With their test information the M2 Ultra seems similar to a (or above a) 6900xt on a mac with the bonus of 128GB VRAM with just over 100W power draw. Comparing to Nvidia, from Apple’s benchmarks I think the internal GPU is better than a 3070, and closer to a 4070. Hopefully more tests come out with clearer information.
@@AchakBrooks in actual real life. Nvidia works much better. Even 3060 will perform better. 3d rendering and everything will test the gpus actual performance. Nvidia will eat up these M2 gpus.
@@aashishnakarmi1690 Apple exclusive techsites have been churning out continuously that the M2 Ultra is only 10% slower than an RTX 4080Ti, so in order to put that into test - that GPU would be the fitting comparison video even if we know Geekbench Metal Benchmarks isn't really gonna translate into real life performance figures.
You didn't have to get 8TB of SSD. That just destroys the price. You could easily have gotten by with just 2TB of storage and not hurt any of the tests you did. And that would have been so much cheaper.
Still 20 times slower than 3 years RTX 3090 in Machine learning and you use upto 8 GPU in workstation. 4 to 5 times slower than RTX 4090 in 3d workloads.
It's how Apple operates. The base model is often not-unreasonable priced. But move up a level or two and the prices become outrageous. For those who are buying a machine for performance-related work, it makes no sense to purchase the base model, as modern Macs can never be upgraded, at all. Customers are forced to purchase all the RAM and storage they think they may *ever* need.
@@HDRPCwell 3d works and gaming rtx will beats, for video editing, 8k and prores, music works, memory thing Ai it can also smokes rtx 4090 with 13900k with half the power the gpu itself consume without any crazy noise, and its super portable as mathew moniz shows the comprasion
@@busybee2148 🤣🤣🤣 his Premiere pro PC scores are 85% lower than same PC which pc world and tech notice has. Matthew got 979 scores in his PC but PC world got 1819 scores in the same RTX 4090 and 13900k. Tech notice got 1790 scores in pugetbench premiere pro.
The reason why the $1k price to get all the GPU cores in the Ultra is much higher than the $200 it takes to get all the cores on the Max is because the Ultra is not made by interconnecting two Max chips taken from anywhere on the wafer. The two Max chips need to be right next to each other on the wafer, where the interconnect between them is made lithographically during wafer fab. The odds that two Max chips which each have all their GPU cores working are right next to each other is low, and the $1k extra price accounts for that. It's much more likely that one or both of the adjoining Max chips that make up each Ultra have a few defective GPU cores, so Ultras with less than the full GPU core count are much cheaper. For the Max, the upgrade to get all GPU cores functioning is much cheaper because all you need is one fully functioning chip, not two of them next two each other.
Yep! If I recall correctly, it has to do with chip binning, which as far as I could tell is a way to not lose too much money and not waste too much materials when making computer chips. I've only heard about it being done with x86-based CPUs and SoCs but not with ARM-based SoCs until now.
I really like the apple silicon so far. M1 was a early "beta" product with limitations on link channels, displays and performance. M2 is a ready product that fixed all of those issues and addressed the gpu core limit that the M1 Ultra had. Can't wait to see what M3 has in stock for us.
I think I will just get a M2 Max 16 inch MacBook Pro plus a PS5 controller o do works and gaming instead of Air 15 Air plus a Steam Deck, just I not sure should I wait for M3 Max since it will have ray tracing hardware as it will be future proof more, but I know game is about the story, but ready spend so much so I do want to have it all, what you guys think? I not sure where I will move to later so a PS5 seem not ideal at the moment unless it is free for me, I also want to emulate other platform of games tho, in addition to the Death Stranding for Mac
How did m2 fix the display limitations? The standard m2 can still only output to 1 external display in laptops and 2 in the mini. Same as the M1. Same story for the M2 pro and max chips - 2 and 4 respectively for laptops. The M1 was incredible and so is the M2 but the M2 isn't revolutionary - it is just a higher clocked m1.
Yeah. I do seem to remember seeing some people I follow on social media continuing to use their x86-based Macs they already had and only making the jump straight to ARM with Apple M2.
I can see you being able to build a custom PC in this form factor, or close. But I can't see you being able to do that and match or beat the performance, and absolutely not be able to match or beat that performance within the same power envelope. This is where this architecture REALLY shines. You'd be pulling 500-600+W from the wall to get those kind of rendering numbers with an x86 system, probably more.
The PC wins... Unless you also constrain the PC to the size the studio has. Or the power consumption of the studio. The comparison is dumb because if you need what the PC brings to the table, the studio makes no sense. But if you use the studio's strengths for your workloads, and are fine with the limitations, I doubt you could go wrong with it. The amount of RAM it has, and the bandwidth in that envelope is already crazy, let alone the CPU performance for the power it sips.
@@lugaidster Disregarding the size aspect of it. The M2/CPU is generations ahead of AMD/Intel. The GPU is also comparable to MID range ones; only limited by the power consumption(due to the mini size) which limits the possibilities of graphics power! I mean if that thing can come close to 70% of a fully specked PC in GPU department then I can’t even imagine what kinda power the Mac Pro Tower will bring in the next 2-3years with all that redundant space!
@@rahbarkarim671 how exactly is it generations ahead if it can't beat last year's parts? The one thing it has going for it is size and power consumption. A benefit it only has against Intel parts. AMD CPUs, especially the laptop parts, are pretty efficient and powerful, to the point that if they were also released with memory on-package, I'd bet they'd win handily in both departments.
A more interesting comparison to see if there is scaling in performance would be to compare the M2 max (or M1 max) to the M2 Ultra and/or the M1 Ultra to the M2 Ultra. This would be more relevant for someone looking into buying the mac studio.
@@haderlumpi When you pay the electricity bill and you have an office full of the things. You'll care. Especially when you're then having to offset the heat output with a 3000W AC unit.
The depressing part for me is , i believed Gurman and all the rumor mill prognosticators that their would be no M2 Studio so i jumped on a maxed out M2 mini pro. Works good, but with my work flow gets hot fast to over 100c , fan goes to 5k rpm, cools down about 5c or so. Would much rather have bought the studio with the much better cooling for not that more than the maxed out mini pro. Oh well, so much for all supposedly all knowing inside rumor guys. 🙄
I'm very curious to see how the M2 Ultra performs in OctaneBench and the Redshift benchmark tool, especially with Apple's supposed toting of 6x performance boost over the M1 Ultra in Octane Render.
I had to log in to my account just to reply. This is the most important question on my mind as well, and could be the deciding factor to purchase one of these for me.
I would love to see a work load that takes advantage of all that RAM, especially the GPU workloads. If you can find something that maxes out the mini’s RAM to see how RAM limits can benefit from a big pool like the studio has
with that much you can actually load an entire game level into substance painter for texture work without having to export individual models. it's pretty nuts.
I'd love to see the performance comparison for World of Warcraft. It's natively compiled for Apple Silicon, runs decently on my M1 mini, but I'd like to know if moving up the line gains me anything.
It is far more cpu based if you turn shadows and water to low. The bigger bump is the m2 clock speeds and single core speeds. I’ve done extensive testing since the original m1 and owned every gpu and cpu variant funny enough until now I’ve tested and recorded wow performance at every step of the way. I’ve marked down to come back here and update with a link to the channel I post for with that info and will update you so you can see everything from air to current cpu and gpu combos after I get the multiple ultra variants in
Is the small performance delta because of Davinci or the x265 output? All other video encode tasks were x264, which likely has more hardware acceleration blocks that scale with core count.
Question regarding the Blender Classroom scene render test. Did you render that scene using the CPU or GPU? By default, the scene opens up to render using CPU. You have to manually switch it over to GPU. When setting the cycles render up for GPU rendering, my M1 Ultra Mac Studio with 128GB of unified memory, 64 GPU and 4TB SSD rendered this scene in 1 Minute and 43 seconds which is faster than the M2 Ultra, which makes me wonder if you were rendering using the CPU instead. Just curious to know how of a speed increase there is with the 76 core GPU vs the 64 core GPU that I have in my M1 Ultra.
Cant believe all your videos is shot on eos r with cropped 4k. Your video quality is up there with those guys that uses cinema cameras in their channel
Alright, let's cut to the chase. Windows 11? It's a nightmare. My beast of a PC with a 3080, 32GB RAM, and an Intel i9 12th is struggling. Premiere Pro stutters more than a broken record. And don't get me started on the daily crashes - it's like the Edge browser's favorite hobby. Then there's my Mac Mini M1. Every software I throw at it runs like a dream, even outpacing my high-end PC. That's why I'm trading up to a Mac Studio for video editing. At the end of the day, raw power means nothing if I'm pulling my hair out due to crashes and lag. Sure, I've got three beefy PCs at home perfect for gaming, but I'm done with Windows for work. I miss the good old Windows 7 days. If there was a way to put Mac OS on a PC legally, I'd have jumped ship ages ago. So yeah, I'm saying hello to a smoother workflow with the Mac Studio. It's about time!
I've got two Windows machines one is a Windows 10 PC with a 3900X, 32GB's of RAM, and a 3060Ti and the other is a laptop with Windows 11, 24GB's of RAM an 8 core 6900HS, and 6700S GPU....I like both but I miss my MacBook Pro for video editing.....I feel like your PC specs need to be 1.5x-2x higher on a PC than on a Mac :/
@@TheShawMustGoOn It was a response to "every software I throw at it [Mac Mini] runs like a dream". Stay focused. Power BI doesn't run on Macbook or Mac Mini, so his statement is false.
I was hoping you were going to mention gaming. I mean even though this unit is made for professional work, there are individuals gamers with deep pockets that would love to have something like this that they can transport from one place to another and play their games on. And I'm betting you that it would be fantastic. I would love to see that. Also, I would love to see you build something that is as fast or as efficient as that box is in a PC form that would be very interesting. As always great review. See you in the next one
I'll say this: the M2 Ultra would definitely be overkill for gaming. The M2 Max version of the Studio would probably be plenty, at least until there are more high-profile games that are Mac native.
@@ehtasam806 It's more that spending that much money won't get you a significant advantage. The M2 Max (or at least, the 60-core M2 Ultra) is most of the way there for much less!
@@kensg The laws of thermal dynamics are a thing. You forgot for a second there. Also this is Apple your talking about. They will sell you something like 3060ti performance for the price of the 4080 and laugh all the way to the bank. Not to mention most games will not even run on Apple/Arm and if they do performance is near 50%. It's really funny to read through the comments.
Super fun, looking forward to the rest of your testing. And yes, a hundred times yes, to building a custom PC close to the same size. This also highlights how f'n good the M2 Mini is for a lot of people.
Who the f realy needs to put an PC on the desk? An completly speced out PC is cheaper and has much more performance and you are even able to play games.
See how many assets/NPC you can spawn in something like GZDoom (which is available on MacOS). Or any sandboxgame. Or how many mods you can run in Cities Skylines. That would push it hard, since it's not ARM native, but apparantly adding mods can eat the RAM pretty well.
The real question is how well does this handle machine learning tasks? From what I've been reading... very, very, VERY well thanks to that insanely fast unified memory. 192GB of ultrafast VRAM and architecture that skips the need to move everything from CPU to cache to RAM to VRAM and instead goes directly to VRAM is just unheard of for a system that isn't sat in a server rack.
This mac m2 ultra is 4 to 5 times slower than RTX 4090 in 3d tasks and also in some specific 3d apps we have dlss 3 frame generation advantage. M2 ultra is 20 times slower in Machine learning and AI compared to 3 years old RTX 3090.
@@logirex m1 ultra 2022 model is already more than 2.4 times slower than 2018 titan RTX. Now we have Microsoft DirectML since 24th May 2023 to further boost performance by over 2 times in ML/AI for all PC GPUs specially for Intel Arc and Nvidia GPUs. 2018 titan RTX is 10 times slower than RTX 4090 in AI/ML. Thatswhy everyone is buying nvidia GPUs for AI/ML and Nvidia stocks are going up and now Nvidia is a 1 trillion dollar company. Intel Meteorlake in September 2023 will have faster igpu than m2 pro and even faster than rdna 3 780m. Meteorlake also has VPU which has faster ML/AI Performance than m2 pro. Meteorlake laptops will also provide better battery than m2 pro MacBook pro.
@@HDRPC why are you on an Nvidia shill tirade? Are they paying you? DirectML ain't gonna be faster than cuda for AI. Also, what are your numbers for this?
Honestly I don't know what you bean by the base M2 chip being enough for productivity and "some creative tasks". I've got an M2 MacBook Air and that thing trades blows with my Ryzen 7 3700X PC when compiling code and rendering - without a huge metal brick and 120mm fans strapped onto it.
The Mac Pros have moved over from Intel chips and use the same chips as the Mac Studio. The only reason to buy a Mac Pro is if you need to use PCI cards to interface with other hardware (audio equipment for example). The new Mac Pros are unable to expand memory like before and have the same 192GB RAM limit as the Mac Studio. The low limits of RAM in their computers, combined with the fact that it's shared with the GPU, caused me to move away from buying Macs. Their maximum amount of RAM in the current iMacs (16GB) is half of the maximum in some Intel versions. When I had an Intel based iMac I had originally bought it with 16GB but I noticed it slow down at times. I bought an additional 16GB and never had a problem again. I couldn't see going back to 16GB again when software requires more memory for all of the new features and bloat.
Can't wait for the Apple Mac Studio M3, Ultra Max Pro, 2mm thin, zero ports, Carbon neutral, GMO free. Oh, and why can't they put the damn power button in the front?
Please mention how the 3d tests are being conducted. There's many different ways of rendering projects, it would be very helpful to make actual comparisons with x86 CPUs and GPUs
One thing that always bothered me and not many reviewers stress about macs is severe lack of repair-ability and upgradability . IF you need more ram or better GPU in future , maybe because your workload increased or change in job , you are stuck with what you bought and have to buy a new MAC ! I haven't seen anyone being able to open the device without harming it , so even if there small problems like too much dust inside the device or need to apply new thermal paste , you can't ! Atleast macbooks can be opened and repaired to some extent .
Companies that buy these kind of machines don't upgrade them or repair them. They replace them. They spec them for what they need and will need (as best they can) for the lifetime of the machine then get rid and buy something new, rinse and repeat. The most they might do is put more RAM in, but even then that isn't all that common these days. Like I said, they spec for their task. They're workstations, they're often used for a single or a very narrow set of tasks and nothing else, and optimised for those tasks. It isn't worth the time and hassle to upgrade with the kind of projects you use this kind of system for to upgrade, because that time being upgraded is time not being used to make money. They'll be paying the person using these kind of systems a crazy amount too in most cases so want to get the most they can out of that persons time. You're coming at it from a single user perspective and not a business perspective. Where I work, they never upgrade or repair things, and we hardly do anything demanding with them. They just replace. It isn't worth it, even though they're just bog standard Dell PC's that can be opened and messed with easily. If something really needs to be repaired or upgraded, that's what support contracts are for. They'll get Apple to do it, or Dell in our case.
Why are the Blender rendering times so bad? My RTX 4080 laptop renders the Classroom scene in 11 SECONDS using GPU. Even my iMac Pro from 2017 renders that scene on GPU in 2 min 29 sec (after loading the render kernels, which happens only once after you open Blender). Did you enable Metal for scene rendering settings, or did you leave it at the default setting, which is CPU? Did you alter the (de)noise or sample settings for the scene?
how size a factor? "For the price and the size" for the creative market? Yeah most sound studios struggle to fit anything larger than a banjo for obvious reasons.... Hope U meant Price and Performance which was the spiel earlier in the video and the prev "ad" that u did using mac mini for studio creative's. Should have compared to similiar price PC for prober comparision imho.
If your workload can exploit the various chunks of application-specific hardware that accelerates certain operations, like media encoding, rendering etc, then the performance improvement will be impressive, as demonstrated in the Maya Arnold rendering benchmark. Triple the performance is great. If, however, the software you're using isn't optimised to make the best of the processor, you will only see mediocre improvements. Generally speaking though, the difference between one computer and another with double the processors will be mind-blowing. It's expected.
I've had a Mac Mini (m2 pro, 12/19/16 cores, 32GB, 1TB) sitting in my basket for ages now, at €2600-ish. The new Mac Studio (m2 max, 12/30/16 cores, 32GB, 1TB) is only €55 more expensive at the moment. Worth it? Or will the mac mini go down in price sometime soon?
@@wrighty338 a 10700k with a 3060ti or 3070 will outperform this, a 7600x with either of those will outperform this, I could go on. The cpu may not be as powerful overall but the gpu is easily 4-20 times more powerful and overall performance for almost anything will be better
@@wrighty338 if you want a pc that can outperform this on top of having a better cpu then you could spend 2000-2500 and get a 7900x with a 4070ti or a 13700k with a 4070to. Those will shit on this
@@hatchell18 Im not seeing any links to your benchmark examples of where these "shit on this" I have both a PC (£2000) and a mac mini M2 (£1049) i want to see numbers. You are talking all the talk and walking non of the walk. Also 10th gen intel and previous 30 series Nvidia are not current components so buying these isn't exactly a walk in the park
Can we please talk more about Mac mini's power consumption? What the fuck 50 watts? My PC needs a 750 watt power supply and probably needs something like 400-600 watts and oh it's fucking slower... what the fuck.
3:07 the $1000 difference between the 60-core and 76-core variant is somewhat justified because they will need to "glue" two full M2 Max together. Apple charges $200 if you go from the 30-core M2 Max to the 36-core M2 Max so they should at least charge $400 extra for the 60-core to 76-core upgrade. The yielding rate for the full M2 Max is probably not that high and if you need two of them, that's exponentially harder. Also, the "glue" process probably has some failures as well. So they charge you $600 extra for binning the chip and the "glue".
@@lugaidster that's true because Apple does not really have competitors if their enterprise consumers just want to use Macs. With that said, if you take a look at pricing structures from the other companies such as Intel, their enterprise grade Sapphire Rapids charges $5171.00 for the 52-core 8471N and $17,000 for the 60-core 8490H. This is CPU cores and their manufacturing process is not the same but the same principle applies here. You want to change extremely high amount of money for the full chip because they are hard to make and you can take the profit for chip R&D, fab improvements (or extra money for TSMC in Apple's case), etc. So we should give our thanks to folks who brought the full M1 Max, M2 Max, and Max2 Ultra so that Apple can produce more chips like this! (At the same time, I will happily use my binned M1 Max for a few more years)
@@lugaidster you have no idea what the cost is and no idea what the yield rate is. I am pretty sure these massive chips with no faulty cores are extremely rare and have super low yield rates, which would make them multiple times more expensive to produce than 60 core versions with bunch of dead cores. I’ d guess the yield on full M2 ultra is 4 or 5 times lower So the question is not if the process of producing one costs 1000usd, which it most likely doesn’t. But I am pretty sure the final margin on the 76 core version is multiple times lower than on the 60 core if you count in the yield
The Gen 2 USB port thing is weird. With X86 manufacturers it happens as well (in laptops.) Alienware, all the USB A ports are gen 1 on all of their top of the line laptops. Razer OTOH, all Gen 2 or thunderbolt. This is the way it should be. I am really surprised with apple here, although I'd likely be using a Tbolt Hub with it anyway and they have tons of Gen 2 ports.
Not my favorite of your reviews. It would have been better to show how much PC can be had for Apple's breath taking asking prices, and how a PC that expensive (Threadripper, Nvidia 4090, 256GB RAM...) would (likely) beat these Apple products in all, or nearly all areas. It's possible for otherwise-good products to be bad products simply because they are overpriced.
Ok but how would you install MacOS on the PC? Reviews should be fair right and not skewed to favor one side, right? Or would it be fair to say that this Mac is not the equivalent to a PC and the comparison not correct?
That's not the point of this video. He wants to compare how much performance you get from mini vs studio fiven the price difference. This is not an Apple to orange comparison
@@o0Donuts0o Most care more about how well a machine can perform a given task, as opposed to a militant brand allegiance. Should car reviewers never compare BMW automobiles to Toyota, because BMW's i-drive operating system is not available on Toyota's vehicles? Macs and PCs are both computers that perform largely identical tasks. They should absolutely be compared. And if a buyer is willing to pay more for a given feature, say Mac OSX, then it should be explained to the consumer that the privilege of OSX comes at 2 or more times the price for the same level of performance that a PC delivers.
@@mluu510 It's not apples to oranges. It's computer to computer. Macs and PCs perform nearly identical tasks. Believe it is incumbent on reviewers to mention more reasonably priced alternatives evaluating products with such breathtakingly high prices.
I'd like to see a video about the case study of the M1 family and M2 family soc's GPU performance scaling (FPS/core) or. (performance/core). Since M1 had that as a bug in the soc design, and I waited LTT to release a review discussing that, but gor a very lazily slapped together rant of the marketing..
You are painting the Mac Studio as this super expensive computer but the base model with M2 Max compared to the Mac Mini M2 Pro with more GPU cores and similar RAM and storage is priced very close to it.
So I'm so confused when techtubers say "this chassis is only x.x liters." Like that really depends on the container. It could be as tall or as wide to accommodate those liters. A US gallon for example has 3.78 liters and it doesn't look anything like a pc case. What happened to WHD?
What is the difference in peformance between the Mac Studio and Mac Pro. They both use the M2 Ultra and have the same max memory of 192GB. Is there a difference in clock speed, or is the main difference the number of ports and the presence of PCIe slots in the Pro? Can we expect better benchmarks from the Pro?
I still do not understand why nobody is talking about that damn flickering plaguing mac ´s since M1. This problem still persists even on M2s. Benq has released even note that they do not have any solution for this problem and it is problem that Apple has to resolve. I went through 3 LCD displays until enough was enough and I have left Apple for custom build PC.
You sounded like you wanted to cry every time you mentioned the price of the M2 Mac Studio Ultra. My wallet and bank account empathizes with your financial pain. The M2 Mac Studio Ultra is only for people running a full-on production company. Otherwise use the M2 Mac Mini or M2 Pro Mac Mini for individual needs or jobs.
I would expect nothing less for a system that has like 176Gb more unified memory. We need to remember the unified memory part here and understand the amount of that memory is used by the - how many more cores - GPU. I feel it is impossible to draw any conclusions here because this is a David and Goliath comparison. Both are computers and both are made by Apple, the memory bandwidth alone will account for a huge chunk of the improvements. I would expect nothing less from the Studio, in fact I had hoped for even better numbers.
Future calling.. you’ll also be surprised by the difference between M2 & M3, M3 & M4.. etc I mean what’s the point of these videos when technology keeps improving and the previous video becomes ironically wasteful and completely absent of insightful content, plus the self awareness of such developments to come year after year.. Comparing the old to the new and keep leveling up to create what actually? Is there any real purpose of just spending on a faster more efficient machine to only create videos that don’t require such power .. just a thought.
