M4 Mac Announcements

[NotEntirelyConfused]

10P+4E for the Mac Mini with M4 Pro

Yes, that and the 273GBps memory are better than even optimistic predictions.

And what about TB5??? I would have bet against that. The chip is a tour de force.

Now, as for the Max, I see four options:
1) Just more GPU, same mask as Pro, uncut. I hope not, but plausible. Same idea as M1.
2) Just more GPU, same mask... but, the sixth P core in each cluster is enabled. So, 12P + 4E.
3) Another P cluster, more GPU. 15 or 18 P, 4 or 6 E. 30-40 GPU cores. Still the same mask.
4) Different mask, all bets are off.

 

[Nycturne]

The M4 line looks like a worthy upgrade for my M1 MBP.

The question is what do I want to do here? I don’t think I’d need the extra oomph from the M4 Max, especially as I don’t really push the limits of the M1 Max GPU (I have a gaming rig for that). I also tend to use my 16” MBP at the desk more than ever, so a Mac Mini with the M4 Pro would still be a pretty nice upgrade. Or perhaps I just drop down to the 14” MBP with the M4 Pro and get more useful portability so I’m more likely to use the MBP away from the desk instead of reaching for the iPad? Decisions, decisions.

And what about TB5??? I would have bet against that. The chip is a tour de force.

DisplayPort 2.1 as well. From the specs, I’m not sure the display engines can feed such a beast at full clip, but good to see this is in the pipe.

Now, as for the Max, I see four options:
1) Just more GPU, same mask as Pro, uncut. I hope not, but plausible. Same idea as M1.
2) Just more GPU, same mask... but, the sixth P core in each cluster is enabled. So, 12P + 4E.
3) Another P cluster, more GPU. 15 or 18 P, 4 or 6 E. 30-40 GPU cores. Still the same mask.
4) Different mask, all bets are off.

My guess is along the lines of #3: Another P cluster (similar to M3 Max), 12/15 P and 4 E. I’d be surprised if the M4 Pro is using a mask with 12 P cores on it. Unless the MBP shows up with a 16 core M4 Pro option.

 

[leman]

Seems pretty obvious to me.

M4 Pro has LPDDR5X-8533 mated to a 256 bit memory bus.

= (8.533 ÷ 8) × 256
= 273 GB/s.

LPDDD5X-8533 has been in the market for quite a while now. For example, Intel's Lunar Lake has LPDDR5X-8533 mated to a 128 bit memory bus, which gives it 136 GB/s of bandwidth.

Thanks, you are right, that would do it!

Do you happen to know a good resource or guide for converting the RAM specs into more useful metrics? For example, why are you dividing the MT/s by 8? I always get confused by these things.

 

[Cmaier]

Thanks, you are right, that would do it!

Do you happen to know a good resource or guide for converting the RAM specs into more useful metrics? For example, why are you dividing the MT/s by 8? I always get confused by these things.

8533 is megabits per second, so get to bytes by dividing by 8?

 

[leman]

8533 is megabits per second, so get to bytes by dividing by 8?

I thought this number is supposed to meant "megatransfers per second"? I always found the concept confusing. Does it simply mean that this is the effective data signaling rate per bit of the bus, and you have to multiply it by bus width to get gigabits/sec?

 

[theorist9]

The M4 Mini with the upper-end Pro chip (14 CPU/20 GPU), 64 GB RAM, 2 TB SSD and 10 Gb Ethernet is $2900. The M2 Studio with the lower-end Max chip and 64 GB/2 TB/10 Gb is $3000.

So if you're thinking of the getting an upper-end M4 Pro Mini, the M4 Max Studio is likely to be the better value* (Max chip, more ports) (assuming it's priced the same as the M2; and it might even have more RAM for the same price).

*Unless your metric is computing power/($ * cm^3).

 

[Nycturne]

Things get a bit more interesting at the 2000$ price range, IMO. Depends on what the M4 Max looks like, and if the Studio gets refreshed this week or not.

Pro + 1TB + 48GB RAM vs Max + 512GB + 32GB RAM for 1999$.

 

[NotEntirelyConfused]

I thought this number is supposed to meant "megatransfers per second"? I always found the concept confusing. Does it simply mean that this is the effective data signaling rate per bit of the bus, and you have to multiply it by bus width to get gigabits/sec?

Yes, exactly that.

 

[theorist9]

Things get a bit more interesting at the 2000$ price range, IMO. Depends on what the M4 Max looks like, and if the Studio gets refreshed this week or not.

Pro + 1TB + 48GB RAM vs Max + 512GB + 32GB RAM for 1999$.

Agreed. As you drop down in price, the Pro becomes more appealing. I'd take the Pro in that comparison.

I'd be pleasantly surprised if we see an M4 Studio this week. Gurman, who's been reasonably accurate about their release schedule, said there would be no M4 Studio until mid-2025.

 

[casperes1996]

10P, 4E is not at all what I expected, but I'm all for it!

Means we have no pattern at all to go on though.
M1 Pro: 10+2
M2 Pro (same)
M3 Pro: 6+6
M4 Pro: 10+4

 

[NotEntirelyConfused]

3) Another P cluster, more GPU. 15 or 18 P, 4 or 6 E. 30-40 GPU cores. Still the same mask.

My guess is along the lines of #3: Another P cluster (similar to M3 Max), 12/15 P and 4 E. I’d be surprised if the M4 Pro is using a mask with 12 P cores on it. Unless the MBP shows up with a 16 core M4 Pro option.

If you are on #3, you can't get 12/15P. It would have to be 15/18, as I said, since clusters are either 5 or 6P, and you're adding a cluster to the existing 10P (or 12P if it's 6P with one disabled).

BTW I'm guessing it's an actual 5P cluster, not 6P with one fused (so 15 P cores in the Max, if Apple picks door #3). I think the latter would be too much wasted good silicon. Though I can, just barely, imagine them using fused/binned Pros in the mini, and keeping the full 6P/cluster parts for the MBPs.

 

[casperes1996]

Also; Intel naming...

 

[NotEntirelyConfused]

You know what would be a really nice option for the Mini (and Studio, and even iMac)? 10Gbps Ethernet, *not* on copper. I wouldn't want actual connectors on the back - just an SFP+ port. I'm a little surprised they haven't done that. It's bigger, yes, but not that much bigger. It ought to fit.

 

[theorist9]

When do the review embargos typically end? Is it on the ship dates (Nov. 8)?

 

[Nycturne]

Means we have no pattern at all to go on though.
M1 Pro: 10+2
M2 Pro (same)
M3 Pro: 6+6
M4 Pro: 10+4

Small correction:

• M1 Pro: 8+2
• M2 Pro: 6+6
• M3 Pro: Same
• M4 Pro: 10+4

If you are on #3, you can't get 12/15P. It would have to be 15/18, as I said, since clusters are either 5 or 6P, and you're adding a cluster to the existing 10P (or 12P if it's 6P with one disabled).

3 clusters of 5P cores for the "full fat" Max, 3 clusters of 4 in the low-end variant by fusing off a core in each cluster. Hence 12/15. This matches what the M4 Pro is: 2 clusters of 5P, with 2 fused off in the low-end variant.

I'm not sure how you get 18 performance cores from clusters of 5 as the high end variant.

 

[theorist9]

You know what would be a really nice option for the Mini (and Studio, and even iMac)? 10Gbps Ethernet, *not* on copper. I wouldn't want actual connectors on the back - just an SFP+ port. I'm a little surprised they haven't done that. It's bigger, yes, but not that much bigger. It ought to fit.

For those of us not familiar: Do you mean you want the 10 Gb option to be a combined fiber optic + copper ethernet port? And what's a connector—is that the termination on a fiber optic cable that allows it to plug into such a port?

 

[Nycturne]

For those of us not familiar: Do you mean you want the 10 Gb option to be a combined fiber optic + copper ethernet port? And what's a connector?

SFP+ moves the transceiver to a daughterboard that plugs into the port. So you can either use a transceiver that exposes RJ-45, or an optical connection for standard optical cables, or even fully integrated cables. I bet data centers hosting Macs would love it.

 

[NotEntirelyConfused]

Small correction:

• M1 Pro: 8+2
• M2 Pro: 6+6
• M3 Pro: Same
• M4 Pro: 10+4

You're both wrong. M2 Pro was 6/8P and 4E.

3 clusters of 5P cores for the "full fat" Max, 3 clusters of 4 in the low-end variant by fusing off a core in each cluster. Hence 12/15. This matches what the M4 Pro is: 2 clusters of 5P, with 2 fused off in the low-end variant.

I'm not sure how you get 18 performance cores from clusters of 5 as the high end variant.

Oh, I see what you mean. I wasn't talking about low/high configurations. I was talking about possibilities. That is, 15P if the cluster is truly 5P. 18P if it's 6P with one fused off in the high-config Pro, but not in the Max. But I expect it's a true 5P cluster, as you also think.

 

[theorist9]

And what about TB5??? I would have bet against that. The chip is a tour de force.

Yeah, the TB5 is exciting.

 

[Nycturne]

You're both wrong. It was 6/8P and 4E.

4 things were listed (all as the "full fat" variants), you need to be more specific on which "it" you are referring to here. I have to assume you mean M2 Pro here?

Oh, I see what you mean. I wasn't talking about low/high configurations. I was talking about possibilities. That is, 15P if the cluster is truly 5P. 18P if it's 6P with one fused off. But I expect it's a true 5P cluster, as you also think.

Considering the M4 Pro already has two variants: 8+4 and 10+4, I'd be very surprised if it was a cluster of 6. That would mean fusing off 4 cores for the 8+4 Mac mini being offered for pre-order right now.

Agreed. As you drop down in price, the Pro becomes more appealing. I'd take the Pro in that comparison.

So would I, but I'm also starting to get into the realm of astro photography where my working data set for a project can be in the hundreds (yes, plural) of GB. The extra storage matters. But for a pure code machine? I might take the Max to churn through data a bit faster.