G3 Vs G4

[Mr. MacPhisto]

That article tells me that Intel's Fab11X is the top plant for .13 process chips. It is speculation if they're pressing at .09.

And a recent major delay would be Itanium - delayed by two years and then not delivering the promised good when released.

IBM is fabbing 970s at .13 as we speak. 750s will be taken to a .10 process and 970s will be at .09 by the end of the year.

Other delays at Intel - their copper technology that IBM developed over 5 years ago. They can't seem to mass produce it. Some sources say their yields are way under 60% on the current P4 line. Most people in the industry I talk to just say Intel is tops because they have no really good, really big name competition. I'll do some digging, but almost everywhere I've read people saying how far advanced IBM is. Intel does have the best tech for a PC desktop manufacturing plant, but that will change. Wait for their 2004 issue when both IBM and AMD have passed Intel by. Their chips are built on archaic tech (my bro is in the industry and he and several others with advanced engineering degrees all see serious flaws) outside of Itanium. Problem is, Itanium was delayed and because of that delay we have the Power4 and Power5. Trust me, in five years Intel will be struggling or out of the CPU business entirely - especially with IBM assisting AMD on its design and manufacturing process.

As for secret plants, that is very much how IBM works. No one in the public sphere can say exactly what is done at all of their plants. IBM develops so much stuff and makes so much that they prefer as much secrecy as possible.

 

[ddtlm]

Mr. MacPhisto:

That article tells me that Intel's Fab11X is the top plant for .13 process chips. It is speculation if they're pressing at .09.

Speculation is only acceptable when it favors IBM and Apple? Keep that question in mind for a few minutes...

And a recent major delay would be Itanium - delayed by two years and then not delivering the promised good when released.

Not related to process technology.

750s will be taken to a .10 process and 970s will be at .09 by the end of the year.

Speculation.

Other delays at Intel - their copper technology that IBM developed over 5 years ago.

Please demonstrate that they have tried to use it, and if so, that they are behind scedule.

Some sources say their yields are way under 60% on the current P4 line.

Please provide even one reputable source with this cllaim.

Intel does have the best tech for a PC desktop manufacturing plant, but that will change.

Speculation.

Wait for their 2004 issue when both IBM and AMD have passed Intel by.

Speculation.

Their chips are built on archaic tech (my bro is in the industry and he and several others with advanced engineering degrees all see serious flaws) outside of Itanium.

This appears to be a (classic, worn out) assault on x86. However I can only wonder how, if your claims are true, that Intel and AMD continue to field top-performing x86 processors. Where are the "superior" processors? What an embarrasment to the makers of chips based on "superior" ISA's.

Trust me, in five years Intel will be struggling or out of the CPU business entirely - especially with IBM assisting AMD on its design and manufacturing process.

Speculation.

 

[mattmack]

Originally posted by ddtlm
Mr. MacPhisto:



Speculation.



Speculation.


Speculation.




Speculation.

This is an opinion and rumors site. If you don't agree with an opinion that is fine, but I don't think he was stating anything outragous or out of bounds. The x86 processor is a fine processor, but any one who thins that intel is still at the top of their game might consider looking a little more critically. Also he did say that AMD as well as IBM was going to give Intel a run for their money with competative chips

 

[ddtlm]

mattmack:

Hey he called some pro-Intel stuff speculation, so he opened up his whole speculative outlook to attack.

 

[Mr. MacPhisto]

Originally posted by ddtlm
mattmack:

Hey he called some pro-Intel stuff speculation, so he opened up his whole speculative outlook to attack.

Not really. I have inside-IBM info and it is not speculation, but it is unsubstantiated at the moment because of the sensitivity of my sources. However, please do note that Intel's plant won the big award this year before IBM's own 300mm plant opened in East Fishkill, NY. This is FACT and public as is the 90nm process and the FACT that IBM will be the first to press at 90nm:http://www-916.ibm.com/press/prnews.nsf/printwc/BE30E0C347D45AD285256C91005458D6

Yes, it is not the 970 or 750, but that data will be forthcoming. IBM at .09 is not speculative.

Copper is actually in most current Intel chips, although the process does not produce the efficiency results of IBM's tech.


Why do Intel and AMD make performance processors based on x86? Because much of the world is on the x86 architecture. The difficulty of migration means that they can't force one, especially with competition in their own market. There are many, many who feel the PPC architecture is superior for the long run. It has been misused in Moto's hands, but IBM will show its true potential. By the end of 2005 we'll see a .065 chip at 8GHZ, hyperthreaded and dual core. Methinks Intel will have problems matching that.

 

[ddtlm]

Mr. MacPhisto:

Well your claims still strike me as way out there, but I don't see either of us changing our positions anytime soon, so I'll try to leave it at this.

 

[Cubeboy]

Originally posted by Mr. MacPhisto
That article tells me that Intel's Fab11X is the top plant for .13 process chips. It is speculation if they're pressing at .09.

It explicitly stated that Fab11X will transition to .09 micron process in the beginning of the second half of 2003.

Originally posted by Mr. MacPhisto
And a recent major delay would be Itanium - delayed by two years and then not delivering the promised good when released.

No I'm afraid this is not recent, Itanium 2 has already been out for a while and the next Itanium (Madison) is set to be released this month. We're nearing three versions of the Itanium, so far their have only been four Pentiums.

Originally posted by Mr. MacPhisto
IBM is fabbing 970s at .13 as we speak. 750s will be taken to a .10 process and 970s will be at .09 by the end of the year.

Which goes with my statement, until IBM sells the majority of it's .13 micron 970s, it won't even begin to put .09 on the market. If IBM isn't even done producing .13 nm 970s, than we won't see .09 nm 970s in quite a while. Simple business logic

Originally posted by Mr. MacPhisto
Other delays at Intel - their copper technology that IBM developed over 5 years ago. They can't seem to mass produce it. Some sources say their yields are way under 60% on the current P4 line. Most people in the industry I talk to just say Intel is tops because they have no really good, really big name competition. I'll do some digging, but almost everywhere I've read people saying how far advanced IBM is. Intel does have the best tech for a PC desktop manufacturing plant, but that will change. Wait for their 2004 issue when both IBM and AMD have passed Intel by. Their chips are built on archaic tech (my bro is in the industry and he and several others with advanced engineering degrees all see serious flaws) outside of Itanium. Problem is, Itanium was delayed and because of that delay we have the Power4 and Power5. Trust me, in five years Intel will be struggling or out of the CPU business entirely - especially with IBM assisting AMD on its design and manufacturing process.

Let's see, the Pentium 4 3.00 is the fastest desktop processor out today, the dual Xeons is the fastest workstation processor, and until quite recently the Power4, despite being a huge and enormously expensive dual core processor, has been largely behind Itanium 2 and Alpha for most of it's existence. Intel has all but one of the original Alpha team working for it, nearly all of the veteran HP team responsible for HP PA-RISC server chips, a large part of the Motorola team responsible for the original G4 (which was quite good at the beginning), and quite a few of the MIPS team responsible for the venerated MIPS server chips. Since before the arrival of the original Athlon, I don't think it's been in a better shape.

Originally posted by Mr. MacPhisto
As for secret plants, that is very much how IBM works. No one in the public sphere can say exactly what is done at all of their plants. IBM develops so much stuff and makes so much that they prefer as much secrecy as possible.

Purely speculation as DDLTM put it. We'll have to wait and see.

 

[Cubeboy]

Originally posted by Mr. MacPhisto
Not really. I have inside-IBM info and it is not speculation, but it is unsubstantiated at the moment because of the sensitivity of my sources. However, please do note that Intel's plant won the big award this year before IBM's own 300mm plant opened in East Fishkill, NY. This is FACT and public as is the 90nm process and the FACT that IBM will be the first to press at 90nm:http://www-916.ibm.com/press/prnews.nsf/printwc/BE30E0C347D45AD285256C91005458D6

Yes, it is not the 970 or 750, but that data will be forthcoming. IBM at .09 is not speculative.

Copper is actually in most current Intel chips, although the process does not produce the efficiency results of IBM's tech.

Why do Intel and AMD make performance processors based on x86? Because much of the world is on the x86 architecture. The difficulty of migration means that they can't force one, especially with competition in their own market. There are many, many who feel the PPC architecture is superior for the long run. It has been misused in Moto's hands, but IBM will show its true potential. By the end of 2005 we'll see a .065 chip at 8GHZ, hyperthreaded and dual core. Methinks Intel will have problems matching that.

As I've said in one of my previous post those are FPGA chips, no where near as complex as a modern desktop processor. It'll be a while before we see 90nm PPC970s from IBM. In fact we've yet to see the announcement launching 130 nm PPC970s much less 90 nm PPC970s.

The first X86 processor was made by Intel in 1978, since then we've had the same group of people predicting that it is obsolete it will die "any time now". It's been 25 years since then, I've yet to see single sign of x86s death or any sign of weakening. Right now, the current line of Pentium 4s and Athlons are actually superior to most of the expensive risc "new instruction" server chips. Did you notice the Opteron launch? It outperformed nearly every server chip in multiple server areas such as high performance computations and transactions/minute and it was several times cheaper than any of the "new" server chips. Please, I've heard this a couple hundred times now, I've yet to see one person prove himself right.

 

[ddtlm]

Cubeboy:

You have so much more patience than me.

 

[MacBandit]

Originally posted by Cubeboy
As I've said in one of my previous post those are FPGA chips, no where near as complex as a modern desktop processor. It'll be a while before we see 90nm PPC970s from IBM. In fact we've yet to see the announcement launching 130 nm PPC970s much less 90 nm PPC970s.

The first X86 processor was made by Intel in 1978, since then we've had the same group of people predicting that it is obsolete it will die "any time now". It's been 25 years since then, I've yet to see single sign of x86s death or any sign of weakening. Right now, the current line of Pentium 4s and Athlons are actually superior to most of the expensive risc "new instruction" server chips. Did you notice the Opteron launch? It outperformed nearly every server chip in multiple server areas such as high performance computations and transactions/minute and it was several times cheaper than any of the "new" server chips. Please, I've heard this a couple hundred times now, I've yet to see one person prove himself right.

Ah, but in the latest revissions of the AMD chips they more closely resemble a RISC processor (Reduced Instruction Set Computer) then they do a CISC (Complex Instruction Set Computer). CISC being X86. So in fact AMD has amitted the failings of the X86 and has gone to RISC and is emulating an X86.

 

[ZeeOwl]

Er.... G3 vs. G4 anyone?

Getting back to the subject of the thread, here's some interesting field info...

I just ran some tests on my two computers (a 350Mhz Blue & White G3, and a 700Mhz iMac G4) using CineBench 2003 R8.1. I ran the speed tests both with the L2 cache enabled and disabled on both machines. Here are the results.

G3 350 with cache: 738.4 seconds.
G3 350 without cache: 901.0 seconds.
G4 700 with cache: 432.9 seconds.
G4 700 without cache: 632.1 seconds.

Now the two machines have some architectural similarities and differences.
Similiarities: 1 GiB CL2 RAM, 100 MHz FSB.
Differences: G3 - 512 kiB cache @ 233 Mhz, G4 - 256 kiB cache @ 700 Mhz.

Now, of course, the G4 is running at 7x the FSB, and the G3 only 3.5x. But even then, the cache-dependency of the G4 seems to be much higher than the G3's. Especially since the G3's cache is twice the size of the G4's (albeit, running at a slower speed). That would tend to indicate that the G3 is a superior, more efficient design than the G4. Makes the proposed IBM 750GX+AltiVec even more exciting, for Apple's consumer line.

I just recieved a Powerlogix 800 MHz G3 upgrade for my G3, and wanted to try it at 700 Mhz to compare G3 vs. G4 performance (the upgrade has a 512 kiB cache running at full processor speed). Theoretically then, the G3 should have slightly outperformed the G4, having a larger cache, and all other things being equal. But unfortunately my upgrade card was DOA. So I'll have to get back to you on that test.

 

[ddtlm]

MacBandit:

This RISC core in x86 clothes thing has been going on since the days of the Pentium Pro! And yet the x86 bashers march blindly on...

 

[MacBandit]

Originally posted by ddtlm
MacBandit:

This RISC core in x86 clothes thing has been going on since the days of the Pentium Pro! And yet the x86 bashers march blindly on...

How do you mean? Are you saying the current Pentiums are RISC with X86 emulation?

 

[ddtlm]

MacBandit:

How do you mean? Are you saying the current Pentiums are RISC with X86 emulation?

Yes. All of them, since and including the Pentium Pro. I think the AMD K6 was too but I don't intend to check. When reading (in-depth technical) articles on the modern x86 chips you'll a lot of talk about instruction decoding, which is the translation stage. Yeah, they knew a pure CISC processor was a bad idea some time ago.

 

[MacBandit]

Originally posted by ddtlm
MacBandit:


Yes. All of them, since and including the Pentium Pro. I think the AMD K6 was too but I don't intend to check. When reading (in-depth technical) articles on the modern x86 chips you'll a lot of talk about instruction decoding, which is the translation stage. Yeah, they knew a pure CISC processor was a bad idea some time ago.

So why the argument over RISC and why does Microsoft continue to right code for CISC when they could do RISC and Intel and AMD could drop the emulation?

Sounds to me like CISC is dead and the only thing keeping it is Windows.

 

[ddtlm]

MacBandit:

The RISC vs CISC thing is, in my opinion, just not important anymore. In fact, the instruction set which encodes my programs is almost irrelevant as well. All that matters is that programs run fast on a lot of computers. With ever higher-level (and therefore more wasteful) langauges like VB and Java being used to code more and more things, noone is going to notice the small performance hit that the x86 ISA probably takes.

I think that all anyone needs to know about PPC vs x86 is that they are different.

 

[MacBandit]

Originally posted by ddtlm
MacBandit:

The RISC vs CISC thing is, in my opinion, just not important anymore. In fact, the instruction set which encodes my programs is almost irrelevant as well. All that matters is that programs run fast on a lot of computers. With ever higher-level (and therefore more wasteful) langauges like VB and Java being used to code more and more things, noone is going to notice the small performance hit that the x86 ISA probably takes.

I think that all anyone needs to know about PPC vs x86 is that they are different.

Sounds like anymore they are different but the same.

 

[mattmack]

Originally posted by ddtlm
mattmack:

Hey he called some pro-Intel stuff speculation, so he opened up his whole speculative outlook to attack.

True I just don't want this to turn into a flame war so I was just trying to be reasonable. Everyone is entitled to an opinion and those are the hardest things to convince people to change. I am all for respect of whatever opinion one holds and thank both of you for letting this remain a friendly discussion

 

[Cubeboy]

Originally posted by MacBandit
Ah, but in the latest revissions of the AMD chips they more closely resemble a RISC processor (Reduced Instruction Set Computer) then they do a CISC (Complex Instruction Set Computer). CISC being X86. So in fact AMD has amitted the failings of the X86 and has gone to RISC and is emulating an X86.

And can you tell me exactly how much revision was done to the latest AMD chips as to call it a RISC chip. The fact is, as ddtlm stated that the CISC have gone through a very evolutionary change. The current x86 cpus are to some degree RISC processors but the external layer is still CISC, their still isn't any parallel pipelining in any of the current AMD and Intel processors, an current x86 instruction is on average 1.5 micro-ops which is still larger than the average RISC instruction and most importantly that they still have significantly larger instruction sets than any RISC processor with the instructions varying in length, hence they are still refered to largely as CISC processors. It's not abandoning as you put, but integration, today RISC chips support as many instructions as previous CISC chips certainly alot more than their predecessors, likewise, today's CISC chips use many techniques formerly associated with RISC chips.