Thanks for sharing this. I don't take all of JEDEC's info about their own memory at face value, but understand that lower speeds could be used for less power. Still, I recommend you take a look at this study, especially Section 5.3. In broad strokes, you'll see HBM2 using much more power than LPDDR4. As HBM2E uses more power than HBM2, and LPDDR5 uses less power than LPDDR4, this gap will widen in 2021. Let me know if you still think this can be surmounted with lower speeds.View attachment 26709
HBM2/E is still the overall lower power option.
There is also multiple speed bin options available; Speed bins: 0.5, 1.0, 1.5, 1.6, 1.8, 2.0, 2.4, 3.0, 3.2 Gbps
LPDDR5 is 6.4 Gb/s => (128 * 6.4 ) / 8 => 102.4 GB/s
HBM2E to do the above => ( 102.4 * 8 ) / 1024 => 0.8 Gb/s
Samsung's 110mm2 HBM2E 16GB archieves 5 Gbps @ 1.1V and SK Hynix's 110mm2 HBM2E 16GB achieves 4 Gbps @ 1.14V.
If Samsung/SK Hynix ever come out with a Low-Power 512-bit it only needs 1.6 Gb/s to beat LPDDR5. With the Low-Power & High-Bandwidth Memory probably being even lower power than HBM2/HBM2E is at it's higher speed.
I think the Infinity Link style solution is likely if not simply necessary at some point. I doubt Apple wants to print lots of 200+ mm2 dies. Granted, they could use some more sophisticated smart interposer, but Infinity Link is simple and within immediate reach.I think it depends on the form factor. Design considerations for the MacBook Air are VERY different than the MacBook Pro.
It’s still amazing to me that the iMac Pro essentially has a 1U server running quiet hanging off the back of a display.
I expect there will be some form of deviation from everything being SOC when we look at say the Mac Pro. They could theoretically develop their own version of AMDs Infinity Fabric Link to maximize the ability for separate SOCs to communicate. The MPX Connector could lead the way to a separate card that has multiple purpose built chips like large blocks of Apple designed GPU cores. Granted they could have their own memory dedicated to them.
Moreover, Apple has always found creative ways to reuse parts. I don't think they want to design a zillion APUs when they could conceivably get away with 4 parts on different chiplet packages.
A14: Goes in the iPhone and Air.
A14X: The Bloomberg APU, 8 perf cores and 8-16 GPU cores. Goes in the iPad Pro, MBP13, MBP16, Mac Mini, iMac.
GPU: Die with only GPU cores. One goes on package with with the A14X in the MBP16 and upgraded iMacs. Two get slapped on a package together for (i)Mac Pros.
CPU: Die with only CPU perf cores, ~16. (i)Mac Pro part; can slap two on a package for high end configs.