What is Digg?23 Comments
- KibibyteBrain, on 10/10/2007, -2/+5Yikes, what's next, x86 ISA based microcontrollers...now that will be the day.
This story brings me mixed feelings. Now we can have a mature platform that has very mature compilers and that many people know how to program for. The platform is also mature in the way that the engineers at Intel have super optimized just about every subsystem that goes into an x86.
However, that being said, x86 is outdated. It can still beat the pants off new academic-constructed architectures that are in theory better because of the decades of optimization it has both on the compiler and hardware side of things. But that doesn't mean that these new architectures like IA-64 don't offer great advantages over x86 that would be worth investing in for the long term. Many of these advantages center on providing code optimized for n-way parallelization which seems to be one of the bigger problems right now too. That same silicon you have in your computer right now could do much more... But of course, Intel and AMD aren't going to pour billions into the R&D of making this happen until they have to.(rightfully so from a business perspective) - KibibyteBrain, on 10/10/2007, -0/+3The PPC vendors are used to competing against tough products. I once worked for a printer company. We kept on wanting to move to other platforms that better suited our needs, but any time we tried, our PPC vendor kept coming up with deals for us on components we could not refuse. If Intel is willing to play this game, they might win a ton of business.
- Fordi, on 10/10/2007, -0/+3ARM (the main competitor in this market) is hardly beta.
Anyway, I think the major problem I have with ARM-based architectures, from a programmer's point of view, is the old bus communication problem. Most devices designed for ARM have all components on a single bus, and if that bus is doing something busy - say, reading a large file off a flash card - the whole system gets paused out. For things like games, this is a problem (rendering, loading, processing and video refreshes happening constantly, resulting in the single-bus model becoming an overall performance bottleneck).
Say what you like about x86 arch. I think the RISC instruction set is by far superior. Still, the idea of having an ultramobile DMA stack is quite appealing. - KibibyteBrain, on 10/10/2007, -0/+2Heh, I know that. What I meant was in terms of manhours spent optimizing ARMs for high speed usage. The fact is, they really aren't used much in that area, so it creates a chicken and egg issue. I haven't worked much with ARMs, but I know you can get DMA controllers with MIPS products.
- mmastrac, on 10/10/2007, -0/+2I thought the PXA-series ARM controllers from Intel already did DMA.
- geminitojanus, on 10/10/2007, -0/+2"Yikes, what's next, x86 ISA based microcontrollers...now that will be the day."
Scary how close they are. Especially with Tolapai, the SoC Pentium M variant. - daveisfera, on 10/10/2007, -0/+2The new Core 2 architecture is a "short and fat" pipeline, instead of the "deep and narrow" pipeline of the P4 architecture. The "fatness" of the Core 2 architecture should lend itself better to a virtual dual core type of idea like HyperThreading, but obviously we'll have to wait and see.
- Fordi, on 10/10/2007, -0/+2I wonder if one could devote a small section of the ARM's main bus to DMA control - sort of simulate an interrupt functionality - and make an ARM's topology a little more like a switch or router than a hub. Essentially, this would require the equivalent of a BIOS to bootstrap the chip.
- HonoredMule, on 10/10/2007, -0/+2The main reasons that multithreaded programs aren't more widespread isn't limited time for adoption, as developers have had virtually forever, but laziness, difficulty and added complexity, and often, simply a lack of benefit. There's no point in running multiple threads if you can't keep them working simultaneously without getting in each other's way. Apache, for example, starts 250 worker threads on startup, which can all serve up data and handle requests simultaneously, but what's notepad.exe going to do?
x86 as a whole is far less energy and material efficient than ARM, and the recent pushes into mobile space are based on traction, not technical merit. That said, Intel and AMD will I'm sure give us a good show courtesy of the expertise of their x86-versed employees. Sometimes technically better has to bow to "overall" better. - KibibyteBrain, on 10/10/2007, -2/+3I also forgot to mention that many of the architectures that Stokes mentions that this move by Intel will kill offer many advantages over x86, especially in these more RISC-adapted applications. However, the manufacturers who implement them are quite behind Intel from a fabrication and implementation technology point of view. It would be sad to see them killed off. This is almost like a beta version line of jet engines going up against a really otto-cycle based gas propeller prop engine. Sure, the propeller planes are going to kick the struggling first attempts of the competition for a while, but once the competition reaches critical mass...we can finally reach some read advancements in hardware besides just clocking up and adding instructions.
- kahrn, on 10/10/2007, -1/+2Interesting.. silverthorne supports HyperThreading, yet HyperThreading was considered to be a dead feature (intel abandoned it, no?) which actually used too much electricity relative to the performance benefit it gave.
Anyone able to clue me in as to why intel has suddenly changed it's mind about HyperThreading? - Turdmoe, on 10/10/2007, -0/+1The Intel ISA is actually called IA-32 or I guess you could call it IA-64 now, but x86 is a micro-architecture not an Instruction Set Architecture!
- OrangeTide, on 10/10/2007, -2/+3I would prefer PowerPC. But Intel wants to compete in a market they dominate and compete against the likes of AMD who do not have good fabs. Wouldn't it be terrible for Intel if PowerPC were mainstream and they had to compete against IBM's fabs? yow! IBM might actually have a real reason to keep their cool silicon R&D going. (ie it wouldn't be losing money)
IA-64 is basically PA-RISC reborn, and was supposed to compete with IBM POWER. But Intel couldn't do it, I blame Microsoft's poor support of the architecture that prevent many early business adopters that could have help spur a speed boost and lower costs. Nobody put out a good 16-bit/32-bit x86 emulation package for it, which in my mind was Microsoft's responsibility. (just as Apple pushed out 68K and PowerPC emulation on their OSes). Also Microsoft's PE could have (but they didn't) been adjusted to support "Universal/Fat" binaries. Intel tried very hard to get MS to buy into it, but it was mostly a: what's in it for us? You just want to make some big iron chips for servers that our OS can't currently scale to.
Also it's important to note that x86 is pretty RISC-like inside with the way pipe lines, caching and the microcode works. But with a terribly ***** MMU that was designed for some theoretical OS far stranger than Windows NT or Unix. I prefer simple software maintained TLB cache like on PPC, MIPS, etc. - stoppedcode12, on 10/10/2007, -1/+1I always thought hyper threading was Intel's sign to the world that they will no longer be producing single core processors. Because hyper-threading allowed developers to multi-thread their programs without buying specialized hardware.
- peterA650, on 10/10/2007, -0/+0So who does Intel disrupt next? Microchip and Atmel of course!!!
- JohnnyXmas, on 10/10/2007, -1/+1"I'm malicious like viruses; old school like ball mices and ISA devises. . ."
-MC JohnnyX - HonoredMule, on 10/10/2007, -1/+1Huh? Multithreading has been handled by the OS (and capable within a single program, supported by such an OS) for a VERY long time. With the advent of ubiquitous multi-core systems, the multithreading is...still handled by the OS anyway. Hyperthreading was merely a failed attempt at improving the perceived responsiveness of a system, without actually having extra cores to dedicate to higher priority stuff.
The chance that you were using an operating system without at least one multithreaded process running on it even ten years ago is virtually nill. - inactive, on 10/10/2007, -4/+3reminds me of the ISA slot, *old skool*
- kethraal, on 10/10/2007, -2/+1"Because hyper-threading allowed developers to multi-thread their programs without buying specialized hardware."
Maybe... but HT sucked so, so badly that most developers stayed the hell away from it. Fair enough. Actually, HT with multiple cores is pretty cool -- but Intel ***** up the idea with the P4 HT's, giving it a bad name. Oh well. - JohnnyXmas, on 10/10/2007, -2/+1Don't be distressed. They're using it in what seems to be mainly trivial devices. Besides, you elitist bastards will be able to install Linux on them even easier.
- stoppedcode12, on 10/10/2007, -3/+1I am still waiting for my 'n' th PC.
http://www.engadget.com/2005/04/27/the-ultra-mobil ...
http://www.winsupersite.com/reviews/winhec2005_tpc ...
http://www.winsupersite.com/images/reviews/winhec_ ... - shahruz84, on 12/17/2008, -8/+5old news. this was all at idf last week.
btw, if anyone was watching the extreme gamer challenge in the showcase, yeah, its me, maddpaky. 2nd place
The Digg Toolbar for Firefox lets you Digg, submit content, and keep track of Digg even when you're not on the Digg site. Download the official 
© Digg Inc. 2009
— Content created and posted by Digg users is 