Wouldn't Thunderbolt benefit from more/faster PCIe lanes? Seems like they would be able to offer more full speed ports that don't have to share bandwidth.
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.
User predictions for Apple Prosumer Silicon (M1X, M2 etc.)
- Thread starter leman
- Start date
- Sort by reaction score
I think the way Apple designs their chips is first decide how much I/O they want and then just implement it. Its unlikely that they will integrate more PCI lanes just for the sake of it.
That is fair.
EDIT: My prediction is that Apple is going to add HW RT and will use the open source FSR (which is more of a Metal thing than hardware but I digress).
Thunderbolt 4.0 is 40gbps/second. A PCI 4.0 x4 is more than enough to satisfy it.
Interestingly, news came out today that Samsung's Exynos mobile SoC will feature AMD's RDNA2 with ray tracing. This is ray tracing for a phone SoC.
So competition is officially on for RT on mobile devices. I think Apple is set to announce RT hardware soon - maybe as soon as WWDC.
I menan they will kind of have to as the only GPU maker that doesn't support it in hardware. 11.4's support of 6000 series cards felt pointless since Metal doesn't have any hooks to use the RT hardware present. I hope that changes next week.
Which is quite funny considering that AMDs raytracing performance is lackluster even on the biggest Navi 2 GPUs. What kind of RT performance does one expect from downsizing them to a phone chip?
Gurman "has now clarified that Apple's next-generation chip will be an iteration on the M1 chip, featuring a 10-core CPU with eight high-performance cores and two energy-efficient cores, with 16-core or 32-core GPU options."
And it uses transistors. If you're going to dedicate a significant fraction of your GPU for that, it'd better be worth it.
I hope so. Do you know what the 5nm+ node offers? I read the trade-off was somewhere around 17% power reduction or 5% performance
Those sorts of numbers are meaningless, because it depends entirely on the design. The fact that a single transistor can toggle 5% faster or with 10% less leakage current is not all that meaningful outside the context of the design.
N5P appears to be the same design rules as N5, but with some unknown optimizations (often times these are things like changing layer thicknesses slightly, introducing or removing cladding or etch stop layers, etc.). I would be surprised if a given design, unchanged, increased performance by more than a few percent. Since there appear to be FEOL changes, I’m a little more optimistic about power, because often times you can optimize the transistor layers to improve on gate capacitance and leakage currents, but I still would be surprised by a change as high as 10%.
From my understanding the AMD solution uses less transitors (not dedicated from what I can tell) than Nvidia. Not sure if Apple would go the same route.
Keep in mind that 40Gbps is shared between DisplayPort and PCIe data. PCIe over Thunderbolt is limited to PCIe 3.0 x4, at around 32Gbps. That 4.0 x4 link you mention could satisfy *two* Thunderbolt buses.
Without a new spec that does more than what TB4 did, that’s where things are stuck for now.
But bringing it back into topic, yes, there’s not really a whole lot of reason for Apple to use PCIe 5. The fact that they put PCIe 4 on the feature list during the M1 announcement at all was strange. The only thing I can think of is that PCIe 4 lets them feed more TB buses with fewer PCIe lanes, potentially making the PCIe connection to the Thunderbolt controllers easier to layout on the die with fewer traces, but that doesn't seem like a huge win.
The Macrumors article you linked to does say: "Apple plans to launch new 14-inch and 16-inch MacBook Pro models with an improved iteration of the M1 chip as early as this summer, according to Bloomberg's Mark Gurman."
However, the Mark Gurman article (https://www.bloomberg.com/news/arti...ook-pro-macbook-air-revamps-with-faster-chips) does not claim that the new chips are based on the M1. It only compares the new chips to the existing M1, i.e.
"The machines will feature processors designed in-house that will greatly outpace the performance and capabilities of the current M1 chips"
"Now more powerful iterations of the company’s silicon are coming to the Mac line." [ this doesn't necessarily imply it's based on the current M1 ]
"For a redesigned, higher-end MacBook Air planned for as early as the end of the year, Apple is planning a direct successor to the M1 processor. That chip, codenamed Staten, will include the same number of computing cores as the M1 but run faster". [ doesn't explicitly say what chip will go into the MBPs ]
The new core will have a new software-assisted branch predictor based on obj_msgsend, which will speed up some code by up to 15%.
That is a surprisingly specific prediction....methinks @cmaier has access to someone closely involved with Apple!
1) Different letter series for the high-end MacBook Pro chips
2) 8 performance cores, clocks at least 20% higher than the M1's. Agnostic on whether the no. of efficiency cores will be 2 or 4.
3) Significant updates to the specialized computation blocks: Those for image processing, video processing, audio processing, and the neural engine.
4) 16" will be offered with options of both 16-core and 32-core GPUs.
5) Max ram on 16" will be at least 64 GB, since they will at least equal the current model.
6) Monitor support on 16" will exceed current model's 2 x 6K/4 x 4k.
7) If there is a 1080p camera, it will be in array form (i.e., an array containing two or more separate lens/sensor combos, where the images are digitally combined to give 1080p), since it's hard to make a single sensor/lens 1080p module thin enough to fit in a laptop lid.
Also:
1) cupholders. they'll all come with cupholders. we may not think our laptops need cupholders, but Apple's job is to realize our needs before we do.
2) monitor support on all other AS Macs will be reduced from two to zero, because all monitors will be cloud-based. in addition to doing your computations in the cloud, all your viewing will be done there as well.
2) 8 performance cores, clocks at least 20% higher than the M1's. Agnostic on whether the no. of efficiency cores will be 2 or 4.
3) Significant updates to the specialized computation blocks: Those for image processing, video processing, audio processing, and the neural engine.
4) 16" will be offered with options of both 16-core and 32-core GPUs.
5) Max ram on 16" will be at least 64 GB, since they will at least equal the current model.
6) Monitor support on 16" will exceed current model's 2 x 6K/4 x 4k.
7) If there is a 1080p camera, it will be in array form (i.e., an array containing two or more separate lens/sensor combos, where the images are digitally combined to give 1080p), since it's hard to make a single sensor/lens 1080p module thin enough to fit in a laptop lid.
Also:
1) cupholders. they'll all come with cupholders. we may not think our laptops need cupholders, but Apple's job is to realize our needs before we do.
2) monitor support on all other AS Macs will be reduced from two to zero, because all monitors will be cloud-based. in addition to doing your computations in the cloud, all your viewing will be done there as well.
Last edited:
yes A15 will have that. That means M2.
The "M1X" in this years MacBook's Pros likely won't have it.
There will not be an M1X in the MBPs. It will be an M2. For now on, new cores appear first in Macs, then later in iPhones.
I don't think we would already have evidence in iOS/macOS code if these new features were only coming in the A15 this fall.
Mark Gurman has said the next MacBook Pro's will have the next iteration of the M1. While for the redesigned MacBook Air coming later this or early next year will have M2, as the Air will the direct successor to the M1.