3 ms·

Anyone who writes for a platform other than x86 is going to significantly complicate cross-platform ports for their platform. The other competitor would retain

by paulmd 7y ago

Anyone who writes for a platform other than x86 is going to significantly complicate cross-platform ports for their platform. The other competitor would retain commonality with PC. So the days of POWER or Cell or whatever are done. Once you accept x86 as a given, there’s only two companies of any significance on the market. You could hypothetically do an Intel CPU and an NVIDIA or AMD discrete GPU but then you don’t have an SOC and that increases system integration costs. You’d have to talk two competitors into licensing you their IP for integration and it’s just not really either easy or something those companies really do lightly or often. AMD is basically the only game in town for a fast CPU core integrated with a fast GPU on a single package, at least for now. After Intel starts getting serious about XE you could hypothetically see a high performance SOC. As in many areas, the world would be drastically different if patent law didn’t protect a processor’s “API” and x86 could be freely implemented by many companies. For some reason it’s (probably) legal with a software API but not a hardware one, despite the “magic” all being in the actual implementation and the instructions just being “function calls”.

keenmaster 7y ago
Non-technical people can't begin to appreciate this problem, much less bother to solve it or even know how. Engineers need to lobby. Have you considered meeting your senator and creating a powerful coalition of interested parties to bring about legislative change? You can even recruit an economist to estimate the annual loss in economic surplus due to hardware API lock-in. Politicians often cave to narrowly interested groups that complain about a very tangible grievance (e.g. a NIMBY screaming "that building would block my view and increase traffic!") or very large corporations with a lot of resources. It doesn't have to be that way.
- mjevans 7y ago
  You might be able to get the average person on board if approached from a repair specifications angle. What a given part does, what size it does it within, where it links up to other parts (with what sort of links), and how well it does the thing are all facts. After knowing those facts a 'cleanroom' implementation of a generic part shouldn't infringe on anything. Convince the average person of that, and then extrapolate that to making a machine that responds the same exact way to a 3rd party's recipe. Often, given physics and current or near possible tech, there's only a small number of possible answers, and for a given set of criteria one is going to be 'the best' answer. Physics and Math are tools that express how nature works.
ksec 7y ago
>AMD is basically the only game in town for a fast CPU core integrated with a fast GPU on a single package, at least for now. After Intel starts getting serious about XE you could hypothetically see a high performance SOC. Intel could have done it with there Iris Graphics as well if they wanted to. The problem is Intel doesn't like low margin, the are too focused on their high margin they let slip the iPhone SoC and Consoles SoC. Not saying this is a bad thing, financially Intel is still doing great

by paulmd 7y ago

keenmaster 7y ago

Non-technical people can't begin to appreciate this problem, much less bother to solve it or even know how. Engineers need to lobby. Have you considered meeting your senator and creating a powerful coalition of interested parties to bring about legislative change? You can even recruit an economist to estimate the annual loss in economic surplus due to hardware API lock-in. Politicians often cave to narrowly interested groups that complain about a very tangible grievance (e.g. a NIMBY screaming "that building would block my view and increase traffic!") or very large corporations with a lot of resources. It doesn't have to be that way.

mjevans 7y ago
You might be able to get the average person on board if approached from a repair specifications angle. What a given part does, what size it does it within, where it links up to other parts (with what sort of links), and how well it does the thing are all facts. After knowing those facts a 'cleanroom' implementation of a generic part shouldn't infringe on anything. Convince the average person of that, and then extrapolate that to making a machine that responds the same exact way to a 3rd party's recipe. Often, given physics and current or near possible tech, there's only a small number of possible answers, and for a given set of criteria one is going to be 'the best' answer. Physics and Math are tools that express how nature works.

ksec 7y ago

>AMD is basically the only game in town for a fast CPU core integrated with a fast GPU on a single package, at least for now. After Intel starts getting serious about XE you could hypothetically see a high performance SOC. Intel could have done it with there Iris Graphics as well if they wanted to. The problem is Intel doesn't like low margin, the are too focused on their high margin they let slip the iPhone SoC and Consoles SoC. Not saying this is a bad thing, financially Intel is still doing great