IBM Processor Yields

[szark]

Never mentioned Intel.... 🙄 ( <---- ha 😛 )

So, you mean they should switch to AMD64 (x86-64) then... 😛

Of course, AMD is relying on IBM's Fishkill engineers to help them develop their 65nm process, so we'd just end up back in the same place! 🙂

 

[Mord]

That is accurate, however they currently have only been able to produce 3 chips running at 1ghz... apple is in the process of welding these 3 together to make a prototype 3ghz G5 for WWDC.

Wha?

you make no sense you cannot weld cpu's together to acheive higher clockrates otherwise the g5 would be advertised at 4GHz not dual 2GHz somehow i get the fealing you are making this up

 

[mdriftmeyer]

This is the same sort of s*** storm Motorola started rolling out when it talked out its ass about meeting commitments with technologies it had yet to perfect to satisfactory levels.

The main reason AMD is making headway on Intel is it doesn't claim delivery before it can support such claims.

This is also the reason historically Apple always gets s*** press from Wall Street-another product line that looks great but scarcely available.

IBM has zero excuses worth justifying this delay.

 

[stoid]

IBM has zero excuses worth justifying this delay.

Other than the laws of physics. The reports I've read state that the heat the chips produce is melting the CPU before the heat energy even escapes the molecular level. Not an easy problem to counteract if you ask me.

 

[rosalindavenue]

Somebody who knows more about this should answer-- is there something about the PPC architecture that makes it inherently more difficult to fabricate than the x86? I mean if IBM can't do it (disregarding motorola), who can? This is really disapointing...

 

[Frohickey]

Somebody who knows more about this should answer-- is there something about the PPC architecture that makes it inherently more difficult to fabricate than the x86? I mean if IBM can't do it (disregarding motorola), who can? This is really disapointing...

No. There is nothing inherent about the PPC architecture that makes it more difficult to fabricate than the x86. But, here is the big caveat.

With IBM or Motorola, there is the PPC architecture, and the specific chips themselves. The market for these chips is not as large as the x86 market, and so, there is limited resources with which to design and produce the chips.

With the x86 market, Intel can afford to put multiple engineering teams to the design and manufacture of the parts, and can afford to have a few of the engineering teams fail. The sheer number of resources they can put is quite large. You just have to look at the fact that there are at least two different products, a desktop x86, and a laptop x86. There is no such bifurcation of design in the PPC architecture, at least for the leading edge processors for personal computers. (Server Power4 and Power5 chips are a totally different beast altogether.)

 

[dho]

Actually, this isn't as crazy as it seems.

IIRC, AMD's K7 (Athlon) and K8 (Athrlon-64) architectures are actually a proprietary RISC architecture with an x86 emulation layer over the top. It may well be possible for them to make these with a PPC emulation layer instead.

I have no idea, however, how well these would compete against a real PPC. I would be really surprised if they can match a G5, but they might be able to match a G4's performance.

Of course, designing any such a CPU would be expensive and time-consuming, and it would be hard to make the business case for AMD to even consider such a project.

I believe TransMeta's Crusoe chip is similar - and its CPU emulation is loaded as software, so you don't need to spin new hardware. But it has never been known for high performance. Crusoe's big claim to fame is low power consumption. But it's another possibility, at least for low-end laptops.

Im afraid it MAY not be that simple. Not only do PowerPC's have different general archetectures, they have opposite endians. PowerPC is big endian and X86 is little endian. When I say i X86 i include those AMD chips with the emulation layer. In order for them to properly emulate they would need to have the same endian.

From my understanding the best you could do with one of those amd chips would be to make an ALMOST pcc with a different endian. We would end up in the same place with annother VirtualPC type app.

If I am mistaken please ocorrect me, this is just how I have come to understand the differences in chip types.

 

[ingenious]

Wha?

you make no sense you cannot weld cpu's together to acheive higher clockrates otherwise the g5 would be advertised at 4GHz not dual 2GHz somehow i get the fealing you are making this up

duh he was kidding 😛

 

[stoid]

Im afraid it MAY not be that simple. Not only do PowerPC's have different general archetectures, they have opposite endians. PowerPC is big endian and X86 is little endian. When I say i X86 i include those AMD chips with the emulation layer. In order for them to properly emulate they would need to have the same endian.

From my understanding the best you could do with one of those amd chips would be to make an ALMOST pcc with a different endian. We would end up in the same place with annother VirtualPC type app.

If I am mistaken please ocorrect me, this is just how I have come to understand the differences in chip types.

Can you explain the little/big endian difference in laymen's terms?

 

[JFreak]

The market for these (PPC) chips is not as large as the x86 market, and so, there is limited resources with which to design and produce the chips. With the x86 market, Intel can afford to put multiple engineering teams to the design and manufacture of the parts, and can afford to have a few of the engineering teams fail.

this is not so simple.

intel has to have many teams to KEEP their market share, whereas ibm has to have many teams to GAIN the market share they have set as their target for long term. ibm would not have even begun the G5 fabbing if they would have been happy with 3% market share in the beginning, so it is very clear that they have started a beginning of a revolution here. they try to make itself the king of the jungle again, that's for sure, and for that reason they must have AT LEAST as strong r&d department as the competition has.

ibm has the powerX series which i believe dissipates even more heat than intel's chips, so it's not a matter of power consumption. ibm just tends to tolerate errors worse than intel, so what i'm saying is that the two are not even comparable in yield percentages - intel has a BIG error tolerance and sells most of the chips in the wafer, but ibm wants to do things right. it has a strong reputation (regarding manufacturing quality) which it cannot afford to lose. for ibm it's better to sell one chip out of the wafer keeping its fame, rather than selling ten and losing it. ibm makes money in the long term.

go ibm! please make intel cry.

 

[ddtlm]

Modern x86 chips have purpose-built RISC cores, not general purpose ones. They are designed from the ground up to hide behind x86 and would not likely apapt well to hiding behind PPC.

stoid:

Endianness is sorta like reading left-to-right or right-to-left.

 

[ddtlm]

JFreak:

ibm would not have even begun the G5 fabbing if they would have been happy with 3% market share in the beginning, so it is very clear that they have started a beginning of a revolution here

I'm not gona get into a lengthy arguement about what amounts to religion, but just so someone stands up to it, I have to that this idea of yours is just a fantasy. All the 970 is doing is spreading out IBM's R&D cost so that they can keep making ever-more-advanced Power series chips to compete with Intel's money-sponge Itanium. IBM is never going to get a large market share, and they are not putting a whole heck of a lot of R&D into desktop chips. Desktop chips are just "free" money for them.

ibm just tends to tolerate errors worse than intel, so what i'm saying is that the two are not even comparable in yield percentages - intel has a BIG error tolerance and sells most of the chips in the wafer, but ibm wants to do things right

Intel is not churning out chips with errors in them, if they were then the computers based on them wouldn't be stable. I'm sure you've heard of Linux and BSD running on x86, thats some dang stable stuff.

 

[JFreak]

Intel is not churning out chips with errors in them, if they were then the computers based on them wouldn't be stable. I'm sure you've heard of Linux and BSD running on x86, thats some dang stable stuff.

sure they do, not one cpu is 100% error free - not even the big iron server hardware. it's all about specification, you know, and intel seems to drive its threshold lower than the others - hell, you can overclock amd some 30% to have it still stable (keeping the error occurency rate low enough), but try to overclock intels, they just are sold already on the limit.

and ibm is famous for having higher thresholds than other manufacturers.

 

[nmk]

I'm not gona get into a lengthy arguement about what amounts to religion, but just so someone stands up to it, I have to that this idea of yours is just a fantasy. All the 970 is doing is spreading out IBM's R&D cost so that they can keep making ever-more-advanced Power series chips to compete with Intel's money-sponge Itanium. IBM is never going to get a large market share, and they are not putting a whole heck of a lot of R&D into desktop chips. Desktop chips are just "free" money for them.

I believe that IBM will be supplying Microsoft, Sony, and Nintendo chips for their next gen game consoles. At least in the case of MS, they managed to drive Intel out of contention. This is a pretty big market. So no, I don't think that IBM is looking at desktop processors primarily as extra money. Their marketing for the PPC 970 (and its future iterations) would indicate that it's a market they're going after quite seriously. Or do you believe that the entire game cosole market is irrelevant.

 

[Bigheadache]

I believe that IBM will be supplying Microsoft, Sony, and Nintendo chips for their next gen game consoles. At least in the case of MS, they managed to drive Intel out of contention. This is a pretty big market. So no, I don't think that IBM is looking at desktop processors primarily as extra money. Their marketing for the PPC 970 (and its future iterations) would indicate that it's a market they're going after quite seriously. Or do you believe that the entire game cosole market is irrelevant.

Have you seen how much consoles cost?? The console market is extremely price sensitive for componentry. There wouldn't be much margin, compared to selling in the big iron part of the market. At the end of the day, getting into consoles and selling G5s is about amortising the cost of developing the POWER series which is the main game for IBM.

 

[sun-ice]

Wha?

you make no sense you cannot weld cpu's together to acheive higher clockrates otherwise the g5 would be advertised at 4GHz not dual 2GHz somehow i get the fealing you are making this up

what is the base of dualcore ?
I think i was to "put 2 CPU" with a common front bus ... 😕

And rumors about the XBox II deal with triple core PPC CPU at 3.5GHz (so 3 cpu at 1.166GHz) ? 😕

I don't know is someone knows ? 🙂

 

[nmk]

Have you seen how much consoles cost?? The console market is extremely price sensitive for componentry. There wouldn't be much margin, compared to selling in the big iron part of the market. At the end of the day, getting into consoles and selling G5s is about amortising the cost of developing the POWER series which is the main game for IBM.

Granted, the console market is very cost sensitive. I hope you realize that consoles are sold at a loss by the major manufacturers. They make their money of game licenses. Therefore they are certainly paying IBM more than a fraction of the $250 that they will sell the console on the market for. Also, IBMs penetration into this market required a fair bit of marketing. It just goes to show that they are serious about finding customers for the product.

Since neither you nor I are involved in the strategic planning at IBM, I don't think it is wize to make assuptions about their developemnt of the G5. Their actions would indicate that it is a product they are taking seriously.

 

[johnnowak]

Wha?

you make no sense you cannot weld cpu's together to acheive higher clockrates otherwise the g5 would be advertised at 4GHz not dual 2GHz somehow i get the fealing you are making this up

I get the *fealing he was joking.

 

[rosalindavenue]

Have you seen how much consoles cost?? The console market is extremely price sensitive for componentry. There wouldn't be much margin, compared to selling in the big iron part of the market. At the end of the day, getting into consoles and selling G5s is about amortising the cost of developing the POWER series which is the main game for IBM.

I would have to believe that the manufacturers (Sony MS Nintendo) are the ones who take the hits on the components, not the component manufacturers themselves. IBM, as the processor manufacturer, is not going to make money from the software/games ; other than large volume, they have no incentives to decrease their margins. If they weren't making money selling processors to the console manufacturers, they wouldn't bother making them. The losses (if any) won't be taken by the component manufacturers.

 

[Bigheadache]

I would have to believe that the manufacturers (Sony MS Nintendo) are the ones who take the hits on the components, not the component manufacturers themselves. IBM, as the processor manufacturer, is not going to make money from the software/games ; other than large volume, they have no incentives to decrease their margins. If they weren't making money selling processors to the console manufacturers, they wouldn't bother making them. The losses (if any) won't be taken by the component manufacturers.

The console makers do take a hit, but that doesn't mean they don't squeeze their suppliers that extra bit. Refer Microsoft vs Nvidia 2002-2003 Xbox contract dispute.

 

[Chryx]

sure they do, not one cpu is 100% error free - not even the big iron server hardware. it's all about specification, you know, and intel seems to drive its threshold lower than the others - hell, you can overclock amd some 30% to have it still stable (keeping the error occurency rate low enough), but try to overclock intels, they just are sold already on the limit.

Where the hell are you getting that ?

Intel chips overclock very well. (the current Prescott P4 aside, and that's got design issues)

Historically, at least over the last two years or so, Intel chips have overclocked a LOT better than AMDs....

I'm no fan of the architectural direction Intel have taken of late, but what you're saying directly contradicts the reality of the situation 🙂

 

[Telomar]

the cost of developing the POWER series which is the main game for IBM.

The main game for IBM are service contracts. That's why they are pretty platform independent. Doesn't mean they aren't interested in succeeding in other areas and having PPC grow though.

 

[gop007]

The issues regarding yields is what is of concern to Apple. Sure they can make all the 1.8 and 2.0 chips, but right now the PowerMacs are not selling that well. Apple is will proabably place many of those chips into the iMacs, but they need much faster chips for the PoweMacs. If history is a guide, the next upgrade will see 1.8 and 2.0 chips into consumer models, and a very limited supply of 3.0 chips or lower for the PoweMacs. Sure they will announce the 3.0 sometime this summer, but once again they will fail to ship them in adequate supplies until early fall.

 

[ssamani]

Can you explain the little/big endian difference in laymen's terms?

There was a brief explanation as reading left to right or vice versa.

Specifically, under ASCII, characters are stored as an 8 bit number. Therefore in a 32 bit system you can store 4 ASCII characters in each memory value (2 in 16 and 8 in 64). I can't remember which way is which but one out of Intel / Motorola with the x86 and the 68xxx chose to store "Nice" as "Nice" and the other as "eciN". The reason for choosing the latter method is that it makes some string handling easier for a CPU. Emulators have to handle this difference when running on different architectures.

The reference for why this is called big endian and little endian is from Swift's Gulliver's Travels. The Lilliputions went to war over which end of a boiled egg should be cracked first - big end or little end. The choice for byte storage for 16+ bit CPU's, like many things in the technical world, generated a fierce argument thought not quite as bad as the Mac vs PC wars of the late 20th and early 21st centuries.

 

[trippen_raven]

Intellegent

I've heard before, actually it was a interview I read sometime ago last year that Intel has tried pretty hard to crack into Apple boxes. So it's really up to Apple where they want to go. If powermac shipments really drop, heaven forbid, but if they do, because of the lack of performance chips like the G5. They have alternitive options and it seems Intel is one of them. I've even heard that Intel CEO and Jobs are good friends.

I just hate seeing good software and OS X being wasted, because the lack of powerful processors. IBM has a lot on the line here, thier customers, reputation, and lastly investment(notice $ is last,,,,,,,,,customers should come first------------hear that ADOBE). So I'm confident big blue will pull through and rise to the occasion as they are doing so well. 🙂