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.
Mac Studio Performance - Disappointing News
- Thread starter CodeSpyder
- Start date
- Sort by reaction score
At this point, I want to know clearly whether this problem is due to a software issue or unfixable problem.
For everyone who’s experienced all of Macs other major transitions: 680xx to PPC, OS9 to OS X, PPC to Intel, this is just history repeating, though honestly a lot smoother than the previous ones. Real headaches those were. There will continue to be anomalies and bumpy rides for another generation of M processors as Apple and Devs adjust.
This necessarily means, for now, the most powerful models will be less powerful in real life than on paper. But they will still be the most powerful in the line for those that absolutely need the best. But, for sure they will not be a value prop for anyone who doesn’t have the work load to run them at 100% day in, day out.
This necessarily means, for now, the most powerful models will be less powerful in real life than on paper. But they will still be the most powerful in the line for those that absolutely need the best. But, for sure they will not be a value prop for anyone who doesn’t have the work load to run them at 100% day in, day out.
The M1 Max chips were made with a 5nm process. Two M1 Max chips were 'fused' to get the M1 Ultra. I wonder if Apple will combine two M1 Max chip designs into one complete Ultra as the process goes to 4nm and then 3nm.
Glad to hear from someone who has lived thru all the Apple transitions - I am in the same age generation when I had my first Apple IIe, then G3/G4/G5, then Intel (32 to 64 bit) and now currently M1.
Agree that the M1 series are the best transition for Apple.
I don't take YT video reviewers much credence as they are motivated for click bait videos that don't really look at real world use.
Love my Mac Studio + Display Studio for the past 12 great days of use!
I am now learning how to use FCP, Logic and Compressor I bought for $199....
Well Apple could reduce the SKUs to just three by only shipping the "best" version of M1 Pro, M1 Max and M1 Ultra, but that would raise base and BTO upgrade prices significantly because they would be throwing away a significant portion of their production run that they are now making revenue from via offering the binned chips as additional SKUs.
The reason encoding a single video to H.265 doesn’t show different times between the Max and Ultras is because H.265 encoding doesn’t use the CPU cores, it uses the H.265 encoders built into the M1 architecture (like Intel QuickSync). So they’re just testing the speed of the H.265 encoders which are the same across the Max / Ultra chips.
Scaling an 8K video down to 4K should use the GPU some, but the GPU only needs to downscale video as fast as the H.265 encoder can handle. It seems to me that the H.265 encoder would be the bottleneck in that task. (also maybe the storage if their project is on an external drive.) The GPU power draw would naturally be lower when it’s not running at full load. I love these guys but I think they got it backwards, it’s not the wattage that’s keeping the GPU from running faster, it’s that the GPU doesn’t need to run faster and the wattage simply reflects that.
If they really want to see how all the CPU and GPU cores scale, they should render a video to a more exotic codec, say DNxHR, which doesn’t have dedicated encoding silicon on the M1. (edit: they did with RED RAW which was faster on Ultra.) I work in documentary film editing and personally I’ve never once rendered a video to H.265 format for my job, so I think this example is a fairly niche YouTuber workflow, that being YouTubers they have fixated on a bit too much and blown out of proportion. No need for conspiracy theories here.
Personally when I saw all the other benchmark charts in this video, it confirmed my choice of the M1 Ultra 48 as significantly better than the Max and the best for my needs. If anyone else saw these charts and thought they’re better off with the Max… then maybe you’re right? Feel free to cancel your Ultra orders and maybe mine will come sooner!
Scaling an 8K video down to 4K should use the GPU some, but the GPU only needs to downscale video as fast as the H.265 encoder can handle. It seems to me that the H.265 encoder would be the bottleneck in that task. (also maybe the storage if their project is on an external drive.) The GPU power draw would naturally be lower when it’s not running at full load. I love these guys but I think they got it backwards, it’s not the wattage that’s keeping the GPU from running faster, it’s that the GPU doesn’t need to run faster and the wattage simply reflects that.
If they really want to see how all the CPU and GPU cores scale, they should render a video to a more exotic codec, say DNxHR, which doesn’t have dedicated encoding silicon on the M1. (edit: they did with RED RAW which was faster on Ultra.) I work in documentary film editing and personally I’ve never once rendered a video to H.265 format for my job, so I think this example is a fairly niche YouTuber workflow, that being YouTubers they have fixated on a bit too much and blown out of proportion. No need for conspiracy theories here.
Personally when I saw all the other benchmark charts in this video, it confirmed my choice of the M1 Ultra 48 as significantly better than the Max and the best for my needs. If anyone else saw these charts and thought they’re better off with the Max… then maybe you’re right? Feel free to cancel your Ultra orders and maybe mine will come sooner!
Last edited:
But the down-clocking on 64 cores vs 48 cores vs 24 cores is a question that needs answer. This effectively limits the Ultra’s potential (that is, doubling what an M1 Max could do). Both the PSU and thermal envelopes look like they are not pushed to the point that down clocking is necessary yet.
If the Ultra is never going to get to a certain expected efficiency then the price difference above the Max is going to be less justified for many people.
If the Ultra is never going to get to a certain expected efficiency then the price difference above the Max is going to be less justified for many people.
YouTuber's need to make a PayCheck it is always the same first week how great it is then the next week they all try to beat each other posting video's of how bad it is rinse and repeat.
Only trust DaveD
It doesn't make sense to have that much cooling if they're going to throttle down the system to insure the fans run at idle. This makes me lean towards the belief that they need to optimize to get better performance out of it.
I agree that software optimizations could probably unlock even better performance on the M1 Ultra, for certain tasks. Adobe Premiere 2022 for example has shown a LOT of improvement on M1 once Adobe FINALLY buckled down and tried to make better use of Mac hardware compared to their Nvidia-optimized Windows software. Even Apple hasn’t released updates to their own software yet. As this is the most powerful M1 chip yet, there is probably a lot of room for software developers to rearchitect and scale their code up to take full advantage of the additional power. I don’t think a lot of the software Max Tech tested is particularly well-optimized for M1 — Logic, FCP and Resolve are probably on the better side.
In Intel CPUs, Nvidia and AMD GPUs, “diminishing returns” is also at play. Once a GPU is nearing peak performance you can make it draw 50W extra power and only get an extra 5FPS game performance.
It seems like Max Tech is comparing the M1 Ultra against some imaginary computer than can perfectly scale all tasks infinitely. I totally get why someone spending an extra $1,000 on the 64 GPU core Ultra could be disappointed if it’s not “that much” faster on a lot of tasks. But I think the reality is that software rarely scales linearly, many software tasks can’t take advantage of many more cores due to their very nature. I think 8-10 CPU cores is about the most that many common software tasks can really take advantage of, for example the Photoshop tests they ran.
Since it’s been ten years since Apple sold a dual-CPU computer, maybe people don’t know what to expect? My 12-core dual-CPU Xeon Mac Pro 2009 isn’t necessarily 2X faster at everything than a 6-core Xeon Mac Pro 2009, and a 20-core M1 Ultra isn’t going to be 2X faster at everything than a 10-core Max. But in both cases doubling the cores does make certain heavy-duty well-threaded tasks run a lot faster, and whether it’s 2X or not, a lot faster is still a great thing. More cores also allows for more smoothly running several tasks simultaneously, I don’t think Max Tech showed any multitasking tests. You could dedicate some cores to a virtual machine and have plenty of power left over for the host OS. Those are more edge case scenarios, for sure, I do not think that every user needs or could benefit from an Ultra, the Max is great for many people!
If the GPU is bottlenecked by the H.265 encoder, a 64-core GPU wouldn’t need to work as hard to keep up with the encoder, as a 24-core GPU would. Makes perfect sense to me!
I wouldn’t agree that the Studio’s fans run at a pretty constant speed because Apple has limited the Ultra’s performance in order to keep the fans quiet. While they do try to make quiet products, in the recent past they’ve certainly let their computers’ fans get loud when they need to. Ask my i9 MacBook Pro 2019! I think the fans have a slightly high floor due to the large internal power supply, hot components like PCIe 4.0 NVMe modules, 10Gbe ethernet chipset, etc. that need constant cooling, and beyond that the efficiency of the M1 architecture plus the massive Mac Pro-level heatsinks they’re using don’t require the fans to spin much faster under load. I think the Mac Pro 2019 fans run at fairly constant speeds, and I don’t think people accuse Apple of limiting the Xeon’s performance for the sake of the fans. Thermal headroom much appreciated post Mac Pro 2013 debacle.
In Intel CPUs, Nvidia and AMD GPUs, “diminishing returns” is also at play. Once a GPU is nearing peak performance you can make it draw 50W extra power and only get an extra 5FPS game performance.
It seems like Max Tech is comparing the M1 Ultra against some imaginary computer than can perfectly scale all tasks infinitely. I totally get why someone spending an extra $1,000 on the 64 GPU core Ultra could be disappointed if it’s not “that much” faster on a lot of tasks. But I think the reality is that software rarely scales linearly, many software tasks can’t take advantage of many more cores due to their very nature. I think 8-10 CPU cores is about the most that many common software tasks can really take advantage of, for example the Photoshop tests they ran.
Since it’s been ten years since Apple sold a dual-CPU computer, maybe people don’t know what to expect? My 12-core dual-CPU Xeon Mac Pro 2009 isn’t necessarily 2X faster at everything than a 6-core Xeon Mac Pro 2009, and a 20-core M1 Ultra isn’t going to be 2X faster at everything than a 10-core Max. But in both cases doubling the cores does make certain heavy-duty well-threaded tasks run a lot faster, and whether it’s 2X or not, a lot faster is still a great thing. More cores also allows for more smoothly running several tasks simultaneously, I don’t think Max Tech showed any multitasking tests. You could dedicate some cores to a virtual machine and have plenty of power left over for the host OS. Those are more edge case scenarios, for sure, I do not think that every user needs or could benefit from an Ultra, the Max is great for many people!
If the GPU is bottlenecked by the H.265 encoder, a 64-core GPU wouldn’t need to work as hard to keep up with the encoder, as a 24-core GPU would. Makes perfect sense to me!
I wouldn’t agree that the Studio’s fans run at a pretty constant speed because Apple has limited the Ultra’s performance in order to keep the fans quiet. While they do try to make quiet products, in the recent past they’ve certainly let their computers’ fans get loud when they need to. Ask my i9 MacBook Pro 2019! I think the fans have a slightly high floor due to the large internal power supply, hot components like PCIe 4.0 NVMe modules, 10Gbe ethernet chipset, etc. that need constant cooling, and beyond that the efficiency of the M1 architecture plus the massive Mac Pro-level heatsinks they’re using don’t require the fans to spin much faster under load. I think the Mac Pro 2019 fans run at fairly constant speeds, and I don’t think people accuse Apple of limiting the Xeon’s performance for the sake of the fans. Thermal headroom much appreciated post Mac Pro 2013 debacle.
Last edited:
Watching this review, Ultra version suddenly really doesn't seem all that appealing.
Of course, depends a lot on what kind of programs you plan to run.
Processors are quite fast, however, for anyone needing fast graphic performance, this still can't really compete with dedicated graphic cards. Out of curiosity I ran Geekbench Open CL test on a 5 year old iMac configuration and checked result for M1 Max (second most powerful Apple chip atm); it turns out, graphic performance increase is about 30%. Against a 5 year old Radeon 580.
Ultra gives you double the performance of this old Radeon, but still not even close to what today's powerfull graphic cards are delivering, and at a very hefty price tag. Above all, using eGPU is not possible with these new Macs.
Don't get me wrong, very impressive for integrated graphics, but for a pro machine at this price... dunno.
Of course, depends a lot on what kind of programs you plan to run.
Processors are quite fast, however, for anyone needing fast graphic performance, this still can't really compete with dedicated graphic cards. Out of curiosity I ran Geekbench Open CL test on a 5 year old iMac configuration and checked result for M1 Max (second most powerful Apple chip atm); it turns out, graphic performance increase is about 30%. Against a 5 year old Radeon 580.
Ultra gives you double the performance of this old Radeon, but still not even close to what today's powerfull graphic cards are delivering, and at a very hefty price tag. Above all, using eGPU is not possible with these new Macs.
Don't get me wrong, very impressive for integrated graphics, but for a pro machine at this price... dunno.
I think your points are mostly accurate and well-made, except for the characterization of web content / YouTubers as a niche workflow. However you feel about most of the content, I suspect that influencers/streamers/new-media personalities now make up a significant part of all video production volume, and an even larger part of the target market for the Studio.
I'm in Audio. Base Ultra is going to be simply amazing.
Unless you need 128GB RAM…Maybe disappointing if you already bought an Ultra. But....maybe good news if you want to buy a Studio. No need to buy the more expensive option of an Ultra then
Apple does a lot of weird things, but it would be really strange to have these giant coolers (even going to the extra expense of double sided heat pipes and copper in the Ultra) and not take advantage of them. Definitely points to the throttling and thermal optimization for this being in the early days.
I find the hypothesis in the video(s) (e.g. "Apple is throttling power") rather unlikely, or at least it puts the cart before the horse: Writing parallelized software as my day job, I can tell you that doing that is a difficult task (and that difficulty increases the more cores you have to keep busy with *useful* work). Also see Amdahl's Law for example.
So, instead of power-gating resulting in low(er) performance, it might as well be not perfectly parallelized software that leaves units idle and therefore yields lower power *demand*.
So, instead of power-gating resulting in low(er) performance, it might as well be not perfectly parallelized software that leaves units idle and therefore yields lower power *demand*.
I'm surprised that some people take Max Tech seriously. By his own account, Max Yuryev is pumping out about six videos a week. How much rigour and thought do you figure goes into flooding YouTube like that? I think that his shtick is to spin the narrative to fit whatever he thinks his audience wants to hear, and that this is all about his non-stop pitch to his "followers" to get him to 1 million subscribers.
Thanks, but I'll take Marques Brownle and Dave Lee*, who of course were the unstated targets of Yuryev's recent attack on "other YouTube reviewers". He got what he wanted, wall to wall comments about how Max, uniquely, "tells it like it is"
If I was considering a Mac Studio Ultra, I sure wouldn't be making the decision based on the videos that Max Yuryev is pumping out almost daily.
*Not to mention John Gruber, the gang at AnandTech and a few others.
Thanks, but I'll take Marques Brownle and Dave Lee*, who of course were the unstated targets of Yuryev's recent attack on "other YouTube reviewers". He got what he wanted, wall to wall comments about how Max, uniquely, "tells it like it is"
If I was considering a Mac Studio Ultra, I sure wouldn't be making the decision based on the videos that Max Yuryev is pumping out almost daily.
*Not to mention John Gruber, the gang at AnandTech and a few others.
Last edited:
Other YouTube reviewers have a very different focus, mainly on production values and slick takes. Max actually bothered to tear down the Mac Studio, FFS. You may be tired of the comparison videos that compare every product six ways to Sunday, but for someone who is looking to pick a product, these end up being really useful.I'm surprised that some people take Max Tech seriously. By his own account, Max Yuryev is pumping out about six videos a week. How much rigour and thought do you figure goes into flooding YouTube like that? I think that his shtick is to spin the narrative to fit whatever he thinks his audience wants to hear, and that this is all about his non-stop pitch to his "followers" to get him to 1 million subscribers.
Thanks, but I'll take Marques Brownle and Dave Lee*, who of course were the unstated targets of Yuryev's recent attack on "other YouTube reviewers". He got what he wanted, wall to wall comments about how Max, uniquely, "tells it like it is"
If I was considering a Mac Studio Ultra, I sure wouldn't be making the decision based on the videos that Max Yuryev is pumping out almost daily.
*Not to mention John Gruber, the gang at AnandTech and a few others.
Yeah, he tore it down and put it back together himself. Kind of makes me wonder about how that might have affected the “specs” testing, which is superficial and suspect at best.
Yuryev, and Luke Miani, the other guy who's into the Apple Fan Boy/Bad Boy shtick, both did a teardown. They wanted to beat iFixIt and knew that their "scoop" had already been revealed on this forum on March 9th (not that either of them acknowledged that). Miani, at least, knows enough about electronics to take the computer apart without looking like a five year old ripping apart the wrapping on a Christmas present.
Next day, Louis Rossmann and ARS Technica's Andrew Cunningham, who actually do know what they're talking about, pulled the rug out from under both of them. Yuryev has since made a video in which he "clarifies" what he said during the teardown video
I check out Yuryev's videos occasionally because it's so obvious what he's doing and it's interesting watching people buy it hook, line and sinker. I've been aware of him and his "style" for a long time, going back to when he was trying to make it as a motorcycle vlogger, got into tech and then of course had a run-in with a mysterious force at YouTube that allegedly had it in for him for no reason.
Meanwhile, he boasts himself that he's flooding YouTube with about six videos a week. Apparently none of his followers think that this could possibly have an impact on quality. Did you catch the recent video that he went ahead with despite basic, glaring sound problems? Even Yuryev acolytes told him to pull it and fix it. And people are supposed to take him seriously?
Last edited:
So you’re crapping on Max because he uploaded a tear down one day after ifixit did and because he uploads almost daily on YouTube? Calling people who enjoy his content “acolytes”?Yuryev, and Luke Miani, the other guy who's into the Apple Fan Boy/Bad Boy shtick, both did a teardown. They wanted to beat iFixIt and knew that their "scoop" had already been revealed on this forum on March 9th (not that either of them acknowledged that). Miani, at least, knows enough about electronics to take the computer apart without looking like a five year old ripping apart the wrapping on a Christmas present.
Next day, Louis Rossmann and ARS Technica's Andrew Cunningham, who actually do know what they're talking about, pulled the rug out from under both of them. Yuryev has since made a video in which he "clarifies" what he said during the teardown video
I check out Yuryev's videos occasionally because it's so obvious what he's doing and it's interesting watching people buy it hook, line and sinker. I've been aware of him and his "style" for a long time, going back to when he was trying to make it as a motorcycle vlogger, got into tech and then of course had a run-in with a mysterious force at YouTube that allegedly had it in for him for no reason.
Meanwhile, he boasts himself that he's flooding YouTube with about six videos a week. Apparently none of his followers think that this could possibly have an impact on quality. Did you catch the recent video that he went ahead with despite basic, glaring sound problems? Even Yuryev acolytes told him to pull it and fix it. And people are supposed to take him seriously?
I don’t know what your beef is with him, but give it a rest. There’s nothing wrong here.
Read my post again. I said that Yuryev's and Miani's motivation was to beat iFixIt to a teardown, which they did.
That aside, my post speaks for itself, and certainly better than your attempted spin on it
There are plenty of people in this thread who are clearly Yuryev fans. Presumably it's permissible for someone to express a different view.
Last edited: