The rise and fall of AMD

[Abwx]

What was the big innovation in Fusion? They certainly weren't the first IGPs.



Let's see, off the top of my head:
- Invented the microprocessor.
- First commercial DRAM IC.
- Created the entire x86 ISA infrastructure.
- Superscalar processing.
- Clock multiplying.
- Integrated L2 cache.
- L2 cache running at CPU speed.
- OOO execution.
- multicore chips.
- turbo mode.
- integrated graphics.
- temperature sensors.
- numerous RAS features.
- SIMD instructions.
- Numerous other ISA improvements (whether you like it or not).
- hyperthreading.
- instruction fusion.
- countless branch prediction improvements.


How about outside CPUs?

- Northbridge/southbridge chipset architecture?
- High-speed nb/sb links?
- The entire line of USB standards?
- PCI?
- The ATX form factor?
- First mainstream SSDs?

I could find dozens more if I wanted to spend the time.

And what has AMD done that's really been innovative? Not much. They mostly just copy Intel.

Most if not all your list is not due to Intel , google a little...

 

[Exophase]

You can make a supercomputer that hits whatever performance figures you want by scaling it with enough nodes. I suppose we'll also be seeing you make the argument about how Tesla is better than FirePro because that's what Cray chose as well, right? And it's a machine that's for HPC, advertises FLOPs, that vast majority of which are provided by the GPUs, but of course its marketing characteristics are all thanks to those Bulldozers..

 

[Exophase]

Most if not all your list is not due to Intel , google a little...

They're first for Intel if you add the implicit "for an x86 CPU" to it, which mostly limits the running to AMD.

Some of these are still wrong though. Pentium Pro was the first Intel x86 processor to feature OoO, integrated (on-package) L2, and L2 running at core speed. But Nx586 was released in 1994 and had all of these features.

Counting things like "instruction fusion" and branch prediction improvements is kind of petty. If you're going to do that may as well count AMD as first having post-decode macro-ops (K7 vs Pentium M) and at times better branch prediction as well (even K6 professed higher predict rates than P6, despite having a much smaller mispredict penalty). And of course AMD was first to include an integrated memory controller, on-die multicore, first to have FP SIMD, etc..

 

[Abwx]

We could add that the uProceesor did exist years before intel s 4000 series ,
IGP was already present in Cyrix mediaGX , superscalar processing existed
in Cray 1 circa 1976 , Turbo mode ?..my old Athlon XPM work at 500MHZ
and goes at 1500MHZ when utilisation rise , temperature probes ???.a joke ,
it existed long ago , and so on for this list that hold more from urban legends
than actual factual history...

 

[Charles Kozierok]

Most if not all your list is not due to Intel , google a little...

LOL. Is that supposed to be a counter-argument?

Some of these are still wrong though. Pentium Pro was the first Intel x86 processor to feature OoO, integrated (on-package) L2, and L2 running at core speed. But Nx586 was released in 1994 and had all of these features.

Whoops, you're right, I forgot about that little guy. It did have OoO first. I'm pretty sure it didn't have integrated L2, though.

Counting things like "instruction fusion" and branch prediction improvements is kind of petty. If you're going to do that may as well count AMD as first having post-decode macro-ops (K7 vs Pentium M) and at times better branch prediction as well (even K6 professed higher predict rates than P6, despite having a much smaller mispredict penalty). And of course AMD was first to include an integrated memory controller, on-die multicore, first to have FP SIMD, etc..

I'm fine to count whatever people want to count. Like I said, it was just a list off the top of my head -- I'm supposed to be writing.

I think it's pretty self-evident to any objective individual that Intel has been responsible for the lion's share of innovations in this segment of the computer industry.

 

[Exophase]

Nx586 had an integrated L2 controller that run at core speed.. they could have packaged it with L2 cache on the package but they didn't. So count that for 1.5 out of those 2 things if you want (integrated L2, L2 at core clock).

I agree Intel has been responsible for more but I personally consider some of the things AMD and others have done to be really significant.

We could add that the uProceesor did exist years before intel s 4000 series

Probably true, although the details are hazy.

IGP was already present in Cyrix mediaGX

Good point. National Semiconductor, who bought Cyrix before they were transferred to VIA, had the specific goal of making the first x86 SoCs.

superscalar processing existed in Cray 1 circa 1976

Of course this was meant to be x86 only, Pentium really was the first superscalar x86 processor.

Turbo mode ?..my old Athlon XPM work at 500MHZ and goes at 1500MHZ when utilisation rise

That's not turbo. You'd may as well count Intel's speed-step as the same thing then, but it isn't. Turbo was a big deal when multicore took off.

temperature probes ???.a joke it existed long ago

Intel was the first to have on-die CPU temperature diodes for x86 processors. I have no idea if anyone else had it non-x86. But I know I appreciated not having to put that damn thermistor beneath the heatsink anymore (god only knows if this kept heat transfer as efficient)

 

[AtenRa]

What was the big innovation in Fusion? They certainly weren't the first IGPs.



Let's see, off the top of my head:
- Invented the microprocessor.
- First commercial DRAM IC.
- Created the entire x86 ISA infrastructure.
- Superscalar processing.
- Clock multiplying.
- Integrated L2 cache.
- L2 cache running at CPU speed.
- OOO execution.
- multicore chips.
- turbo mode.
- integrated graphics.
- temperature sensors.
- numerous RAS features.
- SIMD instructions.
- Numerous other ISA improvements (whether you like it or not).
- hyperthreading.
- instruction fusion.
- countless branch prediction improvements.


How about outside CPUs?

- Northbridge/southbridge chipset architecture?
- High-speed nb/sb links?
- The entire line of USB standards?
- PCI?
- The ATX form factor?
- First mainstream SSDs?

I could find dozens more if I wanted to spend the time.

And what has AMD done that's really been innovative? Not much. They mostly just copy Intel.



This is a joke. How did they "prevent competitors from innovating"? By being better than they were?

http://en.wikipedia.org/wiki/Superscalar

Seymour Cray's CDC 6600 from 1965 is often mentioned as the first superscalar design.

http://en.wikipedia.org/wiki/Out-of-order_execution

About three years later, the IBM 360/91 (1966) introduced Tomasulo's algorithm, supporting full out-of-order execution.
In 1990, IBM introduced the first out-of-order microprocessor, the POWER1, although out-of-order execution was limited to floating point instructions only.

http://en.wikipedia.org/wiki/Simultaneous_multithreading#Historical_implementations

While multithreading CPUs have been around since the 1950s, simultaneous multithreading was first researched by IBM in 1968 as part of the ACS-360 project.[1] The first major commercial microprocessor developed with SMT was the Alpha 21464 (EV8). This microprocessor was developed by DEC in coordination with Dean Tullsen of the University of California, San Diego, and Susan Eggers and Hank Levy of the University of Washington.

And im sure more of what you typed was not researched by Intel, unless you mean that they introduced them to the x86 ecosystem.

 

[Charles Kozierok]

We could add that the uProceesor did exist years before intel s 4000 series ,
IGP was already present in Cyrix mediaGX , superscalar processing existed
in Cray 1 circa 1976 , Turbo mode ?..my old Athlon XPM work at 500MHZ
and goes at 1500MHZ when utilisation rise , temperature probes ???.a joke ,
it existed long ago , and so on for this list that hold more from urban legends
than actual factual history...

What an incredibly dishonest reply.

There were a couple of chips that maybe could be considered microcontrollers before the 4004. That was the chip that defined the modern microprocessor.

The MediaGX's capabilities nowhere approached a modern GPU.

The frequency scaling on your Athlon XP-M was copied from SpeedStep, which Intel invented years earlier (thanks for an addition to my list).

And the rest of your response is puerile whining.

Where's the great list of AMD innovations?

 

[Charles Kozierok]

And im sure more of what you typed was not researched by Intel, unless you mean that they introduced them to the x86 ecosystem.

Since the subject is Intel versus AMD, yes, that's what I am talking about. Of course nearly everything was first invented on expensive mainframes. But most of that was put into x86 by Intel and copied by AMD. And most of the stuff that was truly new to microprocessors was also created by Intel, and copied by AMD.

 

[mrmt]

Counting things like "instruction fusion" and branch prediction improvements is kind of petty. If you're going to do that may as well count AMD as first having post-decode macro-ops (K7 vs Pentium M) and at times better branch prediction as well (even K6 professed higher predict rates than P6, despite having a much smaller mispredict penalty). And of course AMD was first to include an integrated memory controller, on-die multicore, first to have FP SIMD, etc..

Did you notice that your entire list wasn't developed in-house by AMD, but scooped when AMD acquired those companies?

 

[Exophase]

Did you notice that your entire list wasn't developed in-house by AMD, but scooped when AMD acquired those companies?

Does it matter? No. Maybe you want to selectively count every employee who ever worked there as being true to AMD or not but I don't. Do you want to say that all of Intel's smart developments only came because they hired the right people? Or do you somehow think that acquisitions should be treated differently?

I don't know why people are so hung up on pre-NexGen AMD. As far as I'm concerned, when you're dissing AMD as a company at large over their entire history you're also dissing the talented engineers they acquired from NexGen and DEC. And if you do want to only look at them prior to acquisition then they at least deserve credit for making some smart acquisition choices.

 

[Abwx]

LOL. Is that supposed to be a counter-argument?

Read above , and a few more in this post :

Intel invented SIMD ?..lol ..

The first use of SIMD instructions was in vector supercomputers of the early 1970s such as the CDC Star-100 and the Texas Instruments ASC, which could operate on a "vector" of data with a single instruction.

They also invented multicore CPUs..??.Re lol..

POWER4, the world's first non-embedded dual-core processor, released in 2001.

You are spreading innacuracies in a deliberate attempt to mislead
the general public and make intel look better than it actualy is.

How else could such innaccuracies be posted with the hope
that no one will check..?..

Historicaly , Intel were good at looting other corporate IP
and adapt it in X86 thanks to their monopoly in this market ,
it is well known that they are one of the least innovative
company in semi industry , lergely due to their grip over
the said market.

 

[Exophase]

The list was specific to x86. That was implied then explicitly stated. There wasn't any attempt to mislead anyone into thinking otherwise. Can you guys please stop pointing out how other companies used the technology in non-x86 platforms? It's getting annoying.

 

[Sleepingforest]

So you're saying that bringing these features into x86 didn't require any R&D?

 

[mrmt]

Does it matter? No. Maybe you want to selectively count every employee who ever worked there as being true to AMD or not but I don't.

It does matter. It's completely different to have a vision, hire the right people to develop the tech and get the right product out of the door and do what AMD does, which is spot a nice company with a nice concept needing funding out there and inject money there.

AMD is very good on the later, but does almost nothing on the former. The entire list of AMD innovations weren't developed in-house, they came from AMD acquisitions. If you want to portray AMD as an innovating company, bring something developed inside AMD, not something they bought from others.

Ed: AMD does something akin to EA. Both acquires smaller companies need funding, get good first generation products with the acquired R&D and then screw up when they have to develop something by themselves.

 

[Abwx]

What an incredibly dishonest reply.

There were a couple of chips that maybe could be considered microcontrollers before the 4004. That was the chip that defined the modern microprocessor.

The MediaGX's capabilities nowhere approached a modern GPU.

The frequency scaling on your Athlon XP-M was copied from SpeedStep, which Intel invented years earlier (thanks for an addition to my list).

And the rest of your response is puerile whining.

Where's the great list of AMD innovations?

Dishonnesty ?..

CPU were used in US army jets but where not unveiled..

Mediagx had IGP , that it was not as good as current
IGPUs is just irrelevant in the debate..

Speedstep came with pentium3 while AMD s powerNow existed
since K6 2 , so much for your perception of history based on
wishfull thoughts.

 

[Exophase]

It does matter. It's completely different to have a vision, hire the right people to develop the tech and get the right product out of the door and do what AMD does, which is spot a nice company with a nice concept needing funding out there and inject money there.

AMD is very good on the later, but does almost nothing on the former. The entire list of AMD innovations weren't developed in-house, they came from AMD acquisitions. If you want to portray AMD as an innovating company, bring something developed inside AMD, not something they bought from others.

Ed: AMD does something akin to EA. Both acquires smaller companies need funding, get good first generation products with the acquired R&D and then screw up when they have to develop something by themselves.

Whatever, think what you want but I don't agree with you. Those acquisitions brought existing designs but they also brought engineers who could have left but didn't. Those engineers were AMD. You can't refer to AMD and not be referring to them.

The NexGen acquisition was not about AMD simply providing money. It was about them providing manufacturing and experience with x86 systems (eg Socket 7) and ISA (eg MMX). And the much maligned K5 was actually not that different in spirit than Nx586; there was clearly synergy in what they were both working on. K6 wouldn't have worked without the engineering collaboration of the two parties.

 

[Abwx]

So you're saying that bringing these features into x86 didn't require any R&D?

It does require way more good lawyers than good engineers....

As i said , it s well known in the industry that intel is good
at copying non in house innovations and adapt them in their
CPUs and plateforms...

Last exemple , Quickpath (to innovation...) , a carbon copy
of Hypertransport.

 

[AtenRa]

Since the subject is Intel versus AMD, yes, that's what I am talking about. Of course nearly everything was first invented on expensive mainframes. But most of that was put into x86 by Intel and copied by AMD. And most of the stuff that was truly new to microprocessors was also created by Intel, and copied by AMD.

I’m sorry but you are talking like a fanboy. If AMD copied them from Intel and Intel didn’t invent them, then Intel copied them from someone else. Sorry but most of what you wrote in the list above Intel copied them from someone else, they were not invented by Intel.

 

[mrmt]

The NexGen acquisition was not about AMD simply providing money. It was about them providing manufacturing and experience with x86 systems (eg Socket 7) and ISA (eg MMX). K6 wouldn't have worked without the engineering collaboration of the two parties.

It's completely different to push something from concept to product like Intel does than push from prototype to product, which is what AMD did with those "innovations". K7 brought a lot of things that were ready in DEC to AMD. It's not that there were engineering involved, but it was far less engineering involved than if AMD had to develop the concept and integrate it on their products. With NextGen, they marketed the entire processor because the native AMD development sucked big time.

It's the same thing with node development. Despite IBM having done the heavy lift with 32nm SOI, GLF still had to foot a 1.3 billion R&D bill. It's not that there was no GLF engineering involved, but the engineering had a far smaller scope than the engineering at IBM or Intel for that matter. They are not even comparable, not even in the same league.

The fact that AMD couldn't leverage on the strong line up it acquired after a few years down the road says a lot about the suitability of the environment for R&D there. AMD spent 10 years bolting and tweaking the K7 core to give us... Bulldozer.

 

[R0H1T]

I’m sorry but you are talking like a fanboy. If AMD copied them from Intel and Intel didn’t invent them, then Intel copied them from someone else. Sorry but most of what you wrote in the list above Intel copied them from someone else, they were not invented by Intel.

A fail argument is a fail argument is a fail argument ! You can't wake someone up who's pretending to be asleep, this is a common expression in my part of the world, so lets not waste any effort on doing such a thing !

 

[mrmt]

A fail argument is a fail argument is a fail argument ! You can't wake someone up who's pretending to be asleep, this is a common expression in my part of the world, so lets not waste any effort on doing such a thing !

Fail argument? Think harder.

When Intel - or AMD for that matter - brings something new to x86, they must not violate some patent out there or else they will have problems with the judge.

When AMD copies something from Intel, it must not pay anything or worry about anything, because AMD has access to Intel IP.

So bring something new to x86 isn't the same thing of AMD copying Intel IP.

 

[AtenRa]

AMD spent 10 years bolting and tweaking the K7 core to give us... Bulldozer.

By your metric, Intel spends 10x more on R&D than AMD and they were able to bring a pathetic overall 10% IPC in the span of the last three x86 generations (SB to IV to HW).

 

[AtenRa]

So bring something new to x86 isn't the same thing of AMD copying Intel IP.

Same applies to Intel copying from AMD :whiste:

 

[R0H1T]

It's completely different to push something from concept to product like Intel does than push from prototype to product, which is what AMD did with those "innovations". K7 brought a lot of things that were ready in DEC to AMD. It's not that there were engineering involved, but it was far less engineering involved than if AMD had to develop the concept and integrate it on their products. With NextGen, they marketed the entire processor because the native AMD development sucked big time.

It's the same thing with node development. Despite IBM having done the heavy lift with 32nm SOI, GLF still had to foot a 1.3 billion R&D bill. It's not that there was no GLF engineering involved, but the engineering had a far smaller scope than the engineering at IBM or Intel for that matter. They are not even comparable, not even in the same league.

The fact that AMD couldn't leverage on the strong line up it acquired after a few years down the road says a lot about the suitability of the environment for R&D there. AMD spent 10 years bolting and tweaking the K7 core to give us... Bulldozer.

In that case its even more laudable that AMD is pushing for stuff like HSA, which Intel will piggyback on, Unified memory, GDDR6(perhaps) with such a small budget ! Its disingenuous & condescending to throw away AMD's accomplishments by the wayside just because they acquired some other firms & aren't a monolith like Intel in the semiconductor market ! Its like saying Ford was the one first to market internal combustion engine so Fiat's common rail diesel doesn't count !