3 ms·
Intel lost the high end. They don't seem to know how to make CPUs or GPUs that outperform their competition. But they do have fabs and could in principle make c
by JonChesterfield 3y ago
Intel lost the high end. They don't seem to know how to make CPUs or GPUs that outperform their competition. But they do have fabs and could in principle make chips for other companies.
What they could do is use the equipment they understand really well to make chips for the military. 45nm or 14nm or whatever. Or for automotive. Loads of stuff is useful without being on the bleeding edge.
Instead, they're talking about needing heavy investment in new machines and blaming lack of that for their problems. It's hard to read that as anything other than they're trying to retake the high end crown.
Perhaps the motivating defence angle is AI. The US panicking that TSMC makes better GPUs and tech built on that will end their empire. That's something Intel has decades of failing to execute on.
- wmf 3y agoIntel is great at designing chips but being on 7 nm while everyone else is on 4 nm has been holding them back. (I know nm are fake but not fake enough to invalidate my point.)
- JonChesterfield 3y agoThe process misstep was an error. Sticking with monolithic dies well past their EoL was an error. Missing the low power chips for mobile was an error. Everything they've done in discrete GPUs was an error. Intel make really great WiFi chips and C2D and Nehalem were spectacular. Since then they ran a fantastic profit margin. It's not remotely obvious that they're still great at designing chips.
- magicalhippo 3y ago> Sticking with monolithic dies well past their EoL was an error. Recalling how Intel tried to argue their multi-chip approach was better than AMDs single chip approach back in the early multi-core days, I find this one so hilarious. It's like the tables have turned.
- hypercube33 3y agoDidn't Intel exit the HBM market a few years back and now it's soaring? Also Intel is being a dick about their wifi on AMD systems when that's one of their markets honestly - everyone else's wifi chipsets sucks honestly at least under windows so they should really consider partnering with their enemy instead of trying to fight them. One of the coolest Intel products was that NUC with Intel CPU, igpu, HBM and AMD vega.
- everfrustrated 3y agoTricky when their advertising for like 20 years or more has been "Intel inside". Selling an AMD device with Intel WiFi gets mighty confusing with that as your brand strategy. But yeah, it's still short-sighted.
- scrlk 3y agoIntel was never in the HBM market - you might be thinking of their 3D XPoint/Optane memory developed in collaboration with Micron and discontinued in 2022.
- wmf 3y agoSticking with monolithic dies well past their EoL was an error. I'm going to disagree with this one. Monolithic dies are more energy efficient and usually faster. I think Intel, Nvidia, and Broadcom were right to stay monolithic as long as possible. AMD has saved a lot of money by using chiplets and that strategy worked for them as the underdog but companies with better reputation and high margins can afford monolithic.
- LarsDu88 3y agoIntel's ARC chips are actually pretty good, and would be pretty competitive with NVIDIA for the gaming market if not for the drivers... NVIDIA has three decades of DirectX driver support for many, many old games. Intel's ARC GPUS are infamous for rendering big magenta materials for many games, particularly old ones. ARC however has absolutely nothing to do with Intel Foundries however, as the Intel ARC GPUs are fabbed by TSMC! Intel will certainly get to the point where they fab great GPUs, but its actually a question of whether the software will catch up, not the hardware!
- bmicraft 3y agoAutomotive doesn't pay, like at all
- demondemidi 3y agoNo. Auto pays a 2-10x premium. They have higher temperature demands, high safety demands, and lock in a 10 year design cycle. Automotive is resistant to changing suppliers due to development environment stability needs which allows even more premiums.
- SilverBirch 3y agoThis just isn't how the semiconductor industry works. You invest in leading edge and then you try to find demand for the older stuff. There's tiny margins for the older stuff- because the people choosing to use it are literally choosing it for cost reasons and there's no competitive advantage - they can shop around to any of the fabs. The capital investment is all done and you're just trying to make the most of the sunk costs. It would also be a choice to basically not be a company in 10 years. The low end stuff is the stuff that was bleeding edge 10-15 years ago. If you decide to focus on that older stuff and don't chase after the bleeding edge then in 10-15 years you simply won't have anything competitive at any level, your company will basically be done - your 45nm process will be competing with TSMC's 16nm (which will now be old tech) and you'll just slowly move up and up out of competitive markets.