I wouldn't bet on intel unless some convincing data emerges. Biased and very opinionated person here, but I have seen first hand their underhanded tactics in stifling nvidia's innovation. They released code claiming their offerings were 'better' (i.e. GPU raytracing was trash), while being extremely difficult to reproduce. Back in the day (circa 10-12 years ago), intel was still king, and nvidia was value. As an insider, I knew this. nvidia invested in CUDA and the ecosystem. Intel spent time denying this. It was lucky of course that machine learning emerged to make favourable use of GPUs. Nonetheless, intel absolutely deserves their fate. Back then, we were told that CUDA is hard to learn (yes it is, but its pure joy), and no one will use it. Supercomputer users were scared to use CUDA, and orders went to intel (truth be told, these 'legacy' code users are also a retrograde lot - fluid dynamics has seen zero innovation in ages and are still running legacy fortran code, while the rest of the world has moved on). Intel was largely responsible for crippling PCIe speeds so that transfers to GPU were capped.
See this paper published in 2010. While some of it might have been technically sound, we can clearly see that they were going backwards.
Here’s to hoping! Intel has been a massive value trap for me to date (since 2021/2022) but I agree with your points. I'm not selling but I've stopped accumulating after averaging down the past few years.
I wouldn't bet on intel unless some convincing data emerges. Biased and very opinionated person here, but I have seen first hand their underhanded tactics in stifling nvidia's innovation. They released code claiming their offerings were 'better' (i.e. GPU raytracing was trash), while being extremely difficult to reproduce. Back in the day (circa 10-12 years ago), intel was still king, and nvidia was value. As an insider, I knew this. nvidia invested in CUDA and the ecosystem. Intel spent time denying this. It was lucky of course that machine learning emerged to make favourable use of GPUs. Nonetheless, intel absolutely deserves their fate. Back then, we were told that CUDA is hard to learn (yes it is, but its pure joy), and no one will use it. Supercomputer users were scared to use CUDA, and orders went to intel (truth be told, these 'legacy' code users are also a retrograde lot - fluid dynamics has seen zero innovation in ages and are still running legacy fortran code, while the rest of the world has moved on). Intel was largely responsible for crippling PCIe speeds so that transfers to GPU were capped.
See this paper published in 2010. While some of it might have been technically sound, we can clearly see that they were going backwards.
https://users.ece.utexas.edu/~ljohn/teaching/382m-15/reading/lee.pdf
I am happy to see intel go lower and lower (possibly bankrupt) and will buy if more data arrives, or if we are sure it is a global minimum.
Here’s to hoping! Intel has been a massive value trap for me to date (since 2021/2022) but I agree with your points. I'm not selling but I've stopped accumulating after averaging down the past few years.