Quote from: Ava Mee on March 09, 2021, 17:19:39
Apple is crazy, selling phone SoC board at premium price, and saying exceed speed of most Intel chips on market, Intel is following Moore laws, and has not been advancing much since 10 years ago. If you want to compare speed tell us how much your clock speed in terms of Teraflops per sec. A SoC board less than 60 dollars can score 0.5 TF, 700 dollars SoC can score 6 TF + 512 GPU cores + 32GB ram ... seriously what competitive adv. do Apple have to demand for that kind of price? It just asking for too much, GREEDY !!

Umm, the M1 is faster than intel i7 11th gen laptop cpus...

Apple is crazy, selling phone SoC board at premium price, and saying exceed speed of most Intel chips on market, Intel is following Moore laws, and has not been advancing much since 10 years ago. If you want to compare speed tell us how much your clock speed in terms of Teraflops per sec. A SoC board less than 60 dollars can score 0.5 TF, 700 dollars SoC can score 6 TF + 512 GPU cores + 32GB ram ... seriously what competitive adv. do Apple have to demand for that kind of price? It just asking for too much, GREEDY !!

Quote from: Someone (not a fanboy) on March 08, 2021, 12:09:47
Doubling High-Performance core would give only ~60%.
High Performance (4) cores have a frequency of 3.2Ghz and Energy (4) have a frequency of 2Ghz. If all cores can run at maximum clocks (20.8). The share of high performance cores will be 61% which is far less than stated 80%.

The high efficiency Icestorm cores don't have just a lower frequency. AFAIK they're a different ("narrower") design with lower IPC.

Quote from: Someone (not a fanboy) on March 08, 2021, 11:46:51
There is no way that 16 core M2 or M2X will be "4x faster" than M1 (multi core). M1 has 4 high performance cores, M2 or M2X would have 12 high performance cores. Even if we are optimistic, M2/M2X will be ~3x faster than M1, because cores doesn't scale liearly.
The only way it's possible is if it has multi threading support, which CAN give it a boost in multi core performance.

If it was just a question of high performance cores, it would be possible. Just. M2(X) is going to be the next generation. And Apple has seen really nice improvements generation after generation (I believe Apple was averaging around 25 %). In order for it to be 4x faster in a benchmark that scales practically perfectly, they'd have to improve performance of a single core by a third. That's a lot. But not completely impossible. Even "just" 25 % improvement would yield 3.75x increase. Close enough, if you're rounding numbers.

The problem is that it's not just a question of the high performance cores.

The thing is, Apple has so far used numbers in their designations to indicate generations. And letters for core configurations. So, M2 should be a next generation processor in a base configuration. What that base is can change over time. But given that all of this was planned before M1 was released to market, I think it's a fairly safe bet that M2 will end up in a MacBook Air. And it will probably still be a 4+4 chip. They should have their naming scheme all worked out. Whichever way they go about it, M2 should be a direct successor to M1. Meaning, it should be at the lower end of the planned performance spectrum.

Quote from: Jan Onderwater on March 08, 2021, 11:42:49
Apple's roadmap and predictions on how fast the M-Series are going to be.


4 GPU Cores
So, when we double the number of cores, and assume the doubling of High-Performance cores increases the performance by a very conservative 80% (likely to be more like 90%)

M1 already has 8 GPU Cores.
Doubling High-Performance core would give only ~60%.
High Performance (4) cores have a frequency of 3.2Ghz and Energy (4) have a frequency of 2Ghz. If all cores can run at maximum clocks (20.8). The share of high performance cores will be 61% which is far less than stated 80%.

If we have a 12 core "M1X" with 8 high performance and 4 energy cores, it will score ~ 12,100 NOT 13,500. And this would be hard to achieve as CPU cores doesn't scale linearly.
GPU score (Metal) would be +50% (IF it scales well.) as it will have 12 GPU cores instead of 8 in M1.
The score will be : 32,973 NOT 43,964.

There is no way that 16 core M2 or M2X will be "4x faster" than M1 (multi core). M1 has 4 high performance cores, M2 or M2X would have 12 high performance cores. Even if we are optimistic, M2/M2X will be ~3x faster than M1, because cores doesn't scale liearly.
The only way it's possible is if it has multi threading support, which CAN give it a boost in multi core performance.

Apple's roadmap and predictions on how fast the M-Series are going to be.

Let's do some calculations
What do we know:
M1 Series
The M1 4 High Performance Cores, 4 Energy Cores, 4 GPU Cores
Geekbench:
Singe core: 1.709
Multi Core: 7.398
Metal Score: 21.982
Cinebench:
Single Core: 1498
Multi-core: 7508
So, when we double the number of cores, and assume the doubling of High-Performance cores increases the performance by a very conservative 80% (likely to be more like 90%) GPU Scale much better so we double them
M1X 8 High Performance Cores, 4 Energy Cores, 8 GPU Cores (hypothetical)
Geekbench:
Singe core: 1.709
Multi Core: 13.300 (+80%)
Metal Score: 43.964 (Double M1)
Cinebench:
Single Core: 1.498
Multi-core: 13.500
M1Y 16 High Performance Cores, 4 Energy Cores, 16 GPU Cores (hypothetical)
Geekbench:
Singe core: 1.709
Multi Core: 23.940 (+80% M1x)
Metal Score: 78.649 (Double M1x)
Cinebench:
Single Core: 1.498
Multi-core: 24.300
M1Z 32 High Performance Cores, 4 Energy Cores, 32 GPU Cores (hypothetical)
Geekbench:
Single Core: 1.709
Multi- Core: 43.000
Metal Score: 150.000
Cinebench:
Single Core: 1.498
Multi-core: 43.000

M2 Series
M2, The next generation, build on 3NM technology, According to TSMC this will bring: 10–15% (1.10–1.15×) increase in performance. Let's say 10%. So, include another 10% for design improvement, and we are at 20% improvement.
4 High Performance Cores, 4 Energy Cores, 4 GPU Cores (hypothetical)
Geekbench:
Single Core: 2.050
Multi: 8.877
Metal Score: 26.378
Cinebench:
Single Core: 1.800
Multi-core: 16.200
M2X 8 High Performance Cores, 4 Energy Cores, 8 GPU Cores (hypothetical)
Geekbench:
Single Core: 2050
Multi: 16.000
Metal Score: 52.756
Cinebench:
Single Core: 1.800
Multi-core: 29.000
M2Y High 16 Performance Cores, 4 Energy Cores, 16 GPU Cores (hypothetical)
Geekbench:
Single Core: 2050
Multi-core: 29.880 (+80% M2x)
Metal Score: 105.512 (double M2x)
Cinebench:
Single Core: 1.800
Multi-core: 52.000
M2Z High 32 Performance Cores, 4 Energy Cores, 32 GPU Cores (hypothetical)
Geekbench:
Single Core: 2050
Multi-core: 53.800 (+80% M2y)
Metal Score: 210.000 (double M2z)
Cinebench:
Single Core: 1.800
Multi-core: 92.000
M2 ZZ High 64 Performance Cores, 4 Energy Cores, 64 GPU Cores (hypothetical 2023) Geekbench:
Single Core: 2050
Multi-core: 96.800 (+80% M2z)
Metal Score: 840.000 (double M2z)
Cinebench
Single Core: 1.800
Multi-core: 165.000

But these are all (extrapolated) benchmarks scores, real world performance will depend a lot on how much RAM is available. I have seen several tests of the performance difference between the 8GB M1 and the 16GB M1, and only for very heavy use there is a noticeable difference,  so how much impact will there be with 16, 32, 64 or even more on SOC RAM, and since RAM uses a lot of dye space, it is going to be expensive.  And what about upgrading the amount of RAM.
Well, a M-Mac Pro could have the following solution, (example) put in 4 or 8 RAM sticks in parallel on a motherboard in a Quad or Octo Channel configuration and use these as a RAM disk for swapping instead of an SSD. The Memory controller should then:
A: Keep all the active program files on the SOC
B: Swap only data files when needed
C: Assure that when the RAM disk Swap file is full, prioritize which files to swap to the SSD.

5nm GoFLO but X3D/MCM ?

Hope in Qualcomm proc for miniPC.