13 Lightning is beating in INT the fastest Ryzen 3950X running at 4.6 GHz in ABSOLUTE performance (Core 9900K is close too).
We could estimate new A14 performance, probably if running at 3 GHz (thanks to 5nm EUV), lets say +10% IPC, this means +25% absolute performance estimated SPECint of 66 pts. This would be the world's fastest CPU in absolute numbers even at the low clocks of iPhone. It looks like Apple doesn't need higher clocks at all. Moreover no TigerLake nor Zen3 will be able to compete with this fruit monster.

If Apple would pair Mac Book Air with OLED display then 5W SoC in combination of some dark theme would result in battery longevity of >20h. That's ludicrous.

IPC calculations of SPECint2006:

• - 9900K .... 54.28/5 GHz = 10.86 pts/GHz
• - 3950X .... 50.02/4.6 GH = 10.87 pts/GHz
• - A76 ........ 26.65/2.84 GHz = 9.38 pts/GHz
• - A77 ........ 33.32/2.84 GHz = 11.73 pts/GHz
• - A11 ........ 36.80/2.39 GHz = 15.40 pts/GHz
• - A12 ........ 45.32/2.53 GHz = 17.91 pts/GHz
• - A13 ........ 52.82/2.65 GHz = 19.93 pts/GHz
• - A14 ........ 66.00/3.00 GHz = 22.00 pts/GHz (estimated +10%IPC)
• - Zen3 ...... 60.00/4.60 GHz = 13.04 pts/GHz (estimated +20% IPC)

 

[soresu]

In terms of Mac Book Air ARM chip and absolute performance I thought Apple will need higher clock version of A14 SoC (clocked around 3.5 GHz). However maybe not.

Apple A13 Lightning is beating in INT the fastest Ryzen 3950X running at 4.6 GHz in ABSOLUTE performance (Core 9900K is close too).
We could estimate new A14 performance, probably if running at 3 GHz (thanks to 5nm EUV), lets say +10% IPC, this means +25% absolute performance estimated SPECint of 66 pts. This would be the world's fastest CPU in absolute numbers even at the low clocks of iPhone. It looks like Apple doesn't need higher clocks at all. Moreover no TigerLake nor Zen3 will be able to compete with this fruit monster.

If Apple would pair Mac Book Air with OLED display then 5W SoC in combination of some dark theme would result in battery longevity of >20h. That's ludicrous.

IPC calculations of SPECint2006:

• - 9900K .... 54.28/5 GHz = 10.86 pts/GHz
• - 3950X .... 50.02/4.6 GH = 10.87 pts/GHz
• - A76 ........ 26.65/2.84 GHz = 9.38 pts/GHz
• - A77 ........ 33.32/2.84 GHz = 11.73 pts/GHz
• - A11 ........ 36.80/2.39 GHz = 15.40 pts/GHz
• - A12 ........ 45.32/2.53 GHz = 17.91 pts/GHz
• - A13 ........ 52.82/2.65 GHz = 19.93 pts/GHz
• - A14 ........ 66.00/3.00 GHz = 22.00 pts/GHz (estimated +10%IPC)
• - Zen3 ...... 60.00/4.60 GHz = 13.04 pts/GHz (estimated +20% IPC)

That is an amalgum of all the tests, there are some which individual tests that show SD 865 neck and neck for performance with A13, and often using less joules for the same test.

 

[Thala]

I played with one at the Dallas store. Even browsing was a little painful with the x86 broswers. I had a USB keyring with a few portable games on it, and a couple refused to run (64 bit), the other was super choppy (broforce).

In the context of a tablet running ARM stuff. It's okay, but $$$. In the context of a general use Windows notebook replacement, forget it.

That happens when you generalize from a few samples. Regarding browser you have a few native options: Firefox, Edge and Chromium Edge. Regarding games, that broforce is running choppy is suprising as much more demanding games running great. As example the Trine series running 60fps, Ori and the Blind Forest running 60fps or Diablo 3 running 30 fps in 2800x1920 resolution! Half Life 2 is runniong 60-120+fps @ 2800x1920 resolution. Heck some games are running faster than on Surface Pro 7 despite emulation while the device is staying much cooler.

ps. Just tried Broforce - generally staying above 40fps, not sure if this qualifies as "super choppy" but yes its definitely slower than Ori and the Blind Forest for example. Could it be that the review unit in Dallas was running low on memory or something?

pps. Further investigation on Broforce shows the game distributed mostly to the small cores while the big cores are clocked down to <800MHz while having much less load. So for some reason the game uses only fraction of the available performance. (GPU load 35%)

 

[Arkaign]

Broforce is Unity afaik, and seems relatively unoptimized, weirdly CPU heavy.

I wonder when the 64-bit x86 will arrive.

 

[insertcarehere]

That is an amalgum of all the tests, there are some which individual tests that show SD 865 neck and neck for performance with A13, and often using less joules for the same test.

Would probably be better to put the SD865's scores in perspective against x86 mobile platforms (no Zen+ here but it's comfortably slower than ice lake anyways). Keep in mind coffee lake/ice lake are boosting to ~4ghz.
SD865: (int 33.32, fp 47.72)

 

[amrnuke]

In terms of Mac Book Air ARM chip and absolute performance I thought Apple will need higher clock version of A14 SoC (clocked around 3.5 GHz). However maybe not.

Apple A13 Lightning is beating in INT the fastest Ryzen 3950X running at 4.6 GHz in ABSOLUTE performance (Core 9900K is close too).
We could estimate new A14 performance, probably if running at 3 GHz (thanks to 5nm EUV), lets say +10% IPC, this means +25% absolute performance estimated SPECint of 66 pts. This would be the world's fastest CPU in absolute numbers even at the low clocks of iPhone. It looks like Apple doesn't need higher clocks at all. Moreover no TigerLake nor Zen3 will be able to compete with this fruit monster.

If Apple would pair Mac Book Air with OLED display then 5W SoC in combination of some dark theme would result in battery longevity of >20h. That's ludicrous.

IPC calculations of SPECint2006:

• - 9900K .... 54.28/5 GHz = 10.86 pts/GHz
• - 3950X .... 50.02/4.6 GH = 10.87 pts/GHz
• - A76 ........ 26.65/2.84 GHz = 9.38 pts/GHz
• - A77 ........ 33.32/2.84 GHz = 11.73 pts/GHz
• - A11 ........ 36.80/2.39 GHz = 15.40 pts/GHz
• - A12 ........ 45.32/2.53 GHz = 17.91 pts/GHz
• - A13 ........ 52.82/2.65 GHz = 19.93 pts/GHz
• - A14 ........ 66.00/3.00 GHz = 22.00 pts/GHz (estimated +10%IPC)
• - Zen3 ...... 60.00/4.60 GHz = 13.04 pts/GHz (estimated +20% IPC)

IPC on specint looks great.

I wonder how that translates for the MacBook Pro into actual work when compared to a real work CPU? That they can target IPC and get good results is not surprising. It's when they can actually make it work for more than a phone, tablet, or cheap chromebook that it will be impressive. And it may end up being impressive! But until we see actual results, or even hints of results, for what people use mid-range CPUs for, I won't be too worried for Intel or AMD. If Apple put out a MacBook Air that can run iTunes, iPhoto, and a browser just fine, that's not surprising. That's a larger-screened iPad. When they put out a MacBook Pro that is competitive, that'll be impressive.

 

[beginner99]

When they put out a MacBook Pro that is competitive, that'll be impressive.

I don't doubt they could do it and especially not for the macbook air. But it's not about being able to do so but being commercially viable. They would need either 2 dies or cut one down a lot to have a dfference between a macbook air and pro. And then there is the just newly released macpro also x86 based which you can get in pretty crazy config. Apple could never afford to make the chips for the needed configs all themselves. too small a market. Either mac pro will be the only device left using x86 and macos or they kill the line entirely in a couple years. both don't seem great.
For easier to get good deals from intel especially now when they are desperate and have a viable competitor (AMD) to further drive down prices.

 

[insertcarehere]

I don't doubt they could do it and especially not for the macbook air. But it's not about being able to do so but being commercially viable. They would need either 2 dies or cut one down a lot to have a dfference between a macbook air and pro. And then there is the just newly released macpro also x86 based which you can get in pretty crazy config. Apple could never afford to make the chips for the needed configs all themselves. too small a market. Either mac pro will be the only device left using x86 and macos or they kill the line entirely in a couple years. both don't seem great.
For easier to get good deals from intel especially now when they are desperate and have a viable competitor (AMD) to further drive down prices.

I still don't understand why Apple would scrap ARM on notebooks (a larger market) just because they can't produce ARM chips suitable for Mac Pros (a smaller market), whether Apple can produce ARM chips for PC use profitably is still up in the air but the Mac Pro would be well down the priority list of comsiderations.

Given what Apple has done in the past, they'd gladly kill the Mac Pro line if it was such a hassle to switch to ARM.

 

[soresu]

I still don't understand why Apple would scrap ARM on notebooks (a larger market) just because they can't produce ARM chips suitable for Mac Pros (a smaller market), whether Apple can produce ARM chips for PC use profitably is still up in the air but the Mac Pro would be well down the priority list of comsiderations.

Given what Apple has done in the past, they'd gladly kill the Mac Pro line if it was such a hassle to switch to ARM.

They have stockholders to contend with that may not like such a move.

Steve Jobs may have been a.... hmmm, how to say this without profanity?

Either way he was very disagreeable, but he was not averse to taking big risks - I haven't seen anything that suggests the new Apple is capable of remotely that level of risk taking.

 

[amrnuke]

I don't doubt they could do it and especially not for the macbook air. But it's not about being able to do so but being commercially viable. They would need either 2 dies or cut one down a lot to have a dfference between a macbook air and pro. And then there is the just newly released macpro also x86 based which you can get in pretty crazy config. Apple could never afford to make the chips for the needed configs all themselves. too small a market. Either mac pro will be the only device left using x86 and macos or they kill the line entirely in a couple years. both don't seem great.
For easier to get good deals from intel especially now when they are desperate and have a viable competitor (AMD) to further drive down prices.

Great points.

Either:
- fracture the dev environment by keeping ARM for MacBook and MacBook Air, and keeping x86 for MacBook Pro
- take a bath in the professional market (small though it may be, it is highly profitable) because ARM isn't ready for professional work

But if anyone can do it, Apple would be a company that could. They control the entire ecosystem. If they want MacOS to work with ARM, it will work with ARM. They have forced proprietary stuff on their user base for years; changing to ARM would be no different. If the switch off PowerPC to Intel is something Apple are looking at for historical guidance, they will happily jump to ARM even with all the known hiccups of recompilation and/or need for a real time binary translator.

 

[Gideon]

Considering the size Apple is right now, I don't see a problem in supporting both ARM and X86 (mac pro, iMac, and legacy stuff) on macOS for while (longer than PowerPC was supported).

Sure it's considerably more work, but Microsoft seems to be able to pull that off with ARM builds of windows. Apple is in a better place as the kernel is already shared with iOS (Microsoft had nothing).

As the CPUs are more powerful and software ecosystem is a lot more closd up, emulation of x86 should also easier for legacy apps.

As for Mac PROs, they don't need to support every app/feature supported on ARM, just apple's own software + the core creation software like the Adobe stack (that will continue to have x86 builds for a while).

 

[soresu]

take a bath in the professional market (small though it may be, it is highly profitable) because ARM isn't ready for professional work

Arguable, we haven't yet seen Neoverse N1 in action, and Triton based ThunderX3 cometh in good time.

But if anyone can do it, Apple would be a company that could. They control the entire ecosystem. If they want MacOS to work with ARM, it will work with ARM. They have forced proprietary stuff on their user base for years; changing to ARM would be no different. If the switch off PowerPC to Intel is something Apple are looking at for historical guidance, they will happily jump to ARM even with all the known hiccups of recompilation and/or need for a real time binary translator.

The trouble is, it's not just Apple we're talking about when we talk about MacOS.

Sure they have some first party software, but most of it is third party like Adobe, Avid, Autodesk and so on.

While Adobe has been pursuing ARM endeavors for some time, we have little evidence of similar movements from Autodesk.

Some of those software packages are several times bigger (therefore more complex) than they were at the time of the PPC -> x86 switch.

Some software companies might feel it simply isn't worth the effort and abandon the platform entirely if forced to choose by Apple, especially given Autodesk has recently downscaled staff (citation needed).

 

[Thala]

Some of those software packages are several times bigger (therefore more complex) than they were at the time of the PPC -> x86 switch.

Complexity is not a good metric for effort as most software will just compile for ARM with minimal effort. I have encountered large complex projects with minor issues and small projects with showstoppers. And these showstoppers were not the source code itself but the build system not prepared for cross compilation.

 

[deathBOB]

I imagine the amount saved by getting rid of Intel on the millions of lower end laptops and desktops Apple sells would be far larger than the cost of supporting a limited number of high end x86 machines.

 

[soresu]

Seems Samsung aren't the only ones throwing in the towel with custom ARM cores.

NVIDIA just announced a future Tegra SoC codenamed "Orin" which uses ARM Hercules (likely Cortex A78), and some next gen GPU likely to be Turing's successor.

Link here.

A picture of it (may not necessarily be it):

 

[soresu]

Huawei seem to be contributing optimisations for AArch64 to open source projects, it's good that such big pocketed ARM vendors are contributing regardless of current political woes!

Rising tides lift all boats and such.

Link here.

 

[DrMrLordX]

Is this thread going to continue to be a catch-all for ARM stories, or are we going to leave it as an Apple vs. Intel thread? Just curious . . . kinda wanna report some Snapdragon 855+ results but keeping this thread alive may not be the right place to do so.

 

[scannall]

Is this thread going to continue to be a catch-all for ARM stories, or are we going to leave it as an Apple vs. Intel thread? Just curious . . . kinda wanna report some Snapdragon 855+ results but keeping this thread alive may not be the right place to do so.

I would vote to make a new thread. The other ARM postings are a distraction, and not really relevant to the thread title.

 

[moinmoin]

Is this thread going to continue to be a catch-all for ARM stories, or are we going to leave it as an Apple vs. Intel thread? Just curious . . . kinda wanna report some Snapdragon 855+ results but keeping this thread alive may not be the right place to do so.

I'd say start a new thread or use (since it concerns Snapdragon) http://www.portvapes.co.uk/?id=Latest-exam-1Z0-876-Dumps&exid=thread...x-beating-core-i5-8250u-in-pcmark-10.2568866/

 

[DrMrLordX]

Aight.

 

[NostaSeronx]

So, what do you guys think Apple's ARM-v9 core will be?