Apple's Graphics Performance Claims Proven Exaggerated by Mac Studio Reviews

[Garrus]

We aren't talking about GTX 3070, dear child. Nor was Apple. Do try to keep up.

This is why there's no point in talking to anyone on the internet. I gave you useful data showing the M1 Ultra crushing the RTX 3070, which means it is only a bit slower (5-15 percent slower) than the RTX 3090. It's to counteract a one sided interpretation you'll receive by only comparing 5 year old ported software to x86 Macs.

Ultimately though, it is Apple's fault for setting incorrect expectations. Nothing is stopping them from providing a reviewer's guide with the benchmarks for people to independently explain and verify. They didn't. That's not good. Clearly Apple is doing some bad marketing here. But it is also not true that somehow it isn't in the RTX 3080's ballpark.

Apple's original slides were against a TDP limited RTX 3090 (300W or so) and a TDP limited 12900K (160W or so). In that context it might be fairly accurate with properly written software. It is up to Apple to build a gaming console or through some other means get properly written software on the Mac.

 

[DeathtoGnomes]

relying on the reality distortion field

Lets create a value to benchmark the Reality Distortion Field on a scale of 1 to 10. Reviewers could use this to compare PR and leaked PRs to what they are reviewing.

In this case would it safe to say Apples white lie is a 6, or better?, rating on the RDF?

 

[samum]

This is why there's no point in talking to anyone on the internet. I gave you useful data showing the M1 Ultra crushing the RTX 3070, which means it is only a bit slower (5-15 percent slower) than the RTX 3090. It's to counteract a one sided interpretation you'll receive by only comparing 5 year old ported software to x86 Macs.

Ultimately though, it is Apple's fault for setting incorrect expectations. Nothing is stopping them from providing a reviewer's guide with the benchmarks for people to independently explain and verify. They didn't. That's not good. Clearly Apple is doing some bad marketing here. But it is also not true that somehow it isn't in the RTX 3080's ballpark.

Apple's original slides were against a TDP limited RTX 3080 (300W or so) and a TDP limited 12900K (160W or so). In that context it might be fairly accurate with properly written software. It is up to Apple to build a gaming console or through some other means get properly written software on the Mac.

RTX 3090 scores ~50,000 in Wild Life Extreme. Thats 40% faster than the M1 Ultra.

3DMark Wild Life Extreme Hall of Fame

The 3DMark.com Overclocking Hall of Fame is the official home of 3DMark world record scores.

www.3dmark.com

 

[mama]

I am hard pressed to find a company that does not cherry pick data for marketing. In this case, Apple claimed that their GPU is faster than the RTX 3090, and it is indeed the case, only not for gaming. At least on their site below, nowhere did they say that it offers 2x gaming performance over the RTX 3090. Reviewers seems very fixated at the 2x RTX 3090 claim and take it as 2x gaming performance. To them, GPU = gaming, but is It just about gaming since Apple don’t really market Macs as gaming machines?

Apple unveils M1 Ultra, the world’s most powerful chip for a personal computer

Apple today announced M1 Ultra, the next giant leap for Apple silicon and the Mac.

www.apple.com

Again, I will wait for the full independent reviews on performance in production workloads which seems to be the target market. My point is it is not helpful or instructive or honest to make generalised hyperbolic claims on performance. This is a pattern with Apple who have made no secret about wanting to make inroads into the profitable 'gaming' market but have thus far failed to a large extent. The generalisation of the performance claims is more sinister in that context.

I am not an Apple basher. I have used their desktop products for most of my life. What I have an issue with is the trajectory of the honesty in marketing with Apple. Why the vague graphs in presentations? Why compare to a 3090? Just why?

 

[Minus Infinity]

Colour me surprised .... said no one ever. I think they were comparing a 3090 in 2D idle mode to their M! GPU at full tilt.

Look I don't care what brand makes what claims because I always assume it's BS until proven otherwise. Most sites suck because they basically just post the marketing brochure without any questions asked. Good to see this site not just lapping up the coolaid.

 

[Deleted member 24505]

Who cares anyway, its a MAC

 

[Sybaris_Caesar]

I thought 3D performance was mostly benchmarks and video editing/rendering tools where GPU performance is also shown in fps. Did apple actually say it was game fps? Granted I didn't watch their Apple Direct or whatever it's called.

 

[watzupken]

The internet is making up their mind based on Apple's marketing statement, not edge use cases and "if only they used mac optimised software".

When you said edge cases, that may be the case for Windows. But if you are looking at Macs, I actually think gaming is more like the edge cases. If I were to get a Mac, I won’t be expecting to use it for gaming primarily. In fact, I will be glad the game runs.

 

[lexluthermiester]

Did you even bother reading my post?

I did. Made me smile. Yes, yes.

Apple's inflated performance claims are getting old and tired and the company really needs to do better.

Right? Let's be truthful, Apple lied, it's that simple.

 

[user556]

I'm no Mac fan, but yeah, anyone comparing games, or game like benchmarks, has got it wrong right there.

 

[TheLostSwede]

I thought 3D performance was mostly benchmarks and video editing/rendering tools where GPU performance is also shown in fps. Did apple actually say it was game fps? Granted I didn't watch their Apple Direct or whatever it's called.

Unfortunately, they're not doing well in the Geekbench 3D tests either, in fact even using Metal, they're about half the speed of the 3090 using OpenCL (which is a bit slower again on the Mac).

When a company makes a big, bold claim like this, they need to be a bit more specific, as their presentation suggests all GPU workloads, which clearly isn't true.

I'm no Mac fan, but yeah, anyone comparing games, or game like benchmarks, has got it wrong right there.

See above, they're actually doing worse in other GPU benchmarks.

 

[Garrus]

RTX 3090 scores ~50,000 in Wild Life Extreme. Thats 40% faster than the M1 Ultra.

3DMark Wild Life Extreme Hall of Fame

The 3DMark.com Overclocking Hall of Fame is the official home of 3DMark world record scores.

www.3dmark.com

Well you need to be a bit careful there. That's not a 320W RTX 3090 like in Apple's slides, that is a "hall of fame" unlocked result. Probably the 3090 running at 500W. Point is, it isn't that much slower at 3DMark Wildlife, and personally I'd take that as a decent benchmark, not Geekbench (which btw they themselves said that Geekbench results for Apple silicon are not valid as the test is completed too quickly). Clearly the 3090 is faster though.

IF you want to bench, we need to bench World of Warcraft, Total War Rome Remastered, Myst, Tunic, 3DMark Wildlife, or Baldur's Gate 3. Those are the only major native games on Mac Silicon (hard to believe I know, but gaming on the Mac is quite dead).

Tunic is great! Been playing it on my Xbox all day!

Apple deserves the criticism. They have native games. They could have put them in their reviewer's guide. But they prefer misleading TDP limited "industry standard compute benchmarks" over actual truth. Apple is at fault. But it is not true that the M1 Ultra is as slow as people are suggesting. It's about RTX 3080 level. Beats the 3070 easily.

 

[lexluthermiester]

Well you need to be a bit careful there. That's not a 320W RTX 3090 like in Apple's slides, that is a "hall of fame" unlocked result. Probably the 3090 running at 500W. Point is, it isn't that much slower at 3DMark Wildlife, and personally I'd take that as a decent benchmark, not Geekbench (which btw they themselves said that Geekbench results for Apple silicon are not valid as the test is completed too quickly). Clearly the 3090 is faster though.

IF you want to bench, we need to bench World of Warcraft, Total War Rome Remastered, Myst, Tunic, 3DMark Wildlife, or Baldur's Gate 3. Those are the only major native games on Mac Silicon (hard to believe I know, but gaming on the Mac is quite dead).

Tunic is great! Been playing it on my Xbox all day!

Apple deserves the criticism. They have native games. They could have put them in their reviewer's guide. But they prefer misleading TDP limited "industry standard compute benchmarks" over actual truth. Apple is at fault. But it is not true that the M1 Ultra is as slow as people are suggesting. It's about RTX 3080 level. Beats the 3070 easily.

You are missing the point others are trying to make. The MacPro is not a gaming system. Few who buy it will be gaming as a primary activity. MacPro's are productivity, power-user systems. In that context, Apple's claim is patently and deliberately false.

 

[Logoffon]

Is there any reason for testing SoTR only on M1U and not on RTX 3090 at 4K?

 

[TheLostSwede]

Is there any reason for testing SoTR only on M1U and not on RTX 3090 at 4K?

You'll have to ask the Verge about that.

I'll admit that none of the reviews so far, are especially techy and there's also a shortage of "proper" benchmarks for MacOS. As such, this has to be taken for what it is, but my point is still the same, companies need to stop making up silly comparisons in their presentations. The Mac Studio will sell to the target audience without silly comparisons.

 

[lexluthermiester]

Is there any reason for testing SoTR only on M1U and not on RTX 3090 at 4K?

It seems you need to take a closer look at that graph...

You'll have to ask the Verge about that.

Good luck with that one.. LOL!

 

[Garrus]

You are missing the point others are trying to make. The MacPro is not a gaming system. Few who buy it will be gaming as a primary activity. MacPro's are productivity, power-user systems. In that context, Apple's claim is patently and deliberately false.

We're not arguing over whether or not something is a gaming system. We are talking about how fast the GPU is. And I can guarantee you the GPU in the M1 Ultra is about 60 percent faster than the 3060 Ti. That is a good result. That is a fast GPU. It might even be faster than any Intel GPU about to release. It might not be an RTX 3090 but the "reviewers" are not technical and don't know what they are doing. Ars Technica is the only one that even did a half decent job. We need World of Warcraft, Baldur's Gate 3, Myst and Runic, and Total War Rome Remastered benchmarks. 3DMark Wildlife is also a pretty good choice (Ars used it) and it showed the M1 Ultra as ~60 percent faster than a 3060 Ti. That's fast. Not 75 percent faster like the RTX 3090, but still fast.

Apple needs a portable gaming device or console, as energy efficiency is key there.

 

[lexluthermiester]

We are talking about how fast the GPU is.

No, that is where you missed the point. The M1Ultra is a glorified SOC(SystemOnChip). The CPU cores and GPU cores are unified on the same die. Because of that and how they're so tightly integrated, we can not compare just the GPU or the CPU by themselves easily. For now, they can only be compared on a whole-system context. That will likely change as benchmarking software devs learn the platform particulars.

3DMark Wildlife is also a pretty good choice (Ars used it) and it showed the M1 Ultra as ~60 percent faster than a 3060 Ti. That's fast. Not 75 percent faster like the RTX 3090, but still fast.

Agreed. But what you need to understand is that the integration of the CPUs & GPUs through their interconnecting circuitry on the same die makes those tests somewhat inaccurate. They are a good ball-park type test and we can easily say that M1Ultra is a graphics performer, but how can we compare exact numbers?

This fudginess is where Apple thought(mistakenly) that they could get away with their deception.

Now I'd like to be clear, M1Ultra is an impressive piece of tech, there can be no doubt of that. However, it is FAR from the "World's most powerful chip for a personal computer." For one, the systems the M1Ultra is available in are NOT in the budget of the "Personal" computer range. They are exclusively Prosumer and workstation systems. Second, they are not the most powerful single systems in the world, whether Personal or Professional/Workstation sectors.

Regardless of how impressive M1Ultra is, Apple knowingly lied. That is something the industry and the user base can not allow or tolerate.

 

[TheLostSwede]

Ars Technica has some better benchmarks, but Apple is still not competitive against a 3090, as they're struggling against a 3070...

Review: The Mac Studio shows us exactly why Apple left Intel behind

Part Mac mini, part trashcan Mac Pro, the Studio is one impressive mini desktop.

arstechnica.com

 

[usiname]

companies need to stop making up silly comparisons in their presentations. The Mac Studio will sell to the target audience without silly comparisons.

I don't think this is even close to true. I know so many people which are stupid to point that they believe on everything in presentations on their favourite brands and they cannot comprehend how absurd the claims are. They don't believe on 3rd party reviews and do not accept any other facts. Now multiple the percentage of this people which at least have someon tо give them common sense by 100 and you will get close to the number of people which purchase without any extra info beside the ad on TV or somewhere on the web. Apple can say in their next presentation that their new cpu/gpu is 100x faster then the current tech and 99% of their sheeps will belive in this, not just that, people who do not use their products and are not even interested also will belive in this. the 1% will complain for their lies but who cares?

 

[TheLostSwede]

I don't think this is even close to true. I know so many people which are stupid to point that they believe on everything in presentations on their favourite brands and they cannot comprehend how absurd the claims are. They don't believe on 3rd party reviews and do not accept any other facts. Now multiple the percentage of this people which at least have someon tо give them common sense by 100 and you will get close to the number of people which purchase without any extra info beside the ad on TV or somewhere on the web. Apple can say in their next presentation that their new cpu/gpu is 100x faster then the current tech and 99% of their sheeps will belive in this, not just that, people who do not use their products and are not even interested also will belive in this. the 1% will complain for their lies but who cares?

Maybe you need hang out with different people? I don't know anyone like that.

 

[ModEl4]

It seems to me in gaming would be like a 2250MHz (theoretical) Radeon 6800 in 4K with extremely bad performance scaling in lower resolutions (6800 would be -5% in 4K, 6700XT would be -5% in QHD and 6600XT -1% in FHD)
Much more interesting is the Max version, which without the scaling problems seems to be at RX 5700 level in FHD, a little bit better than RX 5700 at QHD and a little bit worse than 5700XT at 4K which is excellent result for a 1275-1300MHz 4096 cores GPU, exceeding Polaris performance/terraflop efficiency and only -15% vs a RDNA 5700XT (-20% vs RX 5700) regarding performance/terraflop efficiency.
The real problem is the transistors needed to achieve this performance level (taking account also the frequency differences/scaling of course)

 

[DeathtoGnomes]

Again, they didn't lie. Benchmarks are being twisted.

I dont see the difference, twisting the truth is still a lie. My morals are likely different than yours.

 

[samum]

Here's my conspiracy theory:

Apple switching to x86 hardware made it much easier to compare performance between a Mac and a Windows or Linux system. No longer could Apple hide behind PowerPC and software that didn't run on other systems.

Apple's x86 hardware was generally underwhelming and a generation or two behind. Mac Pros had some high end stuff, but they would go years without updates.

Apple also had their falling out with Nvidia over the faulty laptop chips. In the years when Nvidia dominated GPU performance, Apple was stuck with AMD.

Intel's 10nm issues and the quality of what they delivered didn't help, but I think Apple's true motivation for developing the M1 is so that they can again hide behind a different architecture.

Apple did not fare well with x86 hardware.

The M1 is a very well designed and implemented platform and it gets a lot out of a few watts, but it's not the performance monster Apple makes it out to be. Apple is back to their "40 times faster than PC!!!!!" claims with a lot of missing asterisks and fine print. The M1's biggest advantage for Apple isn't performance, it's obfuscation for marketing.

 

[JustBenching]

This is why there's no point in talking to anyone on the internet. I gave you useful data showing the M1 Ultra crushing the RTX 3070, which means it is only a bit slower (5-15 percent slower) than the RTX 3090. It's to counteract a one sided interpretation you'll receive by only comparing 5 year old ported software to x86 Macs.

Wow, the bias with this one. Your graph shows m1 ultra beating a 3070 by 15%,and you called it crushing. Yet when the 3090 is 15% ahead you called it "only a bit slower". Make up your mind, is the 3090 crushing the m1 ultra or is the 3070 just a bit slower? Both cant be true