• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

AMD prepares SOI shift in PC processors with initial 0.13-micron process; .13 micron Athlon 4's THIS YEAR!!!

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.


<< &quot;AMD also revealed product plans for the third quarter. The chipmaker will introduce two new mobile Athlon 4 chips, at 1.1GHz and 1.2GHz, during the quarter, Sanders said. A 1GHz version of the Duron mobile processor is also on tap for the quarter, as are 1.5GHz and 1.6GHz versions of the desktop Athlon. A 1.73GHz desktop Athlon is set to follow in the fourth quarter. &quot;

I read this here.

They may be referring to the desktop Athlon 4...I'm not sure though... >>


He is talking about the Desktop Palomino.
 
Right now I have $350 put away for a nForce mobo and a nice Palomino. But damn, it better be this year.
 
That is the first article that I have read that adequately explains the SOI relationship between AMD and MOT and IBM. Good find.
 
SCUBA:


<< ok guys lets stop thinking about those old 32bit cpus and start thinking of the new 64bit ones. they will just duble speed
>>




uh, what are you talking about? The move to 64bit CPUs means no such thing at all. In fact, the only main benefit of 64bit CPUs is that they can address >4GB of memory. You don't use 4gb, do you?

There are architectural changes that may mean good things, but on the whole, using 64bit values instead of 32bit ones is useless.
 
Scuba: Hehe, well, I just pulled that out of thin air, but as I was figuring the Tbird certainly won't make it that high, but the Palimino will make it most of the way there(about 2ghz). Another tweaked Athlon core will likely get close if not actually arrive at that speed. So, yes, I think 3ghz is possible.
 
Just to clear some things up...
  • Intel is using copper in the .13um process.
  • Intel's .13um process is more than a die shrink, especially with the P4.
  • AMD is starting volume production at the end of the year. That means it'll be months before those cpu's hit the street. Q2 if all goes exactly as planned. Q3/Q4 if there are any delays.
  • SOI wafers are much more expensive than the wafers currently used. The benefits are arguable, as some say that advances in insulating layer techniques have most of the same properties without the added cost.
  • SOI does have some drawbacks, as pm pointed out in this thread.
&quot;Certainly sheds new light on the &quot;Intel will have it 6 months before anyone else&quot; line I've seen people lauding around here&quot;

Pabster... I, for one, haven't been saying &quot;6 months&quot;. I've been saying 12 months. 😛 (Still saying that, btw. 😉 )
 
Remnant2
4gb of memeory ????

as far as i know a 486 can adress 4 gb of memory coz it has a 32bit adress bus man
what are u talking about
a 64 bit processor will process 64 bit pipeline in a single clock and that means
a 64bit processor =2(32 bit processors)at the same clock

if PM was here he would made things clear
 
intel is hardly dead. They are a pretty diverse company now, and could survive a while if cpu's didn't make them money. Things shall be getting interesting now.
 


<< SOI wafers are much more expensive than the wafers currently used. The benefits are arguable, as some say that advances in insulating layer techniques have most of the same properties without the added cost. >>


Of course new techniques are gonna be expensive, but I'm SURE that IBM has researched the benefits/problems of going with a SOI process or else they wouldn't be embarking on such a venture at this early state.

Audi uses aluminum for the A8's frame and body. It is VERY light-weight and strong compared to steel. Yes, it's expensive. Yes, it's hard to repair if you get into an accident. But it also makes the A8 lighter than its steel counterparts and it gives Audi the experience needed to introduce the process on lesser models (A6, A4, TT, etc.).
 


<< SOI wafers are much more expensive than the wafers currently used. The benefits are arguable, as some say that advances in insulating layer techniques have most of the same properties without the added cost. >>



I'm thinking the price for them wouldnt be quite as high because of the amount they would buy. They might be expensive now because it's not exactly a mass produced thing that everyone uses. But I'm thinking if AMD is going to buy a ton of them they would be making more and price would drop. AMD knows it has to keep its prices low to gain market share on Intel... I doubt they would go SOI if it would make it so expensive that AMD couldnt keep their prices lower then Intel.
 
SCUBA - Being 64-bit means that 64-bit integers are available, along with 64-bit addressing. It does NOT mean 2 32-bit values at the same time.

Infact, unless you're using 64-bit values for integers, moving to 64-bit processesing can be slower. It also leads to program size inflation. I've heard estimates that if all else were equal, beinga 64-bit architecture is ~5% slower than being a 32-bit architecture.

Part of the reason for this is that all pointers are 8bytes, instead of 4 (that's part of the code dilution).
 
In any CPU, there are functional units. Some load and store values. Others do work on integer values, and others work on floating point values. We're talking about integer values for now.

Lets say that we have a 64-bit machine with the ability to process one 64-bit integer value.

General idea:

a 32-bit integer in binary:

00000000000000000000000000000001 (that represents the integer 1).
00000000000000000000000000000010 (represents 2).

a 64-bit integer:

0000000000000000000000000000000000000000000000000000000000000001 (represents integer 1)
0000000000000000000000000000000000000000000000000000000000000010 (represents 2).

This 64-bit processor can do additions, multiplies, divides on integer values at &quot;full speed.&quot; A 32-bit processor is technically capable of doing the same kind of work, but it requires software (I don't know how it works, but it's along the lines of &quot;Big Int&quot; -- someone correct me where I'm wrong here), and it's ugly and slow.

This 64-bit processor has only one functional integer unit, and can't compute two 32-bit values at the same time. If it could, we'd call it a SIMD machine (Single Instruction, Multiple Data) -- which is what iSSE, 3dNow!, iSSE2, and Enhanced 3dNow! are. &quot;Regular&quot; 64-bit processors don't do SIMD on 32-bit values (meaning they don't do work on 2 32-bit values at once, though instructions and the ability to do so can be added).

With 64-bit processors, &quot;Regular&quot; integer values are 64-bits, and so if you want to (1 + 2), you'll do :
0000000000000000000000000000000000000000000000000000000000000001(which is 1) +
0000000000000000000000000000000000000000000000000000000000000010(which is 2)

On a 32-bit machine, it would be
00000000000000000000000000000001 (which is 1) + 00000000000000000000000000000010 (which is 2).

I'm not sure if I'm helping or confusing....
 
<< &quot;AMD also revealed product plans for the third quarter. The chipmaker will introduce two new mobile Athlon 4 chips, at 1.1GHz and 1.2GHz, during the quarter, Sanders said. A 1GHz version of the Duron mobile processor is also on tap for the quarter, as are 1.5GHz and 1.6GHz versions of the desktop Athlon. A 1.73GHz desktop Athlon is set to follow in the fourth quarter. &quot;

&quot;He is talking about the Desktop Palomino. &quot; -- NFS4

Actually, he's talking about new mobile Athlons to compete against Intel's Tualatin procesors.



&quot;SOI does have some drawbacks...--Wingznut

I believe you are referring to the &quot;floating body effect&quot; and cost. Here is an interesting article on SOI. While SOI does appear to be a difficult beast to tame, the problem of &quot;float&quot; can actually be used as an advantage if engineered properly. As far as cost, recent advances in SIMOX-SOI wafer production should cut incremental processing costs of SOI to only 3%-5% over the cost of the bulk versions.

Either way, it's still probably a better investment than some RAM technologies I've seen recently.



...some say that advances in insulating layer techniques have most of the same properties without the added cost.


What insulating layer techniques? The only insulation layers I've seen under developement are oxide based in otherwords SOI.


 


<< << &quot;AMD also revealed product plans for the third quarter. The chipmaker will introduce two new mobile Athlon 4 chips, at 1.1GHz and 1.2GHz, during the quarter, Sanders said. A 1GHz version of the Duron mobile processor is also on tap for the quarter, as are 1.5GHz and 1.6GHz versions of the desktop Athlon. A 1.73GHz desktop Athlon is set to follow in the fourth quarter. &quot;

&quot;He is talking about the Desktop Palomino. &quot; -- NFS4

Actually, he's talking about new mobile Athlons to compete against Intel's Tualatin procesors. >>




Ummm, unless I can't READ, that quote says 1.5GHz, 1.6GHz, and 1.7GHz DESKTOP Palominos. The statement was made earlier about 1.6GHz and 1.7GHz desktop Thunderbirds which is highly unlikely given their heat production. That's why Insane3D provided the link b/c he wasn't sure.

There is no way in hell that 1.5GHz, 1.6GHz, and 1.7GHz mobile chips will be released this year.
 


<< [*]AMD is starting volume production at the end of the year. That means it'll be months before those cpu's hit the street. Q2 if all goes exactly as planned. Q3/Q4 if there are any delays.
[*]SOI wafers are much more expensive than the wafers currently used. The benefits are arguable, as some say that advances in insulating layer techniques have most of the same properties without the added cost.
 >>



OK, based on the fact that the faster chips won't be out until Q2/3/4, and that they might be more expensive, it would still make sense to purchase a 1.4ghz T-bird in the near future (Aug, Sept) if I wanted to upgrade... they wouldn't go obselete that quickly, right? I don't want to wait until next July to upgrade, but I also don't want to buy a chip that will me gone and forgotten by Christmas.

Thoughts?
 


<< i dont understand
processing 64bits per clock is faster than 32bits per clock right? >>

If we are talking about a memory bus (given both are at the same clockspeed) then yes a 64bit bus is faster than a 32bit bus.

But not in processors.
As BurntKooshie said.

Take a 32bit CPU with 1 integer unit, it can do a single 32bit add at a time or a single 16bit, or a single 8bit, or a single 4 bit, but it can not do 2 16bit adds.
Take a 64bit CPU with 1 integer unit, it can do a single 64bit add at a time or a single 32bit or a single 16bit, or a single 8bit, or a single 4 bit, but it can not do 2 32bit adds, or 4 16bit etc etc.

All that moving from 32bit to 64bit gives you is bigger numbers.
Given the same number of functional units and CPU speed it doesn't make anything faster.

And yes a 486 can address 4GB, you'll notice he said >4GB.
 
Your right NFS4, I must have blurred over that last part, funny what a hangover will do to your eyes/mind!

Hey, but it was a great battle o' the bands last night!
 


<< intel is hardly dead. They are a pretty diverse company now, and could survive a while if cpu's didn't make them money. Things shall be getting interesting now. >>



Well, I dunno. Only thing Intel made money was their CPU's. Everything else (networking, consumer products etc. etc.) lost them money big time.
 
You guys really believe in these optimistic talk? Their roadmap shows Q4 of 2002 for .13um SOI, yea, like they're going to beat their roadmap by a year.
 
<<AMD is starting volume production at the end of the year. That means it'll be months before those cpu's hit the street. Q2 if all goes exactly as planned. Q3/Q4 if there are any delays.>>

Why would it take months? I'm assuming that if volume production is happening then, prototypes are out and production in small quantities to go to OEMs is already happening.

Regardless, even Q2 2002 is ahead of their current roadmap which says 2nd half of 2002.

I hope it's true, but you never see roadmaps changed in that direction. Usually they are the most agressive timings to feed to their investors/analysts.

 
Fozzie, I was only referring to AMD/Intel. I personally have never said &quot;anyone else&quot;.

drewski... First off, it takes ~2 months to fabricate a cpu from wafer start to packaging, in normal High Volume production. Fabs &quot;ramp&quot; up production by having few wafer starts at first, and then adding in more month-by-month. When you implement an advanced process like .13um, your first several batches of wafers aren't going to yield and bin well. Process obsticles are found, improvements are made, and quality increases.

Starcutter, there are MANY advanced in insulating layers that you aren't going to hear about. How often do you hear about advances in Etch techniques? Actually, how often do you hear about anything other than Lithography... Pretty much never. Right now, between Thin Films (insulating) and Etch technologies, there aren't any higher proprietary techniques that are kept tightly under wraps.
 
BurntKooshie

ok 64bit programs are more complex right so if we had SEE for a 64bit processor with 64bit programs we will get much more horse power right
i mean it is like pentium pro with 16bit programs it sucks but when we use 32bit programs the pentium pro shines right
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top