What is Digg?52 Comments
- Bostocks, on 02/06/2008, -4/+32So long AMD, it was fun while it lasted.
- geminitojanus, on 02/06/2008, -0/+20Absolutely, this chip's smaller than some ARM variants, clocks in at astonishingly low power usage, and is ridiculously fast. Every ARM vendor on the planet crapped their pants when they saw this chip's potential to destroy their high-end market.
Interestingly, this chip would probably work well for ultrathin/ultralight notebooks (EEE, OLPC, ...) as well as MIDs, especially at 2GHz. Its single thread performance be slower than chips like the Core (NOT Core 2) processor (due to it being two issue instead of three issue, and being in order), but it'll be more than fast enough for devices like the iPhone, and Intel hopes by making it SMT, it'll more than make up for some of the loss of speed per core.
Some people are worried about the size of Poulsbo + Silverthorne, but combined they're no bigger than any other three chip Blackberry's layout. There's also the potential for a single package variant of Menlow which would jive well with ARM vendor's willingness to squeeze as many dies as possible into their packages. (The iPhone itself does this with its Samsung ARM and PowerVR graphics accelerator). Just to give you an idea of how ridiculously small these chips really are:
http://images.anandtech.com/reviews/cpu/intel/menl ... Silverthorne (45nm, 25mm^2).
http://images.anandtech.com/reviews/cpu/intel/menl ... Poulsbo (65nm, approx. ~50mm^2). - championchap, on 02/06/2008, -0/+13I get the feeling he likes performance.
- reginaldino, on 02/06/2008, -0/+9well, i used to be an AMD fanboy too. That was until the C2D. AMD have gotten so far behind in performance and innovation (i'l give them amd64 was good) its shocking
- championchap, on 02/06/2008, -0/+7So you don't care if your iPhones processor is a bit *****?
Personally, i'd love to see what they can pull off with the boost that one of these bad boys would give them! - Dgen_X, on 02/06/2008, -0/+7replace that IMO with the word "currently"
not that I know what AMD has coming, but like Intel...they're not just going to roll over and die when the competition has them beat - fluidfoundation, on 02/06/2008, -3/+10dammit, I was hoping this was about that show back in the 80's Small Wonder.
- inactive, on 02/06/2008, -1/+5I don't buy the ***** that Intel is a 'monopoly', or engages in those acts. They make the best processors IMO.
- TomFrost, on 02/06/2008, -1/+5Yep, the same way my 2.4Ghz Core 2 Duo gaming computer works the same way as my 486DX gaming computer!
- SirZRX, on 02/06/2008, -0/+3come on AMD wake up!!!
- HonoredMule, on 02/07/2008, -0/+2A modern CISC CPU is really a RISC CPU with the controller translating standard x86 instructions to platform-specific internal micro-ops in platform-optimized order. In other words, it has all the here-and-now performance benefit of RISC architecture made accessible and future-flexible by abstraction behind the stable and here-for-the-next-three-decades x86 'API.'
...and before anyone says something stupid, remember that those on-the-fly translations are being done on PREFETCHED instructions, with little to no effect on branch prediction effectiveness. There's a trifling impact on energy consumption and transistor count (and by trifling I mean in the .1% ballpark), and no /negative/ impact whatsoever on performance.
RISC was inherently the more efficient CPU design by virtue of offloading most of the optimization and clever design to the compilers, but now it's irrelevant, and just less able to adapt new ideas into existing infastructure. x86 will eventually erase ARM's advantages in the mobile sector, and surpass ARM altogether. - MattiX, on 02/06/2008, -3/+5RIP AMD. Fond memories.
- etandrib, on 02/06/2008, -5/+7I've read quite a few articles about this and most seem to think this is the next iPhone processor while Ars says that is won't fit. Does anyone know if this chip really is an iPhone candidate?
- gremos, on 02/06/2008, -0/+2what's up with 3 arstechnica articles on tech front page?
- stoanhart, on 02/06/2008, -0/+2Why do we care? Because we're geeks. Why do gear heads care what type of pistons your engine has? It works either way.
- SSCrow, on 02/06/2008, -0/+2I'm holding a penny.
And I am shocked at how small this thing is. - nblsavage, on 02/06/2008, -0/+2Soleil Moon Frye was Punky Brewster.
- redmaxx, on 02/06/2008, -1/+3Oh, I'm sure that when they announce the next iPhone they'll have Paul Otellini up there describing how they didn't think they could shrink Silverthorne for Apple but they somehow managed to anyway.
- geminitojanus, on 02/07/2008, -0/+1Intel sold their ARM lines to Marvell.
- beatbox32, on 02/06/2008, -0/+1"She's fantastic...made of plastic. Microchips here and there..."
- NoCt1, on 02/06/2008, -1/+2I would love this to start getting into phones. As long as they can work out power consumption.. that would be awesome.
- Zephik, on 02/08/2008, -0/+1I've been using a Pentium 4 processor since as long as I can remember and it has always met my needs. If you need something more than that, then you should be looking elsewhere.
That said, you are really missing the point of this processor. So please re-read the article, as the point(s) was made fairly clear.
Now, if you are arguing that it won't take its market by storm like most other processor developments from Intel have been doing lately, then you have a valid argument. But, this little number isn't complete yet and already its been talked about as serious competition. The final release might become the new standard for portable devices. At the very least, it will be good competition because of a few key features, which again, were pointed out in the article. - bnolsen, on 02/07/2008, -1/+2This is just a CPU. The ARM folks have long moved on to whole SOC (sytem on a chip) systems. This processor is JUST THE PROCESSOR. Something like the OMAP3430 contains all core logic, including north bridge & south bridge functions & integrated video plus a 16 element SIMD unit which burns less than 1W at full power.
Intel has some serious catching up to do...however historically the ARM release cycle has been several years.
Competition is good. - Bostocks, on 02/07/2008, -0/+1Nope, I hate monopolies. But when was the last time you saw an awesome new AMD cpu on digg? That thing is the size of my fingernail... I have nothing against them, I just thing AMD is screwed.
- iXneonXi, on 02/06/2008, -2/+3Dear Nintendo,...
- Zephik, on 02/08/2008, -0/+1Also, if I could get an Eee PC that is just as powerful as most of the computers that I have owned, but is more energy efficient, costs less and is portable? (as the Eee PC is a laptop afterall). I would be a very very happy person!!
Heck, I could just plug in my Monitor and my Wireless Desktop and I would have what I have now, but ultra portable AND ultra mobile! Thats a very exciting thought! :) - Scaryclouds, on 02/06/2008, -0/+1Hopefully their no CPU/GPU's save them. Competition = better products
- stoanhart, on 02/07/2008, -0/+1Architecture has very little to do with buggyness. The source code for most applications doesn't need to be changed at all, it just needs to compiled with the right compiler. Using x86 will simply open up the door to all the software on your PC right now (as longs as the system is fast enough to handle that bloatware).
- DrDabbles, on 02/06/2008, -0/+1...intel owning StrongArm. Keep that in mind. Also, xscale is a rather impressive package, so don't discount that.
- NanoStuff, on 02/07/2008, -0/+1I find performance per watt far more impressive than raw performance alone. Eventually when we're running one million cores, that's going to matter. 0.5W @ 1GHz through 50 million transistors is borderline spectacular.
- blackoper, on 02/07/2008, -0/+1wouldn't mind getting ahold of about 4 to 6 of these for a linux server... 12 to 30 watts for 12ghz of performance would be pretty sweet. Wonder if anyone will release a motherboard for this chip.. something like a blade server would be pretty awesome.. 64 bit chips that do multithreading and speeds up to 2ghz. Since it doesn't appear this will be sucessful in the ultra mobile market yet this should be focused on low power server market
- fudgebrown, on 02/06/2008, -1/+2I know a guy with a Small Wonder. Poor bloke is currently single..
- jamesmudgett, on 02/07/2008, -0/+1*****
- djphilos, on 02/07/2008, -0/+1Everyone always complains about companies like this becoming a monopoly. The whole united states (and world) is controlled by the largest monopoly of them all, Big Oil, and we dont do a thing about it. I'd rather a company monopolize by releasing small, better and cheaper products then by jacking up all their prices.
- Scaryclouds, on 02/06/2008, -0/+1Because if the iPhone adopts and x86 processor that means more apps that are less buggy because their source code does not haven't to be changed (as much) from there desktop variants. Also Silverthorne would far outperform the current iPhone processor (ARM11?).
- jetblackz4, on 02/06/2008, -1/+1I wonder what ever happened to her and that slutty neighbor chick who wanted to bang Jamie all the time.
- inactive, on 02/06/2008, -7/+7So you like monopolies do you?
- Exophase, on 02/27/2008, -0/+0Or at the very least Intel would love it if you believed that. To their credit, it is half true. x86 CPUs are indeed translating to and running RISC code. What isn't mentioned is the quality of this RISC-like code, which is a subset of x86 code. The x86 instruction set architecture is already generally less powerful than ARM, so when you have to use multiple opcodes to execute single x86 instructions things are just getting worse, not better.
No, the benefit in RISC isn't intrinsic to the instruction set, it's in allowing more efficient opcode execution. ARM is executed as efficiently (per pipe, of course) as x86 micro-ops, and more or less always has been. And the quality of these opcodes is in a number of ways better - three address opcodes, conditional execution, folded shifts, and with everything they've added in the last couple revisions (v6 and v7) the list just goes on. Not to mention a superior vector FPU instruction set as well. The Core 2 architecture potentially undoes some of this with macro-op fusion, but that technology has a long way to go (and is far from present in something like Silverthorne anyway)
Ars Technica also seems to heavily disagree with your comment on energy consumption/die count:
"Because a huge chunk of Silverthorne's die area is cache, the percentage of total die area that the new processor spends on x86 instruction decoding won't be as high as the original Pentium's 40 percent, but may well be in the double digits. Like the original Pentium, this relatively large number of transistors spent on x86 decode hardware will put Silverthorne at a performance and performance/watt disadvantage compared to a leaner RISC design like ARM, which can spend more die area on performance-enhancing cache."
I'd like to know exactly how you think decoding (which you think is being done at the prefetch stage, something I find to be kind of a joke) is only taking up such a tiny portion of die size. It also does impact performance because it does in fact tend to mean more pipeline stages which does mean bigger stall penalties. The Pentium 4's trace cache might have mitigated this but again, you won't be seeing that here (in fact, are we seeing that anywhere anymore?). If it's so inconsequential I'd also like to know why Silverthorne is still rated to use much more power than the fastest ARM chips (aside from all that L2 cache, which is probably going to be much higher latency than what you get if you paired a high-end ARM with say, 128KB of L2)
Really, what you said about RISC might be true, but this isn't about CISC vs. RISC, it's about x86 vs ARM, and after all these years ARM is still a more efficient ISA.
I really don't see why x86 belongs in handhelds in the first place, nor do I see why Intel sold off their ARM chips (maybe because they were being panned for some mistakes they made with XScale). It's pretty ironic, given how much Intel was trying to move away from x86 in the 80s only to be tethered by legacy code, but now that things are much more platform agnostic they seem to be the ones forcing it on the rest of us. In fact, in the ways that platform agnosticism hasn't won ARM really has the lead, because mobile devices have been ARM for years. What's even worse is that an in-order statically scheduled CPU like x86 is going to need to have applications compiled specifically for it to really get good performance, so where is the legacy advantage there? Why bother with x86 at all? It's just because it's Intel's calling card and they want to control it.
And from all indications not only is the Cortex-A8 a superior design than the Silverthorne (perhaps only limited by clock speed, but that'll change when it too moves to 45nm), Cortex-A9 is significantly superior to that.
I also don't think it's really any code that 64-bit is being pushed on handhelds so soon either. I'm sure that's going to take its toll in die size and power consumption as much as anything.
I really hope Intel doesn't steamroll the handheld market by virtue of nothing more than having smaller dies before everyone else, which has helped them all too much in desktop CPUs but is really almost all they have going for them here. - ivandir, on 02/06/2008, -2/+2Long pipeline for such a small CPU.
- Exophase, on 02/27/2008, -0/+0Look at the announce times between Cortex-A8 and Cortex-A9, two CPU designs that, despite their names, are quite different.
The release gap is closing.
Oh, and ARM still just does CPUs (in fact, mainly just licenses them to others who are free to put them into whatever chipsets) - zionKing, on 02/06/2008, -1/+1Coming battle for your pocket? Gross, count me out.
- Jfox386, on 02/06/2008, -0/+0Down with intel!!!!!!!
- SiliconMadness, on 02/06/2008, -1/+050 mm2 is quite an understatement: http://bunniestudios.com/blog/images/menlow_ces08. ...
That or silverthorn isn't 25mm2. Where did you got those figures? - TheVirus, on 02/06/2008, -2/+1So does this mean that the Nintendo DS2 will be using one of these instead of the ARM9/ARM7? One can wish, can't one?
- puddgomez, on 02/06/2008, -1/+0Anyone remember that show Small Wonder?
- typicalusername, on 02/06/2008, -2/+1Solei Moon Fry in pron ftw!
- SiliconMadness, on 02/06/2008, -3/+2It may be, but the Poulsbo chipset is too big now, possibly making the whole platform not feasible for usage in there, yet. Big and power consuming. As soon as they also shrink the chipset and optimize it like they did for the CPU, it will probably be the only design worth considering for the iPhone, and all other smartphones.
- TheMisterC, on 02/07/2008, -1/+0Just had to make a quick comment.
I've known about this processor for a while. I think it is a good start, but this version will be a big disappointment. It doesn't have much cache and it is an in-order 16 pipe line processor - can we say Pentium 4? - dagamer34, on 02/06/2008, -3/+1x86 isn't really the greatest for developing mobile games. ARM CPU instruction set is simple, light, and fast since it's RISC. CISC just wasn't meant for dedicated tasks like making games off of (and before you go talking about the Xbox, note that it has a Windows kernel running beneath all the games for the system).
- Tyr7BE, on 02/06/2008, -6/+3Why do you care what processor is in the iPhone? I mean, it'll work either way, right?
-
Show 51 - 52 of 52 discussions
Browsing Digg on your phone just got easier with our enhancements to the 
© Digg Inc. 2009
— Content created and posted by Digg users is 
