Slashdot Mirror
Is AMD Dead Yet?
TheProcess writes "Back in February 2003, IBM predicted that AMD would be dead in 5 years (original article here), with IBM and Intel the only remaining players in the chip market. Well, 5 years have passed and AMD is still alive. However, its finances and stock price have taken a serious beating over the last year. AMD was once a darling in this community — the plucky, up-and-coming challenger to the Intel behemoth. Will AMD still be here in 5 years? Can they pose a credible competitive threat to Intel's dominance? Do they still have superior but unappreciated technology? Or are they finally old hat? Can they really recover?"
Warning: Link leads to a malicious website.
Hewlett-Packard
IBM
Infineon
Intel Corporation
Lenovo Holdings Limited
Microsoft
Sun Microsystems, Inc That doesn't leave you an awful lot of choice, does it?
"It does not do to leave a live dragon out of your calculations, if you live near him." - Tolkien
Long ago, FUD was the bread and butter of the IBM consultant, what's new?
It is interesting to note that this article is dated February 17, 2003. In other words IBM made this prediction literally 2 months before AMD introduced their first 64-bit processors, the Opteron, in April 2003. Little did they know the impact the AMD64 architecture would have on the industry (Intel cloned the architecture) and on AMD itself (it helped them stay afloat for the past 5 years).
O RLY?
Internet memes aside, Nehalem has been confirmed to have GPU cores glued together in the same package as the CPU. That means you could have a Nehalem chip with an Intel X4500 (or even the memory controller) in one package. Considering Intel is currently the largest producer of graphics processors and seems to be more capable of developing and launching such technologies than relatively-small AMD, I would not be surprised in the least if Intel's technology beats out AMDs Fusion technology to the market.
Site triggers my AV with this attack: http://www.microsoft.com/technet/security/bulletin/ms05-013.mspx
Intel stock is down, too.
See also this January 16, 2008 Bloomberg story: U.S. Stocks Fall on Intel Forecast, Extending Global Tumble.
Quote: "Intel, the world's largest computer-chip maker, tumbled the most in five years in Nasdaq Stock Market trading after saying first-quarter sales will be as much as 6.9 percent below analysts' estimates."
have you done any fundamental looks at AMD's balance sheet, income statement, or CF statements? have you seen how the stock has performed over the last 2.5 years?it's stock has been in a precipitous downward spiral. If you were long AMD for the last 2 years, you have basically been crushed.
now, why is it down? well, look at their earnings. they have done pitiful. turns out in a slugout pricewar, intel can stay profitable while AMD is on the ropes. last year they lost money and continue to show no signs of recovering from their tech deficit they have again built against Intel. Now adays, the fastest AMD chip not on the market yet is slower than what intel already has at full production.
FYI: they last 166 million dollars last year. I'm not sure why this looks like manipulation as compared to just poor performance by the company without much of an end in sight.
oh, and my disclaimer: following my advice will hurt my long position in AMD.
Doesn't Sun market/make some Opteron based servers?
bork bork bork!
I really don't care about how much better it performs in office applications, or whatever other tests AMD did to "prove" that clock speed doesn't matter. You can only stretch the truth so far, when one is doing number crunching a faster clock will get you more performance than faster context switches.
I'm not going to question your personal experience as the others did. It's certainly believable that there's an application where the AMD processor performs 1/4th the speed of an Intel one (and vice versa).
I just want to point out that if the AMD processor actually was 2200 MHz, then you still could have found that the application that interests you performed at 1/4th the speed on that processor than the equivalent Intel one. Meaning that without the performance modeling numbers, you still would have found the equivalent "numbers" to result in an invalid comparison.
Clock frequency is not an automatic benefit for number crunching. If you don't change the architecture, then obviously yes a faster clock helps. But you don't go from a 1600 MHz chip to a 2200 MHz chip in the same time period without changing architectures. Performance is Clock Frequency * Insructions per Cycle, and a wide machine (many execution units) with low memory latency is going to tend to have higher IPC. However, the IPC value varies wildly by benchmark, and a benchmark that reveals certain deficiencies or strengths of the architecture may fall well outside normal, as was your case.
AMDs "modelhertz" or "markethertz" numbers became strained when the new generation of Intel products came out, but for most of the life of the Pentium 4 they were extremely generous to Intel's architecture. It's not just office applications -- go check benchmarks on Tom's, HardOCP, Ace's Hardware, and you'll see the AMD processor outperforming in a wide variety of benchmarks from games to high-performance scientific computing (the true number-crunching benchmarks), even including some media encoding benchmarks though Intel was very strong there and generally dominated.
My point is that if what you're looking for is a singular number by which to compare performance, then there is no "truth" to be stretched. MHz is an actual measurable number, true, but to equate that number with performance is "stretching the truth" to a greater extent than taking an aggregate of a wide variety of benchmark scores and relating that to performance. Marketroids can and do manipulate which benchmarks are chosen, but at least the resulting number means something regarding the performance of those benchmarks. MHz, by itself, means essentially nothing for a cross-architecture comparison.
So next time if you want to get the truth about performance, then the truth is that you have to measure the performance of the application you personally care about on the two processors in question. You can get an idea from reading reviews with benchmarks and looking at the results of similar applications (i.e. media encoding, or games, or what have you), but even that won't get you the real picture.
The enemies of Democracy are