Overview of SoC options


Architecture and instructions are fixed stuff. You can only pile things up on top of it, and tweak what's modular. x86 is a dirty arch.

Well, i'm intuitively right most of the time. I don't feel bad about that.

A paper plane doesn't "fly".
 
Architecture and instructions are fixed stuff. You can only pile things up on top of it, and tweak what's modular. x86 is a dirty arch.
The quality of an architecture is almost entirely dependent on one thing: what the CPUs that run the architecture are like. perf, perf/W, perf/$, and so on are almost everything. Obviously how good the compilers are at generating the code matters, but you can't unchain this from the "perf" part of the metrics I gave. How aesthetically pleasing an ISA is to programmers (or worse, people who don't program in assembly but like to have an opinion about it) only matters ever so slightly. And this is coming from someone who does a lot of assembly programming.

You're wrong anyway. The thing you call x86 has fundamentally changed several times, not merely to pile things on top of it as you claim. 32-bit mode substantially changed the fundamentals of the ISA by making the address modes much more orthogonal. 64-bit mode substantially did it again by expanding the general purpose register set to 16. AVX did it again by replacing existing operations with three-address arithmetic. AVX3 will change the SIMD model again in other fundamental ways. There have been a lot of more minor changes.

Well, i'm intuitively right most of the time. I don't feel bad about that.
If you make a lot of claims it's inevitable that some of them will be right. Then you can look back and talk about the things you were right about while ignoring the things you weren't so right about. Rumor mills love to do this. I remember you made a lot of claims about Microsoft and Windows dying, I'm still waiting on this.

A paper plane doesn't "fly".
Employ whatever wordplay you like, but interpreting a comment about whether or not a product will happen to mean anything more than whether or not it will be released is disingenuous at best.
 
Last edited by a moderator:
Nuclear central produce more energy than wind farm. I can safely say that wind farm is cleaner.

Less and less people uses windows, sony is selling it's PC division, ppl use phones to get internet, etc.

Apart patent racket, flight simulator and banks morons using c#, i don't see m$ going anywhere.

Also i think companies were horrified seeing how m$ (almost) killed Nokia, i'm sure chairmen will think twice when they will renew their IT stuff. Not to mention Snowden revelations about how m$ were the first to collaborate.

About rt, okay, no more wordplay, a born dead is born.
 
Meanwhile, Intel's desktop/laptop CPU sales have been more or less flat YoY. Not great but not dying. Linux continues to not grow very much in this segment. Losing ground doesn't mean dead.

But lolz @ MS hanging on with flight simulator.

I'd still rather have a real discussion about why you insist x86 is no option for mobile.
 
One thing i'm not sure about and will certainly have a weight soon: China and their homemade cpus.

I doubt I said x86 is no option for mobile. I only say ARM is a better/wiser choice.

Intel missed the train here. Maybe they would have been better if they weren't so selfish (as every too big companies are).

https://en.wikisource.org/wiki/The_Hare_and_the_Tortoise_%28de_La_Fontaine%29

ARM is in 95% of phones in 2012.

That said, i don't know the x86/ARM ratio for tablets.
 
Last edited by a moderator:
One thing i'm not sure about and will certainly have a weight soon: China and their homemade cpus.
So far China hasn't shown themselves very formidable in designing CPUs. Licensing CPUs from ARM and putting them in SoCs, sure. Original/custom CPUs, no. What significance Ingenic and even Loongson had has long since been eroded.

I doubt I said x86 is no option for mobile. I only say ARM is a better/wiser choice.
No, you literally said it's not an option.

"I'll also add that x86 is a pretty bad architecture, that's why when you know something else than biased benchmarks on "reputed testing websites" you don't even consider x86 as an option."

I don't know if you only meant for Pyra in particular, but I can't imagine why an SoC would be a feasible option for other mobile devices (phones, tablets) and not something like Pandora.

The fact is, for Pyra, Z3770 and OMAP5 were pretty much the only options, since they're the only ones you can buy from distributors. The former is dismissed because it wouldn't offer backwards compatibility out of the box (and I'm sure other reasons), and I accept that; it's certainly ED's prerogative to prioritize backwards compatibility. And it's not nearly as detrimental as the decisions Nintendo has made to maintain compatibility. But I think outside of ARM compatibility, ease of porting drivers, and some other issues specific to the developers (like having better support contact with TI, more familiarity with their documentation, just more familiarity in general) Z3370 is probably as good or better than OMAP5 in every other way. Single threaded performance would be at least as good if not higher, higher multithreaded performance, probably significantly better perf/W due to being on a superior process and having a more moderate and aggressively optimized core design, 64-bit (does have some advantages even with 2GB of RAM, especially for emulators), open source GPU drivers, x86 compatibility, etc.

I know I've beaten this to death now but I just wish you'd stop saying that it'd be bad because everyone knows x86 ruins everything.

Intel missed the train here. Maybe they would have been better if they weren't so selfish (as every too big companies are).

https://en.wikisource.org/wiki/The_Hare_and_the_Tortoise_(de_La_Fontaine)

ARM is in 95% of phones in 2012.

That said, i don't know the x86/ARM ratio for tablets.
Intel did miss a lot of opportunities that made it a lot harder for them to get into phones. But that doesn't mean that their SoC isn't capable, something you claimed when you said that it only looks that way when it cheats at benchmarks. Intel has participated in benchmark cheating.

x86 does have a bigger but not huge share in tablets, it's expected to grow more, partially because they're subsidizing a lot of developers. I would say Intel is gaining more marketshare here than ARM SoCs have been in servers or traditional laptops/desktops.
 
Last edited by a moderator:
duty_calls.png
 
^ woooah that's a good one. I'd have quoted you but I tend to strip images from quotes and I'd just be quoting nothing then :(

Anyone else feel like posting that's getting just a little old? I bet even Randall Munroe is starting to resent it.
 
Last edited by a moderator:
Getting old? not really... it still makes me smile every time I see it. I do think it was inappropriate in this context though.

- Neelix
 
Getting old? not really... it still makes me smile every time I see it. I do think it was inappropriate in this context though.


- Neelix
Yup inappropriate.

I was reading what exophase was saying and he makes some good points.
 
Even IBM is considering selling its chip business, guess who will buy it ?

So at some point China will be able to kick some asses.

Unless some opto-quantic-whatever tech is around the corner.

I probably spoke about the no x86 for Pyra. What difference with other portable thing ? We have some kind of choice about what "we" put on the mobo.

Indeed ARM is slower on desktop/server market share progression. But let's just wait to see the AMD push on servers, and maybe the next Apple desktops.

As it sounds their new iPhone cpu is violent, i won't be surprised to see it (or the followers) on bigger computers.
 
Last edited by a moderator:
Even IBM is considering selling its chip business, guess who will buy it
That has got me a little worried.. although I do work in the development side of the business there. However these rumors had been floating around several times in the past decade.
 
Back
Top