Corporate users don't typically upgrade every year. A lot of them are still running Windows 7 on PCs that are three to five years old assuming their PC hasn't been replaced by a virtual machine running in a data center somewhere.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
New cores (A15)
- Thread starter cmaier
- Start date
- Sort by reaction score
Corporate users don't typically upgrade every year. A lot of them are still running Windows 7 on PCs that are three to five years old assuming their PC hasn't been replaced by a virtual machine running in a data center somewhere.
The 21" iMac is mainly there AFAIK for education/business customers who need iMac, but are not able to transition to M1 yet for whatever reason. My guess is it will get discontinued within the next year, likely whenever they update the current 24" model (the current 24" model could then get a price reduction and replace the current 21" price). Here is the current line-up:
Notebooks:
MBA - M1
MBP (2-port) - M1
MBP 13" (4-port) - Intel**
MBP 16" - Intel**
Desktops:
Mac Mini (2-port) - M1
Mac Mini (4-port) - Intel**
21" iMac - Intel
24" iMac - M1
27" iMac - Intel
Mac Pro - Intel
** Rumored for next AS update this year
I very much doubt that Apple wold use DDR5 in their mobile phones, that would make very little sense. Personally, I think A14 still uses the same LPDDR4X as it gives more than enough performance with good availability and excellent power efficiency.
A15 GPU talk is up: https://developer.apple.com/videos/play/tech-talks/10876
Short summary:
- doubled ALUs (2x FP32 per cycle, bringing it in-line with M1)
- on-the fly lossy compression for render targets, allowing to save memory with virtually no loss in quality
- sparse depth and stencil textures that allow to implement some advanced rendering techniques more efficiently and with massive memory savings
- new SIMD shift instructions for certain compute workloads, allows much more efficient implementation of certain types of shaders (this brings better feature parity with CUDA and friends)
Basically, straightforward performance updates and some more esoteric features. The new compute shader stuff is going to be big for image processing (I can easily imagine 20-30% improvements for some effects).
No mention of hardware ray tracing or anything more groundbreaking though.
Short summary:
- doubled ALUs (2x FP32 per cycle, bringing it in-line with M1)
- on-the fly lossy compression for render targets, allowing to save memory with virtually no loss in quality
- sparse depth and stencil textures that allow to implement some advanced rendering techniques more efficiently and with massive memory savings
- new SIMD shift instructions for certain compute workloads, allows much more efficient implementation of certain types of shaders (this brings better feature parity with CUDA and friends)
Basically, straightforward performance updates and some more esoteric features. The new compute shader stuff is going to be big for image processing (I can easily imagine 20-30% improvements for some effects).
No mention of hardware ray tracing or anything more groundbreaking though.
Last edited:
what are your hopes for the 14" and 16" ? In details if you have the time please
Do you also think the 14" will have the same SoC as the 16" minus the extra Ram and extra gpu cores?
My hopes are still the same as half a year ago (https://forums.macrumors.com/thread...umer-silicon-m1x-m2-etc.2298203/post-29944590) - a new prosumer-oriented CPU/GPU microarchitecture that is neither A14 nor A15, but a parallel development with shared roots. And I hope that these chips will support SVE and hardware raytracing.
But at this point I find it hard to even speculate whether these hopes are realistic. A15 doesn’t seem to bring anything fundamentally new to the table so far, and it’s really unclear what this means for the prosumer Macs.
The only things I am confident about:
- the future M2 chip (low power silicon) will be an A15 „scaled up“ - this is consistent with leaks, which makes them trustworthy in my book
- the leaks surrounding the Jade Die chip and the upcoming Mac Pro are consistent, mentioning building blocks of 8+2+16/32 clusters. This doesn’t mean that the rumors are true, but I think there is a decent chance that’s it’s what we will see. What does it mean for the capabilities is really hard to say, it could be anything at this point
- the prosumer Macs will share the same SoC, anything else does not make sense economically. The 16“ will have a higher throughput and a faster GPU option
I'm not talking as much about buying as expectation setting. More than a few folks shift their budgets and other hardware buys to the cadence of the iPhone. That isn't always 12 months ( some folks keep the 2 year cadence of old phone contracts. Certainly the financing plans are geared that way. )
That said when have a user base of 100M users if the upgrades come every 5 years that is 20M/year run rate. For the folks, coming off of that 5 year wait having something between +/- 3-6 months of the previous release date.
Apple goes through expensive parallel development efforts to make the phone "pop" out of the pipeline at 12 month intervals when underlying components and tech move at non 12 month rates. Parallel, concurrent design teams , expensive "pathfinding" , etc. That is far, far easier to amortize over a larger revenue base . ( which is how the more broad market Intel/Samsung/Qualcomm vendors do it. ). As go higher and higher in the Mac product space the volume shrinks more and more.
For the Mac, Apple just needs something approaching regularity , not something that is exactly 12 months. That gets them some user predictability without explicitly talking about future roadmaps.
Seems to be gibberish. The A15 cores are clearly not the same as M1. Avalanche is not Firestorm. Nobody cares about ARMv9, and adding ARMv9 support will not require some massive rework of the cores.
Agree with @cmaier. Total gibberish. M2 will be to A15 what M1 is to A14. Vadim (Max Tech) is just making random statements, as usual.
I can’t get my head around what he is suggesting, so yes gibberish.
Correct me if I’m wrong…
A14 was based on 5nm.
M1 was based on 5nm.
A15 is based on 5nm+.
We expect M1X to be based on 5nm.
We expect M2/M2X to be based on 5nm+.
But these are still just marketing names, Apple could build a SOC based on 5nm and still call it M2 if they wanted (that’s my personal expectation).
Our company lets anyone choose a Mac or Windows laptop but you can only upgrade machines every 5 years. I agree that most companies don't turn over their machines very quickly.
I think you are overestimating the percentage of people who upgrade phones annually. That was a thing early on for more people, but beyond the enthusiasts in tech forums, most consumers keep their phones for 2, 3 or 4 years between upgrades. Laptops are on a somewhat longer cycle, even for corporate buyers.
Yeah before my iPhone 12 Mini I was using the other halfs iPhone6 as I’d bought her an iPhone 11 Pro (my iPhone6+ had died). I buy Apple kit most years, sometimes it’s Macs sometimes it’s iPhone, iPad, Watch or AirPods. But Apple kit lasts a long time so only upgrade when I need or really, really want to. My best mate is still using an iPhoneSE (1st gen) as is my son. I will be buying another ASi Mac when the new ones are released though… 
Why wouldn’t it be? A15 is 5nm+ and the prosumer chip is going to be produced in a much lower volume. Plus, they debut at the same time.
What results would stop the world in GeekBench and others for the MacBook Pro 14 and 16? I would be supper impressed with 2100 or higher in single core. 9500-10000 in Multi-Core. And then GPU Metal 35,000 or higher. I think that would shut Intel up for a bit.
It’s unlikely IMO we will see above ~1900 in single core. Apples A-series chips gain roughly 200 each year and current chips are around ~1700. That is unless the high-end Mac chips are a totally different breed than A15.
Id bet we would see above 35,000 for Geekbench metal scores. If you just double the 8 M1 GPU cores, you’d get around 40,000. A 32-core M1-based GPU could clock in around 70,000-80,000.
Good points. It would be sweet to see above 2k for single core. That would make it a killer for day to day tasks and then around 40k for Metal would be close to a 1650 Max Q card.