AMD Back in the Black

XaXXon writes "CNN reports that AMD had a profitable quarter for the first time in over two years. According to the story this is mostly because of their 64-bit line of chips (both Opterons and Athlon-64). AMD has forced both HP and Intel to change long-standing plans of only supporting Itanium, with HP coming out with Opteron-based systems and Intel releasing chips mimicking the 32/64-bit behaviour of the Opteron. According to the story, 64-bit processors are better than 32-bit ones because 32-bit processors 'can't take advantage of more than 4 megabytes (sic) of memory at a time.'"

Comments

Profitable

[0mni]

I guess its easy to understand that AMD would be running in the red, its prices are really quite low. Even with small production prices I couldnt imagine there would be too much profit for them.

Re:Profitable

[Anonymous Coward]

They haven't been that low for the last couple of years. I remember a time when every AMD CPU cost about half of what you paid for a comparable Intel.
Heavy investing and comparably small market share would have more to do with the losses.

Re:Profitable

[betelgeuse-4]

When I went to nanotech lab open day, one of the speakers said that 98%-99% of the chips on each wafer must work for the CPU company to make a profit.

Re:Profitable

[Anonymous Coward]

It depends on the chip and the market. Production yields for ARM CPUs would often be at 95%, a level undreamable by x86 chips.

Re:Profitable

[Loki_1929]

"According to this article in Toms hardware,"

I can't believe anyone still reads that rag. Let me give you a hint: Tom's was bought and sold about 3 years ago. Since that time, it has descended to the ranks of online propaganda host for Intel and a number of other companies. Tricks with driver versions and other such foolery causes them to get benchmark results drastically different from those of almost every other hardware site. No one's bias is more apparent than Tom's himself. Many of the conclusions to their own articles are non-sequitur, and the articles themselves are often little more than a press release for the company doing the most advertising at the time.

The few folks left defending Tom's tend to either be Intelbots or those who like to feel they know something after having read a few dumbed-down for-public-consumption articles from the site. To quote them is to invite laughted upon yourself. You would do well to visit other sites instead for your hardware news. Anandtech, Ace's Hardware, and plenty of other sites provide good, in-depth and trustworthy analysis, as opposed to operating a propaganda machine designed to rake in cash.

"prodution yield is about 30%, it it expected to rise up to 60% after two years of production CPU's are just too complicated to be produced with yield of 98%-99%."

But I say that production yield is about 80%, and is expected to rise up to about 95% after three months. And best of all, I can make up numbers and formulas to make it look very official and correct. When I see Tom in a 'bunny suit' on AMD's FAB floor, I'll believe their 'analysis' of production yields. Until then, he's making up numbers and statstics, adding to the other 73.4% of statistics that are already made up.

The quote from the article, which you have parroted here, is as follows: "However, we doubt that AMD's yield will be any more than 30% - this is based on information from other chip manufacturers that use similar processes."

Now, let's put a little bit of brain power into dissecting this, shall we? First of all, the whole thing is rather vague - using words and phrases like 'we doubt' and 'similar processes'. Secondly, these so-called 'other chip manufacturers' aren't even named. Are they talking about IBM? Or are they talking about 'Phil's CPU FAB', which is run out of a basement in a townhouse in Idaho? Just what are 'similar processes'? Is there someone else making Opteron and Athlon64 CPUs? Someone really ought to tell AMD about that. Or perhaps they're referring to the 130nm 'process', which describes almost nothing about the chips themselves? Maybe they're even talking about the 'process' of getting from wafer to die. Well, so far as I, or anyone I've ever talked to knows, AMD didn't go out and re-write the book on die construction with the K8. The 'process' of getting from wafer to die for K8 isn't that much different from that of K7. I would assume then that Tom's is also asserting that Barton and Thoroughbred yields are also a mere 30%. Or perhaps there's an entirely new made-up number for their yields.

It's amazing how you can throw a few numbers onto a website and everyone will believe you. It's almost as amazing that throwing a few numbers into a post will yield +4 Informative.

Intel, 32x64?

[arivanov]

As far as I recall, Intel has not released anything yet. They put something on the roadmap, but they are still 100% behind Itanic. They released an improved 32bit emulation environment for the latter though

Re:Intel, 32x64?

[sbennett]

See this. Of course, there were the standard rumours going around before Prescott's launch that it was going to have a 64-bit layer, but that didn't happen.

What I find interesting is that Intel said before Opteron's launch that they weren't going to make any form of 64-bit x86 processor, and now it's on the roadmap.

Earlier this week, Intel's President and COO, Paul Otellini, confirmed in a web-cast interview that a move into the 64-bit desktop market was certain, but that the company would nevertheless wait for the arrival of operating system and application support. "You can be fairly confident that when there is software from an application and operating system standpoint, we'll be there," he said.

You mean once the OS and application developers have started using AMD's 64-bit extensions, Intel will come up with something to compete?

Re:Intel, 32x64?

[Loki_1929]

"Of course, there were the standard rumours going around before Prescott's launch that it was going to have a 64-bit layer, but that didn't happen. "

It's entirely possible that 64-bit extensions are within Prescott, but disabled. Intel did this with the P4's SMT for quite a while. Xeons had SMT, and it was enabled, while desktop P4s had SMT, and it was disabled. The 64-bit instructions might not yet be finished, to be finalized and debugged in a later stepping of Prescott, or they simply remain dormant, used only as a preliminary testing grounds for Intel, while they're waiting for viable engineering samples of Tejas. To my knowledge, no one has completely accounted for all the new transistors inside the Prescott chips. The speculative execution enhancements, larger cache, longer pipeline, etc all provide for some of the extra transistors, but certainly not all. There's something about these chips that Intel's not telling us, and 64-bit extensions is as good a guess as anything else.

"You mean once the OS and application developers have started using AMD's 64-bit extensions, Intel will come up with something to compete?"

No, he means that when x86-64bit support is there in software, Intel will have a CPU at the ready to support it. Since AMD's 64-bit extensions are the only game in town, and Microsoft has told Intel to go stuff a second set of x86-64bit extensions, Intel will be forced to either emulate AMD64 (a thoroughly bad idea), or include the instructions as the core of any 64-bit x86 CPU they release. Intel has already licensed the AMD64 technology, and thus will be forced to use its 'little brother's' technology to stay ahead of the curve. The interesting thing about that is that AMD can then choose the direction for future instruction sets. So long as the industry is working off AMD's instruction set, AMD calls all the shots.

Intel's big mistake was continuing to behave like a monopoly, and ignoring the breakout CPUs of its chief rival. Intel was banking on a 64-bit nosedive on x86, choosing to all but ignore the concept until it was too late. Intel knew that x86-64 would force Itanium into a small niche at the upper end, and would send 10+ years of R&D down the drain. Now, even HP is getting over its sunken Itanic - choosing to sell Opteron machines in order to remain conpetitive.

Does AMD have anything to compete with Centrino

[MountainMan101]

They're chasing big boys market at the moment with 64-bit, but do they have something for the laptop market to match Centrino.

Re:Does AMD have anything to compete with Centrino

[phusnikn]

Humm I have an Athlon 64 mobility in my emachine and it rocks... I get about 3 hours of juice 2 1/2 w/wifi and this thing flys.

Re:Does AMD have anything to compete with Centrino

[Brian Stretch]

Centrino: marketing-speak for a Pentium-M CPU (Pentium 3 on steroids), Intel chipset, obsolete Intel 802.11b WiFi chip, and $300 million ad campaign.

AMD: Mobile Athlon 64, variety of chipset vendors, variety of 802.11g chip vendors, no Centrino marketing tax. Thus you can buy eMachines Athlon 64 3000+ widescreen notebooks with high-end video chips from Best Buy for $1300 after the usual rebates. If you're reading Slashdot you'll get great battery life; if you're playing UT2004 you won't, but you'll get better performence than the Pentium-M can deliver.

You can't compete if you're bleeding

[ObviousGuy]

I'm glad to hear this kind of recovery by AMD. Not only for the employees of AMD who won't have their lives disrupted by layoffs, but also for the stockholders who can reap the benefits of a company that is now making money.

What's more, it forces Intel to compete against a competitor that can actually put extra top line money towards research and development. Everyone wins when companies can compete.

Go, Go AMD

[JamesP]

Way to go AMD. Intel is eating dust on this one...

The problem is, Intel went from an Engineering company to a marketing company. Let's just hope it doesnt became a lawsuit comapny...

Re:Go, Go AMD

[AndroidCat]

AMD won't have won until Intel starts rating its processors in "equivalent Athlon64 performance". ;)

Re:Go, Go AMD

[Zak3056]

AMD won't have won until Intel starts rating its processors in "equivalent Athlon64 performance". ;)

I'm assuming you're referring to AMD's "Performance Rating." If you are, you might be interested to know that AMD compares their CPUs to a 1Ghz Duron, and NOT any sort of intel chip.

PR3200+ would be 3.2x faster than a 1Ghz Duron.

32, 64,...

[Anonymous Coward]

What I want to know is, where are the 128bit CPUs?

Re:32, 64,...

[pesc]

It really isn't a FULL 64 bit implementation. In it's current incarnation it supports 40 bit physical and 48 bit virtual address spaces, as I recall. Even the Itanium has only a 44 bit address bus.

Please don't propagate this kind of FUD.

The ISA (Instruction Set Architecture) is fully 64 bits. The pointers really are 64 bits wide. The programs you compile now will be able to fully use a 64-bit wide virtual and physical (will we ever see one?) memory architecture.

This is similar to nearly all previous 64 bit architectures such as Alpha and Sparc (and maybe Power and HP-PA and MIPS?). Most of the actual machines used don't really have 64 bit physical adresses.

You have to distinguish between a ISA and a physical implementation of it. Most motherboards can't host more than a couple of GB memory anyway. But the ISA of the processor is still a true 64 bit architecture.

Re:32, 64,...

[MasTRE]

> What I want to know is, where are the 128bit CPUs?

Which will be able to address..... [zoom in on Dr. Evil's face] 1 MILLION MEGABYTES!!!

Re:32, 64,...

[Hoser McMoose]

The Playstation2 does NOT use a 128-bit CPU!!!

The PS2 has a 32-bit CPU core with 128-bit vector units. The Pentium3 also uses a 32-bit CPU core with 128-bit vector units (SSE), as does the Apple/Motorola G4 chip (with Altivec). There has never been a 128-bit CPU used in ANY gaming console, and I'm only aware of 1 64-bit CPU ever used (Nintendo64).

Of course, the reason for this is that going to more bits makes a CPU SLOWER! All else being equal, a 64-bit is 5-10% slower than a 32-bit CPU, and a 128-bit CPU is 10-20% slower than a 64-bit one. Since games don't need to address more than 4GB of memory, it's totally pointless to use a 64-bit CPU in a gaming console. The only other thing that a 64-bit CPU buys you is an integer range of more than 4 billion, and that's RARELY used outside of cryptopgraphy (how often do you do cryptography on your gaming console?).

Of course, all else usually isn't equal (eg AMD64 adds 8 more general purpose registers and cleans up some cruft when compared to IA32). Also PCs often do need to address more than 4GB of memory (virtual + physical).

PCs do not, however, need to address more than 10^19 bytes of memory, and they definitely don't need more than 10^19 integer range for much of anything, so 128-bit CPUs get you absolutely NO positives but you still would have to deal with the 10-20% performance loss.

4gigs of ram

[phreak03]

you can adress more than 4 gigs of ram with a 32bit prossessor You just need a cludge (kinda expensive/slow) but itspossible speaking of lots of ram, anyone seen those Ram Harddrives they had at CES a couple years ago

Re:4gigs of ram

[AndroidCat]

You can address more than 4 gigs of ram with an 8 bit processor. That doesn't mean that the result is pretty or that you should do it.

Thanks for clearing that up

[Anonymous Coward]

Do you also explain jokes to people after you tell them?

Congrats

[Killjoy_NL]

Congratulations to AMD, they've been more innovative in the CPU market than Intel (which is a big feat in my book)
They've also setup a big solid state memcard department (I'm dutch and can't remember the correct name for it right now) which is running along nicely as well.

I hope they can continue keeping up the good work, they deserve it.

Re:Congrats

[Plammox]

Face it. AMD and Intel need eachother. For me it's a sign of health that Intel's roadmaps are affected by AMD's moves and vice versa.

It's easy to imagine how Intel or AMD products would be more inferior due to lack of competition.

Re:Congrats

[Killjoy_NL]

I don't mean to give them a +1000, Innovative.
+500, sure :D

They designed a cpu that's a much better workhorse than that of their competitor (Athlon), then they did it again (AthlonXP) and again (Opteron/AMD64)

That's innovation from where I'm standing.
I know Intel has a lot more different products, but they use those products more as cashcows, trying to milk as much money from it as possible before the market demands something new.

In related news

[Anonymous Coward]

AMD has made deep cuts in their CPU prices, probably pre-emptively.

The metric system can only simplify things so much

[MukiMuki]

In other news, the DVD consortium has finally approved a standard of blue ray disc drive, which allows optical media to break DVD's 18 megabyte barrier, finally allowing for movies times exceeding 10 seconds.

AMD have been better than Intel for some time...

[mu-sly]

...however, it's not about better products, it's about mindshare of the buyers.

I've been building PCs for quite a few years now, and have nearly always used and recommended AMD processors over Intel. In my opinion, AMDs cost less, often outperform their Intel equivalents, and lead the way when it comes to new innovations.

I guess the reason they don't have a bigger market share is because a lot of the OEM companies only sell Intel, and because Joe Public only knows about MHz as a measure of speed.

Re:Yeah ! AMD64 rulez ! Now if the could just...

[Barbarian]

You know the reason for the product differentiation they have chosen is based on what is most likely to fail in fabrication--so for example if some of the HT links are bad, you can turn them off and call it an Athlon 64.

Main Reason for profitability

[MadX]

As long as they have a product that their rivals cannot compete with, they can keep the prices at a premium.
Hence, until such time as Intel release a competitive product, AMD can enjoy high profit margins.
This will change once Intel do release their competitive product though.

BTW: As was said in the article, the other arm of AMD's fabrication was also responsible for their profits ie: flash memory for cellphones. It's only because they have a majority stake in the joint venture with fujitsu, that they are able to declare the income as part of their overall turnover.

Irony

[Killjoy_NL]

"and Intel releasing chips mimicking the 32/64-bit behaviour of the Opteron"

Does anybody else see the irony in this ?

Re:Why 64-bit is better

[pesc]

the real reason that 64-bit is better is that software should be able to move data around more quickly, typically twice as fast as 32-bit given a well-designed data bus external to the chip.

No.

You can move data around fast if you have a good memory architecture. A wide data bus to external memory. And a bus clocked at high speed. And larger caches.

You can have all of that with both 32-bit and 64-bit processors. The 64-bittness doesn't help here. If everything else is equal (in the memory architecture), I would expect the 64-bit processor to lose slightly since it has wider pointers. That puts more pressure on the caches and uses more memory bandwidth.

64 bit processors are good because they can easily adress more than 4GB virtual memory.

I don't think intel has to worry.

[SmallFurryCreature]

I am currently installing gentoo on a dual opteron server as well as a 1u celeron machine that will contain backups. No suprise that there is a performance difference but the gap is gigantic. Same with better specced dual P3 and P4 machines. Even the dual Xeon P4 is left behind easily.

HOWEVER, the dual opteron contains an intel raid and soon an intel network card. And I must say that installing the pentiums in the past was an awfull lot easier.

Price/performance opteron is currently the clear winner, its giganctic cache and better memory structure heads above the same price Xeons. As far as support and quality of the hardware goes. Intel all the way. Sadly for intel the bubble has burst and web companies cannot afford the Itanium. So Opteron it is.

But AMD has been on top before and they always managed to screw up. Intel screws up to but somehow manages to keep making money during the down times. AMD is not so lucky.

Re:AMD HEAT PROBLEM

[dave-tx]

The K6-II+ and K6-III+ are fantastic processors for low-heat applications, because they were intended to be used in notebooks. I ran my email/web/firewall on a K6-III+, overclocked to 550MHz, for years. It was cool, quiet, and so freakin' stable that I would reboot it every couple of months just for kicks.

On the other hand, I had a dual Athlon-MP machine that was like an oven. Really nice computer, but it had to go, because it made my computer room too hot.

I, too, am looking forward to an Opteron-based system in the future. As a former AMD employee, they'll always have my financial support as long as they continue to produce innovative products.

The reason you don't want socket 940

[ColourlessGreenIdeas]

Socket 940 arranges the pins so that it's easy to lay out multiprocessor systems with a 6 layer motherboard (expensive, but you'll want it in a server anyway for reliability reasons). Sockey 939 (real soon now) will work with 4-layer motherboards and so will result in cheaper systems. Both the Athalon 64 and Athalon FX will soon be socket 939, differentiated by the ammount of cache. Opteron will remain as it is, as otherwise your 4 and 8 way boxes won't work. Given that Opteron 8xx is absurdly cheap compared to any other 64 bit 8-way server, I can't see why AMD would want to lower prices.

Re:AMD have been better than Intel for some time..

[dave420-2]

Using terms like "versatile" and "nimble" to describe a CPU makes me slightly wary of the rest of your point ;) What's next, "majestic" RAM? "enigmatic" GPUs? :-P

Re:AMD have been better than Intel for some time..

[Jarnis]

Overall system/platform stability also matters a lot.

In my experience (repairing computers at a 'white box' shop), AMD has still way more 'oddball' problems with it's chipsets and motherboards.

If you build an Intel box, generally it Just Works. If you build an AMD AthlonXP box, it generally probably works, if you are lucky and you are using just the right brand of memory.

Part of the problem is the HAREBRAINED idea of AMD; 'we are not a chipset company'. They gave keys to their kingdom to VIA, and VIA promptly keeps churning out crap. Only the latest chipsets (KT400A etc) are in my opinion any good, and even there you can find big differences with the quality of the implementation between mobo makers.

Granted - motherboard and chipset maturity seems MUCH better with Athlon64 and Opteron, but I've seen too few systems so far to be sure if the status quo is maintaned when Athlon64 goes mainstream and motherboards get cheaper.

But in any case - if I'd have to build a new high-end gaming rig today, I'd still choose Intel, even with the penalty of higher price. I agree that _right now_ is a stupid time to do so, as AMD is rapidly moving to 940pin, while Intel is going to the new 775(?) pin thingy. So basically everything out there today will be obsolete within 6 months. Of course this doesn't really differ from the norm in reality, but at least you can *hope* that if you go for the first 940pin Athlon64 board, it might be upgradeable with just a CPU swap down the road. No such luck for 745 pin mobos.

I really hope Athlon64 motherboard stability and quality is better in the long run than with AthlonXP.

AMD Athlon Processor Build & Installation Guid

[Futurepower(R)]

Haven't seen any problem with AMD processors. It's necessary to follow the Cooling Guidelines, of course.

Make sure you have a good power supply. We use KingWin 350 Watt supplies that have two fans. (Ignore the language, "Extreme Series". That's there just to appeal to gamers, who expect every product to include some reference to violence or games. There is nothing extreme about them, and they are reasonably priced.)

Note that power supply manufacturers sell power supplies that have 100 Watts more rated power for sometimes close to twice the price. That's to take advantage of the "more is better" people.

'sic' Latin for 'thus' - indicates error in quote

[blorg]

Dictionary.com. It is used when quoting, to indicate that the transcriber has faithfully reproduced a possible mistake in the source. This is because it is considered bad form to modify a quote, at least without indication through square parantheses, which are generally used for explanatory additions due to loss of context, not corrections. And if 'sic' wasn't put in, it would likely look like an error on the transcriber's part. It's certainly not a /. thing ;-)

Re:AMD have been better than Intel for some time..

[grondu]

Yeah, that Prescott is a cool running chip.

Re:AMD have been better than Intel for some time..

[Glock27]

They don't have a bigger marketshare because even though the chips perform well, their construction hasn't (in the past) been up to that of Intel. Take, for example, the cooling required for AMD chips. Compare it to that of their pentium equivalents. When said cooling falls off (or stops working) - the pentiums don't burst into flames. That's the difference - higher manufacturing quality.

Modded +5 Insightful? Now that shows the weakness of the Slashdot moderation system...

Athlon, Athlon 64 and Opteron all have thermal protection, just like the P4s...and have had it for some time.

Further, current P4s dissipate more power than the AMD solutions, due to high clockspeeds that don't equate to better performance except for a slight edge in multimedia codec performance.

In short, at this point AMD is flat out better - and a much better deal to boot. You can pick up an Athlon 64 3000+ for about $210...that's a steal!

Re:this isn't exactly correct....

[Slugworth01]

Neither statement is exactly correct. The economics of semiconductor manufacturing are pretty complex. You typically have a certain set of variables to work with. You can invest in line yield, process quality, new processes, automation, capacity, marketing, new device design, retooling, and any number of other areas. You can build cheaper, less complex fabs, build in locations with cheaper labor rates or lower startup costs. You can outsource some or all your manufacturing to a foundry who can make your designs for a contracted price.

AMD has an approach that says they will "build smarter" than their competition. Their flagship fabs (Fab 30 in Dresden, for example) are highly automated with very tight process control, ensuring the right work gets done at the right time. The focus is on equipment utilization; reduction of tool idle time. Further, they focus on minimizing the number of non-product wafers in the line, which take tool time but don't directly produce any chips that can be sold. The management of all this is done through software.

They also have to focus on fab uptime ... since they don't necessarily have the back up manufacturing capability to allow them to recover if their fab is down. For example, AMD makes about two-thirds their revenue from processor sales according to a recent 10-Q filing. Most recent quarter for which there is data (for the period ending 12/28/2003) shows $1,205M in quarterly revenue. You can estimate around $800M in revenue from their processor lines. Fab 30 make nearly all their processors. If Fab 30 were to go down for one hour, that's one hour in the 730 hours in a quarter that they can't make chips. If they have demand that is greater than or equal to capacity, and they're running at full capacity, they would loose roughly $1M due to potential finished goods that could not be made. A cost of $1M per hour of fab down time is pretty typical in the market where AMD competes and for fabs that compare to Fab 30.

A single tool going down is a problem. The entire fab going down is a huge problem. Things that can bring an entire fab down include utilities (electricity, water, gasses, etc.) contamination of facility-wide services like vacuum line, DI water, and various gasses, labor strikes, natural disasters, fires, and plant-wide software.

When you rely on software to manage your manufacturing to the degree that AMD and other high-end semiconductor manufacturers do, you tend to pay a lot of attention to the software.