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.
[CPU only] Apple M1/2(Max/Ultra) (TSMC 5nm) vs AMD Zen 4 (TSMC 5nm) - Technical Analysis
- Thread starter exoticSpice
- Start date
- Sort by reaction score
A couple of days ago, AMD sent a document with this marketing graphic to the press.
Are those benchmarks fair?
Are those benchmarks fair?
"Are those benchmarks fair?"
Depends on what you count as fair. AMD are matching their top of the line against the bottom of the M2 line.
In terms of cores, Apple has 4 P at the low-end, 6 midrange (Pro) 8 at the high end (Pro and Max). E cores basically catch up the extra performance from hyperthreading. Apple frequencies are lower, but IPC is higher, so single core is essentially a wash; but Apple can sustain that IPC across all the cores whereas AMD will have to dial back the frequency when all cores are working.
In terms of GPU, more or less nV, AMD, and Apple all have the same sort of performance in what they call a core, or SM, or CU. At the low end M2 Apple gives you 8 or 10 of these (M2) at the midrange 16 or 19 of these (Pro). So at comparable points Apple is ahead in GPU. Likewise at comparable points Apple gives you more memory bandwidth, both for CPU (nice) and GPU (essential).
Finally Apple has a very nice (probably unmatched) media engine including really good H.265 encoding and a low power inference engine (which sounds good but remains mostly unused still, on both the Apple side and even more so the x86 side).
And other stuff like the SSD engine and the Secure Enclave (which doesn't seem to hit frequent problems the way these things all seem to on the x86 side – maybe that's unfair to AMD, I don't know the details of which TMP has which problems).
AMD says this matchup (their high end, with 8 cores and 12 CUs) against M2 (with 4+4 cores and 10 CUs), rather than M2 Pro, is justified because they sell in comparably priced devices. This is kinda true, but is also because the Apple devices come with nicer other stuff (keyboard, trackpad, screen, better battery, etc).
So "fairness" is not really an interesting point. Basically things remain the same as they have always been
- if you want cheap, you can check more boxes with x86 (Intel or AMD) but the result will not feel as nice
- if you compare tech rather than some sort of vague "equivalent price points" Apple remains ahead with M2 and M2 Pro.
Depends on what you count as fair. AMD are matching their top of the line against the bottom of the M2 line.
In terms of cores, Apple has 4 P at the low-end, 6 midrange (Pro) 8 at the high end (Pro and Max). E cores basically catch up the extra performance from hyperthreading. Apple frequencies are lower, but IPC is higher, so single core is essentially a wash; but Apple can sustain that IPC across all the cores whereas AMD will have to dial back the frequency when all cores are working.
In terms of GPU, more or less nV, AMD, and Apple all have the same sort of performance in what they call a core, or SM, or CU. At the low end M2 Apple gives you 8 or 10 of these (M2) at the midrange 16 or 19 of these (Pro). So at comparable points Apple is ahead in GPU. Likewise at comparable points Apple gives you more memory bandwidth, both for CPU (nice) and GPU (essential).
Finally Apple has a very nice (probably unmatched) media engine including really good H.265 encoding and a low power inference engine (which sounds good but remains mostly unused still, on both the Apple side and even more so the x86 side).
And other stuff like the SSD engine and the Secure Enclave (which doesn't seem to hit frequent problems the way these things all seem to on the x86 side – maybe that's unfair to AMD, I don't know the details of which TMP has which problems).
AMD says this matchup (their high end, with 8 cores and 12 CUs) against M2 (with 4+4 cores and 10 CUs), rather than M2 Pro, is justified because they sell in comparably priced devices. This is kinda true, but is also because the Apple devices come with nicer other stuff (keyboard, trackpad, screen, better battery, etc).
So "fairness" is not really an interesting point. Basically things remain the same as they have always been
- if you want cheap, you can check more boxes with x86 (Intel or AMD) but the result will not feel as nice
- if you compare tech rather than some sort of vague "equivalent price points" Apple remains ahead with M2 and M2 Pro.
A couple of days ago, AMD sent a document with this marketing graphic to the press.
View attachment 2197934
Are those benchmarks fair?
Probably accurate, if the AMD system is configured to run above its TDP. 8 Zen4 cores at 4.5ghz would do that. You can forget about 15 or 25W operation though.
AMD couldn't make a better matchup. The M2 and 7840U have the same number of cores, an integrated GPU, similar power consumption and the laptops that will have the AMD chip should be priced similar to the M2 MBA.
I happened to notice those two tiny words in the middle of the comparison graphic "up to:". Plugged in, with the fans churning and the battery crying, yeah, they will get those numbers. At the same wattage, probably not.
(And yes, I concede, Apple also uses those two words quite a lot.)
The statement about same number of cores is technically not correct. All cores in 7840U are high-performance cores, so AMD can claim advantage in multicore performance on two accounts: a) running all 8 cores at relatively high clock, which would achieve better multi core performance at the expense of power usage, or b) running all 8 cores at very low clock, which allows them to hit optimal spot at the clock-to-power curve and get good performance with low power usage.
When one is concerned about multi core throughput, especially in trivially parallel tasks such as CB23, AMD is certainly more flexible. You can run that CPU at both 50W and 15W, no such flexibility for Apple. The downside of AMD is lower single-core (and general everyday) performance, plus, they absolutely need those 8 cores to compete with Apples 4+4 config.
If only benchmarks and charts could be regulated so that companies couldn't make misleading charts.
Nothing prevents Apple from having 6+2 or 8+0 on M2. It's a choice Apple makes, just like AMD using x86 ISA instead of ARM ISA.
Of course, it’s SKU vs SKU after all. Just pointing out some of the tradeoffs implemented by the two products. As I said, AMDs design is arguably more flexible if you are after trivial parallel performance (and have a chassis that can handle the extra power requirements).
"same number of cores"???
Jesus wept. When a thread degrades to the level of not understanding the difference between E and P cores (or for that matter the meaning of v-Cores) I'm out. This is no longer a thread safe for people capable of thought.
Technically, AMD is using AMD64, which was their 64-bit extension set grafted onto the x86 ISA (and which Intel now licenses from AMD for their own CPUs due to the colossal failure of Itanium)
It might also be worth noting that those are all SMT cores, which means the 7840U can run 16 threads at once, for about a 10~20% performance gain.
Ah, so the definition of "not dishonest" now includes comparing one device's GB6 score to some other devices' GB5 scores. Good to know.
Last edited:
You're still trying to compare systems with a mix of performance and efficiency cores to systems with performance cores only, and either ignoring or pretending those factors either do not exist or do not affect the results. Ironic given that your own signature line states "MR is a graveyard of misinformation"...
If AMD had managed to create a chip with 8 P cores that consumed similar to an M2 with similar performance (pending third party benchmarks), what's the problem? A core is a core, and it's up to each company what kind of cores they choose.
Exactly. I tried to post something here saying that, but it was deleted for whatever reason.
Kudos to AMD. I look forward to seeing more reviews of devices carrying these chips. I hope we see more of what it's like to run these things all day. No matter, competition is good for all consumers and IMO, both of these competitors looks better than Intel offerings for the time being.
Not to be a defender of Apple, but just as one can't complain when comparisons are drawn between chips with different core choices, one can't complain about the nodes being used to develop those cores. What matters is the performance (per watt for mobile) of what's shipping. And for now, that's comparing an M2 to a 7840U or an M2Pro/Max to a 7840HS. Performance per watt looks pretty close. Apple chose not to use N4P. We'll see if the changes the game in a month.
Isn't the P cores of that have hyperthreading? It should boost the performance a little bit. That said, M2 Air have a benchmark of MS of 9854. That said, I'm not impressed especially the possibility of that AMD chip to go higher in wattage. 28w is the default TDP of the 7840u and listening to the video, the commentator specifically said that it runs at 28w to get that performance of 10K on multithreaded. A ~2% at higher wattage is eh.
Lowering the wattage to 25 or 20w and maybe 15w will reduce the multithreaded performance well below 10K and maybe go lower than 9800 score of M2 especially the 15w.
What are their power consumptions when achieving these scores?
Intel says the 13900k has 24 cores and AMD says that the 7950X3D has 16 cores.
Last edited: