Where would CPUs be today without AMD in the late 90s to mid 2006

[Arachnotronic]

I'm pretty sure that from this list only x64 and HBM might be credited to AMD, the others AMD didn't pioneer. But what is impressive is that AMD managed to not make money in most products sporting these technologies.

It's funny that he credits low level GPU APIs to AMD when Apple was shipping Metal in final production form well ahead of Mantle.

 

[Fjodor2001]

I'm pretty sure that from this list only x64 and HBM might be credited to AMD, the others AMD didn't pioneer.

Which ones did they not?

But what is impressive is that AMD managed to not make money in most products sporting these technologies.

They did not? From 64-bit, multicore x86, console APUs, ...?

Also, regardless, the topic is regarding what technological progress we've seen driven by AMD. Not who profits the most.

 

[nismotigerwvu]

Intel wouldn't have come out with dual cores as quickly, but it would have happened sooner or later. The main reason (aside from competition from AMD) why they went down that route was because their prototype Tejas chips needed something ungodly like 200-250W to clock at speeds even comparable to Prescott, and were apparently even blowing out the VRM circuitry on test motherboards. Out of the options on the table, switching to Merom/Conroe would have been the least expensive option for Intel regardless.

In any case, without competition from AMD, chances are that the Core 2 Quad would have never existed. The main reason why Intel produced that so quickly was pretty obviously to get revenge on AMD for the three year ass-kicking it had sustained, so without the (theoretical) competition from Phenom, odds are they'd have waited for Nehalem to show up before going quad-core.

Tejas engineering samples were certainly tunning red hot, but not quite 200+watts hot.

http://anandtech.com/show/1217

Initial ES chips were 150 watts at 2.8ghz. This basically just means that the 90nm process wasn't likely to be used much for the chip and that the 65nm variant would have likely been the one to market in any meaningful capacity had it not been cancelled. Also, it could have been that the ES chips had borked power management and were just running full tilt 24/7. Or even more simple than that, it could have even been yield issues leading to higher than intended voltages being used to get the chips up and running. I think the bigger issue would have been the performance these chips provided. I don't remember any benchmarks coming out at the time and a quick Google search didn't turn any up, but if anyone has a link to any I'd love to see.

 

[MiddleOfTheRoad]

We would be using IA64 and be better off in all metrics.

He just made my year. That's going to be my new signature.

IA64 was such a massive and complete failure, Intel was forced to license /
implement its adversary's instruction set just to stay competitive. Probably
the most embarrassing moment of Intel's entire history.

Itanium flopped worse than Bulldozer.

 

[CHADBOGA]

Multi-core x86 revolution, 64-bit x86 CPU revolution,

There is nothing revolutionary bringing to x86, cpu architecture features that existed for ages on other architectures. D:

 

[MiddleOfTheRoad]

It's funny that he credits low level GPU APIs to AMD when Apple was shipping Metal in final production form well ahead of Mantle.

Well, considering that 3DFX was shipping metal nearly 20 years before both of them...... I'd say that low level GPU API credit really belongs to them.

https://en.wikipedia.org/wiki/Glide_API

 

[MiddleOfTheRoad]

x86 won due to investment level and legacy. This is also why x64 won the day it was created.

No, IA64 failed because of the incredibly dismal performance running the applications that businesses needed.

 

[Burpo]

Well, considering that 3DFX was shipping metal nearly 20 years before both of them...... I'd say that low level GPU API credit really belongs to them.

I remember the Voodoo days. Doom 2 & Unreal were the shiznit on 3DFX 🙂

 

[mrmt]

Which ones did they not?

First low level API wasn't AMD's, first multi-core APU wasn't AMD's, first APU wasn't AMD's, first ARM server chip wasn't AMD's. You seem to think that AMD is some kind of industry pioneer, a small underdog that dictates the future heading of the entire industry when in fact, AMD isn't much of a pioneer in anything.

They usually get from others what trends there are on the market, try to follow... and fail. The fact that they usually are present in developing a market trend doesn't say much about their competence in developing it, but the financial results they have from these products say a lot about their competence.

AMD had a few brilliant moves, like buying DEC IP in order to build K8 and Bobcat/Jaguar, which exploited a huge hole in Intel product line, but most of the time they have a kind of reverse Midas touch that no other company can surpass.

 

[Arachnotronic]

No, IA64 failed because of the incredibly dismal performance running the applications that businesses needed.

Not defending IA64, but from what I can tell, IA64 was sunk by sub-par micro-architecture and horrible x86 performance when it was needed.

 

[MiddleOfTheRoad]

Not defending IA64, but from what I can tell, IA64 was sunk by sub-par micro-architecture and horrible x86 performance when it was needed.

Yeah, I was referring to the awful performance when emulating x86 business apps.

 

[richaron]

Lol without AMD we'd be stuck with Intel trying to break 7GHz on single core CPUs 😛

AMD owned IPC in K6 and for freaking ages after that, my 500 was winning the "booting into half-life race" at lans over P3 550s. And there's a reason the XP series were marketed with an Intel equivalent number instead of clockspeed, my 1.6Ghz chip was sold as "1900+" because it literally had 15-20% better IPC.

AMD was also pushing dual cores and 64bit architecture when the Intel super friends were arguing there's no point "because software didn't support it" (sound familiar?).

I'd say Intel has learnt a lot from AMD (and vice versa of course). And we're all in a better world because of it.

 

[mrmt]

AMD was also pushing dual cores and 64bit architecture when the Intel super friends were arguing there's no point "because software didn't support it" (sound familiar?).

The point made at the time was that for client computers the 4GB limit wouldn't be challenged for a long time, as it wasn't.

 

[richaron]

The point made at the time was that for client computers the 4GB limit wouldn't be challenged for a long time, as it wasn't.

That was one point, on one of my points. And one could consider my quote as covering that already anyway.

AMD was also pushing dual cores and 64bit architecture when the Intel super friends were arguing there's no point "because software didn't support it" (sound familiar?).

Edit: i.e. the lack of needing >4GB is obviously reliant on software

 

[ShintaiDK]

Which ones did they not?

They did not? From 64-bit, multicore x86, console APUs, ...?

Also, regardless, the topic is regarding what technological progress we've seen driven by AMD. Not who profits the most.

None of the above. Unless you change 64bit to x64. And you can always advocate that no other company makes APUs than AMD.

 

[sandorski]

What would have happened on an alternate timeline is pure conjecture. On the one we have, AMD has been very influential and beneficial to PC Technology, the last 15-20 years.

 

[IntelUser2000]

It wouldn't have mattered too much ultimately.

IA64 - Sure, they wanted people to migrate to it, but even on the server it never got traction. Heck, Intel even admitted they had to neuter their other chips to make it look good! Their x86 chips perform far better. IA64 would have ONLY worked if they were able to make native performance 2-3x better than the *best* x86 implementation. That wasn't going to happen. They'd have tried to bring it to the PC for few years, and find out no one wants it. They'd have canned it, just a bit later than without AMD.

Netburst would have still ran into heat density issues. It's just unavoidable. They had to switch not just because of competition but because it wasn't feasible. People would have stopped buying it eventually. We might have seen Tejas at 150W, and performing barely better at 5GHz than Prescott at 4GHz.

You'd still see them struggling to enter mobile. If ANYTHING, AMD prepared Intel better for mobile, because imagine if 2006 was Netburst based Tejas rather than Core 2, and the Core 2 arch was in 2008. Rather than seeing Atom focus in 2010, it could have been in 2012. Atom Tablets wouldn't have seen its foot in the door.

Imagine a fledgling Intel that paraded Silvermont this year, sure it could have been a 14nm version but until now they'd have stuck with Silverthorne, blech. That IMO is the Intel without AMD.

 

[dark zero]

I'm pretty sure that from this list only x64 and HBM might be credited to AMD, the others AMD didn't pioneer. But what is impressive is that AMD managed to not make money in most products sporting these technologies.

AMD is like Linux: They don't make money, but leave a LOT of technologies for free. Is not bad.

When you goes full propietary HW/SW it will come at some market that is near impossible to enter... just look at MS or Intel and the Phone Market.

Both have a very hard time trying to enter to that market.

 

[Techhog]

We would be using IA64 and be better off in all metrics.

You got 4 cores because this is what the 99% crowd wants and OEMs. Its all about performance/watt and integration. Not because evil Intel is holding back.

Having more than 4 cores isn't innovation in any way. And as long as software isn't better than it is, then its useless for the masses. They would be better off with more cache, better IGP, DSP modules or PCH integration.

Prices would be the same as today. AMD have zero effect on these. And there is a lot more other MPU companies in this world than just AMD. Specially back then.

Ah, the tradition of Shintai being the first to post in every AMD thread...

Yes, because IA64 did so well in the server and HPC space, clearly showing off its advantages, so much so that it got... cancelled?

He's probably upset that it didn't increase the value of the shares he owns as much as he'd like.

 

[leper84]

How much has Intel really innovated or improved in the mid/high end pc market since Bulldozer came out/flopped? Marginally at best. Think that's your answer right there, Intel would have only improved as much as they were forced to.

They're only going to improve as much as they have to in order to keep making profit.

Maybe this is too much of a common person way of thinking, but say Zen comes out and is extremely competitive (however unlikely). I have a suspicion those TIM issues will disappear and we'll all see some amazing jump in IPC or clockspeed from Intel, versus the marginal increments and side-grades we've seen lately.

 

[Yuriman]

How much has Intel really innovated or improved in the mid/high end pc market since Bulldozer came out/flopped? Marginally at best. Think that's your answer right there, Intel would have only improved as much as they were forced to.

They're only going to improve as much as they have to in order to keep making profit.

Maybe this is too much of a common person way of thinking, but say Zen comes out and is extremely competitive (however unlikely). I have a suspicion those TIM issues will disappear and we'll all see some amazing jump in IPC or clockspeed from Intel, versus the marginal increments and side-grades we've seen lately.

High end what?

Laptops? They've come a long way.

Servers? They've also come a long way.

I don't think Intel sells much outside of those two groups, and probably wouldn't be too concerned even if they lost some of the desktop CPU market.

And let's not forget about the least common denominator of computing - Intel integrated graphics. It's in 99% of PCs, and what you get today is far and above what it was just 5 years ago.

I mean, yeah, I'm a desktop PC enthusiast, but even I'm starting to lose interest and I doubt performance gains are what this market is really lacking.

 

[Arachnotronic]

High end what?

Laptops? They've come a long way.

Servers? They've also come a long way.

I don't think Intel sells much outside of those two groups, and probably wouldn't be too concerned even if they lost some of the desktop CPU market.

And let's not forget about the least common denominator of computing - Intel integrated graphics. It's in 99% of PCs, and what you get today is far and above what it was just 5 years ago.

I mean, yeah, I'm a desktop PC enthusiast, but even I'm starting to lose interest and I doubt performance gains are what this market is really lacking.

Desktop chip biz is ~$10B for Intel, not a small chunk.

 

[ShintaiDK]

How much has Intel really innovated or improved in the mid/high end pc market since Bulldozer came out/flopped? Marginally at best. Think that's your answer right there, Intel would have only improved as much as they were forced to.

They're only going to improve as much as they have to in order to keep making profit.

Maybe this is too much of a common person way of thinking, but say Zen comes out and is extremely competitive (however unlikely). I have a suspicion those TIM issues will disappear and we'll all see some amazing jump in IPC or clockspeed from Intel, versus the marginal increments and side-grades we've seen lately.

The goal is performance/watt. Anyone stepping out of this have no chance to compete. If AMD ignores this again, then you can pretty much count Zen sales on your hand. Same reason why they lost so much on the GPU side.

 

[mrmt]

How much has Intel really innovated or improved in the mid/high end pc market since Bulldozer came out/flopped? Marginally at best. Think that's your answer right there, Intel would have only improved as much as they were forced to.

You should ask Oracle and IBM about it. Intel's marginally better at best processors are wreaking havoc on their big iron businesses.

Maybe this is too much of a common person way of thinking, but say Zen comes out and is extremely competitive (however unlikely). I have a suspicion those TIM issues will disappear and we'll all see some amazing jump in IPC or clockspeed from Intel, versus the marginal increments and side-grades we've seen lately.

Just in the next decade. Regardless of how Zen performs Intel line up is set in stone until the end of the decade, this for IC design, I guess the foundry roadmap is already set for the next 6-7 years.

 

[Dresdenboy]

It's funny that he credits low level GPU APIs to AMD when Apple was shipping Metal in final production form well ahead of Mantle.

Strange. Was Metal available in 2006? Fjodor did a mistake here. The argument is not about Mantle (as it isn't AMD's first API in this regard), but about CTM.