The Inquirer features a long article on AMD Opteron and its potential.
AMD’s Opteron: Does it measure up?
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
This is a good analysis of the market situation, but does leave out the other competition in the same server market that AMD is assaulting.
HP still has PA-RISC and Alpha out there. IBM has POWER4 and soon POWER% and PowerPC 970. MIPS, while not the strongest it’s ever been, still is out there. And Sun with it’s SPARC.
I think the market is just getting interesting.
with the Dragon CPUs from China as well. I can’t wait to see if all the politics surrounding American CPU designs can keep up with real progress.
It has been argued that had the Alpha progressed as it was planned it would be wiping the floor with IA64. But politics and capitalism put that one to rest.
Gah! The poor 1.25 GHz Alpha is competitive with a 2.6 GHz Pentium 4 when it comes to floating point performance (and the P4 is no slouch in the FPU department). That’s after getting almost no redesign work for the last half decade. With current process technology and some layout work, I’d guess the Alpha could easily be brought up to 2+ GHz. At that speed, it’d beat anything else on the market in FPU code. Plus, its architecture (IHMO) is much prettier than the PowerPC’s.
“It has been argued that had the Alpha progressed as it was planned it would be wiping the floor with IA64. ”
No need to argue, the 21364 does indeed wipe the floor with Itanium2… HP quietly released it, because it would take sales away from their beloved IA64
. Thus far it gets better specfp scores, and that is without a current process technology. Even the 21264C can be quite competitive when compared with Itanium2.
It is a shame really…. thankfully IBM still keeps the good fight with the Power4/5
On a clock per clock basis the R16K is quite impressive even at .11um. Too bad SGI doesn’t have access to the latest processes either. Still impressive for a machine which is nothing more than an improved R10K. Now that was a beautiful design!!!!
Keep in mind that the Alpha has an enormous disadvantage compared to other architectures: cost. Its die size is enormous due to its ludicrously overcomplicated core (namely the branch predictor) Sure it’s an excellent performer, but Opteron has it beaten hands down in the price performance department.
As far as where Alpha was going (when Compaq was still developing it following the DEC purchase) it really wasn’t that interesting. They were planning on taking it multicore and integrating a (RAMBUS?) memory controller on die. There weren’t that many interesting architectural improvements on the horizon.
Sun seems to be the only one actively researching architectural improvements to their processors. Unfortunately, with Sun outsourcing production of SPARC to TI who is in turn using outmoded production technologies (SPARC is just now moving to a .13u process as other processors prepare to jump to 90nm) Sun cannot keep up with the rapid advancements in manufacturing processes being pushed by Intel, IBM, and AMD.
that was a very good article, now i want an athlon64. I wonder if there are any dual opteron mb’s with an agp slot :p
There was a small article in PCMag about a month ago where they benchmarked a dual Opteron 1.6 ghz against a dual Xeon 2.8 ghz machine and the Opteron beat it soundly in all test. IIRC the margin was at least 30%.
that was a very good article, now i want an athlon64. I wonder if there are any dual opteron mb’s with an agp slot :p
And not will be for time, the only AGP Chipset that you will see soon will be a Nforce 3 but only will support one Processor…
Maybe when the Athlon 64 arrival as early as September (in august will be the paper launch)…
Remember? The release of the first round of K7’s, renamed to “Athlon” at darn near the last second, was just as significant for AMD. They spanked the Pentium II clock for clock and elevated AMD from “second or third rate” to “Intel’s major desktop contender.” I mean, the K6 series was good, but the Athlon got AMD the critical and hobbyist respect it needed to stick around until now. Not to mention the sales. If it weren’t for the Pentium III’s with SSE, Intel could have lost a lot of market share on the consumer desktop.
It seems like the Opteron series is in the same position, possibly a better one because Opteron offers almost everything IA-32 can do (no Hyperthreading), and something it can’t–64-bit computing for free. And it SEEMS like Opteron systems may spank current Xeon systems already in raw performance. Pretty darn good for a first release. Kudos to AMD for taking their time.
–JM
<whine>But I want my 2 way Athlon64MP box noooooow</whine>
Rayiner Hashem, why does it seem that whenever I see your posts, you are ALWAYS offering negative remarks toward anything Apple-related? Your comment above is a perfect example.
(Actually, you and another guy named Remaja)
When looked at independantly, it looks like you are simply offering your perspective, which wouldn’t cause anyone to think anything out of the ordinary, but when coupled with nearly every single post you make, it becomes ever-more apparent what your agenda is.
Is this your subtle attempt and fortum trolling without bringing too much attention to yourself?
was very powerful for the sega dreamcast.
given its compact efficient 16-bit RISC instruction set i can’t but wonder how well it can scale.
Hitachi makes a version of it’s SH series for the higher-end desktop, even has a 64-bit version from what I’ve heard.
Not a bad choice. It’s 16-bit instruction set is an advantage, actually.
” Keep in mind that the Alpha has an enormous disadvantage compared to other architectures: cost. Its die size is enormous due to its ludicrously overcomplicated core (namely the branch predictor) Sure it’s an excellent performer, but Opteron has it beaten hands down in the price performance department. ”
Excuse me? Hello? Alpha’s cost had nothing to do with the “overcomplicated core”, it had everything to do with VOLUME. If you want over-complication pick up any moder X86 processor, it is orders of magnitude more complex than anything Alpha ever was. Most of the size in the Alpha dies (21164 and up) was due to the added funcional units, and more importantly the L2 cache on chip (96K I believe). The “branch predictor” was nothing in comparisson with the cache and the its control logic (tags and whatnot)…
Alpha was a beautiful architecture that could be really pushed, the fact that it was able to obliterate the competition (performancewise) for years without much improvement or investment speaks volumes. Oh, and dual core and memory controllers don’t seem exciting to you, what were you expecting “psychic” photon based quantum computing?
but it was not to be. It’s funny that Intel always used to hold up their superior floating point performance over AMD – at least until the Athlon came along. Yet, they still haven’t matched what the Alpha could do.
When the boys at DEC built something – hardware or software, they did it like no one else.
Yeah!!! Alpha is quite goot. We have 2 nos of Alpha here running our critical server application.
However in my opinion, those with the company that produce Alpha just fail in their marketing effort. Can you imagine that the majority of IT personnel here give a blink eyes when sombody mention about Alpha processor. Without press coverage etc. is why they cannot get enough selling volume to survive.
And one more bad thing is that, few years back (before we buy the Alpha server) I did contact Alpha processor team asking how to get the server, you know what is the answer? NO RESPONSE AT ALL UNTIL NOW. Can you imagine why they fail? Their bad attitude just kill them.
I admire Microsoft very much in term of marketing effort (but not their monopoly practices) because I have experience personally which I think I got cheated into buying forge copy of their Win98 and give them a complaint. The result is that they just register me as vendor and getting free preview and got invited to every product launch that they have. I think this is why MS manage to concour the global software market, GOOD MARKETING STRATEGY.
it’s a pity about what happened with the Alpha from what i read about the EV8 (on realworldtech) and about the proposed EV9 (Tarantula was the the codename) it would have really rock. DEC use to have a fairly big plant here in my hometown but they moeved there hardware plant to Scotland (kept software here though).
From the DEC hardware i’ve messed around with (VAX’s and Alpha’s) i’ve nothing but respect, they sure knew how to make good kit.