Doesn't make sense. It's already server-grade packaging. Perhaps better than other server-grade packaging.
I can think of 3 reasons this would make sense:
1. If Apple is combining their silicon with third party chips from Nvidia or semthing.
2. Alternatively, another way it could make sense is if Apple is looking to add more GPU cores to various iterations of its chips, without increasign CPU core counts. i.e. having multiple sets of max/Ultra chips - one set for laptops, one for Mac Studio, and another for Mac Pro.
or...
3. Appel could be redoing the way it tiers its chip lineup. CPU could be the same, but GPU would be different for each tier.
Interesting to see how this develops.
The bigger the chip is, the more it costs. The price of a chip, other than for R&D expenditures, is for chip area. There is just so much room you have. By breaking the GPU out and putting it in the sub board using ultra fast connections, as RAM does now, would allow several things. One is having much more area for the GPU. The GPU has been criticized compared to Nvidia. Maybe we’ll see a big improvement there.
The second thing, if Apple decides not to not shrink the SoC as a result, there will be a lot of space available on that chip. So possibly bigger, more powerful CPU cores. Maybe more of them. Possibly a much bigger Neural Engine. It doesn’t take up a lot of space now, so possibly it could even double in size. There are a number of things Apple could do here.
Apple is still in conflict with Nvidia, something I just read from a good source. So I highly doubt we’d see a board from them. Apple wants small and efficient. That’s the total opposite of what Nvidia and AMD are doing with their GPUs. Both have top power draw approaching 500 watts. No way Apple will allow that.
Multi-chiplet gives Apple options. The ultras so far have been dual chips of the same type, but that doesn’t have to be the case. And they might be able to do edge connectors on more than one edge. More expensive and higher power though, so mostly for desktops and servers. Imagine one chip with just CPUs, and one with just GPUs… and connectors on two edges instead of one. That would create a lot of permutations — all CPU, 3 GPU + 1CPU, 2+2. Put a core or two on the GPU chiplet and then you could have “GPU only” combination. Lots of options, which makes for some interesting possibilities for the pro desktop/server lineup.
None of which means that consumer machines wouldn’t still be single chip SoCs.
CXL would give more options and space to work with in the machine Including more options 3rd party GPU. Could make sense of Apple nVidia team up. Rumours around model training.
A Mac Pro with MPX modules returning could make sense of a machine designed to be modular but basically not modular due to the pipeline they knew was coming
Multi-chiplet gives Apple options. The ultras so far have been dual chips of the same type, but that doesn’t have to be the case. And they might be able to do edge connectors on more than one edge. More expensive and higher power though, so mostly for desktops and servers. Imagine one chip with just CPUs, and one with just GPUs… and connectors on two edges instead of one. That would create a lot of permutations — all CPU, 3 GPU + 1CPU, 2+2. Put a core or two on the GPU chiplet and then you could have “GPU only” combination. Lots of options, which makes for some interesting possibilities for the pro desktop/server lineup.
None of which means that consumer machines wouldn’t still be single chip SoCs.
CXL would give more options and space to work with in the machine Including more options 3rd party GPU. Could make sense of Apple nVidia team up. Rumours around model training.
A Mac Pro with MPX modules returning could make sense of a machine designed to be modular but basically not modular due to the pipeline they knew was coming
No way. That’s not going to happen. Apple has moved in the opposition. They won’t use a GPU that runs through the PCIe bus as other video cards do. Using that would riyn performance. People need to understand how Apple’s chip work before proposing these “solutions”. The entire point of what Apple does is extremely tight integration. The RAM that’s used with the CPU is used with the GPU. But not the wat IG uses it on Intel chips. Going to a third party GPU would destroy Apple’s entire concept. Apple wants to use their own GPU the way they’ve been using it.
Multi-chiplet gives Apple options. The ultras so far have been dual chips of the same type, but that doesn’t have to be the case. And they might be able to do edge connectors on more than one edge. More expensive and higher power though, so mostly for desktops and servers. Imagine one chip with just CPUs, and one with just GPUs… and connectors on two edges instead of one. That would create a lot of permutations — all CPU, 3 GPU + 1CPU, 2+2. Put a core or two on the GPU chiplet and then you could have “GPU only” combination. Lots of options, which makes for some interesting possibilities for the pro desktop/server lineup.
None of which means that consumer machines wouldn’t still be single chip SoCs.
CXL would give more options and space to work with in the machine Including more options 3rd party GPU. Could make sense of Apple nVidia team up. Rumours around model training.
A Mac Pro with MPX modules returning could make sense of a machine designed to be modular but basically not modular due to the pipeline they knew was coming
No way. That’s not going to happen. Apple has moved in the opposition. They won’t use a GPU that runs through the PCIe bus as other video cards do. Using that would riyn performance. People need to understand how Apple’s chip work before proposing these “solutions”. The entire point of what Apple does is extremely tight integration. The RAM that’s used with the CPU is used with the GPU. But not the wat IG uses it on Intel chips. Going to a third party GPU would destroy Apple’s entire concept. Apple wants to use their own GPU the way they’ve been using it.
I agree, I don’t think Apple will give up the vertical integration they’ve built their systems on. And I don’t think they need to. Modular chiplets combined in a package would give them a substantial boost in GPU performance, and their low power advantage would let them put more in a single chassis. Cooling multiple lower power chips is easier than cooling one monster GPU that burns as much power. There is more communication overhead in the former, however depending on the mix of workloads, that may not be an issue. Having the fastest peak performance is usually not as important as having the best net throughput. Cloud servers most often deal with a steady large flow of concurrent incoming requests, not one at a time.
CXL would give more options and space to work with in the machine Including more options 3rd party GPU. Could make sense of Apple nVidia team up. Rumours around model training.
A Mac Pro with MPX modules returning could make sense of a machine designed to be modular but basically not modular due to the pipeline they knew was coming
CXL brings its own issues and isn’t going to work well with GPU memory demands. It is more intended for elastic cloud re-configuration.
Comments
The second thing, if Apple decides not to not shrink the SoC as a result, there will be a lot of space available on that chip. So possibly bigger, more powerful CPU cores. Maybe more of them. Possibly a much bigger Neural Engine. It doesn’t take up a lot of space now, so possibly it could even double in size. There are a number of things Apple could do here.
Apple is still in conflict with Nvidia, something I just read from a good source. So I highly doubt we’d see a board from them. Apple wants small and efficient. That’s the total opposite of what Nvidia and AMD are doing with their GPUs. Both have top power draw approaching 500 watts. No way Apple will allow that.
CXL brings its own issues and isn’t going to work well with GPU memory demands. It is more intended for elastic cloud re-configuration.