Oh definitely , I wouldn't think it would throttle much at all in your studio , just was asking about the reverse actually. If an M2 Max/ultra in a studio would still be faster over long loads due to not throttling vs a M3 Max in a macbook throttling at times (maybe they don't ever throttle at all in a laptop either , just thought I had seen a few videos where they did when pushed in laptop form)
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 Max Chip Around as Fast as M2 Ultra in Early Benchmark Results
- Thread starter MacRumors
- Start date
- Sort by reaction score
It's unrealistic to expect that the first chip from QC/Nuvia would be able to compete with a Mx Max on all metrics (no matter where a couple of their engineers came from). It would take generations, investment and expansion of QC's chip division.
However, the PC space have been struggling with un-fit for purpose ARM processors that can barely handle a Chromebook depending on cloud stuff, if the Elite can provide a fit-for purpose PC ARM chip that'll run Windows ARM at an actually decent power/performance then they have a good start.
If they compare themselves to Apple it's just marketing drivel, their target is to make a dent in the PC space not Apple.
Last edited:
My 2011 iMac says “count your blessings” 😅
Yet only the chips are being tested. I thought the Mac faithful would've realized we need REAL testing - such as the imbedded SSD drives.
The M2 Max/Ultra Mac Studio SSD's have much faster Read/Write speeds than the M2 Pro chip devices, so I'm curious if the M3 Pro/Max machines also got a Read/Write speed bump?
speed matters to two people in this thread. The rest of us have something that loads our non exciting stuff just fine.
Exactly the reason why I want 4TB or even 8TB.
I don’t see how this isn’t obvious. Only the A17 Pro is on TSMC’s 3nm N3B process. The M3-series are on TSMC’s 3nm N3B process. No other TSMC 3nm process is in mass production besides N3B. Yes, they could have redesigned earlier chips to work on 3nm, but that’s unlikely. Why would they when they already have a perfectly good architecture that works with it? The second proof is that the GPU cores are using the same cores in the A17 Pro with mesh shading and ray tracing, features that don’t exist on any other A-series chip besides the A17 Pro. The M3 series also has an AV1 decoder in its media engine, something that was only added with the A17 Pro. We know without a doubt Apple has based the M1’s on the A14 and the M2’s on the A15. The A16 doesn’t have any of the features Apple mentions for the M3 series.
On what toolchain? That's definitely not the case with .NET (which scales poorly beyond a few cores) or Swift (which does scale even to ten cores, but then drops off significantly).
We have the first OpenCL compute scores for the 40GPU cores M3 Max: 92000
The M2 Max with 38 cores has an OpenCL score of 86000
The M2 Ultra with 76 cores has an OpenCL score of 121000
That means that the M3 Max in OpenCL compute is less than 10% faster than the M2 Max
Something does not add up, since the M3 has 5% more cores and 10% higher clock speed
Btw:
A NVIDIA desktop GeForce RTX 4090 scores 304000
A RTX 4080 Laptop GPU scores 195000
Even the M3 Ultra will be far behind.
Microarchitectures are not tied to process nodes. Apple A9 was made on both 16nm and 14nm. Apple A9X was on a thirdly different 16nm process.
Well, for one, because TSMC seems to have trouble ramping up 3nm production.
I don't personally think that's what happened, but I've seen an article that seemed pretty sure of it.
Yes, like I said: the speculation was that the CPU cores were the A16's, and the GPU cores only were the A17's.
your comments will post faster than ever
Check out the article I was referring to. They ran multiple tests, showing far faster performance. OpenCL isn’t even supported on macOS anymore and hasn’t for years. Tom’s Hardware’s analysis showed a performance between an RTX 4070 Ti and a 4080 for the M2 Ultra.
Actually, they are tied to process nodes. If it happens to work between 16nn and 14nm, that’s fine but Apple would test and fix any issues before sending it to production. Apple would have to revise the cores that were designed for 5nm for whether they would work on 3nm. You can’t just take a design that was created for one process node and assume it’ll work with another process. Apple is also clearly using the A17 Pro’s GPU cores. What possible reason would they have for using the A16 CPU cores when they can simply use the same CPU cores from the A17 Pro? The simplest explanation is usually the correct one. It’s a lot less work to simply use what was already designed for a 3nm process than to have to revise a design that wasn’t.
No, but you can adapt a design to another process. And Apple has done it before.
Yes. For the third time, nobody is arguing against that.
And I also personally think the CPU cores are the A17's.
Yield issues.
My M1 Pro 16” never has audible fan noise even when rendering. I also haven’t seen throttling where the render progress slows.
You must be joking.
The US price for this model is USD3,499. Which is at today's google exchange rate A$5,442, when you factor in our (Aus) 10% sales tax (which is always included in the displayed price) you get A$5,986 (i.e. within $15 of Apple's price in Australia noting that a small change in currency rate would make this positive or negative difference probably every other day). Further, considering that the US sales prices doesnt include any sales tax and that many US states do in fact have a sales tax. It is clear that their is no increase in pricing for our region at all. But thanks to our regulations we do get longer support and warranties.
Also misleading to Americans this talk about freedoms and high taxed, as our conservative party are more in favour of raising our sales tax than any other party.
“ … so you can effectively get the same performance for $500 less ..." with a built-in display for free 😊
With respect to the "Pro" laptops: starting at 8GB in 2023 is terrible
With respect to the iMacs leapfrogging over M2 is weak
My base M2 Mini has TB4, while base M3 has TB3, hardware regression there. Although, 3 & 4 both support 40Gb/s ... if past Apple history proves anything they'll use version 3 versus 4 as an artificial cut off point for Mac OS upgrades in the future...
And then finally the 150/300/400/800 bandwidth of their unified memory is kind of ridiculous. Before it was 85-ish, 100, 200 or 400 throughput in versions one and two. They have the die shrink available to them, but they opted not to push the boundaries, and to me that seems lazy, as they own the whole stack now.
And finally, they rarely compared anything of real value to M2. It was all M1 or Intel stuff, so clearly they just are trying to lure you into an upgrade, all thepeople on old Intel hardware, and that's a money decision, not a technology, decision, and I thought we were told that, "we ain't seen nothing yet." They are just itching to drop support for Intel machines and it's gonna come sooner than later I think.
To me, it's clear that they artificially slowed things down since they were so far ahead of the competition, imo a sad trombone is warranted.
And really minor, they took away support for high impedance headphones... Really just give us a decent dac/amp which is already in my m2 base Mini.
I will have however, cut them slack for Wi-Fi seven, the lack of...
Finally, in the reading of multiple websites, nobody's running out to trade in their M2 machines handover fist, so it seems like initially it's going to be a slow roll until M4.
I absolutely could be wrong though...✌️
Sure, and yet the only real world benefit I would get for my use case would be the extra battery life of M1 -> M3, and all of that came with the M2.
The ONLY thing I would gain from the M3 specifically is Space Black.