Replies to post #63398 on Advanced Micro Devices Inc (AMD)

[UpNDown]

kpf, on comp.arch discussion

I had seen the start of that thread, but haven't followed it up in the last few days. The numbers from Google are generated dynamically, of course, so your post numbers will not stay accurate as new posts enter the thread in subthreads above them. So anyone interested in locating those posts may find them further along than the post numbers you've given.

Thanks for the pointer.

[aleph0]

/and a wild speculation for what still could come./

Do you mean the discussion on "parallel X86" ?

as in :
||
The next evolutionary step is a parallel machine. Dual-cores are only an inefficient stop-gap design that wastes transistors with duplicated or unnecessary resources (eg coherency logic between the core's caches).
||

This guy is obvious very young !

1. Creating a "true" parallel machine would involve changing the software architecture at the OS level for starters . That cuts out Windows, Linux etc. etc.
2. To make "use" of a parallel architecture means also new programming high-level languages ( some exists already - but unheard of by most people ). So forget your C++, Java might just have a chance due to the fact that memory mamagement is handled automatically - i.e. by the Java engine.
3. Critical for parallel architectures is CPU-to-memory speed and size. If you halve the memory-size per CPU, the (properly designed) machine will be almost twice as fast.
4. To go this route (properly) would mean giving up X86 almost completely.
5. Maybe Itanium was a half-hearted digression in this direction - and look where it got !!

Been there already !