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.
Apple M3 Pro Chip Has 25% Less Memory Bandwidth Than M1/M2 Pro
- Thread starter MacRumors
- Start date
- Sort by reaction score
Sounds like MBA is the right level for you
Nope, I think it’s a very carefully planned job to squeeze profits or to scale up better.
M1 MacBook Air is still a helluva laptop, but even more so at the sub-$1000 price-point.And yet still M1 macbook 8gb bought at 800$ kick ass of everything in photoshop, capture one... People buy smart and what you truely need
Whoa. This report seems to show the old switcharoo is back.
Apple's latest M3 Pro chip in the new 14-inch and 16-inch MacBook Pro has 25% less memory bandwidth than the M1 Pro and M2 Pro chips used in equivalent models from the two previous generations.
Based on the latest 3-nanometer technology and featuring all-new GPU architecture, the M3 series of chips is said to represent the fastest and most power-efficient evolution of Apple silicon thus far. For example, the 14-inch and 16-inch MacBook Pro with M3 Pro chip is up to 40% faster than the 16-inch model with M1 Pro, according to Apple.
However, according to Apple's own hardware specifications, the M3 Pro system on a chip (SoC) features 150GB/s memory bandwidth, compared to 200GB/s on the earlier M1 Pro and M2 Pro. As for M3 Max, Apple says it is capable of "up to 400GB/s," which is because the scaled-down M3 Max with 14-core CPU and 30-core GPU has only 300GB/s of memory bandwidth, whereas the equivalent scaled-down M2 Max with 12-core CPU and 30-core GPU featured 400GB/s bandwidth, just like its more powerful 12‑core CPU, 38‑core GPU iteration.
Notably, Apple has also changed the core ratios of the higher-tier M3 Pro chip compared to its direct predecessor. The M3 Pro with 12-core CPU has 6 performance cores (versus 8 performance cores on the 12-core M2 Pro) and 6 efficiency cores (versus 4 efficiency cores on the 12-core M2 Pro), while the GPU has 18 cores (versus 19 on the equivalent M2 Pro chip).
Additionally, while the M3 chip's 16-core Neural Engine has the same number of cores as the one Apple featured in the 3nm-based A17 Pro chip that debuted in the iPhone 15 Pro series in September, it's comparatively weaker on paper in terms of maximum achievable throughput, which is measured in trillions of operations per second (TOPS).
According to Apple, the M3 Neural Engine is capable of 18 TOPS, whereas the A17 Pro Neural Engine is capable of 35 TOPS. It's hard to say for certain, but it is possible that the iPhone 15 Pro requires a higher performing Neural Engine for features like computational photography and Face ID, whereas the M3 can compensate in other areas like machine learning by utilizing its additional GPU cores.
Taken together, it's presently unclear what real-world difference these changes make to M3 performance when pitted against Apple's equivalent precursor chips in various usage scenarios, especially given that the latest processors include new Dynamic Caching memory allocation technology which ensures that only the exact amount of memory needed is used for each task.
This opaqueness is not helped by the fact that Apple chose to emphasize the power of the new M3 Pro and M3 Max chips by repeatedly comparing them to the M1 Pro and M1 Max, rather than its more recent M2 variants, against which performance gains appear more modest. Hopefully we will learn more in time when the first thoroughgoing third-party benchmarks become available.
The new MacBook Pro models are available to order now, and they will begin arriving to customers and launch in stores on Tuesday, November 7. Be sure to check out our MacBook Pro announcement coverage for all the details.
Article Link: Apple M3 Pro Chip Has 25% Less Memory Bandwidth Than M1/M2 Pro
I'm sure a MBP with an M3 Pro Max & 128 GB RAM is a dream come true for some. But Apple made my 13" M1 MacBook Air so robust that my needs are more than met. Can't wait to hear what the M5 Competition has in store (the BMW, not the Mac!).
If game developers port AAA titles to Macs, then consider me a day 1 buyer of Mac Studio with M3 Max (from Mac Mini M1). I don’t mind buying a Mac for $3k because I know it’s longevity (5+ years). I do mind paying that for a PC because they just don’t last. My work laptop (Lenovo) is a testament to that - 2-3 years shelf life.
Or perhaps, more accurately, Apple is hitting the same design limitations that Intel and others in the industry are facing (coupled with losing a fair amount of top chip design talent to Nuvia).
You’ll likely see the much more capable M2 model, drop to sub $1000 for Black Friday.
This is the first time the Pro and Max have been differentiated in terms of CPU performance. Something that slipped past most people is that while the M3 Pro remains a 12 core chip it has a 6/6 performance/efficiency core count versus the 8/4 count of the M2 Pro. Part of me wonders if that, coupled with using triple channel vs quad channel RAM is the reason behind the changes to memory bandwidth.
well hang on, the "receipe" of what goes into a chip is far to complex to be summed up so easily. let's see bottom line performance, I am fine with a tradeoff in one area that is more than made up in another so that the entire CPU/GPU results in a performance gain. If the M3 Pro underperforms the M2 Pro in any real world tasks when tests I'll join you on the "outrageous" bandwagon.
Thanks for the processor technical details. I found all the comparisons to the M1 series very odd and if you read the graphs it's obvious the M3Pro is only about 10% faster than the M2Pro. With the reduced number of performance cores as well as relatively low max memory (36GB limit, really Apple!), and now 25% lower memory BW it's obvious that Apple has intentionally crippled the M3Pro processor to force people to the M3Max.
Overall, I am highly disappointed in the M3 specs since we were all told the M2 was just an incremental processor step and the big performance/energy benefits would be with the M3 line. But Apple has made a lot of design tradeoffs with the M3 and/or the 3nm node is not the be-all and end-all we were marketed. Anyway, I can see why Apple rolled this out in an evening event that only Apple fanboys watched. Since both iterations of the M series have yielded similar mediocre improvements one can expect this trend to continue with future releases.
Overall, I am highly disappointed in the M3 specs since we were all told the M2 was just an incremental processor step and the big performance/energy benefits would be with the M3 line. But Apple has made a lot of design tradeoffs with the M3 and/or the 3nm node is not the be-all and end-all we were marketed. Anyway, I can see why Apple rolled this out in an evening event that only Apple fanboys watched. Since both iterations of the M series have yielded similar mediocre improvements one can expect this trend to continue with future releases.
very interesting how they compare to M1 not M2, I would love to see why this is , so those of us with m2's don't go out and buy an m3 max studio thinking its going to be a big leap, when in reality its microscopic..
I wonder what this means, overall, in terms of performance for the M3 Pro.
Because it's hard to tell from the data we have.
What we know:
So, it has 25% fewer p-cores, but each of them is 15% faster. And it has 33% more e-cores, and each is 30% faster.
My guess is this is a less than 10% performance improvement for most tasks, but a significant battery life improvement.
This also suggests — and Srouji’s artwork showing the three SoCs seems to confirm — that the M3 Pro, unlike the M1 Pro and M2 Pro, is no longer a chopped version of the Max. The non-suffix SoC was always a separate layout, but now the Pro is also separate from the Max. Interesting.
But overall, it sure reads like: the Pro’s performance barely changes at all; if you want that, you now have to upgrade all the way to the Max. And the Max, conversely, will have worse battery life due to fewer e-cores.
Because it's hard to tell from the data we have.
What we know:
- The video says p-cores are 15% faster compared to M2.
- …and the e-cores 30% faster.
- Memory bandwidth is down 25%. (Presumably, this is already factored into the above figures. I'm also presupposing that clock changes, if any, are already factored in.)
- The M3 Pro has 6 p- and 6 e-cores, whereas the M2 Pro had 8/4.
So, it has 25% fewer p-cores, but each of them is 15% faster. And it has 33% more e-cores, and each is 30% faster.
My guess is this is a less than 10% performance improvement for most tasks, but a significant battery life improvement.
This also suggests — and Srouji’s artwork showing the three SoCs seems to confirm — that the M3 Pro, unlike the M1 Pro and M2 Pro, is no longer a chopped version of the Max. The non-suffix SoC was always a separate layout, but now the Pro is also separate from the Max. Interesting.
But overall, it sure reads like: the Pro’s performance barely changes at all; if you want that, you now have to upgrade all the way to the Max. And the Max, conversely, will have worse battery life due to fewer e-cores.
Hmm, lets see what the benchmark & real-world performance is like.
If these end up being slower in anything that an M2 model can do, then shame on Apple.
If these end up being slower in anything that an M2 model can do, then shame on Apple.
I think Apple realized the 3nm node wasn't as great as they were expecting and the die was cast so might as well release these before the end of the year. I think if the M3 had kick-butt performance (30+% improvement) Apple would have held a big daytime event in January to launch these chips and not a nighttime web-only event.
Scary price increases for scary downgrades and no scary fast thunderbolt 5. My wallet says no thank you.
Memory bandwidth is just one metric of a system that determines the overall performance. Does it matter if memory bandwidth has been slightly reduced, if improvements elsewhere result in net increase of performance? Besides, the primary user of that bandwidth is the GPU, maybe the dynamic caching feature of the new GPU saves bandwidth, which means it doesn’t need as much as before.
Heck, the 15" m2 already did last week at Best Buy.
Might be related to the fact that most of the original chip designers of m1 have now left the company :/
You've pretty much answered yourself there. It's a decent upgrade from an M1 or an Intel Mac but only really quite small from M2.
Intel was three generations behind TSMC, so your statement is decidedly less accurate.
I wonder if this will be explained by some as another good move by Apple which the rest of us can't understand. Like the smaller RAM which is so good it breaks the laws of physics, or the integrated GPU (and RAM) which has never been done before (sigh).
I hope people who do get them end up with pleasant surprise with the real world performance.
I hope people who do get them end up with pleasant surprise with the real world performance.