3nm Apple Silicon: What Is It and Why Does It Matter?

[theorist9]

Well here is a simple question: Do 3nm chips have features that are 3/5 the size of 5 nm chips?

It's a good question, and the answer is: Probably not. "5 nm" and "3 nm" are principally marketing terms at this point. For instance, the ratio of the transistor gate pitches for TSMC 3 nm and 5 nm is 45 nm/51 nm = 0.88, and that of the interconnect pitches is 23 nm/28 nm = 0.82; both of these are a far cry from 3/5 = 0.6.

The ratio of the square root of the transistor densities is also about 0.8. So if TSMC 5 nm really were 5 nm in some meaningful way (which it's not), then TSMC 3 nm would be closer to 4 nm. And that's just for the logic portion—the SRAM cells are about the same size.

Thus, while 3 nm is a significant technological achievement, it's not accurately labeled.

 

[zapmymac]

3nm:

no gimped I/O:
every usb port is USB4.0 and TB4 (not an * to be found!)
wifi 7
BT 5.3 LC3 (no *’s)
HDMI 2.1 minimum : 4k 120
DP 2.1 (no *’s)

no gimped pci-express lanes either, you know what I mean…

 

[theorist9]

3nm:
every usb port is USB4.0 and TB4 (not an * to be found!)
no gimped pci-express lanes either, you know what I mean…

I'd like to go further and see 120 Gbps USB4 v2.0, and PCIe 5.0.

TB5 might not be available in time.

 

[Gudi]

The MBP 14/16 already got the M2 Max. Can't possible be a "hand me down" situation when they had them first.

From the MacBook Pro handed down to the Mac Studio. The release date is unimportant. Apple doesn’t need to utilize a chip factory the whole year round, because they don’t own one. Apple only needs to sell as many identical chips as possible to lower development costs per unit. The Ultra is just a by-product of the Max. When they produce the M2 Max, they always get a few M2 Ultra as well. It’s the same wafer cut in different sizes.

 

[avkills]

3. Even if do keep software developer optimization scheduling fragmented between AMD and Apple GPUs the time spent on AMD GPUs optimizations isn't going to help the Intel and Apple GPUs of the mac systems in still in use. The 'win' here is mainly about the needs of the few outweighs the needs of the many. Software optimized to Apple GPUs specifics of unified memory will likely work better up and down the whole product portfolio. Is the objective better software that runs on just one , on vintage/obsolete countdown clock system or the whole line up?

'win/win' is only applicable if doing both.

Apple should do both; at least until Apple Silicon can match or exceed current mainstream GPUs from AMD and nVidia.

I agree that the Mac Pro user base is probably tiny as all get out; but it was also a huge investment that Apple should not take lightly.

And releasing another Intel based Mac Pro would take hardly any R&D. If people can build Hackintosh computers using the latest Intel Consumer CPUs in their basements I am pretty sure Apple can easily make another Xeon based system.

I will gladly eat my words if Apple surprises everyone with a worthy AS Mac Pro; but right now I am not holding my breath.

 

[BrianM_CAN]

From the MacBook Pro handed down to the Mac Studio. The release date is unimportant. Apple doesn’t need to utilize a chip factory the whole year round, because they don’t own one. Apple only needs to sell as many identical chips as possible to lower development costs per unit. The Ultra is just a by-product of the Max. When they produce the M2 Max, they always get a few M2 Ultra as well. It’s the same wafer cut in different sizes.

Apple Silicon Ultra - at least based on the M1 Ultra - is 2 Max chips connected - so it is true that Max production can result in Ultras being made, but there is another step to connect them "Ultrafusion" - and each Ultra takes 2 Max chips away from filling demand for Max based systems. - "Apple’s innovative UltraFusion uses a silicon interposer that connects the chips across more than 10,000 signals" It doesn't look like that is part of the standard Max chip production.

It's possible that they intentionally waited for the M3 for some other advantages before doing the Ultra again. Time will tell.

 

[chucker23n1]

The Ultra is just a by-product of the Max. When they produce the M2 Max, they always get a few M2 Ultra as well. It’s the same wafer cut in different sizes.

You may be thinking of Max vs. Pro. The M1 Pro is basically an M1 Max with the bottom third cut off.

The M1 Ultra, instead, is just two Maxs facing each other.

The M1 without a suffix, though, is a different SoC layout. It was probably created before the Pro/Max/Ultra were ready.

As I recall, the M2/Pro/Max are closer together in their design. The M2 Ultra, of course, doesn't currently exist and perhaps won't.

 

[_Spinn_]

I’m just amazed that they keep finding ways to shrink things further.

 

[theorist9]

I’m just amazed that they keep finding ways to shrink things further.

It ain't easy. A key reason TSMC (and Samsung) have been able to do this, and Intel hasn't, is because the former purchased hugely expensive (~$0.3B each) EUV etching machines from ASML, while Intel has been lagging in their adoption of these devices. That's ironic, since Intel is a part-owner of ASML, and contributed to the consortium that led to the company's formation. For more details, see:

https://forums.macrumors.com/threads/why-your-iphone-mac-will-soon-come-with-intel-manufactured-chips.2371223/?post=31755206#post-31755206

 

[OriginalBaki]

It’s a newer process node. That’s all that really matters.





Can you produce any data for this wild assertion?

Check their quarterly report.

 

[chucker23n1]

Check their quarterly report.

Their quarterly report doesn’t split by high-end, and while numbers are down, they’re up ~30% compared to pre-COVID. 7.2M this year, 10.4M 2022, 9.1M 2021, 5.4M 2020, 5.5M 2019.