OPTIMIZER WANTED FOR A SIX CORE AMD PROCESSOR

ML1
ML1
Joined: 20 Feb 05
Posts: 347
Credit: 86,321,259
RAC: 196

RE: RE: RE: RE: I

Message 99854 in response to message 99852

Quote:
Quote:
Quote:
Quote:
I think the point is that Intel did not only not help AMD, which they have no obligation to do, but they actually worked to impede AMD's progress. THAT is what they did wrong!

OK, let's follow that then. ...

Looks like no "Dick Dastardly" is needed... Looks to me like Intel's dirty deeds have been done, all to the monopolistic detriment to progress and to all our cost:

Dirk Meyer: AMD is NOT Up For Sale!

... "AMD is not for sale, but we are happy to listen to any proposal which is in the interest to our shareholders," Chief Executive Dirk Meyer said...


Well hubris and flawed analogies aside, I'm slightly stunned that no one seems to have come up with an even simpler explanation that fits all of this : Intel attempted to dislodge it's biggest competitor that was getting an extended free ride ... so what? Does anyone believe AMD would have compensated Intel shareholders if they led the market? ...


You seem to have a very short memory ;-)

AMD led Intel by a long way for a long time for performance and design excellence. Remember the AMD K9 and K10 (K0A) cores that stole the show from Intel's silliness with chasing GHz with the Intel P4?

Now... Why isn't AMD the worlds biggest CPU manufacturer rather than Intel?...

I believe it came out in the various litigations that Intel strangled AMD by paying the big suppliers to NOT buy from AMD. I also believe that Intel took steps to restrict the foundry capacity available to AMD... Kinda like how Microsoft bundled IE with Windows to destroy Netscape...

A few years later and Intel are now copying AMD's design features. Intel are also using AMD's 64-bit instruction set!

And so, for all of Intel's design might... Why are Intel more expensive than AMD... I guess because they can afford to be (at our expense).

Note also that the design compromises for AMD assume a smaller available silicon area compared to Intel. That has meant that AMD has been far more innovative and far better at silicon utilisation than Intel where for Intel there's various examples where they just squander vast resources on ever bigger cache memory areas to gain some "turbo boost"!

Obviously, I favour the innovative and very effective designs from AMD. Shame they are being strangled and squashed underfoot from the dominant dinosaur.

Can you follow that? Or is anything and all fair in love and war and be damned for any cooperative alternatives?

Regards,
Martin

See new freedom: Mageia Linux
Take a look for yourself: Linux Format
The Future is what We all make IT (GPLv3)

ExtraTerrestrial Apes
ExtraTerrestria...
Joined: 10 Nov 04
Posts: 770
Credit: 545,644,022
RAC: 182,807

@Mike: "I'd propose this as a

@Mike: "I'd propose this as a constructive rather than a destructive solution ..."

Intel would never agree to this due to 2 reasons:
- with the compiler they'd give very important IP away
- once the compiler belongs to a different party (which serves all markets) it looses the reason for its existance from Intels point of view. It's created to provide optimal performance for Intel CPUs, better than anyone else because they know best how their CPUs work. And quicker than anyone else because they know first how their next CPU is going to work. Neither of this they could afford to give to a 3rd party.

So I think they'd rather withdraw the compiler than spin it off to an independent party. Not saying this is ever going to happen, though.

@Martin: you're right that the P4 was a bad design. But apart from this I can not leave your post commented.

- K8 (Athlon 64) was the one to beat the P4. K9 never made it out of the door (not sure if there ever were prototypes) and when K10 finally arrived it faced a tough uphill battle against Core 2. And didn't catch up to it until K10.5 in 45 nm (Phenom II), but by then we already had Nehalem / Core i7.

- "Why isn't AMD the worlds biggest CPU manufacturer rather than Intel?"
Because besides performance there are other qualities which count. There's software support (and the money you throw at it - e.g. see the compiler discussion and the lack of an AMD counterpart), long term steadiness, platform stability (who'd choose a VIA chipset over an Intel one?) and surely more. These are not the areas AMD traditionally excelled at. More often than not AMDs own short sighted business decisions are their worst enemy, or have been. They have change a bit to the better during recent years.

- "Intel strangled AMD by paying the big suppliers to NOT buy from AMD."
Think about it: if the AMD products were really that much better, smaller suppliers would jump on the bandwagon, crush their competition and become the new big suppliers themselves. Don't think there wouldn't be enough of them trying all the time.
Edit: of course there's been dirty business going on. But I'm convinced that it was a factor, but not a deciding one.

- "I also believe that Intel took steps to restrict the foundry capacity available to AMD"
Now that's rubbish. It's actually us who strangle AMD and limit their Fab capacity. The reason is simple: we may like AMD because they're the underdog and because we like cheap yet solid CPUs. However, what we don't like is giving them a lot of money for their CPUs. It's been years since the top AMD dekstop CPU has had an MSRP of over 300$. Their average selling price is so low that they've always been struggling to earn enough money to pay for development and new fabs (both costs billions).

- "A few years later and Intel are now copying AMD's design features. Intel are also using AMD's 64-bit instruction set!"
The 64 Bit extension was a good and long sigthed idea, whereas Intels 64 Bit plans would have involved getting much more money from us. Integrating the memory controller was also very good - but actually neccessary and obvious. The question was just when and Intel decided to do it later since the IMC would lock them into a memory technology.
But apart from that? The Core i3/5/7 design has been a major upgrade from the Core 2. And take a look at the innovations that go into Sandy Bridge. That's just 2 years after Nehalem / i7 and Intel's still got the upper hand (by far). Yet they advance their architecture so much. both, the change from Core 2 to Nehalem and from there to Sandy Bridge is more than AMD ever changed since the original K8 back in 2003.

- "Why are Intel more expensive than AMD... I guess because they can afford to be (at our expense)."
Of course.

- "Note also that the design compromises for AMD assume a smaller available silicon area compared to Intel."
Phenom I Quad 65 nm: 463 mil transistors, 285 mm²
Core 2 Quad 65 nm: 582 mil transistors, 286 mm²

Phenom II Quad 45 nm: 758 mil transistors, 258 mm²
Core 2 Quad 45 nm, 12 MB: 820 mil transistors, 214 mm²

Phenom II Hex 45 nm: 904 mil transistors, 346 mm²
Core i7 Lynnfield 45nm: 774 mil transistors, 296 mm²

You can't fault Intel for building smaller chips than AMD. And there's more: ever since the original Banias Pentium M (even Sandy Bridge is based on that, although heavily modified and extended) Intels design has been power optimized, and aggressively so since Nehalem. The original Athlon was not and this still shows - just compare notebooks of equal performance.

- "Can you follow that?"
Certainly not. I am very impressed by the improvements Intel made to Core 2 and Nehalem. There I sat, being very impressed by how great Conroe and the Core 2 architecture was, thinking that the only thing it was missing was an IMC and hyper threading. Boy was I wrong.. I honstely didn't expect that there was still so much untapped potential left. Especially after the long time of basically stalling performance per clock and core, ever since the Athlon 64 had been introduced.

AMDs upcomming Bulldozer does look very nice and will be a serious contender for the multi threaded integer performance crown. I do like the design, but it lacks the floating point punch that Sandy Bridge delivers, so t's not going to be "the" BOINC cruncher.

MrS

Scanning for our furry friends since Jan 2002

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6,550
Credit: 289,131,765
RAC: 92,862

RE: @Mike: "I'd propose

Message 99856 in response to message 99855

Quote:

@Mike: "I'd propose this as a constructive rather than a destructive solution ..."

Intel would never agree to this due to 2 reasons:
- with the compiler they'd give very important IP away
- once the compiler belongs to a different party (which serves all markets) it looses the reason for its existance from Intels point of view. It's created to provide optimal performance for Intel CPUs, better than anyone else because they know best how their CPUs work. And quicker than anyone else because they know first how their next CPU is going to work. Neither of this they could afford to give to a 3rd party.

So I think they'd rather withdraw the compiler than spin it off to an independent party. Not saying this is ever going to happen, though.


I was meaning as a requirement of a court order, me being the judge and suitably empowered. The AT&T breakup did much more than this proposal eg. their research arm, so it appears do-able.

@Martin: My message constructions are quite careful here. I try to avoid negative emotional hubris and stay neutrally phrased. I don't mind if you're an AMD fan, though I have to respectfully point out that your logic assumes in advance the very thing you claim to prove. But in the end that's your task to realise/examine/reconcile or not, so I'll leave that be. I'm not really siding with either company, or against you, my concern was a wider one ( of which the compiler judgment is an example ): the disturbing systemic tendency to wreck innovation with legal manouevring for the sake of greed, thus avoiding the hard yards of real invention. That AMD may have successfully clothed itself in the converse is a quite real possibility I reckon here. Indeed that is the real trick isn't it? Companies are not 'block devices' but have complex internals, mostly publicly hidden, and can change in actual character almost daily. So truthfully just because AMD, or Intel, does a 'good thing' once-upon-a-time ( which may be quite impressive for us personally ) doesn't mean they always will. This is the basic fault of barracking and fan-dom.

Of course I may be wrong about any of the matters discussed here. FWIW, this is my distant opinion. Like others I'll withdraw from the matter, but from lethargy alone. Nuff said .... :-)

Cheers, Mike.

I have made this letter longer than usual because I lack the time to make it shorter ...

... and my other CPU is a Ryzen 5950X :-) Blaise Pascal

ML1
ML1
Joined: 20 Feb 05
Posts: 347
Credit: 86,321,259
RAC: 196

RE: ... the disturbing

Message 99857 in response to message 99856

Quote:
... the disturbing systemic tendency to wreck innovation with legal manouevring for the sake of greed. ...

That is becoming apparent with the various patents wars that are unfolding. Especially for the case of the use of 'submarine' patents that are used as a 'trump' card many years down the line... The only people to win are the lawyers arguing the arguments.

MrS has given a good summary of one view of the AMD vs Intel history. (Yep, I was meaning "K8" rather than the mystical "K9".)

The real issue is that to suffer a monopoly or an effective monopoly for any part of industry is always bad for innovation and development, and always expensive for the customers.

Intel have produced some very good designs, as has AMD and others. However, the x86 CPU market is highly distorted by Intel being the one dominant player...

It will be very interesting to see how the "General Purpose" GPUs develop...

Regards,
Martin

See new freedom: Mageia Linux
Take a look for yourself: Linux Format
The Future is what We all make IT (GPLv3)

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.