I sometimes wonder if it would make sense for Apple to include a few FPGA cores in their upcoming M(X) chips (if that is even possible). With those they could do even more specialized accelerated tasks.
On a wild imaginary train of thought; I'm not sure how well a FPGA can be integrated into an existing CPU but it could maybe be possible to define new instructions using them. That way Apple can make the chips forward-compatible with new instruction set extensions. Or if Apple finds a serious security problem with one of the existing instructions (in the unmodifiable silicon) they can make Rosetta transpile all untrusted programs to use a fixed FPGA'd replacement for those instructions.
wow, iirc this is similar to how the original alto was implemented... everything was microcode and new instructions could be added at any time and run at full hardware speed... back to the future?
It will be really exciting to see how Intel will respond to this in the coming months. In the consumer laptop space Dell, HP, and Lenovo seem to be committed to Intel.
As for the gaming/hobbyist market, AMD seems to be winning given that Intel doesn't seem to want to play. Comparing the marketing materials for Ryzen Threadripper and the i9 really highlights the difference in marketing approaches.
"Third Point encouraged Intel to consider separating its chip design and manufacturing divisions, and instead seek a joint venture in manufacturing. It was also urged to divest its failed acquisitions, such as the $16.7 billion purchase of programmable chip maker Altera in 2015."
It's this a sound decision? It seems like it would be intel abandoning smaller nodes and sticking to 14nm whilst they go to Samsung or TSMC to manufacture their chips.
It's losing their competitive advantage. Of course they've been struggling so far with 10nm.
>It's this a sound decision? It seems like it would be intel abandoning smaller nodes and sticking to 14nm whilst they go to Samsung or TSMC to manufacture their chips.
I dont read that is what they are suggesting. Basically they want Intel Foundry and Intel Chip as separate business entity. Intel Chip will continue to use Intel Foundry while Intel Foundry will go full force and compete with TSMC. In investment terms, that is unlocking the potential in their stock price, which is trading at ~10 P/E while all other Tech stock are at 50+. ( Nvidia ~85, AMD ~125 )
Whether that is a sound stratergy depends on execution. There is one assumption that we could consider as truth, Intel will need higher unit volume to amortised the cost of leading edge R&D and lower per unit cost to stay competitive with TSMC. With this precondition, there is only two way around it, either open up your Fab and compete as suggested, or doing what Intel is doing now which is to massively expand their market with NPU, GPU, FPGA, Network Processor and all other reasonable margin product segments.
Credit where Credit's due, Intel's sales and marketing department have always been some of the best in the industry. All of the timely and intentional "leaks" as well as sales incentive and delaying tactics are working exceptionally well. But they dont change the fundamentals, and I am not sure if what they are doing now are radical enough to change course in the long term.
That sounds like terrible advice from a hedge fund interested in short-term profits, not a good plan for the long-term success of a company which depends on technology and engineering.
On a wild imaginary train of thought; I'm not sure how well a FPGA can be integrated into an existing CPU but it could maybe be possible to define new instructions using them. That way Apple can make the chips forward-compatible with new instruction set extensions. Or if Apple finds a serious security problem with one of the existing instructions (in the unmodifiable silicon) they can make Rosetta transpile all untrusted programs to use a fixed FPGA'd replacement for those instructions.