Apple should have released the M4 Ultra chip into the Studio and created an all brand new M4 Extreme chip for Mac Pro.
Last edited:
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 Ultra Chip's Massive 80-Core Graphics Performance Revealed in Benchmark Result
- Thread starter MacRumors
- Start date
- Sort by reaction score
They said that about the M3 last summer…
There also wasn’t one for the M3. What we seem to have is M3.5 since M3 also didn’t have Thunderbolt 5.
I’m deciding the configuration of the Mac Studio that I’m going to buy, and the price difference between the M4 Max and the M3 Ultra is way too high for the small performance increase you get, specially in CPU but also in GPU. I don’t find it justifiable for my case. I’d of course like getting 256GB RAM, but for that you need to pay for not only the RAM, but also the Ultra… so RAM becomes twice the price.
I feel that Apple wouldn't have put a "lower" number chip in its most powerful computer if it didn't beat the m4.
What? 16% graphics performance over the M2 Ultra is pathetic. Surely this ignores the built-in raytracing hardware? Apple suggests a 2x boost over the M2 Ultra using Redshift which is a GPU renderer.
Most of the benchmark scenes in Blender are the same or faster with a single M3 Max over the M2 Ultra, so I would hope for a solid 2x in rendering performance.
Most of the benchmark scenes in Blender are the same or faster with a single M3 Max over the M2 Ultra, so I would hope for a solid 2x in rendering performance.
Then we’d still be waiting for any M4 variant. I think the best we can hope for is the Max Studio doesn’t lag behind the Max MacBook Pro as long just because the matching Ultra isn’t ready, so we don’t get a repeat of the Studio skipping the M3 Max.
Apple has stated that it may not release an Ultra version for every generation of chips.
M-series chips are unironically monsters when it comes to Minecraft. That game is CPU-bound 99% of the time, and Mx chips have CPU power for days.
I used to run a server for the kids in my family off my M1 Mac Mini, but those days are long behind us. Damn kids just insist on growing up.
The only good use case appears to be something where the extra RAM and GPU cores are an absolute must, otherwise the higher per-core performance on the M4 Max wins out.
Previously you got double. M2 Max at $2k, double the price at $4k = double the power. Now, you pay double but don’t get double. Weird positioning
Yes I know, people who need the most need it and it’s worth it. They just don’t get as much more and they pay more for a 1 generation older architecture. It’s a bit odd, don’t try to say it isn’t. Because it is.
Yes I know, people who need the most need it and it’s worth it. They just don’t get as much more and they pay more for a 1 generation older architecture. It’s a bit odd, don’t try to say it isn’t. Because it is.
Just leaving this here:
https://wccftech.com/m3-ultra-chip-chip-design-changed-by-apple/
Edited to show a callout from the article (Last year):
https://wccftech.com/m3-ultra-chip-chip-design-changed-by-apple/
Edited to show a callout from the article (Last year):
The M3 Ultra chip could be a standalone chip, similar to the M3 Max chip, which no longer features the UltraFusion interconnect
Last edited:
What's the power consumption of a 4090 again?
My kid is looking at a "low end" GPU for her PC which has a 500 W PSU. A Radeon 6600 pulls 132 watts by itself. The CPU is another 65 watts.
Apple has chosen to be all about portable. By the units sold numbers that is the right decision, but the workstations have been effectively abandoned.
I'm reserving judgement until actual benchmarks come out.
At this point I will just let M5 lineup shakeout. This feels like Apple just needed to do something and voila M3 Ultra...
Never mind that, how about Crysis?
In all seriousness though, unless developers release class A games on the Mac it won't matter how much GPU performance Apple adds to their chips, it will never be a serious contender for legit high-end gamers. They salivate over NVIDIA RTX boards, not unified RAM SOCs with no expandability
I disagree, Apple is making serious efforts and Control by Remedy drops the 12th, let alone Cyberpunk 2077 this year. I am not saying it is perfect yet but Apple and their SOC designs are getting noticed by devs.
The surprising point is that this situation was caused by not updating the Mac Studio last year. Had they updated it to M3 Max and M3 Ultra, they could have skipped the M4 this year if it won’t have an Ultra version. But somebody had the idea of skipping the update and now they pay for it. Anyway, in my case this has been good for me, as I can now get a M4 Max Studio, more powerful and with more RAM than if they had updated it last year.
Your first statement is a valid wish.
The value/validity of your second statement depends entirely on how much technical knowledge you have about chip architecture. Are you knowledgeable?
It's low-hanging fruit, I wouldn't say it's being noticed, it's just a way to wring money out of franchises that are tapering off- Apple is still not considered a day one release target and that is more telling here. This toe-dip is an easy win because it's very little risk. A big risk would be platform exclusivity for the Mac, and until that happens Apple as a platform vendor gets years-old releases that don't need cutting edge hardware. It's like releasing a game on Nintendo Switch, it's technically there but it's compromised in some respects and not quite the same experience. It's also years late when the world has moved on. Control was great for 2019, but they didn't release Alan Wake II. Cyberpunk is an amazing game, but CDPR is already working on the new Witcher (not targeting any Apple platforms for a day one release). There's literally nothing here but warmed up leftovers.
Props to the hardware team. Software team leadership sleeping. Apple too scattered w/ visionOS, tvOS, watchOS, iPadOS, iOS, macOS.
You can’t update something you don’t have. Apple has stated it takes a long time to produce these Ultra chips. Probably because they have to wait for fab lines to open up on the newer process nodes. It makes much more sense to start with smaller chips (A-series) when nodes are new and volume is low, because you can fit more on each wafer. Then work your way up to larger and large chips as production picks up.
If I had to guess what the issue was, is that TSMC didn’t want to dedicate too many fabs for N3B because Apple was the only client using it. So Apple didn’t have the capacity to produce the M3 Ultra until after [almost] everything was moved to M4. This is made more evident because Apple just released two other products that still make use of that older node; the A16 (iPad), M3 (iPad Air) and now the M3 Ultra in the Mac Studio. Seems they’re trying to squeeze what they can from that node to maximize costs.
Either way, if these benchmarks are true, it is a bit disappointing compared to what everyone was expecting, which would’ve been a beast of an SoC in a M4 Ultra.
I think the M3 Ultra is going to mainly appeal to the A.I. crowd with 512MB memory and 32 neural engine cores.
Last edited: