1) Maybe the simplest way to frame Intel’s current woes @FoolAllTheTime @benthompson
Even if they had won the iPhone, embraced ARM and nailed GPUs with a non x86 Larrabee, they would still be losing share and in deep trouble due to the timing of EUV insertion.
Even if they had won the iPhone, embraced ARM and nailed GPUs with a non x86 Larrabee, they would still be losing share and in deep trouble due to the timing of EUV insertion.
2) i.e. Apple would be moving away from them to TSM and they would have zero share in discrete GPU given a two node disadvantage in GPU with none of the advantages in CPU conferred on them by x86 and Intel specific software optimizations.
3) Alternatively, if they had successfully inserted EUV at the same time as TSM sans iPhone, sans GPU, sans ARM, they would be in a better position today than in the above scenario.
If you are 2 nodes behind, nothing else really matters.
If you are 2 nodes behind, nothing else really matters.
4) Important to remember x86 share loss in revenue $ terms dwarfs any potential GPU share gains (for now).
But yes, ideally they would have embraced the iPhone, designed a good GPU 6ish years ago, created a CUDA like tool chain for AI and executed on EUV.
But yes, ideally they would have embraced the iPhone, designed a good GPU 6ish years ago, created a CUDA like tool chain for AI and executed on EUV.
5) There is hope especially with yesterdays events. Likely CXL dominance makes them the only real LT competitor to NVDA post MLNX/ARM. Data center as a computer is real. AMD will eventually need an answer here. EMIB, Foveros are ahead of TSM. Just need to succeed with EUV.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
