My dual core can beat up your quad core, and other mobile CPU mysteries(icrontic.com)
icrontic.com
My dual core can beat up your quad core, and other mobile CPU mysteries
http://icrontic.com/article/dual-core-faster-than-quad-core-mystery
12 comments
It is interesting to compare the new TI arms with current Intel atoms
Here is previous benchmarks on ubuntu
The dual-core Cortex-A9 1.2GHz on the TI OMAP4460 with the PandaBoard ES is mostly comparable to the first-generation Intel Atom N270 in terms of raw performance.
http://www.phoronix.com/scan.php?page=article&item=ubunt...
There's the beginnings of support for the little core/big core paradigm in Linux: "Operating System Support for the
Heterogeneous OMAP4430:A tale of two micros" http://www.ssrg.nicta.com.au/publications/papers/LeSueur_Rod...
Think they said if it's not in kernel mainline soon, they'll write a usenix paper instead
Think they said if it's not in kernel mainline soon, they'll write a usenix paper instead
There is a good LWN summary of what will probably happen short and longer term here http://lwn.net/SubscriberLink/481055/f5344869dc672690/
reads like an ARM commercial. is that independent writing?
I confess I stopped reading when they said that "Even Apple Macintosh computers now use x86, after nearly 30 years of utilizing a competing and incompatible ISA known as PowerPC (PPC)."
I understand it's all ancient history now, and that I probably shouldn't expect a ton of fact checking from somebody named "Space Penguin", but it was circa 10 years on Motorola 68k processors and then 12 on PowerPC before the Intel switch in 2006. Maybe the rest of the article is solid, but if I'm reading an article on the details of processors, it worries me when they miss a detail about processors.
I understand it's all ancient history now, and that I probably shouldn't expect a ton of fact checking from somebody named "Space Penguin", but it was circa 10 years on Motorola 68k processors and then 12 on PowerPC before the Intel switch in 2006. Maybe the rest of the article is solid, but if I'm reading an article on the details of processors, it worries me when they miss a detail about processors.
If you finish the article, it's not really about any of that at all, he's just writing a flourishing introduction. A detail like that isn't that important in the scope of the article.
But how can I know if I can trust the information in the article that I don't know (the reason I'm reading it), if the information I do know is wrong?
"Maybe the rest of the article is solid, but if I'm reading an article on the details of processors, it worries me when they miss a detail about processors" I think expresses it pretty well.
"Maybe the rest of the article is solid, but if I'm reading an article on the details of processors, it worries me when they miss a detail about processors" I think expresses it pretty well.
It's also surprising that the author credits Intel with the invention of ISAs. Sure, x86 is now the dominant ISA in many markets, but the notion of abstraction from hardware implementation was around for at least 10 years before the 8086 processor (IBM 360).
It's not just surprising -- it's a disgusting lack of respect for history. It's kind of like saying McDonald's invented "takeout food" or Madonna invented "pop music".
My thoughts, exactly.
a computer is not a machine that does computing. it is a primarily a machine that moves information around, and occasionally performs computation on it.
as such, a quad core will saturday the processor - memory bandwidth much sooner than a dual core chip will, which stalls the pipeline. couple that with that four cores will abuse the cache much more than two will and you have a recipe for slowdown.
as such, a quad core will saturday the processor - memory bandwidth much sooner than a dual core chip will, which stalls the pipeline. couple that with that four cores will abuse the cache much more than two will and you have a recipe for slowdown.
s/saturday/saturate/ ?
Perhaps we can use all this extra processing power for a better autocorrect implementation? ;)
Perhaps we can use all this extra processing power for a better autocorrect implementation? ;)