Slashdot Mirror
AMD Preparing To Give Intel a Run For Its Money
jfruh writes: "AMD has never been able to match Intel for profits or scale, but a decade ago it was in front on innovation — the first to 1GHz, the first to 64-bit, the first to dual core. A lack of capital has kept the company barely holding on with cheap mid-range chips since; but now AMD is flush with cash from its profitable business with gaming consoles, and is preparing an ambitious new architecture for 2016, one that's distinct from the x86/ARM hybrid already announced."
But the real fight of a decade ago, when AMD was first to 1GHz, the first to 64-bit, the first to dual core, seemed missing. It's not surprising since the company was facing a real threat to its survival. But with a gravy train from the gaming consoles, it looks like the company is ready for a fresh battle, with a familiar face at the helm.
Uh, wait. No. It was surprising when AMD was the performance leader. It was surprising because they were broke. It's not surprising to see AMD pushing out a new architecture now that they have money. It takes a lot of money to do that. So we start out completely ass-backwards here.
Much elided, then
The most logical move for Keller would be to dump the CMT design in favor of a design with simultaneous multi-threading (SMT), which is what Intel does (and IBM's Power and Oracle's Sparc line).
Wait, what? Why? Why wouldn't it make more sense to just fix the lack of FP performance, perhaps by adding more FP units? Why would it make more sense for them to go to a completely different design? It might well, but there is no supporting evidence for that in the article.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
I was so proud of them when they kicked IA64's ass with their amd64 architecture, beating Intel at their own game by choosing to be x86-compatible when even Intel didn't go that way. Then I was sad when amd64 started getting called x64, since it stripped AMD of the credit they deserved. Go AMD! A world without strong competition for Intel would be very bad for consumers.
Last I looked, Intel's R&D budget was larger than AMD's revenue
That certainly was true (probably still is), but it's misleading. AMD no longer owns fabs and the majority of Intel's R&D spending is on process technology. By spinning off GlobalFoundaries, AMD is able to share that R&D cost with other SoC makers and go to other companies if they happen to be able to do it better at a specific time.
I am TheRaven on Soylent News
Sorry AMD, you're heading in the completely wrong direction. CPUs are already plenty fast. They have been for years. 3D gaming is starting to look like just another "Gold plated speaker wire" guy hobby as everyone moves to mobile devices.
The real winners in the future are going to be the very cheap, very efficient chips. Do you want one very powerful computer to run everything in your house? Or do you want everything in your house to have its own dedicated, highly efficient CPU that does just what that device needs?
Compaq was afraid to use AMD chips given out for free, because Intel would "retaliate", ok?
What kept AMD's market share low was not "clever marketing" of its competitor, it's crime.
Back in P4 Prescott times, Intel's more expensive, more power hungry, yet slower chip outsold AMD's 3 or 4 to 1.
Not being able to profit even when having superior products, it's really astonishing, to see AMD still afloat.
how is spinning off your fabrication capability 'good' in the long run?
I don't work at AMD, but I do work at another company that relies partly on foundries.
Basically, it's economies of sale and competition. Semiconductor fabrication processes keep getting more expensive. Foundries specialize in process development and spread the R&D across many, many customers. Unless you're willing to spend a fortune keeping up (as Intel is), have special requirements, or need a ton of volume, you have little to gain and a lot to lose from rolling your own process. Remember, you don't just have to make transistors, you also have to have good enough yield to turn a profit and good enough reliability to keep your customers. If you fail, you have to spend even more money to fix the fab on top of the money you're losing on the stuff you manufacture. Meanwhile, TSMC is cheerfully cranking out wafers for your competitors.
Visit the
That is true. However AMD were the first to make a 64-bit architecture, which was x86 compatible. And it was also the first 64-bit CPU to be in a price range that was acceptable to average consumer. But most importantly, AMD designed an architecture so successful that Intel decided to make their own AMD compatible CPU. Today Intel probably earns most of its money on CPUs using AMD's 64 bit design.
But if AMD now want to go and build an entirely new design, which is nothing like x86, they may very well be repeating the exact same mistake Intel made to let AMD64 get the lead.
By now it might be safe to ditch all 8, 16, and 32 bit backwards compatibility with the x86 family. But AMD64 compatibility is too important to ignore.
Do you care about the security of your wireless mouse?
Netburst did seem like a reasonable idea, in testing. While it was low IPC, it looked like it would scale bigtime in the speed area. They had test ALUs running at 10GHz.
So I can see the logic: You make an architecture that can scale to high frequencies easily, and that gets you the speed.
Obviously it didn't scale, and wasn't a good idea, but I can see what they were going for. It wasn't like it was completely nuts.