Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
User predictions for Apple Prosumer Silicon (M1X, M2 etc.)
- Thread starter leman
- Start date
- Sort by reaction score
With respect to chips - those ‘Jade C Die’ rumors still seem to resonate. Would it be reasonable to think the ‘chop’ version be the base of the new pros (8 perf core/2 efficiency/16 gpu) with the non-chop (8/2/32) being an upgrade option?
and then the Jade 2C (16/4/64) and Jade 4C (32/8/128) would be Mac Pro-ish?
and then the Jade 2C (16/4/64) and Jade 4C (32/8/128) would be Mac Pro-ish?
Last edited:
I don’t think you are the only one thinking this. I mostly agree, but I would expect the high end iMac to also offer the Jade 2C, personally.
That is the obvious interpretation, yes. But who knows what will actually happen…
The current MBP 16" and MBP 13" chassis have two substantively different levels of thermal cooling capacity. The MBP 13" does not have another 20-55W of discrete GPU to deal with. It also has an incrementally smaller logic board.
'chop' could be used over both the MBP 14" and 16" but the 16" would need two logic boards ( as the "non chop" pretty likely doesn't use the same size SoC package. If have chopped off 25-40mm of die space there is pretty decent chance that some memory controller also gets chopped off along with that.. That would mean a different number of attached RAM packages. ). This is "chopped" not "binned".
Take the smaller board space 'win' with the smaller package for the MBP 14" and toss in incrementally more battery. The M1, fewer port , MBP 13" manages just fine with just one M1 configuration. Can argue whether the fan in the M1 MBP 13" has a huge impact or not, but even the "chop" should bump the SoC much closer to where the Intel version was (at least at base TDP not the turbo. )
If the MBP 16" can drop from two fans to one fan, It would be a large volume clawback. It wouldn't as pressed to come up with the logic board size savings.
Unless "Unleashed" means putting a Jade 2C in a laptop (with no huge bump in battery life or 'almost silent' acoustics ) .
Jade 2C is likely iMac-ish as well (if not the Mini if don't completely toss the current casa). Doubtful that the 'plain full' Jade is going to cover whole range that the iMac did. And again probably talking about two different SoC package sizes.
I think there will be some binned options too, like maybe the 14in comes with the option for a 14 or 16 core GPU and the 16" options could be 16, 28 or 32.
Couldn’t it just be a binned chip, with half the GPU cores enabled/disabled for the 16 vs 32 core options? That, plus different clock speeds to factor for the lower thermal headroom on the smaller notebook.The current MBP 16" and MBP 13" chassis have two substantively different levels of thermal cooling capacity. The MBP 13" does not have another 20-55W of discrete GPU to deal with. It also has an incrementally smaller logic board.
'chop' could be used over both the MBP 14" and 16" but the 16" would need two logic boards ( as the "non chop" pretty likely doesn't use the same size SoC package. If have chopped off 25-40mm of die space there is pretty decent chance that some memory controller also gets chopped off along with that.. That would mean a different number of attached RAM packages. ). This is "chopped" not "binned".
Take the smaller board space 'win' with the smaller package for the MBP 14" and toss in incrementally more battery. The M1, fewer port , MBP 13" manages just fine with just one M1 configuration. Can argue whether the fan in the M1 MBP 13" has a huge impact or not, but even the "chop" should bump the SoC much closer to where the Intel version was (at least at base TDP not the turbo. )
If the MBP 16" can drop from two fans to one fan, It would be a large volume clawback. It wouldn't as pressed to come up with the logic board size savings.
Unless "Unleashed" means putting a Jade 2C in a laptop (with no huge bump in battery life or 'almost silent' acoustics ) .
Jade 2C is likely iMac-ish as well (if not the Mini if don't completely toss the current casa). Doubtful that the 'plain full' Jade is going to cover whole range that the iMac did. And again probably talking about two different SoC package sizes.
The current MBP 16" and MBP 13" chassis have two substantively different levels of thermal cooling capacity. The MBP 13" does not have another 20-55W of discrete GPU to deal with. It also has an incrementally smaller logic board.
'chop' could be used over both the MBP 14" and 16" but the 16" would need two logic boards ( as the "non chop" pretty likely doesn't use the same size SoC package. If have chopped off 25-40mm of die space there is pretty decent chance that some memory controller also gets chopped off along with that.. That would mean a different number of attached RAM packages. ). This is "chopped" not "binned".
Take the smaller board space 'win' with the smaller package for the MBP 14" and toss in incrementally more battery. The M1, fewer port , MBP 13" manages just fine with just one M1 configuration. Can argue whether the fan in the M1 MBP 13" has a huge impact or not, but even the "chop" should bump the SoC much closer to where the Intel version was (at least at base TDP not the turbo. )
If the MBP 16" can drop from two fans to one fan, It would be a large volume clawback. It wouldn't as pressed to come up with the logic board size savings.
Unless "Unleashed" means putting a Jade 2C in a laptop (with no huge bump in battery life or 'almost silent' acoustics ) .
Jade 2C is likely iMac-ish as well (if not the Mini if don't completely toss the current casa). Doubtful that the 'plain full' Jade is going to cover whole range that the iMac did. And again probably talking about two different SoC package sizes.
Going by M1, we can assume roughly 4-5 watt of sustained power per CPU core and 2 watts of sustained power per GPU core.
A 8/2/16 config would therefore require ~50-60 watts for a workload that fully loads up both the CPU and the GPU (these workloads practically don’t exist). Apples historical chassis TDP for 13“ MBP is 30W, which should allow good hybrid performance on such a SoC. The 16“ TDP is around 70W which would enable higher opportunistic overclock of the basic SoC as well as utilizing the 32-core GPU in hybrid workloads.
Using larger SoCs in laptops does not make any sense as you will quickly run into diminishing returns. Same goes for using the 32-core GPU on the 13/14“ MBP.
I don't think we have any figures for this. I didn't take them into account since they are more situational. If the NPU is doing work, the CPU is probably not pushing the limits of the power envelope.
Couldn't that question be asked for any of the ASICs on board (ISP, video encode/decode etc)?
You forgot to take into account the larger SoC means a bigger Fabric , more system Cache , more DRAM capacity and more DRAM B/W , all contributes to additional watts being used across the SoC due to the added Cores , also we dont know the clock frequencies yet , so if they trade off performance for power your numbers will vary.
Idle will also take a hit due to more leakage both on the SoC and DRAM`s , but larger battery will be more then enough to compensate I would imagine.
You look at AMD CPU`s and a BIG chunk of power is being "wasted" at the infinity fabric.
few days to go , and then a few more for early reviewers to get their hands on those laptops and the leaks will come in fast and furious !! I think around 10 days before it reached consumers we had the M1 benchmarks leaked by early reviewers.
No, they wouldn't bin half the GPU cores. There aren't going to be chips with that many defects that still manage to have perfect CPUs.
The largest binning I could see possible is down to 24.
I wouldn't take the concept of "binning" too literally - selling partly defective chips off cheap may have been the original business model, but cynical me suspects that the proportion of chips sold at each price point will be determined by marketing strategy and the factory will, as if by magic, turn out exactly that proportion of lower-spec chips...
...especially for Apple who are selling systems, not loose chips. The prices of the various Mac models are pretty obviously set by carefully chosen price points, not some "bill of materials + costs + margin" formula. Unless you think that the M1 Macs just happened to work out at exactly the same price as the Intel models they replaced, despite being completely different internally...
The whole overclocking thing with DIY PCs was partly based on the fact that a lot of chips sold as lower clock-speed models worked perfectly fine at higher speeds... and I had a perfectly nice dual processor Celeron system for a while despite the fact that Celeron at the time didn't officially support multi-processor configurations.
Agree but the best way to see the W draw is at outlet to see the maximum...My dell xps is drawing over 121W at outlet...i think my M1 is around 65W
To be fair to AMD I believe a big chunk of the reason why their fabric’s power consumption is so bad is because they manufacture the fabric on a substantially worse node than the rest of the SOC which they aren’t necessarily using by choice. For reasons to do with the split they have to use GloFo for X numbers of chips, so that’s how they spend it. Then again, leading edge TSMC nodes are expensive and constrained so they might do it anyway as if you’ve got to choose which aspect to manufacture on a worse node, the fabric is it.
Apple’s connections are currently on die since everything is monolithic and it must be pretty good/efficient regardless though Apple are pretty tight lipped about it.
A 32/8/128 Mac Mini would rock my world, aint gonna happen though
Of course it won't happen. You are talking about 300+ watts in a Mac mini chassis. How would that work?
Well, the numbers I used already include many of those factors, but of course, it kind of depends on what kind of technology (RAM type etc.) they end up using. You can add 10% or so for these things if you want, I don't think it changes the picture too much.
It is kind of optimized for a different task though (it has to route the signals through the substrate outside the chip).
Yes of course they will limit some good working chips to meet the market demand for the cheaper models, but this is Tim Cook he's not going to waste money on large dies to turn half of them off (heck I can't think of anyone in recent memory limiting a chip by 50%). Plus we have already seen the rumors showing as much that there will be a 16 and 32 GPU core chip. The smart bet is on there being 2 separate chips.
You'd have to ask someone into chip fabbing - but I'm sure that there is a significant cost penalty whenever you have to make two different versions of something.
It wouldn’t be 50% of the chip though. It would be 50% of the GPU portion of the chip. I don’t know how the cost breaks down, but it may be cheaper in the long run to just buy 1 chip in bulk than to create and manufacture 2 separate chips. I guess we will likely find out soon though!
Let's see how I did.
Clearly a new architecture but we still don't know if it is closer to the M1 or the A15 for the cores. Also we don't know about the clock speeds either. 8 performance cores on the M1 Pro with up to 10 performance cores as an option. Up to 16 GPU cores on the M1 Pro. With up to 32 GPU cores on the M1 Max. So I was too conservative. Apple outdid themselves with the M1 Max.
Up to 64 GB of LPDDR5 on the M1 Max with up to 32 GB on the M1 Pro. So I got that one. It looks like the two to three 6K displays was correct too with 2 on the M1 Pro and 3 on the M1 Max. It doesn't look like the HDMI is actually HDMI 2.1 so I'm not sure of the point of having a dedicated HDMI port but since I didn't predict anything about that, I'll say its a wash.
Overall, I think I got it about 85% right with some more information still waiting to be found out.
Edit: And I was right about the SoC not being called M1X or M2.
The Pro MAX was not crazy in price.
I paid about the same for my 2019 16" i9 MBP maxed out.
I paid about the same for my 2019 16" i9 MBP maxed out.