Slashdot Mirror
AMD's David to Intel's Goliath
Diox writes "A very insightful article was posted on ye old Tom's Hardware Guide about AMD's and Intel's strategies over this past year.
" Good stuff as usual from Tom's.
← Back to Stories (view on slashdot.org)
As the Grand Prix driver Alain Prost once said, it doesn't matter how fast you are, or how far you are in the lead, it's whether you finish that counts.
So long as Transmeta works hard and concentrates on it's own stuff, it might inherit the lead, by default, with AMD and Intel effectively taking each other out.
That lesson would be best learned by Linux users & developers, too. You can't win a race you don't finish.
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
Bruce
Bruce Perens.
Try sticking a number 1-6 between "www" and ".". For example, www6.tomshardware.com. -jwb
Intel has lost the low end of the Xeon market. For two or fewer processors, the Xeon is a waste of money. They have lost design-ins in that market segment. Everyone switched to single or dual Pentium IIIs.
Mea navis aericumbens anguillis abundat
T1s could be much cheaper than $600 a month if there was true competition in the telephone business. The cost of providing a T1 has dropped like a rock but the telephone companies have kept their rates artificially high. DSL technology has made it much cheaper to provide a T1.
Mea navis aericumbens anguillis abundat
This suggests (but does not prove; definitive answers will have to await the availability of Crusoe systems for more orthodox testing) that the Crusoe is struggling to carry out DVD decoding and would have fewer CPU cycles to spare in a multitasking environment than the PIII.
Seeing as how Transmeta's primary platform for their Crusoe processor is webpads and other portable devices, having fewer CPU cycles to spare while decoding a DVD (assuming he means watching the movie, and not just running DeCSS <grin>) I frankly don't see where having less spare cycles should be a major issue. On a <b>portable</b> device one is not very likely to be doing anything besides watching the movie!
Even if (for instance) one is running some sort of network service(s) such as http or email while watching the movie, these services typically require very little CPU time, and should not be significantly degraded -- Unless of course Rob is going to be trying to run Slashdot off his new webpad!
--Cycon
Your Brain + EEG + LEGO Robots = Brainstorms
needs to take some jounralism classes, you can hardly read the article because his personal bias is laid on so thick. How come everyone is clamouring for the death of Intel? If it weren't for the x86 series of processors computer technology would not be where it is today, at least night in my livingroom. I think Intel did make a major mistake with Merced/Itanium. The VLIW structure works with the Crusoe which isn't touted as a powerhouse processor but it is a joke with a processor as complex as Itanium. I think with the Itanium Intel out to have taken a que from MIPS and designed a specialized core to do a given number of instructions very very fast. Of course doing so would mean no software you have now would work on it but that would be a bonus in the long run. The intelligent thing would be to make only a recompile necassary to run the software, maybe even add a little emulator in the BIOS that would allow for using x86 code. If all of the processor's instructions were SIMD capable they could easily build an ass kicking chip where the compiler would batch processes together and allow for branch prediction to keep the pipeline full. A simple chip architecture would mean alot less heat and alot cheaper chips which would do well to increase the profit margins on said chips.
I'm a loner Dottie, a Rebel.
He's got to be kidding. It's waaay too early about the impact sledgehammer will have on merced. There is no sledgehammer. AMD hasn't taped it out. They certainly haven't sampled it. It's in the vapor stage. Admittedly the release date for the merced keeps slipping (making it the perfect processor for running win2k ;) ) but sledgehammer isn't anywhere close to being released. Can anyone tell me of any active project to modify an OS to use sledgehammer's 64 bit instructions? All the sledgehammer hype sounds like wishfull thinking.
--Shoeboy
I can't believe this crap. AMD comes up with admittedly good x86 implementations, but has historically failed to execute; i.e. it has failed to deliver enough of any processor family in the quantities required on everything up to the Athlon. The Athlon has been the exception so far, but we're only talking about three quarters. Intel's failure to deliver on processors, on the other hand, is measured in one or two quarters and then the supply matches the market demand.
Then there's the talk about Transmeta. I agree with the general thrust of the article concerning the Transmeta chips, but I have an even dimmer opinion of the company and its products. There are too many other low-powered advanced architecture chips for mobile system use, in particular the ARM series of microprocessors. Why go with a questionable unknown from Transmeta when I can pick a variant of the ARM from a number of vendors, and there's even a port of Linux to the ARM, along with a Java port at Blackdown. Linus Torvalds' may have the magic touch with software, but it's not carrying over too well with hardware. If I were he, or one of the Transmeta investors, I'd seriously question what Transmeta delivered for all that money and time.
Not to mention that most software will only use the processing resources of 1 processor at a time, so in the end you'll just be able to run 16 programs at near-full 50 MHz speed rather than 1 at 800 MHz.
Glückwünsche, haben Sie Slashdot ermordet, indem Sie zum korporativen Druck beugten und Subskriptionen einlei
It was funny once
Now it's passe
Just like the Grits, and Natalie Portman thingy.
Enough about the global financial crisis that will occur when intel's failed business plan leads its stock, the NASDAQ, and the global financial system into ruin.
/.. We want the skinny on Transmeta!
This is
Although it is believed that Transmeta initially intended and believed that Crusoe would outperform a similarly clocked PIII, Crusoe fell well short of this. Performance is so poor, that when Transmeta publicly introduced the Crusoe, the company described the performance of its new VLIW processor with a set of bizarre benchmarks....
Why is the Crusoe falling so far short of initial performance expectations? Again, as with the Itanium, the problem can be trace to VLIW. This design philosophy, especially for general purpose CPUs, is in its infancy, so no one has experience to set truly credible performance expectations.
Does this mean that the Crusoe is already a non-factor? Yes, and no. Yes, and Transmeta has already conceded this, the Crusoe will not be a player on desktops anytime soon. But in the portable market, Crusoe has a chance at success. With this greatly diversifying market, Crusoe's low power consumption is attractive by providing reasonable performance where battery life is paramount. It has the potential to be successful in Web tablets and low-priced, long battery-life notebooks.
The key to Crusoe's success will be if Transmeta can rapidly reduce the price of the 700MHz version to below $100 (it was announced at over $300).
What do you folks think? Is Crusoe a disappointment? Do you think that the do-it-in-software transmeta philosophy will pay off in prices as well as power?
Preferential Voting: easy as 1-2-3
The fundamental difference between a processor and an OS, is that a processor can be switched for anthoer one with no change in the operating system or programs running on top of it. This used to be clearer with the old socket seven boards where one could switch a Pentium for a K6 and then just boot the system. Now most processor changes come with at least a motherboard change as well. The thing of it is, that the new system can still run the old code. The processors run the exact same instruction set (minus 3DNow and SSE) and can therefor run the exact same software.
I will agree though that it will take more than better technology to display Intel from its current posisition. Many companies will only buy Intel machines for thier users. Intel has done very well to build loyality with thier Intel Inside program. Dell still refuses to sell AMD parts.
What the future holds for Intel is smaller profits and a more competive market place. Something Intel is not ready to handle as of yet. They'll be taking a beating until they do.
Competition == Good;
M$ == Competition;
M$ == Good;
-----------
"You can't shake the Devil's hand and say you're only kidding."
-----------
"You can't shake the Devil's hand and say you're only kidding."
So now that Microsoft has been proven both good and evil, are they an oxymoron? Or does it work like antimatter/matter annihilation, and most of Redmond just vaporized?
... if that turns out to be true, then i guess m$ really is an oxymoron =)
i'm not sure if this is true, but i heard somewhere that micro$oft used to be called 'unlimited limited' and 'unincorporated incorporated'
_______________________________________________
There is no statute of limitation on stupidity.
I'm talking about the general computer market, for which it is nothing but a PR disaster, not about specialized markets for which the PID has some merit (though software serving the same function, such as digital signatures, would probably be better even in the latter case because of its greater flexibility).
/.
/. If the government wants us to respect the law, it should set a better example.
If they lack the sense to take even obvious steps that don't cost any significant resources (e.g. announce "OK, Big Brother Inside was a stupid idea, and every batch off the line from now one will be burned with all zeros instead of a chip ID"), a "nice reserve" only delays the inevitable.
/.
/. If the government wants us to respect the law, it should set a better example.
The article makes some worrying claims about the Intel business model:
dependent on high margins, unable to reduce costs, long time to
market, losing its best engineers. These suggest that Intel can't
just mark up its recent troubles to experience and get back to old
form. One way or another it does look as if Intel shareholders are
not going to be happy.
circuit, but it was just gate logic. It was Intel who first
assembled all of the components of the processing unit onto one chip.
Did TI do something that I am not aware of?
Intel has been so succesful for so long for good reason. I believe AMD might finally get in the black, and Transmeda might gain some market share... but Intel will continue to dominate.
The difference between these companies is simple:
Transmeda and AMD are good at making great chips.
Intel is great at making *MONEY*
McDonalds does not make the best or the cheapest burgers out there, but it has continued to thrive because of an ingrained, highly succesful business model. The better products often lose.
And while that may be upsetting to all you geeks out there, because in a perfect world, the best product always wins... but that is not how it works.
Last time I checked, Intel had absolutely no debt and over 10 billion dollars sitting in cash. How much do you really think they are going to suffer?
Theres a difference. If Linux was the only OS, there still would be quality-breeding competition. I'm not talking between the different distributions warring it out, but between the developers, the users, etc. I have rewritten many a line of code just to prove my implementations are better than theirs. OSS isn't about money or fame; It's about satisfaction, and showing up the other developers and improving the quality that hits the user are both very satisfying..
.sig: Now legally binding!
Sorry, I like my algo better!
Corruption == Evil
Power == Corruption
Money == Power
Bill Gates == Money
Microsoft == Bill Gates
Microsoft == Evil
So now that Microsoft has been proven both good and evil, are they an oxymoron? Or does it work like antimatter/matter annihilation, and most of Redmond just vaporized?
.sig: Now legally binding!
Tom says...
Huh? As far as I remember, Itanium has always had backward compatibility; it was supposed to have a Crusoe-like translator to change x86 instructions to VLIW instructions. Did this get tossed, or is Tom smoking something?
--
It absolutely baffles me that most people here think that AMD success == Intel leaving the microprocessor business. Talk about blowing things out or proportion!
.18 process (which was the fastest ramp-up in its history). Intel is going to so seriously kick ass when that process is fully mature (and AMD, also, will likely have problems ramping their .18 up).
The only reason AMD was so successful last quarter is because the demand for Intel chips FAR exceeded Intel's ability to manufacture them. This hardly says anything bad about Intel! It says their customers can't get enough of the parts which Intel made. AMD was used, at least by the major OEM's, as a substitute for Intel chips it couldn't get. Basically, AMD got lucky, because they were able to make what may potentially turn out to be long term relationships with the top OEM's as a result. If Intel didn't have manufacturing problems, AMD would have done much worse.
The Athlon succeeded, not because it was faster, but because it was available.
Now, here comes the meat of my point. Intel was not able to satisfy demand the huge demand. Q499 was Intel's record quarter in volume (and sales, and profit...) The reason for this is because the demand for PC's was HUGE that quarter. This is GOOD for chip makers, both AMD and Intel.
A record volume quarter which was hindered by manufacturing problems suggests to me that all Intel needs to do is get their manufacturing back on track, and they will be even better and do another record quarter! Intel's problem is not an inherent problem such as lack of demand, but it is too much demand! Manufacturing problems are temporary, and Intel's are associated with ramping up the
Right now Intel has 84% marketshare, and AMD has 14%. Tom's suggestion that Intel is going to go to 0% marketshare and AMD is going to go to 97% marketshare is ludicrous because:
AMD DOES NOT HAVE THE MANUFACTURING CAPACITY TO SUPPLY THE ENTIRE PC INDUSTRY WITH CHIPS.
Intel has 7 or 8 really big fabs, all of which ALREADY are 0.18. AMD has one, which is in the process or being ramped up. The suggestion that AMD with this one really big fab can out MANUFACTURE Intel with a bunch of really big fabs is just ludicrous.
Of course, Intel does not have the capacity to supply the industry, either. We saw that last year. However, Intel does have much more capital and no debt, and is putting two new fabs up this year. Intel can recover from manufcaturing problems because they have money.
Folks, this is NOT an issue of who has the better chip. There is virtually not performance difference between Pentium III and Athlon (if anything, Pentium III has a significant advantage right now because it can do MP and is available with much bigger and faster caches). All the OEM's want is parts. This is an issue of manufacturing.
Worst case scenario (for Intel) is that they drop to about 50% marketshare. Is that bad for Intel? You bet. Their market cap would go on a free fall.
I do believe that AMD is undervalued and that Intel is ludicrously overvalued. Does that mean Intel will go bankrupt and AMD will be worth 1/3 trillion dollars next year? No, but probably they will come a bit closer in price (perhaps even the same order of magnitude)
I like AMD, Intel annoys me, but I'm not ready to dismiss them. The article is right, Intel is hurting, but while a drop in margins will hurt their valuation, they aren't near bankruptcy.
They can throw HUGE amounts of resources into fixing their woes, which is a nice reserve. They have to accept a LARGE decrease in margins, especially as the Intel name means less and ceases to give them a premium.
Intel has massive manufacturing might and can crank up production on lower cost chips. Corporations aren't ready to adopt Athlon for mission critical systems, although that may change. Microsoft's decisions will also make a key difference. Will Microsoft realize an AMD optimized compiler? Will Microsoft release software (including NT/Win2K) optimized for the AMD system? Will third parties release drivers that don't suck it up under AMD chipsets (see the GeForce fiasco, mentions on Tom's site)?
Realize that for servers, a few hundred dollars in savings is trivial for the Intel guarantee. AMD will have trouble stealing the NT market from Xeon. The Linux market, however, could be theirs for the asking if they contributed optimizations for GCC to make it optimized for their processor. If Red Hat made an Athlon Linux (all the software recompiled to scream on an Athlon), it would hand them that market.
AMD has a HUGE opportunity hear, but the rest of the industry has to play along. A strong AMD and Intel duking it out for the x86 server market is GOOD for the industry, but we'll see what happens.
Alex
I guess that when another OS comes out that can do what M$ does and ALSO allows people to run there legacy M$ programs it will take the dominate place in the market if it is better.
Please don't tell me about wine and linux and vmware, cause they do not completely fit the bill. Wine suffers from the fact that it cannot run all of M$ products and results usually vary from user to user to much at this time. Maybe when corel's changes get in there it may make a difference, but not enought to make people switch.
Yes I also know that there are substitutions for Word and Office, but flat out some people just have to run word, or just want to run word cause there is some feature in M$ product that they like .
As far as vmware goes, when I tried it my cdrom started going to crap. They write there own driver for the system (under Linux atleast). I had other problems with upgrades to vmware not working and not letting me install new programs in a vmware sessions. It also does not handle all the hardware that it should. They need to write a proper program (IMHO) that does not make the kernel modules that it makes. There is already a cdrom driver why reinvent the wheel, so they can do what they want with the cdrom?
I'll look forward to seeing if Netscape can make a comeback with Netscape 5.0......
send flames > /dev/null
Only 'flamers' flame!
It really does go a long way to prove the efficient marketplace theory doesn't it?
And Intel certainly did have a monopoly... I believe they actually lightened up on AMD a lot... just to keep them straggling around so the justice department would get off their back.
What I think is the key here, is that a long-running technology monopoly is near impossible.
The barriers to entry are simply too low, and things change at such a torrid rate.
The monopolies that were able to last had control of static commodities (oil and I would really include phone service (namely, the lines) at that time as a commodity). Where a start-up company would need enormous amounts of capital to even enter the market... let alone fend off a larger company.