AMD Blasts Rumors of Dropping x86 for ARM

Who was saying AMD would drop x86 entirely for ARM? That would be a bit ridiculous. AMD would simply become an ARM manufacturer (like TI, Nvidia, Qualcomm, Samsung, etc..). However, AMD seems to be adamantly denying any desire to create ARM processors, which is silly if you ask me. They need to branch-out, not be tied to x86 and be at the mercy of whatever Intel decides to do next (e.g. another SSE extension). AMD has to hope Intel will even license that tech to them, and then be able to afford it. With ARM, the prices are fair because so many are already manufacturing the chips. AMD has the experience to start making good ARM products and getting those products out there.

Hell, if Intel weren't so invested in x86 (they have to keep their patents alive), even they might start making ARM processors. Thing is, Intel wants to own the market, you can't own a market making the same thing everyone else makes (which is why they so desperately want AMD out of the x86 business). AMD seems to be trying to follow Intel. At this point, AMD should just be trying to make money. Even when they were better, AMD still couldn't beat Intel.

Right now, people are buying ARM, it's an exploding market (more people own cell phones than computers). It just makes sense to be in on that market, rather than trying to push the inefficient, long-in-the-tooth x86 on everyone. Let Intel waste their billions on that folly. If Intel carves a market out for ultra-mobile x86 processors, you can serve up your offering. In the meantime, there's no market there, and, sadly, no one is going to follow AMD into unknown territory.

the mobile space is what amd should be vying for its the only space that intel hasn't conquered and its going to be big in a couple years. Besides they have the graphics devision to blast anyone(except nvidia but without them it wouldnt be fun )out of the water. As hellwig said they should be concerned with making cash not following intel.

I agree with all, but that lil' statement. From the pure "arch" point, you CAN make CISC (x86 in this case) as light as RISC in power consumption and be close to it in performance IMO (not going into details here, but it can be done, I'm sure). But the thin red line of licensing might be on Intel's way of bringing down x86 into that territory. The closer u're on perf/watt to what ARM has now, the closer you are to their licenses, I'd say. Software wise, I'm clueless about which one is easier/cheaper to develop on; that is, thinking about ultra portables. So, there is a market IMO.

This is going to be an interesting battle in the coming years, just like when 8086 won back then, lol.

Cheers!

I think AMD should keep x86 manufacturing and also manufacture ARM processors for low power clients like HTPCs, tablets, cell phone processors and etc. while still addressing the desktop, server and HPC market with x86.

Yes, x86 is ancient and it's time for it to slowly die out for a more modern and open design. Even Microsoft realized this when they announced Windows 8 being ported to ARM. Windows NT has had a run in with RISC architectures before like PowerPC, SPARC and etc.

You are so off the mark on every single point that it's kind of funny.

AMD's entire history is just intel *clone*......

Let me go ahead and analyze your counter argument ... oh right, you didn't present one. Good job!

Nvidia already put their GPU together into ARM chips. Now AMD might do that as well. An ARM chip with low power Radeon GPU? Possible.

This is going to seem stupid, but. Why is everyone saying x86 needs to die? I know it's old, but what's going to replace it in the desktop market? I see that ARM makes sense in the mobile and portable scene, but not in desktops. What I'm thinking is that ARM and x86 should be killed and we need photonic processors lol. But, seriously, what's wrong with x86? I'm new to some of this stuff so please be kind lol.

Well, there are several factors that make "x86" something you would not be very fond of... The one and foremost is Intel's monopoly (basically) on the licensing around it. So, if it weren't for AMD, it would be a real monopoly; back in the day, as I remember, IBM told Intel to license their tech to another company (AMD) so they could still buy processors from them and have a choice of providers. That's what I remember, could be a lil' different though

The CISC arch behind x86 itself ain't a bad thing (in itself) at all. It's just different from the RISC behind ARM, but like people stated around, there's a lil' red line between those 2 anyway. But since Intel holds every software around it's proprietary instructions sets (you now MMX is mandatory nowadays? ), making "improvements" around x86 is REALLY hard if you're not Intel. RISC has its spawning around IBM, ARM and Sun (now in Oracle's claws) if I recall it correctly (plus ARM licensing to third parties), so you have a lil' more competition behind it.

That's how I remember it though, feel free to add facts into my memories XD

So, it's not that "x86 is a bad thing" at all. But (IMO) is kinda boring that x86 is still around unchallenged as a "common" development arch. And that's a Desktop rant anyway

Cheers!

Oh right, VIA has some x86 licenses too... Forgot about 'em, sorry

Cheers!

VIA also makes the most RISC-like x86 CPUs. Basically, fewer dedicated execution units, I think. Which means it takes more cycles to complete the same instructions as a CPU that has more dedicated execution units, because it's interpreting bigger instructions as a series of simpler instructions. VIA compensates for this by including a few specialized execution units for things like encryption, so their CPUs are more attractive for their intended use in security systems, kiosks, and thinclients.

This could have gotten really interesting if AMD was going to do Fusion with ARM+Radeon, like how NVidia did ARM+GeForce (Tegra). First thing I noticed when Microsoft demo'd Windows 8 ARM was it's on a Tegra system.

The low-end could be ARM, with the high-end remaining x86. Unless AMD has little faith in Windows 8 ARM? Granted, they probably know more about it - and most importantly, more about its compatibility or lack thereof with x86 Windows - than we do, yet.

Intel used to make ARM based processors (XScale), but they sold that product line off to Marvell.

x86 isn't an architecture, please stop producing stupid articles that don't know the difference between an architecture and an instruction set. In fact, modern x86 processors don't even execute x86 instructions directly, so their architecture is one level removed from it. Netburst was an architecture. P6(Pentium Pro to Nehalem) is an architecture. Sandy Bridge is a new architecture. K8 is an architecture. x86 is an instruction set with varying architectures based on it. I don't know why this would be confusing. It's quite simple.

The 8086 didn't win, the 8088 did. The distinction is important, because one of the main reasons they chose the 8088 was because it was 8-bit externally, and made it much cheaper to implement, while still maintaining the instruction set and huge addressable memory of a 16-bit processor.

If the 68008 were out sooner, IBM might have made a different choice.

Cellphone tech has been accelerating at an insane rate. I've already read phones due out early 2012 will be 4 core, 2Ghz+ with 12Mp cameras. As for desktops, I truly believe in the not too distant future (less than 10 years) the core processor structure needs to change. Not just going from 64bit to 128bit, but rather a new ground-up design change. To me the CPU function will diminish with the majority of the work load being handled by GPGPUs.

Erm---except for say, a little AMD innovation known as x86-64 which pretty much singlehandedly foiled Intel's plans for Itanium world domination...

x86-64 is what saved AMD from extinction. Intel wanted to drop x86 in favour of their new 64-bit Itanium 'superchip'. By the time they realised it totally sucked, AMD had already gone mainstream with x86-64, forcing Intel to license it from AMD to catch up. Now both companies depend on each other's licenses so Intel can't totally call the shots.

I wonder who at intel decided once upon a time to sell their ARM division?
Their xscale processors were ARM-based, but they were intended as replacement for the old i860 and i960 (used in printers, raid controllers and other peripherals iirc).
That kind of experience combined with current high tech, could have given them a strong advance in mobile market.
So now intel has to make x86 feasible for mobile solutions, which in my opinion is like trying to fit an elephant in a beetle.
AMD has always been more versatile and I really hope they will add ARM to their portfolio.
This would give them more revenues, and a stronger AMD would in the end mean better products from both AMD and intel.

Why did AMD sell ARM in the first place?

or should i say ATI did but they are 2 now.

1* even. wow typos

Microsoft also deserves some credit. They told Intel that they were not going to support 2 instruction sets for Windows.

Well, if you insist. . .


AMD doesn't manufacture anything. They're a design house.



Intel has a perpetual and mutual licensing agreement with AMD. They can use x86 free of charge along with any changes and/or updates we provide. In return, we get to freely use any changes and/or updates AMD comes up with for x86. Please, explain what you mean by "at intel's mercy" and how they have "to hope they can afford" a license that's provided to them free of charge.


The LAST thing Intel wants is AMD to go out of business. And I assure you, the performance gap between x86 and any arm device is the reason we still use x86, not because "we want to hold onto our patents". Where are you pulling this stuff from?



And this is different than the last 30+ years how?



x86 is NOT inefficent. You listen to too many lame-o fanboys on forums such as this one instead of actually trying things out for yourself. Just because something is old doesn't mean it sucks. There's many reasons why x86 is still the king. There may be better suited parts for certain tasks, but x86 can't be beat across the workloads it performs.

And, there is one hell of a market out there just waiting for x86 to go ultra mobile. They just need it to go mobile without sucking down their battery in an hour. Don't worry, you'll see that happen soon enough. As it stands, I'm enjoying this x86 ported version of honeycomb.

AMD is only recently fabless (since spinning off their German facilities into Globalfoundries). ARM is also a design house. Not sure how this affects the argument either way.

x86 has still proven to be too complex and large to ram down into low power platforms. The recent advancements by intel with cedar trail are at the cost of single core performance as well, and considering how the intel atom faired in average day to day tasks (or rather consumer complaints lack thereof) means there will have to be massive steps taken to transform x86 into something more RISC/ARMy then CISCy.

x86 is still dominant on server/desktop and lappies because it's universal and rather easy to develop for. When RISC/CISC were competing many many years ago it was intel's money and power than ultimately sold x86 as THE platform. And at the time they were right, it was far easier to develop for and the universality of x86 has benefited computing tremendously, but nowadays with CPU advancements and risc/cisc hybrids (along with huge advancements in software development and incredibly cheap hardware) it isn't necessary anymore. If you need further proof then just open your android app store.

And Intel has repeatedly stuck to x86 through thick and thin. The itanium project is treated like crap and underperforms in comparison to x86 xeons. Trying to force x86 through low power platforms requires more than 32nm and 22nm, but efficiency AND cost reduction.

http://www.xbitlabs.com/news/other [...] cture.html

apparently AMD disagrees with me. It's not surprising, as both Intel and AMD have stated pretty much the same: Easier to go from x86 to lower power consumption and more processing power (and in turn market domination via x86 license) than it is to scale from ARM up.

also, AMD is nearly done with 28nm and iirc the globalfoundries facilities in construction are ahead of schedule.

http://www.xbitlabs.com/news/other [...] arter.html

they are on schedule with regards to 32nm and 28nm CPUs and APUs.

A complete quod core ARM with graphics costs 20 dollars and draws 2.5watt. A x86 chip is about 5 times larger and costs therefore 5 times to manufacture. A 2+ghz dual/quod core ARM is enough for the wast majority of computer users. Even heavy stuff like video encoding is faster on ARM chips. It is fun to see Imovie encode movies faster on Ipad2 then an intel mac. Within 5 years X86 will be a niche market for "power" users who believe that they need to have a CPU that is in idle 98% of the time.

I think as soon as Windows works flawlessly on ARM, as well as all programs made for 98/xp/vista/7 can work in some emulator, many people will leave x86!

The reason intel is so popular is because of Microsoft!

If MS had done a change to ARM long time ago (eg: In the XP era), Intel would not have even built core i processors.

Regardless, corei processors is about as far as they can go.
They integrated CPU and GPU, and the only thing they still can do is include audio processor, and USB hubs within the CPU die (next to making them on a smaller aka 28 to 14 nm die).

The Moore's law is about to be broken! Once the switch to production of 28nm processors has begun, Moore's law will no longer hold!

ARM will be able to deliver similar performance for processors running at 1/3rd the speed, that is, if they continue building their designs.
At least that way we still have a decade or two where moore's law will be in operation, if only with a sharp fall on transistors with the switch to ARM.

the video encoding has to do with the way ARM processors are structured and is also the reason why PowerPC computers (think apples in 90's and early 2000's) were preferred for media intensive tasks and design studios. it's also in your xbox 360.

But yes, the cost is a big big issue. ARM is making loads of money because the chips are so damn easy to manufacture and cell phone and mobile tech is so competitive and fast-moving. x86, on the other hand, is dominated by 2 corporations that don't REALLY care for innovation all that much. As long as the consumer sees a small improvement from generation to generation they'll get paid. If you think about significant advances in x86 among AMD and Intel I can only name 3-4, and these core i5>corei5 sandy bridge and phenom > phenom II crap are just minimal boosts to keep the cash flow coming in.

Just the very fact that an arm processor shuts down when in idle mode, makes it run cooler.
Currently ARM processors can't even tip core2duo processors, much less corei7, because they have not yet been optimized for games, and other than encoding/office workloads.

I think there's a grand future in ARM, even if they would get a sort of 'turbo boost' implemented on their chips, they would run massively much faster (some say ARM processors could reach speeds of upto 1.5/2x the clock frequency on a single thread on a 2/4core processor, just because they run so efficiently).

ARM is also in order technology if I'm correct (with perhaps some instruction sets that run out of order; I don't know).

I think the only negative sides of ARM is, that it's much harder for software companies like MS to secure their software, because reverse engineering is so much easier on an ARM; thus cracking programs is so much easier. I think that's been the major hesitation for MS to switch sides. It's probably also one of the reasons they are not going to do much about backward compatibility with previous Windows software (presumably).
But regardless, I think the switch to ARM from MS is going to go not without troubles, and a lot of programs will take time to get ported over.
Once MS switched, I'm sure Intel, AMD, Nvidia, ARM, and even Texas Instruments, and a few others will be ready with their ARM processors.

Too sad it'll be the death of some, with most likely Intel, AMD, Nvidia and ARM surviving.

From there on, it'll take at least 6 months to a year before people will start to switch. Noone wants to switch to an operating system running on ARM that has nearly nothing to port, unless they all go Android or so.

But Android is not really a good gaming platform, so it's upto MS to get the latest DX11 graphics cards working on ARM, and optimize them for a good boost in performance.

I think before the ARM market is fully established, both software wise, and hardware wise, it'll be at least a good 3 to 4 years down the road.
Don't buy any more graphics cards if you want an ARM system!
I'm sure they'll find a more suitable system where AMD and Nvidia will plug the graphics card straight into a socked without a controller on board, and graphics latencies will be reduced to virtually zero (or 1) ms. But before there will be an ARM processor that gives enough throughput to feed a Radeon 6970 card or similar, we'd need to sit and wait another decade or so.