Peak Peek - Apple officially announces new SE, new iPad Air, M1 Ultra, Mac Studio and Studio Display

[Video Beagle]

I wonder if the M1 Ultra has it's own hidden UltraFusion™ connectors to link 2 of them side by side, giving us M1 Voltron. (I'm only joking about the name...I wonder how much this tech can do)

[leadeater]

12 minutes ago, Kisai said:

No you silly. They should have just called every part i3, i5 or i7 and short of an actual change in the feature set, dispense with the XXXXX numbers and letters.

The biggest problem Intel has is that the numbers mean nothing, just like with AMD's CPU/GPU's and NVIDIA's GPU's.

 

Here, I'll even spell it out.

 

2010(2nd gen):

i1 = Dual Core, no hyper threading, no boosting (eg what's currently Pentium, Celeron and Atom, and formerly Core Solo/Core 2 Duo)

i3 = Dual Core, hyper threading, no boosting

i5 = Quad Core, no hyper threading, no boosting

i7 = Quad Core, hyper threading, no boosting

i9 = Quad Core, hyper threading, boosting

Xeon E3 = same as i7 part, but ECC memory

 

2017 (8th gen):

i1 = Dual Core (Y/U and Atom/Celeron/Pentium brands)

i3 = Quad Core, no hyper threading (former i5 line)

i5 = Six Core, no hyper threading

i7 = Six Core, hyper threading

i9 = Six Core, hyper threading, boosting

 

2021 (12th gen):

i1 = Quad Core

i3 = Six Core

i5 = Eight Core 

i7 = Eight Core + Hyperthreading

i9 = Eight Core + Hyperthreading + Boosting

Where and how exactly do you put in the product model number the generation or archecture. Also that is literally impossible to whittle down the product models to that few ergo you need more than just simple iX numbering.

 

The numbers in fact do mean a lot, if you don't know Intel literally offers a document/guide on how to read them. Not that you need to, it's extremely obvious.

 

Anyone not willing to figure it out or look at them can and will default to "i7/i9 is best" and just buy the current one, they aren't going to lose out in any way.

[leadeater]

15 minutes ago, Video Beagle said:

I wonder if the M1 Ultra has it's own hidden UltraFusion™ connectors to link 2 of them side by side, giving us M1 Voltron. (I'm only joking about the name...I wonder how much this tech can do)

There is no "M1 Ultra", It's two M1 Max's on an interposer using the known but not officially confirmed chip interconnects. The existence of them has been known for quite some time. There are no more interconnects on the M1 Max dies so the Ultra is the current maximum using M1 Max dies.

 

Also M1 Plaid

[hishnash]

27 minutes ago, leadeater said:

Handbrake, Blender (pure CPU), Matlab, Excel (eh MS), 7zip (it leads and lags).

The issue with comparing between arcs with these tools is they are extremely well optimised for x86, making extensive use of different cpu feature sets and private vendor specific extensions (AVC512 etc). But on ARM they do not make use of much if any of the equivalent optional features. 

 

[leadeater]

3 minutes ago, hishnash said:

The issue with comparing between arcs with these tools is they are extremely well optimised for x86, making extensive use of different cpu feature sets and private vendor specific extensions (AVC512 etc). But on ARM they do not make use of much if any of the equivalent optional features. 

Yep, that's why I said they are "non Apple software"

 

However Geekbench and CB illustrate the same point. Comparing unconstrainted desktop CPUs made to HIT THE NUMBERS to constrained CPUs/SoCs is probably the most flawed thing to do.

[Video Beagle]

14 hours ago, leadeater said:

Devil is in the details.

aka, Marketing

12 hours ago, leadeater said:

Nice. Does that also mean Mac Pro will come in at M2 production generation? Not read the announcement so not sure of the exact wording

They teased the Mac Pro as the final machine left to be transitioned. They didn't mention the chip. As for "they stated the Ultra is the final M1"...It's a marketing presentation, not sworn testimony or laws written on stone tablets. They could introduce the Maxtra, when one day someone carrying an Ultra bumped into someone carrying a max "You got ultra in my max!" "You got Max in my Ultra!"

 

6 hours ago, FakeKGB said:

But then it won't match the Studio Display.

Also, these are professionals. They don't need people to focus on their computers, they need people to focus on what they make.

Yet, they made a big point about colors of the new Magic keyboards and stuff...curious..

 

3 hours ago, saltycaramel said:

LMG coverage of the event and the Mac Studio: it’s not a Steam Deck so let’s generally be debbie downers. 

 

 

linus gonna linus. He'll pout and stamp his feet that Apple ignores him while paying attenion to Marques and Justine, then he'll click bait negative thumbnails to get views. (I haven't watched the videos yet to comment on content). Why the hell wasn't there an immediate reaction stream with Anthony and Horst following the event? The biggest tech company does a presentation, and when people look on youtube for commentary...where the hell is LMG? Off refilling their water bottles? Deciding what t-shirts to wear while doing 69 jokes? I mean, Snazzy Labs didn't have anything up, but he's what, 2 people? How many people are in the production side of LMG?

 

 

13 minutes ago, leadeater said:

There is no "M1 Ultra", It's two M1 Max's on an interposer using the known but not officially confirmed chip interconnects.

 

 

I saw a logo. There's a M1 Ultra.

 

 

13 minutes ago, leadeater said:

There are no more interconnects on the M1 Max dies so the Ultra is the current maximum using M1 Max dies.

 

 

Until they reveal you can put two of them face to face like a sandwhich!

 

[Kisai]

36 minutes ago, leadeater said:

Where and how exactly do you put in the product model number the generation or archecture. Also that is literally impossible to whittle down the product models to that few ergo you need more than just simple iX numbering.

 

The numbers in fact do mean a lot, if you don't know Intel literally offers a document/guide on how to read them. Not that you need to, it's extremely obvious.

 

Anyone not willing to figure it out or look at them can and will default to "i7/i9 is best" and just buy the current one, they aren't going to lose out in any way.

I was responding to your assertation about not needing model numbers, not that they weren't necessary. You look at apple's M1 chips, they do in fact have numbers on them

Nowhere do you see M1 on that

 

Now look at an Intel CPU:

They actually took the clock speed off the chip, 11th and 10th gen have that beside the SRL4H, which Intel calls "Spec code"

 

Intel can, and should just be marketing the "12 gen i9" not "12900", because those numbers have no meaning to anyone, and fails the "bigger number is better" 

 

In fact, if you pay attention to the actual numbers used on Intel parts, the "12th gen i9" is redundant with the 4 or 5 digit numbers to begin with.

12900 = 12th gen i9

[12850] = 12th gen i9

12700 = 12th gen i7

12600 = 12th gen i5

12500 = 12th gen i5

12400 = 12th gen i5

12300 = 12th gen i3

12100 = 12th gen i3

 

So those next two digits? Means nothing. Intel has never used the xx200 on a desktop part, only the U/Y mobile parts. Likewise the xx850 parts have been used for mobile i7 and desktop i9's. They don't make an i4 or i8 logically because these mean death and luck in superstitious cultures. The model numbers don't tell you the core configuration, graphics configuration other than "absent" with a letter F, or speed.

 

If you can swap an 4th gen i7 with a 11th gen i7, then you're in the same performance tier, who really cares if that's a T, non-K or K model unless you need to build a very specific configuration (eg for overclocking.)

 

[saltycaramel]

He now ordered an M1 Ultra, can’t wait for him to repeat the comparison.

No Apple Software involved.

[leadeater]

1 minute ago, Video Beagle said:

I saw a logo. There's a M1 Ultra.

There is an M1 Ultra SoC but there is no M1 Ultra die. Since the die has been x-rayed quite some time ago the interconnects have been known about.

 

2 minutes ago, Video Beagle said:

Until they reveal you can put two of them face to face like a sandwhich!

Past analysis and also TSMC's chip stacking readiness would likely result in that not being a thing. Anyway make far more sense to release M2 and 3nm than persist with M1 and 5nm now, that like choosing to take a step rather than a leap.

[Video Beagle]

Something that leaped out at me when watching, btw, was there was an awful lot of women presenting in the video...only 3 men...most of the Apple people and all of the developers were women....on International Women's Day. I don't know if that was intentional or coincidental, but with Apple, I'd suspect the former. 

 

(I'd seen some memes on facebook about it being IWD right before I watched the presentation, so it was on my mind.)

[leadeater]

5 minutes ago, Kisai said:

12900 = 12th gen i9

[12850] = 12th gen i9

12700 = 12th gen i7

12600 = 12th gen i5

12500 = 12th gen i5

12400 = 12th gen i5

12300 = 12th gen i3

12100 = 12th gen i3

You are aware you just now showed why it's not possible. You have three i5's so this is a non-starter.

 

5 minutes ago, Kisai said:

If you can swap an 4th gen i7 with a 11th gen i7, then you're in the same performance tier, who really cares if that's a T, non-K or K model unless you need to build a very specific configuration (eg for overclocking.)

Literally every DIY builder.. and that's just a starting point.

[Video Beagle]

2 minutes ago, leadeater said:

There is an M1 Ultra SoC but there is no M1 Ultra die. Since the die has been x-rayed quite some time ago the interconnects have been known about.

 

Past analysis and also TSMC's chip stacking readiness would likely result in that not being a thing. Anyway make far more sense to release M2 and 3nm than persist with M1 and 5nm now, that like choosing to take a step rather than a leap.

as we've discussed in the past, I don't understand anything at this level of computers. I saw the video..the things snapped together. That tells me they're like Lego Brand Bricks®.

They're wafer's right? you take a wafer, you put a filling down.. I guess that thermal paste that looks like silver frosting..put another wafer on top of it.. POWER!

[leadeater]

7 minutes ago, saltycaramel said:

 

He now ordered an M1 Ultra, can’t wait for him to repeat the comparison.

Well he should also order an actually modern Zen or Ryzen system, a 5900HX does it in 1 minute 41. And a 5950X does it in 47 seconds.

[leadeater]

11 minutes ago, Video Beagle said:

as we've discussed in the past, I don't understand anything at this level of computers. I saw the video..the things snapped together. That tells me they're like Lego Brand Bricks®.

They're wafer's right? you take a wafer, you put a filling down.. I guess that thermal paste that looks like silver frosting..put another wafer on top of it.. POWER!

Chip stacking requires a whole bunch of stuff that TSMC simply didn't have ready at the time Apple did those die designs so that just makes it unlikely as a possibility. Currently the dies are most likely Lego bricks but with smooth tops and can only sit on top of a flat sheet of Lego (the interposer).

 

Edit:

Anyway roadmap wise M2 just make more sense.

[Paul Thexton]

1 hour ago, Video Beagle said:

I wonder if the M1 Ultra has it's own hidden UltraFusion™ connectors to link 2 of them side by side, giving us M1 Voltron. (I'm only joking about the name...I wonder how much this tech can do)

M1 Dude

 

 

[Kilrah]

52 minutes ago, Video Beagle said:

They teased the Mac Pro as the final machine left to be transitioned. They didn't mention the chip.

I'd expect the Mac Pro to be the first M2-based machine. Gonna need a redesign to have PCIe slots, maybe module-based memory etc...

 

48 minutes ago, saltycaramel said:

He now ordered an M1 Ultra, can’t wait for him to repeat the comparison

I mean he has a 5 year old 1st gen Threadripper, it's been trumped in performance long ago. 

 

47 minutes ago, Video Beagle said:

Something that leaped out at me when watching, btw, was there was an awful lot of women presenting in the video...only 3 men...most of the Apple people and all of the developers were women....on International Women's Day. I don't know if that was intentional or coincidental, but with Apple, I'd suspect the former. 

Intentional of course, they even made a couple of "winks" to it in the presentation.

[Paul Thexton]

38 minutes ago, leadeater said:

Well he should also order an actually modern Zen or Ryzen system, a 5900HX does it in 1 minute 41. And a 5950X does it in 47 seconds.

Depends entirely on what modules you’re compiling as well of course. In this case I believe the thread ripper was cross compiling the kernel for his M1 Linux work.

[WhitetailAni]

1 hour ago, Video Beagle said:

Yet, they made a big point about colors of the new Magic keyboards and stuff...curious..

Yes...

...because those are for the regular consumer market. Y'know, the one that likes having things in fancy colors and want people to look at their things.

[Dracarris]

1 hour ago, Video Beagle said:

Something that leaped out at me when watching, btw, was there was an awful lot of women presenting in the video...only 3 men...most of the Apple people and all of the developers were women....on International Women's Day. I don't know if that was intentional or coincidental, but with Apple, I'd suspect the former. 

it was intentional. And Apple does this for a while now with their presentations. No matter what you think of it, IWD or not - it is a stark misrepresentation of reality. The percentage of female PoC or hijab-wearers in this industry is way lower.

[Spindel]

M1 Ultra is exciting  

[igormp]

1 hour ago, leadeater said:

Well he should also order an actually modern Zen or Ryzen system, a 5900HX does it in 1 minute 41. And a 5950X does it in 47 seconds.

The numbers you have are for a clean kernel build, a build like Marcan's took like ~2m30s in my 5950x.

Here's another point of comparison between a desktop AMD and a M1 Pro:

 

50 minutes ago, Paul Thexton said:

Depends entirely on what modules you’re compiling as well of course. In this case I believe the thread ripper was cross compiling the kernel for his M1 Linux work.

Indeed. Also, just a reminder for the unaware that cross-compiling a kernel makes no difference when it comes to performance comparison as long as you target the same arch in both runs.

[Dracarris]

5 minutes ago, igormp said:

The numbers you have are for a clean kernel build, a build like Marcan's took like ~2m30s in my 5950x.

Here's another point of comparison between a desktop AMD and a M1 Pro:

are there also comparisons against the Max available? With 32 and 16P-threads a comparison against a 6P-thread M1 might be a bit off.

[igormp]

5 minutes ago, Dracarris said:

are there also comparisons against the Max available? With 32 and 16P-threads a comparison against a 6P-thread M1 might be a bit off.

No idea, I don't remember seeing any targeting a linux build specifically. My guess is that it should get close to a 5950x, but still be a bit slower.

[igormp]

1 hour ago, leadeater said:

Well he should also order an actually modern Zen or Ryzen system, a 5900HX does it in 1 minute 41. And a 5950X does it in 47 seconds.

Btw, as a fair point of comparison with a clean config:

 

So yeah, no laptop can match the M1 (specially when you consider the power consumption), and it's really close to desktop numbers at a fraction of the power.

[LAwLz]

51 minutes ago, igormp said:

The numbers you have are for a clean kernel build, a build like Marcan's took like ~2m30s in my 5950x.

Yeah, there are more than one way to build the Linux kernel. You can't comparing building one Linux kernel against building a different Linux kernel.

The variables have to be the same between tests.

 

I'm sorry Leadeater, but the Firestorm cores really does wipe the floor with Zen3, and does it at way lower power consumption.

 

 

Another person replied to Jens that his 9900K did it in 22.75 minutes.

Please note that the M1 Pro that did it in 20.75 minutes uses like 25 watts. The 9900K uses like 200 watts.