Intel at 65-nm in 2005?

[MisterMe]

Re: AMD Athlon-64/Opteron vs Itanium

Originally posted by dguisinger
Everyone likes to bash the Itanium chips but it actually is started to take off, and a great design. You don't see Apple carrying extra luggage from the 6502 or 680x0 processors in the PPC.

If you have ever programmed in x86 assembly you would know that it is complete hell because the x86 instruction set has been extended many, many times.

It has gone from the 4004 (4-bit), to the 8008 (8-bit), to the 8080, 8086/8088, 80186 (flop), 80268 (added memory production), 80386sx/dx (added 32-bits), 80486sx/sx2/dx/dx2/dx4 (added floating point internally), Pentium, Pentium MMX (added basic vector processing), Pentium Pro (Added support for more than 32-bit addresses), Pentium II, Pentium III (Added SSE), and Pentium 4 (Added SSE2, CPUID)....and lets not forget that in either the Pentium or Pentium Pro they added SMP support. And how about 3D Now, a failed instruction set from AMD.

The AMD is a horrific design. It once again adds extends the processor, adding more complexity to decoding and execution, and to assembly language design. No new registerers. The x86 has always suffered from a lack of registers. Modes have to be switched, which adds overhead.

The Itanium 2 fixes many problems from the original Itanium. It is even further reduced than RISC...its designed from the ground up as a modern processor....the compiler gets to define instructions in blocks, and specify how it works across multiple pipelines to correctly fill the processor to its max...instead of extra decoding at execution time. Its a simpler design, but leads to more complicated compilers, a problem that plagued the first processor.

The only part that the Itanium does not deal well with is legacy x86 code. But out of complete seriousness, if you are running 64-bit apps, there aren't many 32-bit apps you should be running at the sametime. The Itanium does very will in benchmarks.

Dell refused to work with the first Itanium, but are now seling a good number of the new Itanium 2s. While competition is good, the AMD 64 bit processors area really bad idea, from the ground up, they are based on a very outdated instruction set, one that has been extended around 10 times. It is time for the x86 architecture to be retired.

Your check is in the mail.

Mr. Michael Edwards, Asst. V. P.
Public Relations
Intel Corporation

 

[dguisinger]

Re: Re: AMD Athlon-64/Opteron vs Itanium

Originally posted by MisterMe
Your check is in the mail.

Mr. Michael Edwards, Asst. V. P.
Public Relations
Intel Corporation

Rofl. I wish.

 

[i_am_a_cow]

hey i have an idea!

everyone should do this:

I'm sending an email to IBM saying

"IBM IS SOOOOOOOOOOO TIGHT. SOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO TIGHT. LIKE THE TIGHTEST COMPANY EVER BESIDES APPLE."


or something along the lines of that

 

[singletrack]

Don't write off Moto yet

Not that I'm a Moto apologist but Motorola claim they'll have 65nm before anyone else according to The Register. http://www.theregister.co.uk/content/3/34171.html

They've also got Dual-core G4+ coming up which could be a kick ass laptop chip even if IBM can't get the 970 suitable for laptop use.

http://www.theregister.co.uk/content/archive/31026.html

If they do manage that, it's quite a change on past performance.

 

[Sol]

AMD with IBM

Originally posted by SiliconAddict
Where's AMD in all of this?!?!

I think I read that AMD is with IBM on this. When IBM gets 65nm chips into production AMD will fabricate their chips there and so will nVidia.

Who knows if Intel will really produce one in time. They will have to do more than shrink x86 compatibility if they want the faster computers. IBM and Apple changed enough of the dual G5 PowerMac to make it a solid upgrade from anything that came out before it. Can Intel and Microsoft do the same thing for PC users?

Turning them towards Apple is not a sign that they can do that. By the time 65nm CPUs come out there will be a mass migration of users going from PC to Apple.

 

[peejay]

techie

don't think SRAM is quite the same as a working processor. However I don't know the industry well enough to know what it takes to get from where they are right now to having a working processor, butÉ they seem to think more than a year. Again my impression is that this is just PR trying to position themselves as technology leaders before the Christmas rush.

If any techies out there want to explain the ins and outs of all this it would be interesting.


Memory and logic are usually produced in different fabs, but the technology is almost all the same to produce it. The most critical stage is the photo process, which is common to both. They'll be using ArF (Argon-Flouride excimer laser) scanners that should print a 35-40nm line in tests, but the realities of production call for some process lattitude. The lens and illumination of the scanner can be fine tuned to favor dense patterns (memory) or isolated lines (logic). Processors are harder to do because they have a few more layers. Both Intel and IBM have been running 65nm in development for some time now.

Motorola is still scratching marks on cave walls compared to either of these guys. Apple is good to be rid of them.

 

[iMeowbot]

Originally posted by SiliconAddict
Where's AMD in all of this?!?!

Oh, them. The intel press release about 65 nm is clearly a bit of damage control to show it isn't falling very far behind the IBM-AMD alliance.

AMD announced last week that it is adding a new fab to its Dresden facility to build wafers at the 65 and later 45 nm processes. The equipment will be pretty much a twin of the 65 nm fab in IBM's E. Fishkill plant. AMD's schedule is about the same as the one intel hinted at, with parts appearing in 2005 and volume production in 2006. IBM should begin a bit earlier.

 

[dho]

Re: Don't write off Moto yet

Originally posted by singletrack
Not that I'm a Moto apologist but Motorola claim they'll have 65nm before anyone else according to The Register. http://www.theregister.co.uk/content/3/34171.html

They've also got Dual-core G4+ coming up which could be a kick ass laptop chip even if IBM can't get the 970 suitable for laptop use.

http://www.theregister.co.uk/content/archive/31026.html

If they do manage that, it's quite a change on past performance.

This I will have to see to believe

edit: good eye though

 

[scem0]

Originally posted by Trowaman
Competition good

Competition results in lower prices So let's kick some butt guys!

quite right...

I'm really hoping IBM will lower prices and invest more time and effort into winning this chip war. God knows it will benifit them.

But in all truth, this is great news for me.

And even if Intel wins this chip war, all is good. Apple has acknowledged that they can run OS X on x86 and they will if Intel offers anything significantly better than IBM/AMD/moto/etc.

I have nothing against Intel. You can't blame a company for making good products.

scem0

 

[dracoleb]

Re: Don't write off Moto yet

Originally posted by singletrack
Not that I'm a Moto apologist but Motorola claim they'll have 65nm before anyone else according to The Register. http://www.theregister.co.uk/content/3/34171.html

They've also got Dual-core G4+ coming up which could be a kick ass laptop chip even if IBM can't get the 970 suitable for laptop use.

http://www.theregister.co.uk/content/archive/31026.html

If they do manage that, it's quite a change on past performance.

Could this be where the rumor of a dual processor G4 Powerbook came from? Perhaps it a dual core G4 if the G5 didn't come in time rather than a dual processor. It would seem more feasible as-well

 

[tortoise]

65nm won't save them

A new process technology won't save them unless the figure out how to build good cores again. By the time Intel has 65nm, most everyone else will have it or will be getting it shortly. And the other folks have clean and scalable core designs.

Intel is just trying to stay in the news by pre-announcing a process technology way ahead of its time, kind of like how Microsoft pre-announces their products years in advance in an attempt to gain some type of mindshare in the absence of a real product.

 

[tortoise]

Originally posted by ITR 81
AMD won't have 90nm until around the middle of next yr compared to IBM which will have them probably by the end of Jan. and at the latest by Feb.

IBM and AMD have a nice fab technology relationship which means that AMD will probably get process technologies around the same time IBM does, though it may take them a bit longer to get them running in their fabs. I don't know the specifics, but I do know they have a technology and fab sharing agreement. This is a pretty good deal all around; IBM uses some of AMDs chip designs and fabs, while AMD gets access to IBMs process technologies. This really plays to the strengths of both companies.

Yes, the G5 contains pieces of AMD technology in it. But at least it isn't Intel.

 

[dongmin]

Re: Don't write off Moto yet

Originally posted by singletrack
Not that I'm a Moto apologist but Motorola claim they'll have 65nm before anyone else according to The Register. http://www.theregister.co.uk/content/3/34171.html

They've also got Dual-core G4+ coming up which could be a kick ass laptop chip even if IBM can't get the 970 suitable for laptop use.

http://www.theregister.co.uk/content/archive/31026.html

If they do manage that, it's quite a change on past performance.

Can't believe people are referencing old Moto roadmaps, especially one cooked up by the Register. (Long-time readers will remember that the Register was predicting G5s two Januaries ago.) Steve himself has said that Apple is working on fitting a G5 in its laptops; as soon as Apple figures out how, it's dropping Moto completely. It's as simple as that.

Moto is desperately trying to spin/sell-off its semiconductor business. It's basically a sinking ship for them. On top of it all, they're focussing more and more on the embedded market, i.e. low-power chips. Dual-cores makes no sense. The article you linked is full of sketchy claims like "The G4+ that appears on the January roadmap will be fabbed at 0.1 micron, the sales sheet say." ???

 

[tortoise]

Re: AMD Athlon-64/Opteron vs Itanium

Originally posted by dguisinger
The AMD is a horrific design. It once again adds extends the processor, adding more complexity to decoding and execution, and to assembly language design. No new registerers.

Uhhh, say what? You are definitely mistaken. When running in AMD64 native mode, it uses a different register model from ia32 style x86, and has a hell of a lot more registers. The "x86-64" moniker is slick marketing, but it is a clear break from the x86 ISA lineage. It can get away with it because it fakes the ia32 ISA better than the real thing. AMD64 has the compiler efficiency of RISC-like register models, with the core optimization of CISC-like opcodes, which is an appropriate design choice for modern CPU cores.

In AMD64 native mode, you have twice the number of registers as IA32 x86, and they are all general purpose (unlike IA32). The Opterons may be IA32 compatible, but the native AMD64 ISA is a very efficient and fairly elegant design that is optimized for modern CPU cores. In native mode, all the really annoying baggage of IA32 x86 is eliminated.

 

[MacsRgr8]

Re: Re: Don't write off Moto yet

Originally posted by dongmin
Steve himself has said that Apple is working on fitting a G5 in its laptops; as soon as Apple figures out how, it's dropping Moto completely. It's as simple as that.

G5 in PowerBooks.
Not sure when the G5 will come into the iBooks (or eMacs). I think the G4 will stay in these products throughout 2004.... <wishing not>.

 

[Analog Kid]

Re: AMD Athlon-64/Opteron vs Itanium

Originally posted by dguisinger
Everyone likes to bash the Itanium chips but it actually is started to take off, and a great design.

Yeah, really taking off... I hear they sold a few thousand over the past year!

Almost a whole wafer!

 

[Analog Kid]

Personally, I see this as a "pay no attention to our troubles at 90nm, 65nm is right around the corner" statement.

Intel better get 65nm fired up on time or they're in real trouble. They can afford to be sloppier than the competition, but not too many times in a row...

Itanic is a money pit. 90nm isn't going well-- and even if they ship what they promise they're still looking at a commodity chip over 100W. If they can't get 65nm together with a well built core and lower leakage then their vast resources are going to be under some strain...

 

[1macker1]

LOL @ Intel being in trouble. They are ahead of everyone else, including IBM.

 

[mrsebastian]

i just want my 20ghz 17" powerbook

 

[stingerman]

Originally posted by Macco
Seeing as IBM is basically the only PPC producer left, and Apple would never switch to Intel, competition isn't really so relevant.

Seeing that the Microsoft, Nintendo and Sony are all now going to use IBM PowerPC Cell processors for their new game systems, I think competition is pretty relevant.

 

[stingerman]

Re: Re: AMD Athlon-64/Opteron vs Itanium

Originally posted by tortoise
Uhhh, say what? You are definitely mistaken. When running in AMD64 native mode, it uses a different register model from ia32 style x86, and has a hell of a lot more registers. The "x86-64" moniker is slick marketing, but it is a clear break from the x86 ISA lineage. It can get away with it because it fakes the ia32 ISA better than the real thing. AMD64 has the compiler efficiency of RISC-like register models, with the core optimization of CISC-like opcodes, which is an appropriate design choice for modern CPU cores.

In AMD64 native mode, you have twice the number of registers as IA32 x86, and they are all general purpose (unlike IA32). The Opterons may be IA32 compatible, but the native AMD64 ISA is a very efficient and fairly elegant design that is optimized for modern CPU cores. In native mode, all the really annoying baggage of IA32 x86 is eliminated.

The native AMD 64 bit mode is nice to have even if it isn't going to be used in Windows for another 3 years if then. But their are Linux builds that can take advantage of it. One never knows, so I would rather have 64-bits than not have it, even if I don't get to use it for a while. But, then again, I would want the latest processor in three years not the one I buy today. So, for most of us the AMD64 Bit mode is more a technology preview. Whereas the G5 is here today running an awesome OS, which will continue to add G5 optimizations.

Eliminate the IA32 baggage and eliminate all those programs that depend on it. If the software don't work, what good is the chip. You can still run DOS and Win16 Apps in Windows, and believe you me that is still a requirement for many companies. Not mine IBM built in special ISA add-on for Apple to migrate their VM to 64-bits and maintain compatibility with 32-bit OSX and even Classic, while using the 64-bit ISA natively. That s what you call a partner.


"AMD64 has the compiler efficiency of RISC-like register models, with the core optimization of CISC-like opcodes,"

Hmmmm. I'm going to need to puke. CISC-like opcodes is badddd not goodddd. It is much harder to optimize CISC op codes since each can do so many different things.

 

[Malic]

Yeah, IBM will be there too

According to New Scientist...

Other chip makers say they are also on track to release 65-nanometre technology soon. "I think we're all marching to the same timeframe," IBM spokesman William O'Leary told the San Francisco Chronicle.

http://www.newscientist.com/news/news.jsp?id=ns99994424

 

[tortoise]

Re: Re: Re: AMD Athlon-64/Opteron vs Itanium

Originally posted by stingerman
IBM built in special ISA add-on for Apple to migrate their VM to 64-bits and maintain compatibility with 32-bit OSX and even Classic, while using the 64-bit ISA natively. That s what you call a partner.

This "feature" is nearly useless, and has been supported by Intel for many years (PAE). Having more than 4Gb of available RAM is worthless for most purposes if you can't address more than 4Gb memory within a process. It is a hack that makes for good marketing but has almost no use in real applications. I run large memory applications that need lots of physical RAM, and I can't use a G5 for those purposes unless I am running Linux on it. Since LP64 compatibility pretty much defines "64-bit OS", I am somewhat annoyed that Apple has been selling a 32-bit OS on 64-bit hardware as a "64-bit" system without any roadmap to ACTUALLY delivering a real 64-bit system.

Originally posted by stingerman
Hmmmm. I'm going to need to puke. CISC-like opcodes is badddd not goodddd. It is much harder to optimize CISC op codes since each can do so many different things.

Actually, this is not true. It is harder to optimize traditional CISC register models at a software level. It is easier to optimize CISC opcodes in the CPU core. AMD64 redid the classic x86 register model (which WAS pretty ugly) so that compilers could optimize software in the way they can for RISC ISAs. They used simple CISC-style opcodes to retain core performance optimization. A bit of a hack for compatibility's sake, but a very smart engineering choice as well. RISC ISAs do have some disadvantages on modern silicon cores (though this was not nearly as true many years ago).

Of course, I would point out that the extremely high core efficiency of Opterons running AMD64, even in the absence of an AMD64 specific optimizing compiler, is a fine testament to their engineering design choices. We are a mixed shop here, but the Opterons (running AMD64 native) are the fastest systems we have for our high-performance codes, including the G5s. I prefer (and use) OSX workstations, but Opterons are generally faster for our purposes clock-for-clock. Which is all I really care about -- I use whatever works best.

 

[Jonathan Amend]

Re: Re: Re: AMD Athlon-64/Opteron vs Itanium

Originally posted by stingerman
The native AMD 64 bit mode is nice to have even if it isn't going to be used in Windows for another 3 years if then. But their are Linux builds that can take advantage of it. One never knows, so I would rather have 64-bits than not have it, even if I don't get to use it for a while. But, then again, I would want the latest processor in three years not the one I buy today. So, for most of us the AMD64 Bit mode is more a technology preview. Whereas the G5 is here today running an awesome OS, which will continue to add G5 optimizations.

Microsoft will release a completely 64-bit OS for the AMD64-line next year (supposedly in H1, betas are already out there). Apple has yet to say anything about such an OS for the G5. I don't know where you got 3 years from, unless you thought Longhorn would be the first AMD64 OS.

 

[Analog Kid]

Originally posted by 1macker1
LOL @ Intel being in trouble. They are ahead of everyone else, including IBM.

Not sure if this was a reply to my post or not, but if it is I suggest you read what I said...

If they continue to screw up and the competition stays strong they, like any other company, will have trouble.

It's not clear to me that they're so far ahead of the pack any longer. It's quite possible that the 90nm G5 will ship in volume at the same time as if not before Intel manages volume production of Prescott.

It's neither easy, nor cheap, to build a mass market machine around a 100W chip. OEMs will deal with it in the short term to continue selling Intel chips, but if it keeps up you can be sure they'll look around for alternatives.

Prescott has to be the last of it's line, and Intel has to offer a workable chip at 65nm-- on time.

I think Intel itself realizes this, which is why they're eager to press release any advances on future technology a year in advance. This wasn't a talk at a technology symposium, this was a PR move.

Yes, Intel has a strong reputation, and impressive engineering history, but that doesn't make them immune to failure. Itanic was what, 2 years late? I had several friends working on that chip, and it was a fiasco. It left a lot of people wondering if Intel was still capable of designing entirely new cores.

Yes, Intel has a lot of resources and experience so they can afford a few goofs. My point was simply that if they don't get it together the inefficiency is going to cause them trouble.