wow, that is very very bad news. Either Nvidia is setting itself up to dominate its competitors or it is setting up ARM to have a vastly reduced market space. If this goes through I think we will see massive upheavals in this space.
Yes. Big upheaval. Apple just made a huge bet on ARM. 90% of the world's smartphones use ARM.
The best outcome would be a shift to an open-hardware architecture that no one owns. RISC-V is one possibility, but it is based on MIPS and has architectural limitations. A new clean design by an independent foundation would be best.
A new clean design by an independent foundation would be best.
Because designs produced by committee have such a great track record.:-(
The VAX instruction set was designed by a committee, and it had a pretty good track record.
The IBM System/360 instruction set was probably designed by a committee, and it also did pretty well.
The VAX instruction set was designed by a committee, and it had a pretty good track record.
I enjoyed working on various vaxen, but I wouldn't use VAX as an example of success. It had a brief moment in the sun, but then it died. Completely died. If the VAX architecture is discussed today, it's usually as an example of what not to do.
The IBM System/360 instruction set was probably designed by a committee, and it also did pretty well.
I agree that (1) S/360 was a hugely successful and influential architecture, and (2) it was probably designed by a team rather than a single architect. But a "team" isn't the same as a "committee". Even assuming good faith by all participants -- which a questionable assumption -- industry standards committees are orders of magnitude slower than internal company teams. For every success, they produce are at least two camels ("A camel is a horse designed by committee").
The VAX instruction set was designed by a committee, and it had a pretty good track record.
I enjoyed working on various vaxen, but I wouldn't use VAX as an example of success. It had a brief moment in the sun, but then it died. Completely died. If the VAX architecture is discussed today, it's usually as an example of what not to do.
The IBM System/360 instruction set was probably designed by a committee, and it also did pretty well.
I agree that (1) S/360 was a hugely successful and influential architecture, and (2) it was probably designed by a team rather than a single architect. But a "team" isn't the same as a "committee". Even assuming good faith by all participants -- which a questionable assumption -- industry standards committees are orders of magnitude slower than internal company teams. For every success, they produce are at least two camels ("A camel is a horse designed by committee").
Almost all computer instruction sets eventually die, and I have not seen the VAX instruction set discussed as an example of what not to do. As far as I know, it is still considered a good example of a Complex instruction set.
I don't know what the difference is between a "team" and a "committee". The people who designed DDCMP, an ancestor of Ethernet, called themselves a "committee", even though we all worked for DEC. Ethernet itself was mostly designed by people from Xerox, Intel and DEC, then turned ove
Almost all computer instruction sets eventually die
Sure, most architectures eventually die. But some influence what comes after them, while others simply go extinct. CISC in general, and VAX in particular, went extinct.
As far as I know, VAX is still considered a good example of a Complex instruction set.
VAX is a great example of CISC architecture. My point is that such architectures are now considered a Bad Idea. That doesn't mean VAX was bad idea at the time it was created -- it just means that the world has moved on. No new designs similar to VAX have been created in a very long time.
I don't know what the difference is between a "team" and a "committee".
I assume you've never participated in an industry standa
Almost all computer instruction sets eventually die
Sure, most architectures eventually die. But some influence what comes after them, while others simply go extinct. CISC in general, and VAX in particular, went extinct.
As far as I know, VAX is still considered a good example of a Complex instruction set.
VAX is a great example of CISC architecture. My point is that such architectures are now considered a Bad Idea. That doesn't mean VAX was bad idea at the time it was created -- it just means that the world has moved on. No new designs similar to VAX have been created in a very long time.
The optimum degree of complexity of a computer instruction set is based on the ratio of memory speed to logic speed. For a givem memory speed,
the faster the logic speed the more complex the instructions should be, so that both memory and logic are doing useful work all the time, with neither
waiting on the other. As that ratio changes with advancing technology, the optimim degree of complexity changes. Currently memory is so fast compared to logic that simple instruction sets are best, but that will cha
bad news to start the week (Score:5, Insightful)
Re: (Score:4, Insightful)
Yes. Big upheaval. Apple just made a huge bet on ARM. 90% of the world's smartphones use ARM.
The best outcome would be a shift to an open-hardware architecture that no one owns. RISC-V is one possibility, but it is based on MIPS and has architectural limitations. A new clean design by an independent foundation would be best.
We live in interesting times.
Re: (Score:4, Insightful)
The best outcome would be a shift to an open-hardware architecture that no one owns.
I agree! And the good news is that it's already here -- it's called RISC-V.
RISC-V is one possibility, but it is based on MIPS and has architectural limitations.
To some extent, all RISC processors are "based on MIPS." So what?
Can you describe the specific "architectural limitations" in RISC-V that you are concerned about?
A new clean design by an independent foundation would be best.
Because designs produced by committee have such a great track record. :-(
instruction set designed by committee (Score:2)
A new clean design by an independent foundation would be best.
Because designs produced by committee have such a great track record. :-(
The VAX instruction set was designed by a committee, and it had a pretty good track record. The IBM System/360 instruction set was probably designed by a committee, and it also did pretty well.
Re:instruction set designed by committee (Score:2)
The VAX instruction set was designed by a committee, and it had a pretty good track record.
I enjoyed working on various vaxen, but I wouldn't use VAX as an example of success. It had a brief moment in the sun, but then it died. Completely died. If the VAX architecture is discussed today, it's usually as an example of what not to do.
The IBM System/360 instruction set was probably designed by a committee, and it also did pretty well.
I agree that (1) S/360 was a hugely successful and influential architecture, and (2) it was probably designed by a team rather than a single architect. But a "team" isn't the same as a "committee". Even assuming good faith by all participants -- which a questionable assumption -- industry standards committees are orders of magnitude slower than internal company teams. For every success, they produce are at least two camels ("A camel is a horse designed by committee").
Re: (Score:2)
The VAX instruction set was designed by a committee, and it had a pretty good track record.
I enjoyed working on various vaxen, but I wouldn't use VAX as an example of success. It had a brief moment in the sun, but then it died. Completely died. If the VAX architecture is discussed today, it's usually as an example of what not to do.
The IBM System/360 instruction set was probably designed by a committee, and it also did pretty well.
I agree that (1) S/360 was a hugely successful and influential architecture, and (2) it was probably designed by a team rather than a single architect. But a "team" isn't the same as a "committee". Even assuming good faith by all participants -- which a questionable assumption -- industry standards committees are orders of magnitude slower than internal company teams. For every success, they produce are at least two camels ("A camel is a horse designed by committee").
Almost all computer instruction sets eventually die, and I have not seen the VAX instruction set discussed as an example of what not to do. As far as I know, it is still considered a good example of a Complex instruction set.
I don't know what the difference is between a "team" and a "committee". The people who designed DDCMP, an ancestor of Ethernet, called themselves a "committee", even though we all worked for DEC. Ethernet itself was mostly designed by people from Xerox, Intel and DEC, then turned ove
Re: (Score:2)
Almost all computer instruction sets eventually die
Sure, most architectures eventually die. But some influence what comes after them, while others simply go extinct. CISC in general, and VAX in particular, went extinct.
As far as I know, VAX is still considered a good example of a Complex instruction set.
VAX is a great example of CISC architecture. My point is that such architectures are now considered a Bad Idea. That doesn't mean VAX was bad idea at the time it was created -- it just means that the world has moved on. No new designs similar to VAX have been created in a very long time.
I don't know what the difference is between a "team" and a "committee".
I assume you've never participated in an industry standa
Re: (Score:2)
Almost all computer instruction sets eventually die
Sure, most architectures eventually die. But some influence what comes after them, while others simply go extinct. CISC in general, and VAX in particular, went extinct.
As far as I know, VAX is still considered a good example of a Complex instruction set.
VAX is a great example of CISC architecture. My point is that such architectures are now considered a Bad Idea. That doesn't mean VAX was bad idea at the time it was created -- it just means that the world has moved on. No new designs similar to VAX have been created in a very long time.
The optimum degree of complexity of a computer instruction set is based on the ratio of memory speed to logic speed. For a givem memory speed, the faster the logic speed the more complex the instructions should be, so that both memory and logic are doing useful work all the time, with neither waiting on the other. As that ratio changes with advancing technology, the optimim degree of complexity changes. Currently memory is so fast compared to logic that simple instruction sets are best, but that will cha