Battery life certainly can use improvement, at least from my M1 Pro laptop. It’s good, but not great. Also 30% isn’t a marginal improvement over the standard M3. From the Geekbench scores, there’s a 30% difference between M3 and M3 Pro and a 39% improvement between the M3 Pro and the M3 Max. Remember the goal was to put space between the Pro and the Max. In the prior model the M2 Pro and M2 Max had the same CPU scores. If Apple had made the Pro 7 p-cores and 5 e-cores and gotten it to 35% and 35% (just a guess), you’d still be complaining because they cut down on the p-cores.
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.
M3 Pro Chip Barely Faster Than M2 Pro in Unverified Benchmark Result
- Thread starter MacRumors
- Start date
- Sort by reaction score
They aren’t making machines annually for the same people to buy every year. No one is on the same update cycle and there will always be buyers all year round. Each year, they’re selling to different people. If you don’t like a one-year update cycle, then you personally can buy every two years. Problem solved. But others will always need one at different times than you will. You’ll notice every other computer manufacturer is doing the same thing, annual updates.
On this, I agree with you. The M3 Pro is a poorer value than the M2 Pro, just as the iPhone 14 was a poorer value than the iPhone 13. That has to happen with at least one product when you rebalance your product lines. It is inevitable because they can’t all move up higher at the same rate and actually rebalance the line. Value-wise, one has to go up (M3 Max) while the other must go down (M3 Pro) when you try to differentiate them. If the M3 Pro maintained its value proposition, it’d still be too close to the M3 Max. You’re essentially arguing you can do both at the same time when you can’t really.
Ask Samsung. They’ve been caught doing that many times where their OS detects a benchmarking tool being run and then switches their hardware into maximum performance mode, something that they cannot sustain in regular use. I think Geekbench even removed Samsung’s eligibility to be benchmarked because of that cheating.
If you don’t care what it costs people, well, it’s ignorant and classist and stunted. It’s not that hard to care about both.
Battery life is only something people here are bringing up as a post-hoc justification, it wasn't anything that apple advertised on, nor is it, nor was it, how the Pro SoC was advertised relative to the standard SoC. Yes battery life will probably be better, but, again, I doubt you complained about M1 Pro battery life, you're now here using it, only after, the announcement, to justify the decisions of Apple.
Serious question:
If Apple had not lowered the P-Core, GPU, and Memory Bus would you be complaining that they didn't give you more battery life?
Also note:
M3 Pro (15173) vs M3 (11894) = 27.5%
30% is a smaller difference than M2 and M2 Pro
M2 Pro (14256) vs M2 (9744) = 46%
If the only point was to make more of a difference between the Pro and the Max, well they would still have a pretty big difference if they hadn't made the Pro worse.
M2 -> M3 = 21% multicore improvement
lets assume
Hypothetical M3 Pro that is better:
M2 Pro -> M3 with say 15% multicore improvement (lets be conservative)
= 16681 multicore score
that is still 30% slower than the 21,000 score of the M3 Max...
Still plenty of gap between then two.
The M3 Pro that wasn't worse would still have 200 GB/s memory bandwidth and would have 20 GPU cores in my world, and that M3 Pro would still be half as powerful from a GPU perspective as M3 Max. Again plenty of distance.
Apple did not need to make the M3 Pro worse to add distance between it and thee M3 Max.
I would complain at 7 + 5 as well because I don't like it when companies make a product worse (relative to the other products in the lineup) than it was in the previous generation.
I keep saying, sure the M3 Pro is great, but that it is a worse value proposition than M2 Pro was.
Why must it go down?
If it maintained its position it would still have 30% lower multicore than M3 Max, it would have half the Gpu performance of M3 Max, half the memory bandwidth... support fewer display out, fewer Pro Res encoders...
At this point the 5+6 core variant is pretty sad, having only a single additional performance core vs the M3.
Your confusing die size and transistor cost.
The die size didn't shrink 8% the transistor count did. The die size is likely far smaller than M2 Pro because logic density increased 1.7x compared to N5 (TSMC Numbers)
Sorry, but your opinion doesn’t make it objective fact. The facts are that the M3 Pro is 30% faster than the M3 in CPU and close to 80% faster on GPU. You didn’t think that was enough. That is a opinion. Your entire argument is about value proposition while acknowledging the M3 Pro is an improvement over the M2 Pro. Value propositions are opinion-based. So yeah, both of us are expressing opinions.
I also wasn’t justifying anything. I’m telling you what their strategy is. That is fact. Their strategy was to rebalance their lineup, not to individually ramp up every chip by the same percentage. There will always be one product that loses in the value proposition along with at least one product that gains in its value proposition whenever one of these rebalancings happen. It isn’t possible to maintain value propositions for every product in the product line when doing that, just as the iPhone 14 suffered in the iPhone rebalancing. The Pro just happens to be the unlucky one.
Next year, I expect you’ll complain one of the iPads sucks because of a lower value proposition after Apple rebalances the iPad lineup. It is inevitable that one product won’t improve as much as other products do within the lineup.
It’s also true the A17 Pro doubled in cost over the A16. The transistor count did not double while the yield was down to 60%. You’re forgetting yield. While the 5nm process was at 90% yield, the N3B process had horrendous yields. As a result the M3’s are going to cost a heck of a lot more than the M2’s. The costs will come down once they go to N3E, though transistor densities will suffer as a result. Remember that 3nm is just a name. It doesn’t actually mean the traces are 3nm in width. Some parts of the chip will be 3nm, but others will be 5nm, others will be 14nm, all on the same chip.
Thanks for the information about TSMC but I was referring to Apple’s total gross margins, not just the Mac business. They have been rising steadily this year.
Last edited:
It’s not ‘true’ it’s a guess by analysts.
N3B yields are horrendous, also a guess.
I have no idea why you brought up the actual measurements because the marketing numbers have been meaningless for years. TSMC claims 1.7x density increase for N3 vs N5, even if we take analysts guess at 30% increase cost as accurate this still means that per transistor costs went down.
we all knew this would be coming, developing and producing custom cpus for an shrinking market is expensive.
off the shelve parts aren't that bad and if we look around with what qualcomm is coming up with oh boy apple's lead is dwindling scary fast.
off the shelve parts aren't that bad and if we look around with what qualcomm is coming up with oh boy apple's lead is dwindling scary fast.
It’s not opinion that the M2 Pro was 46% faster in CPU than M2
46% > 30%
Therefore
Value per dollar in moving from M2 to M2 Pro greater than value per dollar in moving from M3 to M3 Pro
Edit:
M2 to M2 Pro also offered 2x memory bandwidth, m3 to m3 pro is only 1.5x
Again 2 > 1.5
M2 to M2 Pro GPU 1.9x
M3 to M3 Pro GPU 1.8x
1.9 > 1.8
Price did not go down therefore value, for money of M3 pro vs m3 is lower than value for money for m2 pro vs m2
The numbers are objective fact.
Xdollars used to get you a greater percent increase over base than it does today.
Last edited:
I’m going to laugh myself silly when real world tests are done and M3 and M3 Pro chips turn out to be pretty respectable after all the booing and hissing around here.
Rebalancing can be accomplished in more than one way. You can improve the top, lower the middle or both. Apple did both. The balance Apple chose is not objectively better than any other balance point.
If next year they make the iPad Air worse relative to its current position, sure I’ll complain, why shouldn’t I? They have made something worse value for money relative to its siblings than the product it replaced.
It may be nice for Apple that they did this that doesn’t make it good or praiseworthy.
Actually I did complain about M1 Max battery life. I still complain about it. Keep in mind that’s why they doubled the e-core count on the M2’s. I don’t have an M2 Pro or Max to complain about since I don’t how the battery life is. Remember the M1 Pro/Max had only 2 e-cores. If Apple had kept the M3 Pro at 4 e-cores, the same as the M2 Pro, no I wouldn’t be complaining about it because I still wouldn’t have bought it. I only complain about things I have and don’t like, not things I don’t have and don’t have any experience with.
I do find it odd that Apple went for a 3-chip memory array on some models and 4-chip array on others, hence the difference in memory bandwidths (similar to the old problem of using one NAND storage chip versus two that has since been rectified in the M3 Macs). I’m curious if this has to do with bad chip yields where a memory controller was damaged. This weird memory array also exists on the M3 Max, but I’m not complaining about it because I don’t know the reason for it.
You’ll notice the binned M3 Max has only 3 memory chips, lowering its memory bandwidth, while the unbinned version has 4 memory chips for the full 400GB/s. This is why the M3 Max binned supports only 36GB and 92GB of RAM using 3 modules of 12GB and 32GB. Meanwhile the unbinned version supports 48GB, 64GB, and 128GB using four modules of 12GB, 16GB, and 32GB. I’m curious why the binned version doesn’t support 48GB using 16GB modules.
Let me ask you this: is there an objective best way? It seems someone’s going to complain no matter what they do. This is how they decided to do it, but in every rebalancing, at least one product will win while another loses. Apple improved all three, just not all three to the same degree. Some call that a win, while you call it a loss. Again, there is no objective criteria except all three of them cannot improve at the same rate while achieving a rebalance. You say that they should have jacked up the Max even more than they did. That comes at a cost. The M3 Max is already more expensive than the M2 Max objectively. If they raised the p-core count even further because they jacked up the M3 Pro, that would raise costs even more. Which is better? Again, there is no objective answer.
So many responses going off the rails. Perhaps you missed the part of the article that says "This is only a single benchmark result." It very could well be that other stuff was going on while the benchmark was running, so the result can't be trusted. Someone at Apple might be having a laugh watching you all squirm over this.
Seems if you're after a 14" the two best options are a Refurb 14" M2 Pro or splurging for an M3 Max.
At their price points neither the M3 or M3 Pro 14s make sense compared to a Refurb 14" M2 Pro; especially if you're getting the M3 with 16GB RAM.
At their price points neither the M3 or M3 Pro 14s make sense compared to a Refurb 14" M2 Pro; especially if you're getting the M3 with 16GB RAM.
I didn’t say the max needed to be jacked up more. But that if they hadn’t made the pro worse (relatively) then the max would still have been very very well differentiated with 33% more multicore, double the GPU and double the memory bandwidth.
The rebalancing could have been done differently so that no product lost transistors and value vs the previous generation. The m3 max with 12 p-cores would have pulled away from the pro even if the pro had 8 p cores.
Multicore performance is very little affected by the initial spotlight indexing, just look at the m3 scores where the single core score went up from 3000ish to 3150ish the multicore score only increased by about 100 points. These results are quite likely fairly representative.