Slashdot Mirror
More Drooling Over The Opteron
bradv writes "I havent heard much about the new 64bit chips from AMD lately and was excited to find this article to satisfy my appetite for a little while longer. Probably more info than most people will ever care about. "
How can I not love a 64 bit processor that for some reason makes me think of the Transformers...
Fire in the hands of the village idiot is no tool, but a weapon of mass destruction
SPEC results linked from The Inquirer ... here
I think this is accurate because of the architecutural choice AMD made--instead of going with an all-new architecture, ala Itanium, they instead blew out the x86 system to 64 bits. That level of division in the CPU market at this time feels like it will have a very significant effect on the balance of power.
I dont know why this needs to be continually restated but... AMD (Advanced Micro Devices, not Advanced Micro Processors) makes more than just PC processors! PC processor's are the tip of the iceberg when you consider embedded processors. TV's, VCR's, phones, dishwashers, etc. are all using embedded processor's now, and the # of products that use them in the future will continue to grow. Absolute worst case scenario, AMD will bow out of the CPU market. AMD is a huge company. One, two, three, or even four flopped products are not going to phase AMD. Look at AMD's website. PC processors are the most mainstream product they make, and thus the highest profile, but processors are just a cog in the AMD wheel. (And the same thing goes for Intel).
It is a common assumption that 64-bit is for servers only. I am working on a quite widely used medical imaging & physics application that is suffering from the 2 GB (and even 4 GB) barrier at the client side. The CT/PET/MR image data with symbolic images, triangle meshes, dosimetric data, etc. are just too much for the 32-bit memory space. Our db servers are fine with 32-bit memory space, but the clients must be upgraded pretty soon now.
-- Imperial units must die --
"Now, this is cool tech, so it's a fun read. But is anyone really holding there breath for this thing? This thing is doomed without support from Microsoft, and they are going to be in bed with Intel as usual. AMD should stick to what it does best, emulating Intel's CPUs, until it can amass enough market share and forge enough partnerships with OS makers to strike out on their own."
Actually, Microsoft has endorsed AMD's 64 bit platform. They are also pressuring Intel to adopt a x86 approach as well. Why? Easier to code for.
The Itanium has been nothing but a disaster so far.
The Opteron could be a real turning point, with Intel for once forced to clone AMD designs...
Corporatism != Free Market
Yes, but as far as the PC market is concerned the original poster is correct -- if the Athlon64 and Opteron do not significantly increase AMD's market share then they're gone -- as in out of the market. Which means we essentially go back to a single chip maker monopoly for the PC market.
If AMD wasn't around to spur Intel on (and vica versa) do you think we'd have a 3 GHz CPU available to the general public right now? Yes, you can question the need for one, or you can whine about the price, but the reality is that competition has significantly improved both prices and features.
Will AMD, the company, go under? Doubt it. But they can't stay in a losing market much longer, and right now the x86 market is a massive loss leader for them.
Of course, the Opterons haven't been shown at full speed yet. By all means keep posting flamebait and disinformation, though. Estimated SPEC scores have been available for a while. Here is the relevant snippet:
A single Opteron core running at an actual clock speed of 2.0-GHz with registered PC2700 memory yielded a SPECint2000 score of 1202, and a SPECfp2000 score of 1170, Weber said. He did not formally disclose whether the chip was a "Clawhammer" or "Sledgehammer" chip.
The scores for a Dell 3.06 P4 are 1084 SPECint, 1092 SPECfp. Not bad for 2/3 the clock speed...and much faster on integer performance than Itanic. :-)
Do you really think AMD's new .13 micron chip will top out at 2.0 GHz. in the near term?
The other beauty of Opteron is the ease of building multiway systems up to eight-way...as opposed to HYPErthreading. Personally, I prefer multiple real processors if I'm going to pay additional license fees...
Galileo: "The Earth revolves around the Sun!"
Score: -1 100% Flamebait
Itanium has lots of cool new features that compilers could be using and people could be taking advantage of, but it doesn't have good backwards compatibility, and therein lies the problem.
Lots of data-intensive applications desperately need more than 2Gbytes of RAM. If Opteron can deliver that for only a modest premium over regular Athlon-bsaed PCs, it will be a huge success. And if it can run existing binaries in 32bit mode and work with existing drivers, that's icing on the cake. There is just nothing else like it out there.
As soon as they come out, assuming Linux does run reasonably well on them and there are no unexpected show-stoppers, we are going to buy half a dozen of them. We want a Beowulf cluster of these.
Electron Migration? what are you talking about. Processors continue to get faster and faster due to improved processing technology and increased parallelism. Leakage and electromagnetic interference from the clock signal are major problems today but who knows what scientists are working with nowadays.
> By doubling the word length to 64-bits, you can reduce the clock rate of the chip, and will still be able to perform more instructions per second than your top-of-the-range Athlon/Pentiums.
That is absolutely not true. Having 64 bits allows you to access a larger amount of total memory, and it lets you put more information in each instruction. The amount of data you can work on in any given clock cycle is proportional to the cache access and bandwidth and the register size (Neither of which inherently need 64 bit long instructions).
To perform more instructions per second (or instructions per clock cycle) you need instruction level parallelism (ILP). This has been a major goal of processor manufacturers for many years now. Intel had two main ways of trying to increase ILP.
Itanium has lots of cool new features that compilers could be using and people could be taking advantage of
Yes, and Intel would really appreciate it if someone would develop a compiler that takes full advantage of the Itanium. Really. Please. Because their own compiler is still struggling with the problems inherent in VLIW... yes, it's much, much, much better than it was a couple years ago, but it's still nowhere near where it needs to be.
And it costs a fortune. But, hey, if you can afford a $9000 chip, you should be able to afford the compiler too.
I like and respect Intel... I've grown beyond the newbie EE stance of "it sucks because it does", and recognize that they have some of the best minds working there, and that their fab processes are second to none. But Itanium has been a massive disaster for them, and they're now caught between a rock and a hard place. They can continue developing future revisions of IA64 and hope that someday their engineers figure out how to make it work well, work cheaply, and work fast with legacy code, or they can commit corporate hari kari and adopt x86-64 from AMD. Or they could do something similar, but different, to x86-64 on their own and just piss off everyone. Bad choices all around.
The only chance Itanium has is if AMD flubs the Athlon64/Opteron launch. AMD will probably pull out of the market shortly after and Intel can gradually increase profit margins to the point where throwing cash at a losing proposition (IA64) remains viable. And eventually force everyone to transition, like it or not (which, admittedly, would probably be a good thing in the long run, but the short term would suck).
The K5/K6 were Pentium/Pentium II clones, but the K7 was basically AMD's coming out in the Microprocessor realm, and has been extremely successful!
Nope.
Around 1995, AMD was really struggling to build a Pentium class processor. In 1994, the first company to ever reverse engineer an intel processor and create a functional equivalent was NexGen and their Nx586. This processor utilized a RISC core and a translation unit to get 386 instructions into RISC form. I actually owned two of these, a Nx586 66 and a Nx586 100. They were pretty funky. FPU's were optional, but most of the Nx586 100's had FPUs on board. There was always talk of putting a FPU socket a la 487 on the boards, but it never happened.
Neadless to say, AMD purchased NexGen in late 1995 and released the K5, a clone of the Nx586. The K6 was the first processor released by AMD which was faster than the current Intel processor, a oft forgotten fact. For about 3 months before the Pentium II was released, the K6 233 was 5% or so faster than the Pentium Pro 233.
The K6 and Athlon lines of AMD all utilized the same internal RISC core with a translation unit.
So, you are wrong, no one ever cloned the Pentium or Pentium II. A lot of nasty history between Intel and AMD in the 386 days made sure that would never happen ever again. I wish I could find some fun links on the Nx586 for you, but even on google it seems to have been forgotten.
I don't read or respond to AC posts