Mac Studio M1 Ultra GPU will be beaten by AMD and Nvidia in just 6-7 months.

[aevan]

Also, you can actually find very competitive laptops to MBP. A lot of high end alder lake based laptops coming out perform equal to or better than macbook pros in quite a few benchmakrs/scenarios and also in build quality,

Some of them outperform the MBP only when hooked up to power and at really high temperatures. I honestly don't think these performance gains are worth it (at those temperatures and battery life, Apple could've just put an M1 Ultra in a MacBook Pro and be on top, but what would be the point?). They might as well be desktops. And build quality is relative, I honestly haven't seen a single Windows laptop that matched MacBooks in build quality, with the exception of, maybe, the Surface line.

I still think there is nothing in the Windows world that can compete with current MacBook Pros as a general productivity laptop, at any price. Of course, there are specific situations - you might need some software that requires Windows, you might need a specific hardware capability (like stylus support) and, for sure, you might be looking into gaming - so there are situations where a Windows laptop might be better. But in general terms, no, I disagree there are better laptops in terms of performance on battery, heat and build quality. And as a package - nothing comes close in Windows world. This may change at some point, but for the majority of people, not counting those with specific requirements as I mentioned, one of the MacBooks is the best recommendation.

You can find equal or better non apple laptops though if you are ok to sacrifice a bit on the battery life part.

If by "a bit" you mean almost twice, then yeah. Also, remember how much these laptops throttle on battery power. So it's not a bit, and you're getting way less performance. As I said, if you don't care about these things and plan on working plugged in, you might as well compare these laptops to a Mac Studio.

 

[throAU]

Lol. in how many watts?

 

[Pressure]

It seems the Mac Studio is reusing the iMac Pro power supply at 370 Watt.

 

[throAU]

It seems the Mac Studio is reusing the iMac Pro power supply at 370 Watt.

Haha. So that's less than a single 3090. For the entire system.

 

[oz_rkie]

Some of them outperform the MBP only when hooked up to power and at really high temperatures. I honestly don't think these performance gains are worth it (at those temperatures and battery life, Apple could've just put an M1 Ultra in a MacBook Pro and be on top, but what would be the point?). They might as well be desktops. And build quality is relative, I honestly haven't seen a single Windows laptop that matched MacBooks in build quality, with the exception of, maybe, the Surface line.

I still think there is nothing in the Windows world that can compete with current MacBook Pros as a general productivity laptop, at any price. Of course, there are specific situations - you might need some software that requires Windows, you might need a specific hardware capability (like stylus support) and, for sure, you might be looking into gaming - so there are situations where a Windows laptop might be better. But in general terms, no, I disagree there are better laptops in terms of performance on battery, heat and build quality. And as a package - nothing comes close in Windows world. This may change at some point, but for the majority of people, not counting those with specific requirements as I mentioned, one of the MacBooks is the best recommendation.



If by "a bit" you mean almost twice, then yeah. Also, remember how much these laptops throttle on battery power. So it's not a bit, and you're getting way less performance. As I said, if you don't care about these things and plan on working plugged in, you might as well compare these laptops to a Mac Studio.

Again, as I've tried to explain many a times in various threads in this forum. No one is arguing or disputing the unmatched efficiency of the m1 chips at the moment. It is true that neither intel or amd currently has a chip that can perform the same as an m1 in the same power envelope. BUT there are laptop chips that can perform at or even well above even the m1 max performance at a higher power envelope.

This may not be worth it to YOU but it does not mean it everyone will have the same use case as you. A LOT of people use their laptops plugged in majority of the time. As an example I fall in this category as well. I am a software engineer working from home and my laptop (which btw is an m1) is basically plugged in 95% of the time. You can't just make blanket statements like 'oh it only can do this when plugged in and therefore its not worth it'. The first part of that claim might be true, the second part is true or false depending on who you are. While the m1 chip is more than enough performance for me, if I needed more performance I would switch to a high end alder lake based laptop very easily as in my use case I don't mind that my laptop is plugged in most of the time. For those few times I need to go into the office, I can take my laptop and plug it in at my office desk, easy.

The key is - what is your use case.

 

[dieselm]

dieselm said:
It'll be years to never before NVIDIA or AMD can make something with equivalent performance which will sit quietly on a desk in an 8"x8" square.

source?

What's the performance of the next closest CPU/GPU which can quietly dissipate heat in a 8"x8"x4" enclosure?

 

[oz_rkie]

dieselm said:
It'll be years to never before NVIDIA or AMD can make something with equivalent performance which will sit quietly on a desk in an 8"x8" square.

What's the performance of the next closest CPU/GPU which can quietly dissipate heat in a 8"x8"x4" enclosure?

No idea, but I was mainly curious to the source for It'll be years to never before part of the claim. Is it your gut feeling?

 

[Spindel]

It seems the Mac Studio is reusing the iMac Pro power supply at 370 Watt.

Not really a good metric.

My M1 mini has a PSU of 150 W and never draws more than 35-40 W from the wall.

But doing rough math the Ultra, based on Max figures, should draw around 180-200 W when both CPU and GPU go full throttle at the same time.

From the Max we then can conclude that the Ultras CPU should take 65-70 W. This leaves GPU power draw at 115-135 W at full load.

 

[dieselm]

No idea, but I was mainly curious to the source for It'll be years to never before part of the claim. Is it your gut feeling?

There's no integrated cpu/gpu soc in the pipeline for these companies at these power/performance levels. M1, yes. But not the Max/Ultra.

If they started now, it would take 3 years for it to come out the other side. Even if AMD/NVIDIA could get the foundry capacity at the latest process node, which even now they can't.

There's a reason why Apple makes the iMac so thin, the macbooks so thin w/long battery life, and the Mac Studio small with good acoustics. Competitors can't fit in the power budgets without severe technical and business model compromises.

 

[Pressure]

Not really a good metric.

My M1 mini has a PSU of 150 W and never draws more than 35-40 W from the wall.

But doing rough math the Ultra, based on Max figures, should draw around 180-200 W when both CPU and GPU go full throttle at the same time.

From the Max we then can conclude that the Ultras CPU should take 65-70 W. This leaves GPU power draw at 115-135 W at full load.

Remember TB also needs to supply power. I’m not saying it will draw 370 Watt but just that it’s what the power supply is capable of.

 

[oz_rkie]

There's no integrated cpu/gpu soc in the pipeline for these companies at these power/performance levels. M1, yes. But not the Max/Ultra.

Of the companies you mentioned, only AMD really makes SOCs. While nvidia does have some lower end/embedded SOCs they mainly are a gaming and data center GPU company, so obviously they would not have any m1 competing SOCs in their pipeline

If they started now, it would take 3 years for it to come out the other side

Do you have a source for the 3 year timeline? Or just pulled it out of your ass? Chip makers have several future products already in various stages of production at any given time (design, fabrication, testing, market etc). They don't wait to release a product and then 'start' work on the next product.

 

[uller6]

It basically means nothing nowadays (not much in older days either). Below are the TOPs in INT8, from Xilinx (now AMD) VCK5000 press release. Real-world performance differs largely from that theoretical number.

View attachment 1970831

Dark silicon is a HUGE issue and has been for the past 15 years when Dennard scaling hit the brick wall of physics. Now that it's physically impossible to run all transistors on the CPU at once (due to power density limitations), we've been pushed us into the era of heterogeneous CPUs. The problem will only get worse as we continue with smaller process nodes.

Apple actually has a very clever workaround to somewhat mitigate the power density issue by including the power cores as 4X assemblies in the M1 pro and M1 max CPU dies. 2 separate 4X assembles help spread out the CPU power density across the chip surface, allowing better power core performance.

 

[Unregistered 4U]

Not really a good metric.

My M1 mini has a PSU of 150 W and never draws more than 35-40 W from the wall.

But doing rough math the Ultra, based on Max figures, should draw around 180-200 W when both CPU and GPU go full throttle at the same time.

From the Max we then can conclude that the Ultras CPU should take 65-70 W. This leaves GPU power draw at 115-135 W at full load.

Apple’s charts look to show that the Ultra hits 60W when stressing the CPU and 110W when stressing the GPU. Once they’re in testers hands, if Apple’s right, the Ultra peak draw should be in the range of 170W. (Their Max charts appear to be accurate within a few W.)

 

[MayaUser]

Guys calm down, even the OP realised and asking for mods to delete this

 

[MayaUser]

Apple’s charts look to show that the Ultra hits 60W when stressing the CPU and 110W when stressing the GPU. Once they’re in testers hands, if Apple’s right, the Ultra peak draw should be in the range of 170W. (Their Max charts appear to be accurate within a few W.)

so around 200W total package , thats not bad at all

 

[mcnallym]

Software has to be written to use specific Apple APIs to make use of that neural engine, though, right?

And how do you think that a computer makes use of a GPU.
You write code that uses the API to make use of be GPU, or write code to access the GPU directly if don’t use the API provided,
What was the point you trying to make with this?

 

[Macative]

So what?

People choose Mac because they want Mac.

The people who choose Mac ask Apple, "can we please have a decent GPU too?"

Apple has delivered above and beyond expectations.

It makes zero difference what AMD or NVIDIA are doing.

 

[cmaier]

Of the companies you mentioned, only AMD really makes SOCs. While nvidia does have some lower end/embedded SOCs they mainly are a gaming and data center GPU company, so obviously they would not have any m1 competing SOCs in their pipeline


Do you have a source for the 3 year timeline? Or just pulled it out of your ass? Chip makers have several future products already in various stages of production at any given time (design, fabrication, testing, market etc). They don't wait to release a product and then 'start' work on the next product.

As someone who worked at AMD and has watched the trajectory of each of these companies, it’s going to be quite awhile.

 

[OnawaAfrica]

the reason they do is cause their desktops are very expensive

yes they are expensive cause most of their desktops are not targeted at the average consumer. They are Targeted and Designers Editors etc. so best option for a desktop is Mac mini, or the Cheaper imac

 

[diamond.g]

As someone who worked at AMD and has watched the trajectory of each of these companies, it’s going to be quite awhile.

When you were there did AMD build chips that OEMs didn't ask for?

Or I guess put another way if no one is asking for M1 Ultra performance/efficiency in a AMD soc would they make that soc anyways?

 

[cmaier]

When you were there did AMD build chips that OEMs didn't ask for?

Or I guess put another way if no one is asking for M1 Ultra performance/efficiency in a AMD soc would they make that soc anyways?

We made chips we thought we could sell. Sometimes that meant we did a chip that the OEM was not asking for - nobody asked us for AMD64.

 

[diamond.g]

We made chips we thought we could sell. Sometimes that meant we did a chip that the OEM was not asking for - nobody asked us for AMD64.

Do you think AMD would have buyers interested in the Series X APU with DDR5 and Zen 3?

 

[dysamoria]

This is the GPU tech for which Apple did a new deal with Imagination. I think it looks pretty good.

IMG CXT GPU - Imagination

Ray tracing for mobile. Bringing high-performance, desktop-quality ray traced visuals to the palm of your hand.

www.imaginationtech.com

Thanks. I didn’t see any news about Apple licensing this. There’s a lot of marketing on that page I don’t want to sift through. Is this hardware that Apple need to implement, hardware they have implemented, or software? I recall reading that Metal has raytracing in the API, but I’ve yet to see any statement of hardware specifically dedicated to ray tracing in Macs.

 

[diamond.g]

Thanks. I didn’t see any news about Apple licensing this. There’s a lot of marketing on that page I don’t want to sift through. Is this hardware that Apple need to implement, hardware they have implemented, or software? I recall reading that Metal has raytracing in the API, but I’ve yet to see any statement of hardware specifically dedicated to ray tracing in Macs.

Hardware BVH traversal and intersection testing would be ideal.

 

[cmaier]

Do you think AMD would have buyers interested in the Series X APU with DDR5 and Zen 3?

I dunno. I’m an engineer, not a salesperson.