I hope I'm wrong, but even seeing the incredibly modest A14->A15 leap I think it's the direction we are headed. That didn't have a node shrink of course. We probably will see slightly bigger bumps with A16/A17 on 4nm and 3nm.
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.
Leaked Intel Roadmap Reveals More Efficient Chip Than M1 Pro and Max to Launch Within Two Years
- Thread starter MacRumors
- Start date
- Sort by reaction score
I hope I'm wrong, but even seeing the incredibly modest A14->A15 leap I think it's the direction we are headed. That didn't have a node shrink of course. We probably will see slightly bigger bumps with A16/A17 on 4nm and 3nm.
By the time those Intel CPU's are ready, Apple will be producing SoC's on a two nanometer process.
It's proof that competition is a great thing and keeps companies innovating, yes it could be argued that it's too little too late but if Apple hadn't made such an impact then Intel wouldn't have bothered. If you're dependent on x86_64 code then having a better CPU that doesn't need its own fusion reactor to power it is a good thing.
A leaked roadmap by Intel suggests that the company is developing a new lineup of CPUs that are targeted at outperforming Apple's 14-inch MacBook Pro with the M1 Pro and M1 Max chips by late 2023, or early 2024, which would be almost two years after the new chips and laptop made their debut.
The roadmap by Intel, initially leaked by AdoredTV and interpreted by Wccftech, explicitly states that Intel wants to compete with Apple's 14-inch MacBook Pro with its Arrow Lake series. According to the roadmap, Intel's 15th generation Arrow Lake processors could be ready to ship by late 2023, or early 2024 with a priority on delivering high-performance while using minimal energy.
The roadmap also says that Intel will utilize TSMC's 3nm process. Apple currently utilizes the 5nm process for its latest chips and is expected to adopt the 3nm chip architecture in 2023 with the "M3" Apple silicon chip and A17 chip in the iPhone 15.
Intel has already beaten Apple's M1 Max chip on paper if you ignore high-energy consumption and poor battery life. Benchmarks show that Intel's latest Core i9 processors received a higher score than Apple's M1 Max chip in tests, but that 4% increase in performance is offset by a marked reduction in battery life compared to Apple's chips. Tests show that a laptop with Intel's latest i9 Core chip only lasts six hours for video playback. In comparison, Apple advertises the latest 16-inch MacBook Pro as getting up to 21 hours of battery life for offline video playback.
Ever since Apple announced its transition away from Intel during the summer of 2020, it has been slowly transitioning its Macs to custom-made chips. So far, Apple has released four laptop computers with Apple silicon, alongside two desktop computers. In just a few weeks, Apple is expected to announce at least one new Mac with Apple silicon, with possibilities being a new high-end Mac mini and an update to the low-end 13-inch MacBook Pro.
Article Link: Leaked Intel Roadmap Reveals More Efficient Chip Than M1 Pro and Max to Launch Within Two Years
I find that a bit hard to judge from your chart, given that the y-axis isn't linear, but I wouldn't say so. The A9 and A10 were a bit of an outlier.
First chart shows single- (green) and multi-core scores in Geekbench 5.
Second chart shows YOY improvement of single-core scores. The A15 is, depending on how you look at it, either somewhat concerning or simply more focused on battery life. But for the most part, we've had multiple years of around 15-25% improvement.
(edit)
OK, if we add scores from Geekbench 4, I think your hypothesis that the pace used to be higher is true.
Same charts:
The average YOY in the Geekbench 5 era (A7 through A15) is 27.15%.
The average YOY in the Geekbench 4 era (A5 through A12) is 56.31%.
Last edited:
By the time we see an Apple Silicon Mac Pro, this will be addressed (maybe not to the tune of 32GB per core, but certainly we’ll see an option higher than 64GB). This is hardly an intractable issue with Apple Silicon, otherwise Apple wouldn’t have made the jump. Remember, the M1 is roughly equivalent to an i3 or i5, with the M1 Pro being roughly equivalent to an i7 and the M1 Max being roughly equivalent to an i9. We have yet to see an Apple Silicon chip intended to compete with Intel’s Xeon chips, the ones we have seen are chips oriented to consumer and prosumer models.
Not sure if you actually think this, but it’s not how Apple or any other company operates. Products that take years to develop can’t be driven by designers / engineers watching what the competition is doing. Intel has watched Apple develop their A series chips over the past ~10 years, getting to a point where they’re both more efficient and more powerful than many laptop and desktop systems, and it took Intel this long to figure out they need to do more than simply running faster, and less efficient, chips in order hit certain performance levels.
If as you say, that it takes competition for Apple to stay on their toes and innovating, why would Apple have created the A series chips and the M1’s when there was literally zero competition and hasn’t been for years. They’re in a league of their own and focused on developing systems that provide the best UX they can, per price point, while also managing for efficiency and thermals. That’s not going to change going forward and it won’t have anything to do with what Intel is talking about releasing at some point in the future.
It's not that simple. As you say, it takes many years to iterate on CPUs. It was around 2015 that Intel realized they needed to course-correct, at least as far as their 10nm strategy was concerned. That's why Cannon Lake essentially only shipped in prototype form.
Intel Lakefield was their first attempt at a heterogenous core concept, and Alder Lake expands that to the general line-up. (Alder Lake is also the first time, embarrassingly, that Intel can roll out 10nm to desktops as well.) So we're only now starting to see what Intel has learnt from ARM and Apple taking this heterogenous approach.
Right, I forgot to mention the y-axis was log-scaled! My bad. You're right that the A9 was a uniquely big jump, though with my data (using Geekbench 4) we're also seeing the rapid improvement from the 4S to the 5 (284 to 754 single core, a 265% improvement) and from the 5 to the 5S (754 to 1266, a 168% improvement), hence the claim of a drop-off in year-over-year improvement relative to *that* (which I doubt anyone would expect of Apple). Their steady ~20% gains in recent years are still very impressive, and I'll be very happy if they can keep that up.
Maybe we're seeing Apple's CPU team migrate to an old Intel-style "tick-tock" cycle, where they have performance gains one year and efficiency gains the next. Very excited to see what the A16 has to offer.
This is definitely true for now, though I'm interested to see whether it holds over the next few years. Once's Microsoft's exclusive deal with Qualcomm expires we should see more efforts from other ARM chipmakers, and Apple forcing the software industry's hand with ARM Macs has already resulted in more native Windows ARM ports than we've seen in years. If x64-on-ARM emulation on Windows keeps improving, we'll very soon be at a point where most casual users won't notice a difference.
As ever, Intel tries to build a reputation on what they're GOING to do (maybe) in the future. Meanwhile, Apple glides across the ice, skating not to where the puck is but where it is GOING to be.
Yeah. I've added a Geekbench 4 equivalent chart.
Indeed, the YOY was a fair bit higher at the time. As you say, the A6 vs. A5 change was stunning.
(edit) although I just remembered that the A6 was the first chip to actually use an Apple design. The A4 and A5 were ARM Cortex-A8 and ARM Cortex-A9. So really, the A6 is quite a different beast anyway.
Agreed. If the A16 is another single-figure change like the A15 was, I'd be concerned. But we've outliers like it before, such as the A8 over the A7. If it's in line with previous ~20% YOY gains, I think that's fine, and hard for competitors to keep up with.
Quite possibly.
Last edited:
One thing I'm finding lately is that for my typical usage, internet bandwidth is really the biggest bottleneck. I wish gigabit offerings were more affordable.
FWIW it's a huge noticeable difference to me to go from 200-300Mbps down to 50-100Mbps now that I'm used to it. Extremely bloated app-like websites are some of the causes of this. M1 chips do awesome at javascript processing at least.
FWIW it's a huge noticeable difference to me to go from 200-300Mbps down to 50-100Mbps now that I'm used to it. Extremely bloated app-like websites are some of the causes of this. M1 chips do awesome at javascript processing at least.
It doesn't matter how many ARM chipmakers there are. That may actually make things worse, as they compete for a tiny platform and drive prices and products to rock bottom, thus destroying the platform. But it all doesn't matter if Microsoft doesn't try to make Windows on ARM viable. And currently they have done very little. Microsoft's flagship IDE does not run natively on WoA and has no plans to do so. Microsoft is only shipping 64-but emulation in Windows 11, even though WoA began in 2012 (with RT). Meanwhile Apple ported all their software, including Xcode, Final Cut, and Logic, by WWDC, and made sure to work with important software vendors (Microsoft and Adobe) to get their software ready ASAP, and shipped a great emulator in the meantime. One company put in the effort to make their new platform a success, and it has been one by all accounts. Extrapolating from that, that another company who has put in minimal effort will also succeed is bizarre.
I'm glad Intel is starting to feel the heat from Apple and AMD. Intel has been stagnant for far too long.
IF Intel can execute to their roadmap, great - competition is wonderful.
History has shown Intel does not execute ...
Time will tell
History has shown Intel does not execute ...
Time will tell
I... what??? They haven’t made their own development tools native on a platform they’ve been selling for years? Wow, okay. That’s worse than I thought.
Maybe Intel’s worry is that if they don’t stop the “ARM is the future” narrative now, Microsoft might actually start taking the architecture seriously instead of dragging their feet.
Nah, Intel will have access to the same TSMC N3 process as Apple will have (which is rumored to have yield issues, so maybe both Intel and Apple will be delayed). If anything, it will be interesting to see how x86 and Apple's version of ARM really compare when on equal footing with regard to the manufacturing process.