The whole point of SemiAccurate is to report "true rumors" before they become official, so pretty much everything they publish is officially denied. So either this article is completely wrong or it's right but AMD isn't willing to admit it yet.
I don't know, but considering the transition would happen within 2 years, I can see how they wouldn't want people to find out 2 years earlier about it.
Of course they're going to deny it. Even if completely accurate, it's going to be a staged transition and they'll try and salvage as much as they can from current Steamroller R&D. I guarantee that AMD will do at least one more high end CPU. Without the high cost of R&D charged against them, AMD can much more easily make money on its chips, and it will do so as long as it can.
The chip might look more like Piledriver than expected, but it will have some Steamroller bits and will probably be called Steamroller.
I've been thinking that Win RT and Surface are the real future of Windows - ARM based, no more backwards compatibility. This pretty much seals it. Apple and MSFT were already moving to ARM, ChromeOS and Android are obviously already there. AMD is going there now too. Many/most machines we use in 3-5 years will be ARM powered, not x86 powered.
Intel of course is going to resist ARM and low power computing for one huge reason - profit. There are only so many machines sold each year and switching to ARM doesn't sell more processors for Intel. So, switching to lower perf and lower power processors like Atom means Intel can only sell them for $20-30 vs. $100-300 for a core i3/5/7. That is why Atom hasn't been pushed by Intel on mobile, it would destroy their whole business model. You can't drop billions into manufacturing if you don't have the revenue and profits to fund it.
AMD's future is going to look a lot more like nVidia's current plan than Intel's.
Even in an all ARM world, Intel would still be very much able to charge $100-300+ for desktop chips, and $1000+ for server chips.
With their fabs and engineering talent, they'd still be strong leaders in terms of IPC, core counts, computing density, performance per watt, etc.
The dynamics would be similar to AMD vs Intel today. Charge double for a 50% lead, 5x as much for a 100% lead, etc. When marginal performance is important, the premium is gladly paid.
While they'd have countless fabless design companies competing with them for the low end, few if any companies will be able to compete toe to toe at the cutting edge. Maybe IBM - who else?
While a shrinking percentage of the market is fitting in that high performance bucket, I would hazard a guess that the absolute numbers are still going up, if only because of server chips and gamers. Though I could not find any data on this and would be happy to be proven wrong.
That is an interesting assertion. While it is true that Intel has great fabrication chops, their biggest competitor in that space, Samsung, makes ARM chips. There is an outside possibility that AMD gets into the ARM space, gets some traction and Samsung brings their manufacturing muscle to bear and captures the 'high' end.
It is interesting to understand the 'Intel monopoly' [1] and how it came to be. I worked at Intel in the mid-80's and basically IBM 'created' a personal computer market by bringing their reputation to the market, Microsoft provided the software, and rigid compatibility was demanded by the market to maximize software re-use (software and training has always been the biggest 'cost' of switching instruction set architectures). Early on Intel was cross licensing that 8086 architecture with partners to get market share against the 68000 architecture but later they clawed back a lot of the licenses.
The requirement that is be 'windows compatible' created a giant moat around the x86 instruction set architecture. That moat wasn't officially breached until the early 2000's when AMD announced their Sledgehammer (aka Opeteron) AMD64 architecture. Here was an architecture that was not Windows compatible (windows didn't really support 64 bit chips until Windows 7) and yet AMD managed some big design wins in servers because of Linux and OEMs who could adapt quickly to the new extensions. Intel finally caved and shipped the AMD extensions in their own chips, even though they knew it was probably the end of the Itanium as a viable product.
Now AMD announces an ARM server chip, they have some expectation that it will gain some market share because they did it with Opteron. They are even calling it an Opteron perhaps as a nod to that history. They will be able to innovate in the peripheral chips because all off the 'bets' are off in terms of what has to be in the 'south bridge'. They can do more creative memory interconnects because they aren't constrained by folks expecting the Intel stuff.
And here is the interesting thing, there are easily 10x the ARM licensees than the x86 licensees. That pretty much guarantees a lot of competition and price pressure.
The strategic questions are these:
Can AMD establish a 'server' standard? Something like the PC spec but for ARM chips, a minimum set of things that will always be there and a way to find out what optional things are there. This enables shipping a generic kernel that boots on everything and customizes itself to the environment.
Can AMD survive the ingress of a number of other players earlier in the life cycle rather than later. If they are getting traction expect competitive chips in 12 months rather than 24 or 36 months.
What is Intel's "Thermonuclear war" strategy and would they be willing to pull that trigger. Unlike Apple's Android antagonists, AMD is not well positioned to survive a huge litigation punch.
We'll see of course, the game is afoot as Sherlock Holmes would say. I'm hopeful that we crack this thing open and start innovating again.
I don't think it's true that gamers will always drive an appetite for faster chips. At some point the real choke point will the content production which in turn will be choked by how much gamers will pay per game. Sure you can get intro procedural stuff, and faster chips allow lazier implementations and higher level programs... but at some point price, power and portability do win over speed.
I tend to think, that Intel will be able to leverage their process advantage also against ARM. For one thing, Atom based cell phones are at the moment rather unremarkable.[1] This means, that Intel can fully leverage the advantages of the 22 nm fin FET process. (Since the cell phones are unremarkable, one will get a good phone by just putting a better processor into the phone, as opposed to the risk of developing a viable platform for general use.) On slightly higher powered devices, Intel has a huge ecosystem and the marketing advantage of building stripped down versions of higher powered processors. For even higher performance, Desktops and Servers, Intel is already close to a monopoly and there is no one to change this. And in addition Knights Corner/ Xeon Phi seems to be a rather good competition for GPUs in HPC applications. ( They just don't have a good competition for IBMs Power processors.)
So in conclusion I think that Intel is in a good position to beat ARM in their home field (cellphones, tablets). Can easily shield the low power desktop/ netbook market from ARM and has of course the dominant position for high powered processors. So I simply do not see a bright future for ARM based (consumer) computing. ( That is not to say, that there are not niches for ARM, but Intel has the stronger position in everything the average consumer thinks of as computers.)
Normally, this would've been the right strategy - if done 2-3 years ago. But at that point whoever was running AMD was still being dismissive about ARM and the mobile market. Now they realized they shouldn't have been, but it may be too late.
This is why it pays to have a visionary CEO, one that can feel where things are moving years ahead of the others, instead of a mere "manager". Transitions to disruptive markets can literally kill your company if you're not doing it on time. By the time a "pragmatist" manager - one that looks only at present data - realizes what is going on, it's already too late.
Flawless execution for the switch to ARM may not be enough. They may need a good differentiation strategy, too. If they knew they would do this, it may have even paid off to be the ones buying the MIPS architecture. The ARM market is starting to consolidate right now, and I'm not sure there will still be room for AMD's stock ARM CPU's (and probably stock GPU's too?) by 2014.
Exactly. The hindsight bias that is present when discussing CEOs, failed companies, successful companies and business in general is absolutely stunning.
Well no one on the ARM market other then AMD has the expertise in Server CPU. Nvidia may be able to do a very High Performance ARM CPU for HPC but it is completely different matter on Server. There will also be NAS, Small Server, etc and whole not of other market where a big name ARM player could help. Although most of that wont happen till 2015. I am not sure if there is enough time for AMD.
And given the recent sales figures that 70-80% of CPU sales are coming from APU, the APU is rasiing, CPU is a shrinking business, it makes sense for AMD to consolidate and make that 80% better and slightly more profitable rather then trying to complete with an aggressive Intel.
IIRC they did have this sort of visionary CEO, but they sacked him because he was spending too much money developing new technologies.
OTOH I expect AMD will do pretty well in the lightweight thread server market. They've got server experience and infrastructure, they've got a decent low-power x86 core, and they've spent of lot of time thinking about having cores with different instruction sets living together.
You're correct about the server market. AMD acquired SeaMicro earlier this year who specialize in high-density, lower-power x86 servers. They've already stated their intention to produce ARM servers.
And newcomers can't enter the market because they'd need a patent cross-licensing agreement with Intel, like AMD had. And a newcomer is unlikely to get such an agreement.
I'm sure AMD would be more than happy to sell their position to any would-be competitor, if it's transferable.
I think the bigger problem is just that it's extremely difficult and expensive to compete with Intel on x86. And all of a sudden the x86 market has taken a turn toward lower growth and lower margins. So, really, why would anyone bother when it's cheaper to get into the mobile/low-power ARM race and those markets have so much more up-side and room to make advancements that aren't contingent on billion dollar foundry investments?
AMD is still developing their Atom-class processors, so technically there is still competition. It is possible that their next generation of netbook CPUs is showing such good performance that they decided that they could succeed in taking it upmarket. I doubt it, but it is possible.
It's a very strange move for AMD. The A15 chips they're developing are roughly in the same performance & power markets as modern Atom-class chips.
Another way to think of this move is that AMD is betting the farm on the bet that mediocre CPU + fast GPU is a more compelling play than fast CPU + mediocre GPU. That might be a good bet, but it's too bad they no longer have the resources to take both sides of the bet.
apart from market economics, it's frustrating to not have political choice. intel invests heavily in israel and for some of us, that was reason enough to buy amd (and yes, i know they recently opened a centre in tel aviv). really frustrating news.
Oh please spare us. You buy [just about everything] from china, your car/bus runs on gas from [some dictatorship], your home is powered by coal. You don't want to buy from a multinational because they employ Israelis? You need to get out and meet some israelis, nice people. Probably worried about a rocket landing on thier heads, but nice people.
actually, i do know some israelis (and you are right - they are nice people). in general they are also as frustrated as i am with current politics and are quite happy with my attitude (in my experience, it's the american jews i know who get more annoyed, but some of them are cool too). in short, it's a complicated world, and takes all sorts. you're welcome to not share - and even belittle - my opinion, but i am, thankfully, free to state it...
Andrew, I decided to delete my rant because we are not understanding each other and you dont deserve to be ripped into publicly, no matter how stressed I am. I think my comment above is enough of a response.
your arguments don't make much sense to me, and i suspect everyone would be happier if we took this to email. so feel free to contact me at andrew@acooke.org if you think there's something worth rationally discussing here - all i am getting is that you disagree with my politics and that makes you mad.
(i don't blame you for being mad - i am angry too, but from a different viewpoint, obviously.)
I've always been an AMD fan, but their recent cores have left much to be desired when compared to Intel. I'll really be looking forward to what they do with ARM servers.
Maybe 2600k@5GHz is practical and theoretical limit for silicon based chips. Scientists were talking about this nearly two decades.
BTW: AMD makes low-end chips, maybe it is more perspective line for future development (power efficiency). Something like that happened with Pentium4 and Pentium3 Mobile.
One of the saddest things about this move is that they're moving into a market currently dominated by a competitor that they directly enabled. One reason that Qualcomm is doing so well is due to their purchase of the Adreno platform from AMD in 2008.
Like ElliotH mentioned, having a monopoly is not illegal. What is illegal is using your market power to drive out competitors and crush upstarts. If Intel is guilty of this[1], then they will face anti-trust hearings. If they're not, having a monopoly on x86 processors will not be enough for them to be punished.
On this one, it seems to be AMD causing their own downfall by poor management decisions leading to bad engineering practices. There's a lot of talent over there that is going to waste.
[1] I believe AMD successfully sued them for this in the past, but that seems to be over and settled.
Microsoft has a desktop monopoly with their 90% desktop market share. They aren't currently being sued. It was Microsoft's illegal tactics in the 1990's that got them in trouble. These days I don't believe they are doing anything illegal.
No. 2500K or 3750K clocked to 4.8 GHz is more than enough for now.
Intel are not in a good position to be able to price gouge the market. There is a point that will be hit in the near year or two where an arm device with lots of ram will be worthy desktop environment. If intel tightens the screws it will just come faster. And although not a trivial task if the pain becomes way to much for the enterprise market we could see some massive efforts into x86 emulation on ARM.
"Updated 11/19/2012@10:15am: AMD contacted us with an official denial of the story and stated that Kaveri and the big cores are still on track."