M4 Mac Announcements

[dada_dave]

New analysis post in the Strix Halo thread comparing the Halo and M4 (sadly only one Halo data point so far, but still some things we can say about it):

D

Post in thread 'Strix Halo - AMD'

Feb 18, 2025

Click to expand. Things are getting a bit scrunched up, haven't had time to prune the data points (unfortunately Numbers isn't great here, change the order or remove a data point and the entire graph's style has to be redone).

With respect to the Strix Halo in Asus ROG Flow (AI Max+ 395) I have to admit this a little worse on the CPU side than I thought it would be. I expected higher power/lower efficiency for the ST but not this bad and I thought there would be better performance than the Strix Point. Instead it is just same performance, much worse efficiency - not totally...

• dada_dave

• 

Updated:

D

Post in thread 'Strix Halo - AMD'

Feb 28, 2025

A more complete Strix Halo (and Point analysis) versus the M4:

Notebookcheck added extra TDP figures to its Halo analysis. Unfortunately they did not measure wall power at each TDP. However, while TDP doesn't not correspond well to wall power in an absolute sense, the Strix Point analysis where they did indeed measure both seems to confirm that relative TDP values are proportional to the wall power. So, assuming the same is true for the Halo, I can still use the TDP values to estimate wall power assuming 70W TDP for the default state wall power was measured in. I could do the...

• dada_dave

• 
• 

 

[Jimmyjames]

Seems like no Ultrafusion on M4 Max.
Top photo: close up of m4 max where ultrafusion would be

Bottom photo m4 and m3 max for comparison.

 

[Altaic]

Seems like no Ultrafusion on M4 Max.
Top photo: close up of m4 max where ultrafusion would be
View attachment 34058
Bottom photo m4 and m3 max for comparison.
View attachment 34057

Looks like part of the die is obscured by the white frame thing (edit: nvm, visual confusion). I gotta say, though, that is one harsh die pic. Really hard to pick out any features at all. Source?

 

[Jimmyjames]

Looks like part of the die is obscured by the white frame thing. I gotta say, though, that is one harsh die pic. Really hard to pick out any features at all. Source?

It’s from twitter. I thought we couldn’t post links here.

 

[Yoused]

It’s from twitter. I thought we couldn’t post links here.

We can post links, they just do not auto-expand into the original Xcretion like they usedto.

 

[Jimmyjames]

We can post links, they just do not auto-expand into the original Xcretion like they usedto.

I thought we were boycotting it and/or it was discouraged.

 

[Cmaier]

Seems like no Ultrafusion on M4 Max.
Top photo: close up of m4 max where ultrafusion would be
View attachment 34058
Bottom photo m4 and m3 max for comparison.
View attachment 34057

I can’t tell shite from that photo

 

[Jimmyjames]

Errr...

So Apple is gonna put an M4 Max into the base Studio and an M3 Ultra into the high end one, and the base model will have

Faster single core performance
Faster memory.
Thunderbolt 5
Faster Neural Engine.

Make it make sense. This can't be correct?

Right?

 

[MacPoulet]

Weird. It would make more sense for both Studios to be on M3 than have an M4 Max and M3 Ultra...

Unless it's a brand new chip based on the M3 that's like overcharged or something.

 

[Vincent Hannah]

That makes zero sense. And the M3 Max does not have UltraFusion to even make an M3 Ultra.

 

[casperes1996]

Errr...
View attachment 34061
View attachment 34062

So Apple is gonna put an M4 Max into the base Studio and an M3 Ultra into the high end one, and the base model will have

Faster single core performance
Faster memory.
Thunderbolt 5
Faster Neural Engine.

Make it make sense. This can't be correct?

Right?

I’m certainly betting against it. Sounds insane to me.

 

[Jimmyjames]

That makes zero sense. And the M3 Max does not have UltraFusion to even make an M3 Ultra.

Yes. So they’re making a new die on a node they were apparently desperate to get off of (N3B). Zero sense.

 

[MacPoulet]

Yes. So they’re making a new die on a node they were apparently desperate to get off of (N3B). Zero sense.

Why were they so desperate to get off N3B? I wonder if it has any advantages for an Ultra-level chip that isn't two Maxes glued together...

 

[Jimmyjames]

Why were they so desperate to get off N3B? I wonder if it has any advantages for an Ultra-level chip that isn't two Maxes glued together...

It was the first N3 node and expensive, plus a bit hot!

 

[dada_dave]

It was the first N3 node and expensive, plus a bit hot!

I do think the N3B troubles were slightly exaggerated by the press and especially "community" since it wasn't up to TSMC's usual standards and Gurman this close in is usually accurate, but yeah ... I'm struggling to see why they would do this. I mean the GPUs between M3 and M4 weren't terribly different, but even so for all the reasons you and everyone else mentioned to release both together is nuts. Also, they went back an added ultra fusion to the M3 Max? Huh? Why? Ultra Studios are low volume products, they can't make enough M4 Maxes to satisfy both chip lines? I would imagine making an M3 Ultra now would be cheaper than to do it last year and almost certainly cheaper than M4 Ultras might be, but again ... really? You can tell, I'm trying to come up with justifications, but this is odd if true. Very odd.

 

[dada_dave]

Errr...
View attachment 34061
View attachment 34062

So Apple is gonna put an M4 Max into the base Studio and an M3 Ultra into the high end one, and the base model will have

Faster single core performance
Faster memory.
Thunderbolt 5
Faster Neural Engine.

Make it make sense. This can't be correct?

Right?

Re-reading this. The tweet is actually very not confident about what chips will be there - the phrase "There are signs these will come with an M4 Max but that its new Ultra chip will actually be an M3 Ultra" make it seem like he's hedging his bets here (unlike the last time he made an almost equally puzzling prognostication that the M4 iPad Pro was launching soon that turned out to be true and he was fairly adamant it was happening). So he simply may be not understanding what the "signs" mean. Whatever those "signs" are.

Kremlinology Gurman edition.

 

[Jimmyjames]

Re-reading this. The tweet is actually very not confident about what chips will be there - the phrase "There are signs these will come with an M4 Max but that its new Ultra chip will actually be an M3 Ultra" make it seem like he's hedging his bets here (unlike the last time he made an almost equally puzzling prognostication that the M4 iPad Pro was launching soon that turned out to be true and he was fairly adamant it was happening). So he simply may be not understanding what the "signs" mean. Whatever those "signs" are.

Kremlinology Gurman edition.

I am sure you are correct. He’s guessing.

 

[The Hardcard]

That makes zero sense. And the M3 Max does not have UltraFusion to even make an M3 Ultra.

It’s going to be the Hidra, whatever the configuration. Even at the old speed it would still be 800 GB/s memory bandwidth to the M4 Max’s 546 GB/s, plus they can put the faster memory speed on it as well.

To me, the biggest naysayer would be Thunderbolt 5. There’s no way the top Mac Studio can not have TB5. And if they put TB5 on it, why not use the nearly year old M4 cores at least. M5 cores are already in production.

 

[dada_dave]

For their part, MR is putting forward that maybe why an M3-based Ultra would be released now is that it is a new monolithic die. Maybe?

 

[Jimmyjames]

For their part, MR is putting forward that maybe why an M3-based Ultra would be released now is that it is a new monolithic die. Maybe?

Then why not do it last year? If it’s a new die then they could presumably do it for the M4.