Skyscraper wrote:

Threads about Microsoft, Intel or both have with time a 100% probablity of ending up in hilarious offtopic discussions.

Im just going to namedrop Netburst, Vista, Atom and Clippy to see if things get even better! 😀

Hehe, I remember clippy! Which version of Office had clippy again? 😀

I think clippy is very preshott 😁

Scali wrote:

Tetrium wrote:

So I guess you think RAMBUS isn't memory then and you think chipsets have nothing to do with memory? I'm sorry, but you're just being silly here 😀

Clearly you were commenting on what I said about IA64 earlier, about how it was never consumer-oriented.
Now you are mixing it up with RAMBUS. You make no sense.

You mean i820 actually supported IA64 chips? 😲
Imma gonna get me a Itaniumket right away 🤣!

I'm going to bow out of this before the discussion degenerates further. It's strayed far enough from the original assertion (Microsoft's interests and Intel's interests don't necessarily coincide with consumer's interests) that I don't see any real hope of it circling back.

Tetrium wrote:

You mean i820 actually supported IA64 chips? 😲
Imma gonna get me a Itaniumket right away 🤣!

You're the one bringing up i820 in the first place, then somehow trying to couple that to IA64.
You know, you could just admit you misread my earlier post when you responded, but by now it just looks like you're deliberately trolling. Hence I will report you.

Scali wrote:

Tetrium wrote:

You mean i820 actually supported IA64 chips? 😲
Imma gonna get me a Itaniumket right away 🤣!

You're the one bringing up i820 in the first place

No, it was brought up here:

gdjacobs wrote:

IA64 and the RAMBUS saga, two prime examples. Unlike many other companies, Intel had enough runway to recover.

Perhaps you should just get your facts straight before making wild statements which are simply not even true, it would help a lot I think as the way you present your arguments isn't really helping here.

gdjacobs wrote:

I'm going to bow out of this before the discussion degenerates further. It's strayed far enough from the original assertion (Microsoft's interests and Intel's interests don't necessarily coincide with consumer's interests) that I don't see any real hope of it circling back.

^same
This was actually the main reason I didn't respond in this thread in the first place 🤣

Tetrium wrote:

No, it was brought up here: […]
Show full quote

Scali wrote:

Tetrium wrote:

You mean i820 actually supported IA64 chips? 😲
Imma gonna get me a Itaniumket right away 🤣!

You're the one bringing up i820 in the first place

No, it was brought up here:

gdjacobs wrote:

IA64 and the RAMBUS saga, two prime examples. Unlike many other companies, Intel had enough runway to recover.

Perhaps you should just get your facts straight before making wild statements which are simply not even true, it would help a lot I think as the way you present your arguments isn't really helping here.

gdjacobs wrote:

I'm going to bow out of this before the discussion degenerates further. It's strayed far enough from the original assertion (Microsoft's interests and Intel's interests don't necessarily coincide with consumer's interests) that I don't see any real hope of it circling back.

^same
This was actually the main reason I didn't respond in this thread in the first place 🤣

Lolwut?
Firstly, I don't see i820 mentioned there.
Secondly, apparently everyone but yourself sees that the post you responded to was aimed solely at IA64.

Scali wrote:

Whatever, clearly RAMBUS completely ruled at the time, and was the obvious choice for high-end systems.

That's only you opinion and a high end system of the time would have been dual Tualatin or for P4 on an i860 chipset.
What I mentioned was the i850 chipset. Athlon and even single Tualatin walked all over Willamette so don't even try to call it high end.

Scali wrote:

As for slot processors, the obvious reason was that Intel wanted to put large amounts of L2 cache close to the CPU. The Pentium Pro with 1 MB L2 cache proved to be prohibitively expensive to manufacture, hence the introduction of the Pentium II in slot-form, where cache-chips could be put onto a PCB together with the CPU.
Clearly technical limitations.
AMD also went to its own slot processor system for the first generation of Athlons.

None of this (were it even true) makes what I said invalid.
Perhaps you don't know that socket 7 didn't accept Pentium Pro CPUs.
.
The Pentium Pro was succeeded by the Pentium II Xeon.
The Pentium socket 7 succeeded by the Pentium II.
None of the PII's achieved as high a clock speed as AMD managed on socket 7.
- Yet Intel's reason for going slot was that (according to them) socket 7 had no more head room.
.

PCBONEZ wrote:

That's only you opinion

No, it is not. It is a fact that RAMBUS delivered the highest bandwidth of any available mainstream memory technology at the time.
It is also a fact that high-end systems benefited from a lot of bandwidth.
So from these facts we draw the logical conclusion that RAMBUS is the obvious choice for high-end systems (or for other bandwidth-hungry systems, such as game consoles).

PCBONEZ wrote:

Athlon and even single Tualatin walked all over Willamette so don't even try to call it high end.

Don't bring CPU performance into a conversation about memory bandwidth.
You can clearly see that once the initial cache-benefits run out, the Pentium 4 was in a league of its own in terms of bandwidth: http://www.anandtech.com/show/661/15
http://www.anandtech.com/show/661/23

Scali wrote:

Perhaps you don't know that socket 7 didn't accept Pentium Pro CPUs.

What is that even supposed to mean? You brought up slot processors. The first of which was a Pentium II. And if you had done your homework, you would know that it was based on the P6 microarchitecture of the Pentium Pro.
It was basically a Pentium Pro with MMX in a slot package.

PCBONEZ wrote:

None of the PII's achieved as high a clock speed as AMD managed on socket 7.

Which is completely irrelevant.
AMD just stuck with socket 7 longer, big deal (because they are always trailing behind Intel on technology). That is completely arbitrary.
My point was that the reason to switch to slot was because of external L2 cache chips on a PCB, not because of any clockspeed issues.
Heck, we went back to sockets quickly, and then Intel proceeded to slap everyone silly with clockspeed with the Pentium 4, remember?

Scali wrote:

No, it is not. It is a fact that RAMBUS delivered the highest bandwidth of any available mainstream memory technology at the tim […]
Show full quote

PCBONEZ wrote:

That's only you opinion

No, it is not. It is a fact that RAMBUS delivered the highest bandwidth of any available mainstream memory technology at the time.
It is also a fact that high-end systems benefited from a lot of bandwidth.
So from these facts we draw the logical conclusion that RAMBUS is the obvious choice for high-end systems (or for other bandwidth-hungry systems, such as game consoles).

PCBONEZ wrote:

Athlon and even single Tualatin walked all over Willamette so don't even try to call it high end.

Don't bring CPU performance into a conversation about memory bandwidth.
You can clearly see that once the initial cache-benefits run out, the Pentium 4 was in a league of its own in terms of bandwidth: http://www.anandtech.com/show/661/15

Scali wrote:

Perhaps you don't know that socket 7 didn't accept Pentium Pro CPUs.

What is that even supposed to mean? You brought up slot processors. The first of which was a Pentium II. And if you had done your homework, you would know that it was based on the P6 microarchitecture of the Pentium Pro.
It was basically a Pentium Pro with MMX in a slot package.

PCBONEZ wrote:

None of the PII's achieved as high a clock speed as AMD managed on socket 7.

Which is completely irrelevant.
AMD just stuck with socket 7 longer, big deal. That is completely arbitrary.
My point was that the reason to switch to slot was because of external L2 cache chips on a PCB, not because of any clockspeed issues.
Heck, we went back to sockets quickly, and then Intel proceeded to slap everyone silly with clockspeed with the Pentium 4, remember?

The memory bandwidth is pretty irrelevant when the complete system gets stomped by an earlier technology.

I brought up socket 7 and you responded with irrelevant info about Pentium Pro.

And no. It's not irrelevant.
The discussion is about Intel making poor decisions and Intel though socket 7 was dead.
AMD proved them wrong.
.
.

PCBONEZ wrote:

The memory bandwidth is pretty irrelevant when the complete system gets stomped by an earlier technology.

Except that wasn't the point.
The point was that RAMBUS delivered way more bandwidth at the time, which is clearly true.
Hence it is clearly a good decision to go with that.

PCBONEZ wrote:

I brought up socket 7 and you responded about Pentium Pro.

You brought up slot processors. So I brought up Pentium II (the first slot processor), and its predecessor, Pentium Pro.
What part about that don't you get?
Socket 7 was indeed dead. You are referring to "super socket 7", which was a poor hack on an outdated platform, which wasn't able to compete with Intel's newer platforms. Clearly you have no idea what you're talking about. No wonder I didn't get your point. You're not making any sense.

Scali wrote:

PCBONEZ wrote:

The memory bandwidth is pretty irrelevant when the complete system gets stomped by an earlier technology.

Except that wasn't the point.
The point was that RAMBUS delivered way more bandwidth at the time, which is clearly true.

No that's not the point.
The point is Intel made poor decisions and locked themselves into a platform that couldn't even keep up with older technology.

Scali wrote:

You brought up slot processors. So I brought up Pentium II (the first slot processor), and its predecessor, Pentium Pro. What pa […]
Show full quote

PCBONEZ wrote:

I brought up socket 7 and you responded about Pentium Pro.

You brought up slot processors. So I brought up Pentium II (the first slot processor), and its predecessor, Pentium Pro.
What part about that don't you get?
Socket 7 was indeed dead. You are referring to "super socket 7", which was a poor hack on an outdated platform, which wasn't able to compete with Intel's newer platforms. Clearly you have no idea what you're talking about.

No I did not. What I said was Intel decided Socket 7 was dead when it clearly was not. - AMD proved it.
Yet another Intel poor decision.
And it's no more a hack on an outdated platform than evolving a P-1 or a P-II out of a Pentium Pro.
.

Well, I don't mean to derail this any further, but something piqued my curiosity. Scali: did the Netburst architecture have any merits regarding memory speed, or was most of it due to the platform (quad-pumped bus)? I could look up some Pentium M benchmarks but I'm not really sure this would be a fair comparison, as a high-end Pentium 4 from the same time frame would run at a higher bus speed and use non mobile-oriented chipsets.
Also, was SS7 that bad? Or was most of its problems related to the fact neither VIA, ALi or SiS had the same expertise as Intel in making stable and fast chipsets with good AGP implementations.

PCBONEZ wrote:

The point is Intel made poor decisions and locked themselves into a platform that couldn't keep up with older technology.

RAMBUS wasn't the cause of the performance issues, on the contrary.

PCBONEZ wrote:

No I did not.

Yes you did:
"Further back, Intel abandoned socket 7 at 233 MHz and went to slot processors claiming that socket 7 (and socket processors in general) had no more room for growth due to technical limitations."

So I pointed out the REAL reason, which was not FSB speed (which should be obvious).

PCBONEZ wrote:

AMD proved it.

In the extremely rose-tinted view of an AMD fanboy perhaps, completely ignoring the fact that AMD's super socket 7 platforms were completely outpaced by Pentium II/III systems in every way... FSB, memory performance, caching, CPU grunt...

PCBONEZ wrote:

And it's no more a hack on an outdated platform than evolving a P-1 or a P-II out of a Pentium Pro.

It is actually.
Besides, get your facts straight for a change. The Pentium was the architecture BEFORE the Pentium Pro, and Pentium Pro was hardly an evolution of all that went before it. It was a revolution, and is pretty much the blueprint for x86 CPUs even today (decoding x86 to micro-ops, scheduling these micro-ops out-of order with a superscalar execution backend, and then retiring the instructions in-order).

alexanrs wrote:

did the Netburst architecture have any merits regarding memory speed, or was most of it due to the platform (quad-pumped bus)?

It seems that Pentium 4 was all about stream processing. SSE2 combined with the large caches and high bandwidth of RAMBUS made it extremely fast at tasks such as video encoding/decoding and other types of image processing.

alexanrs wrote:

Also, was SS7 that bad? Or was most of its problems related to the fact neither VIA, ALi or SiS had the same expertise as Intel in making stable and fast chipsets with good AGP implementations.

Well, the lousy chipsets were one problem. But the other was the poor performance of the on-board caches, as I said. Intel wanted L2 to be next to the CPU, for lower latency and higher bandwidth. They did this with the Pentium Pro, by placing two dies into a single package, the CPU and the cache.
But this proved to be too expensive for mass-production at the time. Which is why they came up with the slot system: have the cache in separate chips, but put the cache chips and CPU on the same PCB, so you could still have a very low-latency high bandwidth connection with the CPU.
Eventually manufacturing matured to the point where L2 cache could easily be integrated with the CPU on a single die, so the slot arrangement lost its purpose, and we went back to the cheaper and simpler socket solution.

alexanrs wrote:

Well, I don't mean to derail this any further, but something piqued my curiosity. Scali: did the Netburst architecture have any merits regarding memory speed, or was most of it due to the platform (quad-pumped bus)? I could look up some Pentium M benchmarks but I'm not really sure this would be a fair comparison, as a high-end Pentium 4 from the same time frame would run at a higher bus speed and use non mobile-oriented chipsets.
Also, was SS7 that bad? Or was most of its problems related to the fact neither VIA, ALi or SiS had the same expertise as Intel in making stable and fast chipsets with good AGP implementations.

A Pentium M is basically an updated Tualatin with a higher clock and more cache. - One of the few good decisions Intel made.
.

.
Thus illustrating what a huge mistake NetBurst was.
.

PCBONEZ wrote:

One of the few good decisions Intel made.

Wow, bias much?

Scali wrote:

PCBONEZ wrote:

One of the few good decisions Intel made.

Wow, bias much?

In addition to having your facts wrong you obviously don't even remember what the argument is.
All of this banter and citing examples is in response to your earlier claims that Intel doesn't make mistakes.

FYI: I haven't had an AMD CPU in my main rig for 15 years.
However despite the fact that I prefer Intel CPUs I'm not going to ignore the BLATANTLY OBVIOUS fact that Intel makes LOTS of mistakes.
.

I never said that Intel didn't make mistakes.
However, claiming that they only made 'a few good decisions' is completely ridiculous.
Erm hello?
Intel is the company that invented the microprocessor in the first place. They've been at the forefront of chip design and manufacturing for over 40 years now, and they have crushed all but a few of their competitors over the years. They are known as "Chipzilla" because of how huge they are compared to every other player in the industry.
You don't get where Intel is by making "only a few good decisions".

Scali wrote:

I never said that Intel didn't make mistakes.

Yes you did and I'm only one of 4+ people that don't agree with you.

Scali wrote:

Intel is the company that invented the microprocessor in the first place.

No they weren't first. They were no sooner than third.
The AL-1 was in 1969 and the MP944 was in 1970. The 4004 wasn't till 1971.

Scali wrote:

They've been at the forefront of chip design and manufacturing for over 40 years now, and they have crushed all but a few of their competitors over the years. They are known as "Chipzilla" because of how huge they are compared to every other player in the industry.

None of that means they don't make mistakes and everyone except you knows they do. Some big ones in fact.

PCBONEZ wrote:

Yes you did and I'm only one of 4+ people that don't agree with you.

Quote or it didn't happen.

PCBONEZ wrote:

None of that means they don't make mistakes and everyone except you knows they do. Some big ones in fact.

I know they made mistakes, I never said otherwise. Heck, the whole x86 architecture is one big mistake. I just pointed out that people are listing some things that *aren't* mistakes, and explained what things really are like.
To somehow paint me as some kind of Intel-fanboy because of that says a lot more about you than it does about me.

As for the AL-1, see here: http://www.computerhistory.org/revolution/dig … gic/12/282/2291
The 4004 was the first microprocessor to be sold as a component.