Sign in with
Sign up | Sign in
Your question

How did AMD got the SSE2?

Last response: in CPUs
Share
June 15, 2004 9:47:17 PM

How did AMD got the Intel SSE2 and the SSE instruction set?

I mean, if AMD can get those sets, why can't Intel get those 3D now sets?

And why AMD didn't develop a new instruction set for their 64 series?

More about : amd sse2

a b à CPUs
June 15, 2004 9:59:25 PM

Cross liscenseing. Intel could use 3d now but they choose not to.

I aint signing nothing!!!
June 15, 2004 10:13:29 PM

What's cross licensing? Did AMD pay Intel for the SSE2?
Related resources
June 15, 2004 10:21:00 PM

No, not specifically, but they did pay intel for the right to make x86 chips, as well as use all intel's x86 extentions as a result of a lawsuit many years ago where intel tried to revoke AMD's right to produce x86 compatible chips.

BTW, its not really a cross licence, its really a x86+extention licence AFAIK, . For instance, AMD can not just implement HT the same way intel did; amd and intel do not share each others patents. They just share each others x86 instruction set implementations/extentions, which for instance allows intel to implement AMD64, but thats it.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
June 15, 2004 11:00:27 PM

Quote:
And why AMD didn't develop a new instruction set for their 64 series?

Acutally the "64bit" support is an extension of x86 32bit. It's why AMD64 is 100% compatible with all the actual x86 software. AMD extended these instructions with 64bit adressing capabilities, etc...

--
It's tricky to use words like <b><font color=green>AMD</font color=green></b> or <b><font color=blue>Intel</font color=blue></b> in a signature some users could think your are biased.
June 16, 2004 2:44:22 AM

LOL! AMD must have great lawyers! That cross license was for the 8086 and 8087 coprocessor chip dating back to the 1970's. It was before PCs existed when there were just the chips.

<b>56K, slow and steady does not win the race on internet!</b>
June 16, 2004 5:28:46 AM

AMD was granted the cross-liscensing capabilities for reasons of compatibility. They didn't want the market flooded with 20 different extensions each not compatible with the other.

"We are Microsoft, resistance is futile." - Bill Gates, 2015.
June 16, 2004 8:54:13 AM

>that cross license was for the 8086 and 8087 coprocessor
>chip dating back to the 1970's.

Like I said, I don't think they have cross licence. Up until (I think) the 386, AMD actually got their masksets from intel, they just fabbed the thing as a second supplier to IBM, for the 486 they where required to design their own microcode (or rather, reverse engineer Intels :)  and after that, design their own cpu's, and even chipsets/busses after the Pentium 1. They only thing "shared" now is the instruction set.

If you look at it in this context, AMD has actually done incredibly well in a relative short ammount of time:
* 286/386/486 just cloning intel designs
* K5 was their first own x86 design, wasn't that terrible, but mostly much too late and didn't clock high enough.
* K6 was far superior to the P5, and not far behind the Pentium Pro/P2/P3, except in FP and chipset/fsb support. Granted, K6 was mostly a nextgen design, but the core was competitive on integer workloads, and that was against the best core intel ever designed
* K7 was without a doubt the best x86 design around at its time. It totally owned the intel cores (P3/willamette), and definately corrected K6's biggest weakness: FP. But it wasn't followed up fast enough by the:
* K8 which was a few years late again, but even so further extended AMD's lead.

On the chipset/fsb/ side of things, over the same time, it went from pin compatible (k6), piggy backing pentium chipsets, to using its own (licenced) FSB (EV6) which was better than intels, to using a whole new architecture (hypertranspot and ODMC) which currently defines state of art.

On the ISA front, they went from cloning x86 and its extentions (like MMX) to designing the first SIMD extentions (3DNow) to redefining the ISA (AMD64).

In short, they went from zero to hero..

= The views stated herein are my personal views, and not necessarily the views of my wife. =
June 16, 2004 10:51:18 AM

Quote:
I mean, if AMD can get those sets, why can't Intel get those 3D now sets?

Because 3DNow! is worthless if you have SSE. It's inferior.
June 16, 2004 3:26:50 PM

Quote:
Like I said, I don't think they have cross licence. Up until (I think) the 386, AMD actually got their masksets from intel, they just fabbed the thing as a second supplier to IBM, for the 486 they where required to design their own microcode (or rather, reverse engineer Intels :)  and after that, design their own cpu's, and even chipsets/busses after the Pentium 1. They only thing "shared" now is the instruction set.

AMD would have been in bad shape if Intel had been successful in preventing AMD from using the x86 and FP87 instruction set. I still maintain that this is due to a cross technology licence (circa 1978).

As you said Intel subcontracted AMD to produce early chips, up to 80286 and 80287. AMD leveraged the agreements to make 386 clones, which of course were not subcontracted by Intel. AMD has been using the instruction sets ever since despite Intel patenting all subseqent technologies and copyrighting their trademarks.

I believe the first court challenge occured during the 286 and 386 days.

I have no idea how AMD finagled the 32-bit instruction set, MMX, SSE, SSE2, which were not part of that original contract.

Quote:
If you look at it in this context, AMD has actually done incredibly well in a relative short ammount of time:
* 286/386/486 just cloning intel designs
* K5 was their first own x86 design, wasn't that terrible, but mostly much too late and didn't clock high enough.
* K6 was far superior to the P5, and not far behind the Pentium Pro/P2/P3, except in FP and chipset/fsb support. Granted, K6 was mostly a nextgen design, but the core was competitive on integer workloads, and that was against the best core intel ever designed
* K7 was without a doubt the best x86 design around at its time. It totally owned the intel cores (P3/willamette), and definately corrected K6's biggest weakness: FP. But it wasn't followed up fast enough by the:

...just a quick thought here. I think Tualatin (I never spell that right) might have been the best core of the day but set aside, so Intel could ramp up Willamette (yuck).
Quote:
* K8 which was a few years late again, but even so further extended AMD's lead.

On the chipset/fsb/ side of things, over the same time, it went from pin compatible (k6), piggy backing pentium chipsets, to using its own (licenced) FSB (EV6) which was better than intels, to using a whole new architecture (hypertranspot and ODMC) which currently defines state of art.

On the ISA front, they went from cloning x86 and its extentions (like MMX) to designing the first SIMD extentions (3DNow) to redefining the ISA (AMD64)

Granted it must have required a lot of work for AMD to reverse engineer the processors starting with 80386 not to mention later utilizing their own new technologies. However they still capatilized on Intel's earlier (and very expensive) R&D and Marketing.

By the way, didn't AMD start innovating with the DX4-100 with the 4X multiplier?

Funny, I don't recall K6 being an effective challenger to the Intel counterpart. I may have (and probably do have) a distorted memory but I seem to remember K5 and K6 both continually trailing Intel on the clockspeed front by six to nine months. That and weak floating point units meant the processors were insignificant competition, except in the mobile department. In any case, none of the AMD processors ended up in any high-end OEM machines.

Despite performance lead changing hands a few times, AMD works for me. I've mostly used AMD processors over the years due to lower cost (to me), sometimes sacrificing performance sometimes not. Lower pricing and the availability of heavy discounts to the individual is what made AMD processors attractive to me. It still does.


<b>56K, slow and steady does not win the race on internet!</b>
June 16, 2004 3:53:17 PM

Quote:
How did AMD got the Intel SSE2 and the SSE instruction set?

You see, AMD was smart enough to hire an elite squad of mutant ninja carp that resulted from a top secret military project to feed the homeless with mind-altering substances to create an army to run Walmart and McDonalds if ever a draft was needed to call everyone else into military service. Armed with Fabreeze the ninja carp broke into Intel's secret underground laboratory and stole Intel trade secrets without leaving a trace of their actions. (Because that's what ninjas do.) And AMD implemented them into their products just because they were tired of answering all of the tech support phone calls asking when they were going to implement SSE.

Quote:
I mean, if AMD can get those sets, why can't Intel get those 3D now sets?

Oh, Intel tried. They sent in their own squad of 4691 irradiated haggis (god help anyone who recognizes that reference) disguised as lunch into AMD's caffeteria and managed to psychicly steal 3DNow from the minds of AMD's employees. Much to Intel's chagrin though, the raw power of the 3DNow sets was so beyond the abilities of a Pentium to perform that Intel had to abandon the project to include these into the Pentiums, lest it create a transdimentional vortex that would allow Satan to steal their souls.

Quote:
And why AMD didn't develop a new instruction set for their 64 series?

You know, I think that they tried. They really really did. But in the end all that they could come up with was the useless AMD64 instruction set, and who the hell wants to base a CPU on <i>that</i>?

<pre><b><font color=red>"Build a man a fire and he's warm for the rest of the evening.
Set a man on fire and he's warm for the rest of his life." - Steve Taylor</font color=red></b></pre><p>
June 16, 2004 4:16:18 PM

>As you said Intel subcontracted AMD to produce early chips,
>up to 80286 and 80287. AMD leveraged the agreements to make
>386 clones, which of course were not subcontracted by
>Intel. AMD has been using the instruction sets ever since
>despite Intel patenting all subseqent technologies and
>copyrighting their trademarks.

I don't know the details of the licencing either, in fact, i don't think they where made public at all, but I assure you there is no broad cross licence between AMD and intel. AMD can not use any of intel's patents, whether it is about architectural implementation tricks, process technology, software, FSB infrastructure, nothing. The opposite is just as true. Only a court order or mutual agreement between Intel and AMD assured that the <i>instruction set</i> and its extentions could be used by both (even though I think AMD pays intel money for it).

>I have no idea how AMD finagled the 32-bit instruction set,
>MMX, SSE, SSE2, which were not part of that original
>contract.

The agreement has already been extended twice AFAIK. Its always a 10 year agreement.

>...just a quick thought here. I think Tualatin (I never
>spell that right) might have been the best core of the day

Well, intel is/was ahead on process engineering, but I'm not sure a 130nm tualatin would have remained competitive with 130nm Athlons cores given similar processing technology. You really think it would have scaled to 2.3 GHz like Athlon ?

>However they still capatilized on Intel's earlier (and very
>expensive) R&D and Marketing.

I think you'd be surprised how cheap the 8088/286/386's where to develop when you compare it with a modern cpu. But yes, they obviously capatilized on it nevertheless. BUt I think IBM is more to credit (or rather, blame) for making x86 a standard in the first place. As for piggy backing intels marketing, I don't think so. intel didn't really start marketing itselve until the 3/486 days with their "intel inside" campaings, and I don't see how AMD profited from that. "Intel Inside" and subsequent Pentium branding efforts has done some tremendous damage AMD still suffers from.

>By the way, didn't AMD start innovating with the DX4-100
>with the 4X multiplier?

That was IBM afair with their "blue lightening" 486, they where the first to introduce a multiplier. Actually Cyrix was much more innovating back then, or a bit later with their 6x86 chips, even though HOT, borderline overclocked and requiring overclocked chipsets (83 MHz FSB) where very much ahead of intel chips from an architectural POV. Have you seen Bob Colwell's presentation ? I remember a quote from him saying something like "Cyrix where really usefull back then, they would go out and implement all sorts of advanced techniques, and we got to look at it and say:'hey look, that actually works'. we didn't know that". Bob Colwell is ex-intel fellow, lead architect of PPro in case you didnt know.

>Funny, I don't recall K6 being an effective challenger to
>the Intel counterpart.

The K6 was the first time AMD ever stole the performance crown from intel, even if only for a week or so :p  It was released at up 233 (or 266?) MHz speeds when Pentiums topped out at 200 (or 233). About a week later intel intro-ed the Pentium 2 though and set the record staight even though the 300 Mhz chips where a pure paper launch (yes, even back then) with a $2.000 price tag.

I also seem to remember reading then that one of Intel high brass had sworn that would have been the last time AMD (or any other x86 contender) would outclock intels fastest chips. the athlon must have hurt, I hope he doesn't burn in hell for it ;) 

= The views stated herein are my personal views, and not necessarily the views of my wife. =
June 16, 2004 6:13:02 PM

>As you said Intel subcontracted AMD to produce early chips,
>up to 80286 and 80287. AMD leveraged the agreements to make
>386 clones, which of course were not subcontracted by
>Intel. AMD has been using the instruction sets ever since
>despite Intel patenting all subseqent technologies and
>copyrighting their trademarks.

I don't know the details of the licencing either, in fact, i don't think they where made public at all, but I assure you there is no broad cross licence between AMD and intel. AMD can not use any of intel's patents, whether it is about architectural implementation tricks, process technology, software, FSB infrastructure, nothing. The opposite is just as true. Only a court order or mutual agreement between Intel and AMD assured that the instruction set and its extentions could be used by both (even though I think AMD pays intel money for it).


<font color=blue>You are right. Today "Licensing agreement" is a more accurate term of the arrangement. However, originally there was a cross technology agreement. (I'll look for it in one of the companies' history). I'm not sure how/why/when it went wrong but Intel did sue for infringement.

I'd love to see how much AMD is paying for a license. Perhaps Intel is laughing all the way to the bank.</font color=blue>

>I have no idea how AMD finagled the 32-bit instruction set,
>MMX, SSE, SSE2, which were not part of that original
>contract.

The agreement has already been extended twice AFAIK. Its always a 10 year agreement.

>...just a quick thought here. I think Tualatin (I never
>spell that right) might have been the best core of the day

Well, intel is/was ahead on process engineering, but I'm not sure a 130nm tualatin would have remained competitive with 130nm Athlons cores given similar processing technology. You really think it would have scaled to 2.3 GHz like Athlon ?

<font color=blue>No I don't believe it would scale as far but I think Intel had the ability to bring Tualatin to the public before AMD could bring Tbred.

We could have had a year of Tualatin instead of Willamette (just speculation).</font color=blue>

>However they still capatilized on Intel's earlier (and very
>expensive) R&D and Marketing.

I think you'd be surprised how cheap the 8088/286/386's where to develop when you compare it with a modern cpu.

<font color=blue>I might be surprised if I knew how they all cost to develope, but I don't. LOL</font color=blue>

But yes, they obviously capatilized on it nevertheless. BUt I think IBM is more to credit (or rather, blame) for making x86 a standard in the first place.

<font color=blue>Isn't that the ultimate marketing goal of any product? To make yours the standard?</font color=blue>

As for piggy backing intels marketing, I don't think so. intel didn't really start marketing itselve until the 3/486 days with their "intel inside" campaings, and I don't see how AMD profited from that. "Intel Inside" and subsequent Pentium branding efforts has done some tremendous damage AMD still suffers from.

<font color=blue>True but Intel didn't have to do the marketing. IBM and Microsoft were doing it all.

AMD was marketing clone chips. It's only when AMD threatened to take a big enough share that Intel started the famous campaign.

Do you really think AMD would have had any market at all if they did not have the x86 instruction set?

What chance do you think they would have had with their own unique processor?

Could they have competed against Intel on technology alone? They were doing well with Flash memory technology but a new ground-up CPU would be a hard sell.

IMO, the "Intel Inside" campaign was more an anti-AMD stunt than it was a marketing campaign for the merits of Intel products.

The demand was for PCs. AMD did what it had to do, fit it's chips into that market.

The Intel Inside campaign capitalized on FUD (fear, uncertainty, doubt). Without saying so, without stating any superior qualities of Intel CPUs (real or not), Intel implanted in the minds of some consumers that there was something wrong with AMD processors. They didn't have to prove which processor performed better all they had to say was "look for the 'Intel Inside'".

Brilliant in my opinion and not a very expensive campaign considering what the return probably was.</font color=blue>

>By the way, didn't AMD start innovating with the DX4-100
>with the 4X multiplier?

That was IBM afair with their "blue lightening" 486, they where the first to introduce a multiplier. Actually Cyrix was much more innovating back then, or a bit later with their 6x86 chips, even though HOT, borderline overclocked and requiring overclocked chipsets (83 MHz FSB) where very much ahead of intel chips from an architectural POV. Have you seen Bob Colwell's presentation ? I remember a quote from him saying something like "Cyrix where really usefull back then, they would go out and implement all sorts of advanced techniques, and we got to look at it and say:'hey look, that actually works'. we didn't know that". Bob Colwell is ex-intel fellow, lead architect of PPro in case you didnt know.

>Funny, I don't recall K6 being an effective challenger to
>the Intel counterpart.

The K6 was the first time AMD ever stole the performance crown from intel, even if only for a week or so :p  It was released at up 233 (or 266?) MHz speeds when Pentiums topped out at 200 (or 233). About a week later intel intro-ed the Pentium 2 though and set the record staight even though the 300 Mhz chips where a pure paper launch (yes, even back then) with a $2.000 price tag.

<font color=blue>I forgot about the K6-233 and the others. I don't think I ever knew the facts in the first place. Thanks for the info!

I didn't and will never believe that any of the K6/K6-2/K6-III lines were ever a match for PII. (Did I mention K6 and K6-2's were cheap, though? <b>LOL</b>

I stuck with a Socket 7 motherboard longer than I wanted because PII and Slot 1 was expensive)</font color=blue>

I also seem to remember reading then that one of Intel high brass had sworn that would have been the last time AMD (or any other x86 contender) would outclock intels fastest chips. the athlon must have hurt, I hope he doesn't burn in hell for it ;) 

<font color=blue>Give him a Mendocino for eternity! LMAO!!!</font color=blue>

<b>56K, slow and steady does not win the race on internet!</b>
June 16, 2004 7:00:11 PM

AFAIK (pieced together from a reprinted <i>Microprocessor Report</i> article, and reading <i>Upgrading and Repairing PCs</i>, AMD originally made Intel clones b/c the united states military was buying large shipments from Intel, and at that point the military always wanted 2 suppliers to insure they weren't paying inflated prices. Then, AMD just started reverse engineering Intel's processors (I think AMD may have created the "White Box" reverse engineering technique that is lawsuit-proof.). Then AMD just made their own procs that supported all of Intel's instructions. Now, Intel and AMD have an agreement that basically amounts to not suing each other for copyright infringement. And Intel, in a startling change, reverse-engineered the AMD64 instruction set for their own 64-bit procs(although there are a total of, I think, 2 or 3 instructions that are different for the different extensions).

"If I owned this place and hell, I'd rent out this place and live in hell" - Toombs
June 16, 2004 7:12:18 PM

>I'd love to see how much AMD is paying for a license.
>Perhaps Intel is laughing all the way to the bank.

I don't think its a stellar ammount, the agreement really is in both companies interest.

>No I don't believe it would scale as far but I think Intel
>had the ability to bring Tualatin to the public before AMD
>could bring Tbred.

Oh yes, intel has always been ahead on new process nodes. that is not surprising, intel has god knows how many fabs, AMD just one really, so intel starts early with one fab, irons out any issues, and then slowly rolls it out across it s other fabs. AMD otoh migrates more in a big bang style. Typically 6-12 months later, but much faster. Anyway, my original point was that the design of the Athlon core was vastly superior over the ppro/2/3, not that the resulting products where always better.

>I might be surprised if I knew how they all cost to
>develope, but I don't. LOL

A lot. To give you an idea, it is estimated intel and HP spent around $5 billion on dollar on developping Itanium initially, and intel currently spends $4-$500 million per year just on ongoing development cost. I would be extremely surprised if the 8088 and 8086 development combined costed more than 5 million inflation corrected dollars, could be even significantly less.

>True but Intel didn't have to do the marketing. IBM and
>Microsoft were doing it all.

definately not true. If intel's brandname and its pentium trademark is now almost as universaly known as Coca Cola or Ford, that is all to intel's credit. They designed the marketing campaigns and convinced (read: paid) the OEM's and ISV's to put on the "intel inside" and "pentium optimised" stickers, no one else did it for them.

> It's only when AMD threatened to take a big enough share
>that Intel started the famous campaign.

I have a different theory on that, I don't think AMD had much to do with their reasoning. The campaign was not specifically directed against AMD, it was designed to establish intel's name brand as a world leader, a status which would allow it to expand beyond the PC CPU market. I think Intel had Motorolla, IBM, DEC, TI and HP in its sight, more than AMD.

>I didn't and will never believe that any of the
>K6/K6-2/K6-III lines were ever a match for PII.

The cores where definately a match, but they lacked the superior memory bandwith and chipsets of intels 440LX/BX chips, as well as floating point performance. Have a look here:

<A HREF="http://www.anandtech.com/cpu/showdoc.html?i=902&p=12" target="_new">http://www.anandtech.com/cpu/showdoc.html?i=902&p=12&lt;/A>

K6's got a bad name since they sucked at gaming (due to weak FP) at a time where FPU's where still doing most of what is now done either by the GPU or SIMD units (SSE1/2) hence, AMD's release of 3Dnow which wasn't used a whole lot though. But on integer workloads and business apps, the K6 was fairly competitive with the P2/3 clock for clock .

= The views stated herein are my personal views, and not necessarily the views of my wife. =
June 16, 2004 10:05:52 PM

My God!! I have an old CoCo3, Colour Computer 3 from Tandy Computers (No longer in business) that has a Motorola 6800 chip in it for CPU.

Damn thing still works everything including the peripherals. In fact they used a DOS Emulation software program so we could simulate the programming and use of the 6800 micro processor in MS Windows computers in the lab/classroom.

Barton 3200+ 400MHz
A7N8X Deluxe
Air Cooled 45C
2x512 Corsair DDR 400 PC3200
GeForce FX5900
Two Maxtor 40Gig 8MB cach 7200rpm
SONY RW 52x/24x/52x
SONY DVD 16x/40x
June 16, 2004 10:26:31 PM

I still have a tandy TRS-80 model 1 somewhere in the basement, powered by a Z-80 and with an excessive ammount or RAM (48Kb !! instead of the 4Kb standard size) I doubt it still works though, even at the time I had to hit it (literally) and reset it regulary to keep it working. That machine was about as robust as a cardhouse.

BTW, I recently found a windows port of a game I used to play on it:<A HREF="http://jeffrey.henning.com/app/paravia/default.htm" target="_new"> Santa Paravia </A> I remember I was pretty addicted to this and I played it for hours with my friends.. Hilarious to play it once more. I recently sent the link to a few friends from back then and within a few minutes I got replies like "that overflow hack still works !". They where playing it :)  I laughed my ass off playing the game in "multiplayer" through MSN remote assistance.

> In fact they used a DOS Emulation software program so we
>could simulate the programming and use of the 6800 micro
>processor in MS Windows computers in the lab/classroom.

The 6800 is still used to learn people the basics of (assembler) programming. Its a neat, simple ISA.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
June 16, 2004 10:36:04 PM

I totally regret most vehemently not going into computer software back in my school days. I instead went into High tech electronics at the manufacturing level as in robotics used for that purpose.

I can only now imagine the advantage of being totally fluent in programming instead of a user of written material and some time programmer of small programs used in robotics.

The electronics industry has gone to POT and has been since the big Nortel crash. I must state I was also a victim more then most as I was an employee of said company may they rot in hell or next to Martha Stewart!!

Barton 3200+ 400MHz
A7N8X Deluxe
Air Cooled 45C
2x512 Corsair DDR 400 PC3200
GeForce FX5900
Two Maxtor 40Gig 8MB cach 7200rpm
SONY RW 52x/24x/52x
SONY DVD 16x/40x
June 17, 2004 3:53:06 PM

>No I don't believe it would scale as far but I think Intel
>had the ability to bring Tualatin to the public before AMD
>could bring Tbred.

Oh yes, intel has always been ahead on new process nodes. that is not surprising, intel has god knows how many fabs, AMD just one really, so intel starts early with one fab, irons out any issues, and then slowly rolls it out across it s other fabs. AMD otoh migrates more in a big bang style. Typically 6-12 months later, but much faster. Anyway, my original point was that the design of the Athlon core was vastly superior over the ppro/2/3, not that the resulting products where always better.

<font color=blue>I was talking about potential capability of the day not future product evolution.

It was my impression that Tualatin was nearly as far along in development as Willamette and could have been put into production concurrently, and long before AMD's Tbred.

I believe (IMO) that Intel chose not to bring Tualatin to the market as when it could have.

Question (just curiosity): Would a full fledged Tualatin at say 1.6 - 1.8 Ghz be a match in performance for a P4 based Celeron 2.2 - 2.4 Ghz?

I'm just sort speculating as to where things would be now if Tualatin had been released early and had evolved into the Celeron line.</font color=blue>

>I might be surprised if I knew how they all cost to
>develope, but I don't. LOL

A lot. To give you an idea, it is estimated intel and HP spent around $5 billion on dollar on developping Itanium initially, and intel currently spends $4-$500 million per year just on ongoing development cost. I would be extremely surprised if the 8088 and 8086 development combined costed more than 5 million inflation corrected dollars, could be even significantly less.

<font color=blue>I can't find any archived annual reports from the 1970's. I really would like to know the true cost.

5 million adjusted dollars seems awfully small. That would be about 50 man-years in today's dollars.

I might believe 5 million in 1976 dollars

In any case, millions vs billions is a big difference. I concede to your argument</font color=blue>

>True but Intel didn't have to do the marketing. IBM and
>Microsoft were doing it all.

definately not true. If intel's brandname and its pentium trademark is now almost as universaly known as Coca Cola or Ford, that is all to intel's credit. They designed the marketing campaigns and convinced (read: paid) the OEM's and ISV's to put on the "intel inside" and "pentium optimised" stickers, no one else did it for them.

<font color=blue>You said earlier that Intel did not really begin their marketing capaign in earnest until the "Intel Inside" campaign.

I was agreeing with you</font color=blue>

> It's only when AMD threatened to take a big enough share
>that Intel started the famous campaign.

I have a different theory on that, I don't think AMD had much to do with their reasoning. The campaign was not specifically directed against AMD, it was designed to establish intel's name brand as a world leader, a status which would allow it to expand beyond the PC CPU market. I think Intel had Motorolla, IBM, DEC, TI and HP in its sight, more than AMD.

<font color=blue>You are talking about the server/workstation markets while the advertisements seemed to be directed at consumers of destops and laptops (IMO).</font color=blue>

>I didn't and will never believe that any of the
>K6/K6-2/K6-III lines were ever a match for PII.

The cores where definately a match, but they lacked the superior memory bandwith and chipsets of intels 440LX/BX chips, as well as floating point performance. Have a look here:

http://www.anandtech.com/cpu/showdoc.html?i=902&p=12

K6's got a bad name since they sucked at gaming (due to weak FP) at a time where FPU's where still doing most of what is now done either by the GPU or SIMD units (SSE1/2) hence,

<font color=blue>OT: I can attest that K6-2 sucked at games. My K6-2 at 400 managed a piddly 20 FPS in Q3 at any resolution (up to 1280x1024 of my monitor) with a Geforce256, severely CPU bound. The Same Geforce256 with a Duron 600 yielded about 55 FPS at 640x480. Nothing to write home about but that made the game playable and it was a big performance gain. Only a 50% difference in clockspeed yielded a 175% boost in framerates.

It would be an anal exercize but I would be curious how a K6III would compare with a PII/PIII using a modern video card.</font color=blue>

<b>56K, slow and steady does not win the race on internet!</b>
June 17, 2004 4:49:35 PM

Clearly stated thoughts. Why do you name yourself Obtuse?

<b>56K, slow and steady does not win the race on internet!</b>
June 17, 2004 7:15:09 PM

>Would a full fledged Tualatin at say 1.6 - 1.8 Ghz be a
>match in performance for a P4 based Celeron 2.2 - 2.4 Ghz?

Are you serious ? It would beat the living the crap out the Celeron. At those speeds, it would be competitive with P4 Northwood on a 400 Mhz FSB.

>5 million adjusted dollars seems awfully small. That would
>be about 50 man-years in today's dollars.

Have a look here:
<A HREF="http://www.intel.com/labs/features/mi06031.htm" target="_new">http://www.intel.com/labs/features/mi06031.htm&lt;/A>
They did it in two years, and I doubt they had more than two dozen people working on it, who knows, I could be dead on of those people made that much money (which I doubt)

>You are talking about the server/workstation markets while
>the advertisements seemed to be directed at consumers of
>destops and laptops

Sure, but (future) IT managers, CFO's and CEO's are also consumers. The effort was to establish its name, period. well, IMHO.

>It would be an anal exercize but I would be curious how a
>K6III would compare with a PII/PIII using a modern video
>card.

Maybe even worse than with a slower videocard. I don't know why, but I remember seeing several benchmarks where using a high end GPU on a low end CPU actually resulted in lower performance than using a slower GPU on the same CPU.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
June 17, 2004 11:25:28 PM

YOU:<<
On the ISA front, they went from cloning x86 and its extentions (like MMX) to designing the first SIMD
extentions (3DNow) to redefining the ISA (AMD64


Reply:

You know that MMX is a SIMD

i need to change useur name.
June 17, 2004 11:42:11 PM

Yeah technically you're right. Not that MMX was usefull, being limited to integer operations and sharing its registers with the FPU. But i'll grant you a point and rephrase than into "first usefull FP SIMD extentions".

= The views stated herein are my personal views, and not necessarily the views of my wife. =
!