And because of the whole Performance Rating thing, it may make more sense to compare on the basis of performance per dollar, rather than performance per Mhz.

Tetrium wrote:

Lets just agree that AMD and Intel steal stuff and ideas from each other 😀

The only reason AMD makes x86 chips today is because they stole from Intel and got away with it.

Scali wrote:

Tetrium wrote:

Lets just agree that AMD and Intel steal stuff and ideas from each other 😀

The only reason AMD makes x86 chips today is because they stole from Intel and got away with it.

I thought amd was the major player in x86 dev and Intel stole it from them. I recall reading that intel and amd worked together on x86 then split up or something.

Jade Falcon wrote:

And I'm amazed that no one said anything about the Am5x86.

Wasn't the Am5x86 more of an upgrade chip for people with older 486s. By Q4 95' when the 5x86 came out I would have thought there would have been focus on modern Pentium class machines with stable PCI implementation.

i486_inside wrote:

Jade Falcon wrote:

And I'm amazed that no one said anything about the Am5x86.

Wasn't the Am5x86 more of an upgrade chip for people with older 486s. By Q4 95' when the 5x86 came out I would have thought there would have been focus on modern Pentium class machines with stable PCI implementation.

Yes, but it was far better then any other 486 class cpu if I recall. Kind of how the k6-ii was better then any socket 7 intel cpu.

Jade Falcon wrote:

I thought amd was the major player in x86 dev and Intel stole it from them. I recall reading that intel and amd worked together on x86 then split up or something.

Of course not. Intel was the leading microchip company since way back when. They started with the 4004, the first successful mass-produced microprocessor. Which then evolved into the 4040, 8080, 8085 and eventually the 8086, which was the first CPU of what we now know as the x86 family.
IBM decided to use the 8088, a budget version of the 16-bit 8086, on an 8-bit bus, for their original PC.
IBM however didn't want to have any supply issues, so they forced Intel to license the 8088 architecture to second-source manufacturers. AMD was one of these manufacturers.
With the 80286 for the IBM AT, AMD once again was a second-source manufacturer.
By the time the 386 came, IBM was no longer relevant in the PC world. There were tons of clone builders to sell their x86 CPUs to, so Intel no longer had to dance to IBM's tune. Hence, the 386 was never sub-licensed to any other manufacturers. This effectively kicked AMD and the other second-source manufacturers out of the x86 world. Up to now, they were happily freeloading on Intel's x86 designs and the success of IBM's PC.

So, AMD decided to just reverse-engineer the 386 and launch their own carbon-copy Am386 CPU. AMD thought they could spin the second-source license in a way that they had the rights to manufacture ANY Intel x86 CPU, even ones that Intel never supplied the designs for, but which they reverse-engineered.
AMD was found guilty of copyright-infringement on the 386 microcode.
However, the court also decided, in a form of anti-trust, that x86 had become so relevant to the PC world, that it would be bad to have only a single manufacturer with a monopoly on the x86. So they forced Intel to license the x86 to others. Which led to the x86 cross-license deal with AMD and a few other manufacturers.
So AMD basically got away with reverse-engineering the 386 (and by then also the 486) and copying the microcode.
Having an x86 license is invaluable. Nearly every other CPU manufacturer became extinct, only x86 and ARM are still relevant today. AMD would be nowhere near the size they are today, if they even survived at all, if it wasn't for that x86 license.

Note also that especially in those days, it was very unusual to license CPU architectures to other companies (especially direct competitors). It was common for every CPU manufacturer to have their own proprietary CPU architecture and ecosystem.
These days it's somewhat more common... but still, could you imagine nVidia having to license AMD to be able to make Pascal-based GPUs, or AMD having to license nVidia to make GCN-based GPUs? Because that's what we're talking about here. We're not just talking about generic patents related to microchip design and manufacturing, we're talking about the actual instructionset, which allows AMD to run all software originally written for Intel CPUs, with 100% compatibility. That would be like being able to run nVidia's GeForce drivers on your AMD GPU, and getting all their game optimizations etc for free.

Jade Falcon wrote:

Yes, but it was far better then any other 486 class cpu if I recall.

The 5x86 was clocked very high, other than that it wasn't very different from a vanilla 486. It had 16 kb of L1 cache, where most 486 had 8 kb.

Wasn't there something the two worked on that Intel backed out of and Amd sued for back in the late 70's early 80's ?

Scali wrote:

Of course not. Intel was the leading microchip company since way back when. They started with the 4004, the first successful mas […]
Show full quote

Of course not. Intel was the leading microchip company since way back when. They started with the 4004, the first successful mass-produced microprocessor. Which then evolved into the 4040, 8080, 8085 and eventually the 8086, which was the first CPU of what we now know as the x86 family.
IBM decided to use the 8088, a budget version of the 16-bit 8086, on an 8-bit bus, for their original PC.
IBM however didn't want to have any supply issues, so they forced Intel to license the 8088 architecture to second-source manufacturers. AMD was one of these manufacturers.
With the 80286 for the IBM AT, AMD once again was a second-source manufacturer.
By the time the 386 came, IBM was no longer relevant in the PC world. There were tons of clone builders to sell their x86 CPUs to, so Intel no longer had to dance to IBM's tune. Hence, the 386 was never sub-licensed to any other manufacturers. This effectively kicked AMD and the other second-source manufacturers out of the x86 world. Up to now, they were happily freeloading on Intel's x86 designs and the success of IBM's PC.

So you don't think that the second sourcering of the 8088 and 80286 actually made the x86 and the IBM PC standard a success? I also guess the licence for 8088 and 80286 weren't free.

Kamerat wrote:

So you don't think that the second sourcering of the 8088 and 80286 actually made the x86 and the IBM PC standard a success?

No, Intel would have managed just fine on their own. They're Chipzilla now, they would only have grown even faster and larger if they didn't have to share with the second-source manufacturers.

Kamerat wrote:

I also guess the licence for 8088 and 80286 weren't free.

I'm not sure if the manufacturers actually had to pay anything upfront or if they just had to pay a percentage of every sold unit.
Either way, given the huge success of x86, it's pretty much a no-brainer that they got a great deal... They had 0 development costs (where Intel had invested billions into developing the x86).
If you just look at where AMD is. They're a huge player in the CPU market, probably second only to Intel. And AMD has never had any success with any of their own CPUs. The x86 license propelled AMD from zero to hero quite literally (in the early years, AMD wasn't the biggest second-source, and they weren't even the most successful x86 clone builder initially... Companies like NEC, IBM, Cyrix and TI initially offered better x86 clones than AMD).

Scali wrote:

No, Intel would have managed just fine on their own. They're Chipzilla now, they would only have grown even faster and larger if they didn't have to share with the second-source manufacturers.

Yeah we've seen how well Intel innovates when there is no competition to drive them.

Like, the last five years have been amazing.

Take Kabylake for example. A prime example of Intel innovation.

...

appiah4 wrote:

Yeah we've seen how well Intel innovates when there is no competition to drive them.

We're talking the period of 1981-1991 though... As in the period before AMD became an independent x86 vendor.
That's when Intel became what it is today.
They developed the 8086, 186, 286, 386 and 486 before anyone ever really competed with them in the x86 market.
Heck, the first two products from AMD as an independent x86 vendor were carbon-copies of Intel's 386 and 486 chips. Who's the non-innovative company here?

The first time Intel was actively pushed because of competition was when the K7 came around. Which was way after this period.

I remember my first Windows XP machine was built around an Athlon XP 1800+ (Palomino) because I had read that the Athlon XP stomped the crap out of Pentium IIIs and early Pentium 4s at the time. And while my graphics cards were always a bit lacking, the CPU always performed at top-notch.... until I upgraded to XP SP2 and SP3...then my 256MB of RAM was woefully inadequate.

I have since rebuilt that PC as a retro gaming build. And while the Radeon 9550xl is STILL crap, the 2GB of DDR-400 I stuck in there and the upgraded Athlon XP 3200+ (Barton) are super-beast with Windows XP SP3. I think it's all due to the 2GB of RAM. I have 2GB of DDR2-533 in my gaming laptop with a Geforce 7800GTX PCI-E video card and that thing is a monster beast at 2006-gaming.

Scali wrote:

gdjacobs wrote:

Considering where much of the talent came from, I think i there was more replication of DEC Alpha technology, starting from the EV6 bus and through the operating units.

No.
It's talking DEC Alpha's operating units and putting then in a P6-like x86 microarchitecture, on an EV6-like bus. So it's like building an x86 on top of a RISC backend, which is basically what P6 also is. They also did it in virtually the same way (roughly the same configuration of pipeline in number of stages, number and type of different processing units etc... main difference is 3 ALUs instead of 2 ALUs, and the FPU being 3-way pipelined, which is where they got their slight gains over the P6, along with the faster bus).

This was an idea introduced in the K5 architecture which featured x86 decoded to and run on an AMD 29k RISC backend. I suspect much of the efficiency of the RISC backend on the K7 was due to the expertise of some of the talent AMD picked up as teams from DEC and such were divested from their original organizations.

Jade Falcon wrote:

One thing most people miss that that there is more to a cpu then raw speed and performance.
lower power draw is a big thing in mobile systems, lower cost is a big deal in budget system too. It is not always about speed, the Duron was a very well remarked cpu in the day do in part to its low cost and overclocking ability.

And I'm amazed that no one said anything about the Am5x86.

You happen to know how Thunderbird and Spitfire compared to Coppermine and the Coppermine Celerons?
I completely overlooked the Am5x86. I'd say it's somewhat about half the MHz of a Pentium.
I don't know about K5, but K6 was something like roughly about the same as a Pentium, the K6-2 should be slightly faster (partially due to the higher FSB) but slower compared to P6.

And a Phenom II 3.2GHz is in about the same range as a Q6600 (running at 2.4GHz).

But again, please correct me if I'm wrong.

gdjacobs wrote:

This was an idea introduced in the K5 architecture which featured x86 decoded to and run on an AMD 29k RISC backend.

Motorola did it even before then, on the 68060. Besides, the P6 followed the introduction of the K5 way too quickly to have lifted any ideas from it.
Please, the world is not x86 only.
Also, the RISC backend is not the hard part of making a fast x86 CPU. The decoding and scheduling of instructions is. The actual ALUs were basically a solved problem by the time of the 486/Pentium: most instructions could complete in 1 cycle. Bascially the same performance level as we still have today.

Scali wrote:

gdjacobs wrote:

This was an idea introduced in the K5 architecture which featured x86 decoded to and run on an AMD 29k RISC backend.

Motorola did it even before then, on the 68060. Besides, the P6 followed the introduction of the K5 way too quickly to have lifted any ideas from it.
Please, the world is not x86 only.

Perhaps Intel stole some ideas from K5 even when it was still in development? So it's very well possible to steal ideas from something like that. Not having been introduced =/ not possible to steal ideas from it 😀

Tetrium wrote:

Perhaps Intel stole some ideas from K5 even when it was still in development? So it's very well possible to steal ideas from something like that. Not having been introduced =/ not possible to steal ideas from it 😀

Not impossible, just not very likely.
I mean, put things in perspective... The K5 is a rather insignificant, low-end CPU, which barely even competed with the 'oldskool' x86 Pentium architecture. It was in fact so unsuccessful that AMD's next architecture was not even based on it, but instead they acquired NexGen (which, although better than the K5, was still no match for the P6 architecture).
The P6 on the other hand was a revolutionary architecture, and is basically the blueprint for x86 as we still know it today: deeply pipelined, and deep out-of-order execution.
If anything, it looks like AMD scrambled to re-use their ill-fated 29000 design ('strap an x86 decoder on it'), rather than designing a proper CPU from the ground up, like Intel did.

Have they both lifted the idea from the 68060? Again, unlikely. Motorola probably wasn't the first to think of it either. They were just the first to get a CPU with this technology to market, so we haven't heard about other ideas or even internal experiments in research labs or wherever.
The only thing we know for sure is that since the 68060 was on the market a few years before the K5, that the ideas in the K5 weren't new to anyone anymore, least of all Intel.

NexGen had also been running RISC cores in X86 compatible chips for a long time. The P6 made it to the market the year before the K5. The K5 was for the most part a flop due to under performance especially floating point performance. The modeling number system was also heavily criticized. After this failure AMD actually acquired NexGen and the nx686 to become the k6. It's interesting people mentioned the 060 as I always liked this chip and used them in video editing systems but sadly it never had the success of the early 68k chips as the architecture was abandoned in favour of PPC. It's also good to see Alpha mentioned another great platform that died a death far to soon.

Bobolaf wrote:

The K5 was for the most part a flop due to under performance especially floating point performance.

The K5 also failed to scale to high clockspeeds. The main reason for Intel to take such a revolutionary step with the P6 is that the complexity of CISC was making it very difficult to scale beyond 200 MHz with the technology of the day, while RISC-like CPUs were quickly going 500+ MHz.
P6's advanced out-of-order architecture could scale much higher while maintaining the same clock-for-clock performance of a Pentium.
And indeed the P6 architecture went virtually unchanged from Pentium Pro to PII to PIII, going from 200 to 1400 MHz over the years, to show just how big a leap P6 was, and how well that goal was achieved.
And even today, the Core architecture is not too different from P6, and scales to 4+ GHz.

Bobolaf wrote:

It's interesting people mentioned the 060 as I always liked this chip and used them in video editing systems but sadly it never had the success of the early 68k chips as the architecture was abandoned in favour of PPC.

The 68060 actually became somewhat of a standard in the demoscene.
The Amiga 1200 has a trapdoor expansion slot, which allowed for CPU accelerators. The machine shipped with a simple 68EC020 at 14 MHz. Over time, 030, 040 and 060 accelerators became available.
The demoscene eventually settled on the fastest 060 boards available (around 50-60 MHz) in the late 90s, and this has been the most popular Amiga platform aside from the original Amiga 500/2000 spec ever since. Hundreds of demos have been made for it over the years, stuff like this: https://youtu.be/W5_NMxW5UfE