Original XBOX designed with AMD

[Hitman928]

Maybe this isn't news to some of you, but I had always wondered about the hardware choices of the original XBOX and found this tidbit interesting. Apparently, the original XBOX was designed with AMD (not just AMD cpu, but with AMD engineers) and then 1-2 days before the unveiling, Andy Grove (intel) talked with Bill Gates and the decision was made to use an intel cpu. Just thought it was an interesting piece that I didn't know and thought others might find it interesting too.

http://www.ign.com/articles/2015/07...bosses-past-and-present-share-stories-secrets

(27:45 mark).

 

[soccerballtux]

wow, that's weird, and too bad. I wonder what CPU they would have used? that Celeron was notoriously weak (128kb cache)

 

[.vodka]

I find it weird that they decided to go at the last minute with a cut down P3. AMD's competition at the time was the Athlon, it'd have been a much better choice. I suppose Intel made MS a deal they couldn't refuse...


Haven't heard the podcast yet, so just speculating on what's noted on the OP

 

[bryanW1995]

I find it weird that they decided to go at the last minute with a cut down P3. AMD's competition at the time was the Athlon, it'd have been a much better choice. I suppose Intel made MS a deal they couldn't refuse...


Haven't heard the podcast yet, so just speculating on what's noted on the OP

Gates was probably concerned about AMD's ability to supply enough cpus.

 

[SPBHM]

wow, that's weird, and too bad. I wonder what CPU they would have used? that Celeron was notoriously weak (128kb cache)

probably K7 around the same speed, if you look at the xbox the nforce chipset in it is similar to what they used on the socket A nforce chipsets anyway, and Nvidia never tried to sell a s370 or s423/478 nforce at that time...

also the Xbox CPU was not weak, even a regular Celeron 128K coppermine was pretty decent (relative to high end CPUs, compare the PS4 CPU with a high end CPU now), and the xbox CPU had 8-way set associative l2 like the pentium 3 (celeron was 4-way)

 

[nismotigerwvu]

probably K7 around the same speed, if you look at the xbox the nforce chipset in it is similar to what they used on the socket A nforce chipsets anyway, and Nvidia never tried to sell a s370 or s423/478 nforce at that time...

also the Xbox CPU was not weak, even a regular Celeron 128K coppermine was pretty decent (relative to high end CPUs, compare the PS4 CPU with a high end CPU now), and the xbox CPU had 8-way set associative l2 like the pentium 3 (celeron was 4-way)

Yeah, the Xbox was far and away the most powerful piece of hardware that generation CPU and GPU wise. Granted it came over 3 years after the Dreamcast and a year and a half after the PS2 so this isn't exactly unexpected. However, the distance it placed between itself and the rest of the pack (even the Gamecube which launched in a similar time frame) is remarkable. Sadly, most titles were just developed to the least common denominator, so most of this muscle went unflexed outside a few notable examples.

 

[mrmt]

I find it weird that they decided to go at the last minute with a cut down P3. AMD's competition at the time was the Athlon, it'd have been a much better choice. I suppose Intel made MS a deal they couldn't refuse...

Probably supply issues. Should be one of the reasons Ruiz was so obsessed in building a new fab for AMD despite all the cash flow analysis at the time pointing to a bad business.

 

[Fox5]

This was known when the Xbox was launching, it was hyped that it was going to use an 800Mhz Duron. A Duron was AMD's version of a Celeron, but performed about on par clock for clock with Pentium 3's of that time, so in general it would have whooped the Celeron. I think a P3 based Celeron may have had a better FPU though, given that it had SSE versus 3dnow on the Duron. (some Durons had SSE later, but I don't know if it would have made it into the Xbox)

Rumor has it that AMD had started taking market share from Intel around this time. Given that, Intel was able to basically dump processors at or near cost onto the market since AMD had taken sales from Intel in the PC market, thus Intel used its surplus to take the Xbox at a cost of around $10/cpu for the first million sold to Microsoft.

Performance wise, the Xbox's cpu wasn't all that far behind the Xbox 360 or current gen consoles in per core performance. The GPU had most of the important features for modern graphics features too, albeit slow and limited. But Halo 2 @ 1080p wouldn't look all that bad against modern games imo.

 

[ShintaiDK]

800Mhz Duron used 50% more power and costed more than the Celeron.

 

[NTMBK]

It's talked about openly by Microsoft staff in Dean Takahashi's book, Opening the XBox: http://www.amazon.com/Opening-Xbox-Microsofts-Entertainment-Revolution/dp/0761537082/

Basically, Microsoft went to both Intel and AMD, and asked them for their best offer. AMD were the technically superior offering, and Microsoft went with them initially, but Intel were willing to undercut them at the last minute. AMD didn't want to drop their price, because they were supply constrained and could sell their chips at a better price in the desktop PC market, but Intel had fabs to fill.

 

[Soulkeeper]

The old plans for the tuxbox console were based on an 800MHz duron and nvidia gpu at around the same time, unfortunately those never happened (similar path to what valve is doing today).
Funny that I can no longer even find old articles/links on them anymore.

 

[BigDaveX]

800Mhz Duron used 50% more power and costed more than the Celeron.

Yeah, but it also performed a lot better, so it was worth the extra money.

Heat might have been more of a problem, especially seeing how Thunderbird and its derivatives didn't have any real thermal control logic, although I doubt it'd have been the major reason why MS didn't go with AMD.

The main issue of the early-2000s Celerons was that they only had a 66MHz bus; the 128KB of L2 cache wasn't as big of a performance killer as it was on the P4 Celerons. The Xbox's CPU on the other hand (IIRC) had a 133MHz bus, which likely made up for the smaller cache.

In any event, back in 2000/01 Intel were in the process of switching from Coppermine to Tualatin and the various NetBurst CPUs, so they probably worked out that selling a few defective Coppermines to Microsoft was a convenient way of reducing their inventory.

 

[ShintaiDK]

Yeah, but it also performed a lot better, so it was worth the extra money.

The price delta was also almost 50%.

And did it?

Remember consoles isnt locked into legacy code.

The Xbox CPU was just a P3 with half the cache disabled.

Anandtech even had an article about it as well. The Duron was considered too power hungry at 35W or so TDP. Funny considering the following consoles.

In order for the Xbox to be taken seriously as a gaming console and not just a PC in a black case it would have to be no louder than a DVD player and put out no more heat than an A/V receiver. It is a widely known fact that the Coppermine core runs significantly cooler and with lower current requirements than the Athlon/Duron cores.

 

[Smoblikat]

Tell us how to make an emulator for the original Xbox and ill be happy.

 

[nismotigerwvu]

The price delta was also almost 50%.

And did it?

Remember consoles isnt locked into legacy code.

The Xbox CPU was just a P3 with half the cache disabled.

Anandtech even had an article about it as well. The Duron was considered too power hungry at 35W or so TDP. Funny considering the following consoles.

Dude you're kidding right? The Duron of that era stomped the Celerons of the time. Here, take a peek at one of those Anandtech articles you're referencing http://www.anandtech.com/show/699/6 specifically the gaming benchmarks, considering that was the usage case for the Xbox. Also, he even says, "And while the Duron is still not producing as much heat as the Athlon, what we are really waiting for is the new Duron based on the Morgan core that should promise much cooler operation." so yes, heat was an issue for the K7 at that point, but the Duron did run cooler than the Athlons you are likely thinking of. OR how about this one http://www.anandtech.com/show/566/18 which concludes, "The Celeron is where the Duron likes to sink its teeth into; Intel's value processor that helped to kill the K6-2 and K6-III sales is now being held at the mercy of AMD's latest creation. At any default clock speed, the Celeron, even with its SSE instructions can't compare to the Duron. Chances are that the "slowest" Duron running at 600MHz would have its way with Celerons clocked at higher speeds as well". I know we're talking about 15 years ago and it's easy to forget, but it is never wise to make claims you can't support with data and worse yet to cite data that actually undermines your own argument.

 

[ShintaiDK]

Dude you're kidding right? The Duron of that era stomped the Celerons of the time. Here, take a peek at one of those Anandtech articles you're referencing http://www.anandtech.com/show/699/6 specifically the gaming benchmarks, considering that was the usage case for the Xbox. Also, he even says, "And while the Duron is still not producing as much heat as the Athlon, what we are really waiting for is the new Duron based on the Morgan core that should promise much cooler operation." so yes, heat was an issue for the K7 at that point, but the Duron did run cooler than the Athlons you are likely thinking of. OR how about this one http://www.anandtech.com/show/566/18 which concludes, "The Celeron is where the Duron likes to sink its teeth into; Intel's value processor that helped to kill the K6-2 and K6-III sales is now being held at the mercy of AMD's latest creation. At any default clock speed, the Celeron, even with its SSE instructions can't compare to the Duron. Chances are that the "slowest" Duron running at 600MHz would have its way with Celerons clocked at higher speeds as well". I know we're talking about 15 years ago and it's easy to forget, but it is never wise to make claims you can't support with data and worse yet to cite data that actually undermines your own argument.

The Celeron in the Xbox isnt Celeron based like the ones you link. But a with a 133Mhz FSB instead of 66Mhz.

You forgot to link all the text:

At any default clock speed, the Celeron, even with its SSE instructions can't compare to the Duron. Chances are that the "slowest" Duron running at 600MHz would have its way with Celerons clocked at higher speeds as well. This is mainly because of the fact that the Celeron's performance is limited by its 66MHz FSB and 66MHz memory bus.

Now this one is twice as fast in the Xbox.

 

[JoeRambo]

Intel overnerfed Celeron. 66Mhz bus and L2 cache that was both smaller and only 4 way associative only were bad combo for general performance.

Xbox had what was P3 with half of cache and was in fact very competitive for its day.

And I am not sure if anyone was buying those Celerons in these forums? The rage was 100Mhz bus coppermines running 140-150 bus speed? Or others turned to dark side and used Durons/K7 Athlons.

 

[nismotigerwvu]

The Celeron in the Xbox isnt Celeron based like the ones you link. But a with a 133Mhz FSB instead of 66Mhz.

You forgot to link all the text:


Now this one is twice as fast in the Xbox.

Twice as fast of a bus perhaps, but that prediction that it was choked by it's FSB turned out to be untrue. http://www.anandtech.com/show/694/5 Granted that was a hop to 100 and not 133, but the Celeron gained basically nothing from it and even the 800mhz model was slower (by a small degree) than the 600mhz Duron. Cranking the bus up further wasn't going to help much man.

 

[ShintaiDK]

Twice as fast of a bus perhaps, but that prediction that it was choked by it's FSB turned out to be untrue. http://www.anandtech.com/show/694/5 Granted that was a hop to 100 and not 133, but the Celeron gained basically nothing from it and even the 800mhz model was slower (by a small degree) than the 600mhz Duron. Cranking the bus up further wasn't going to help much man.

From you link when being mainly CPU limited:

The Celeron 800 still manages to provide a 24% increase in performance over the Celeron 766, in spite of only boasting a 4% actual clock speed increase (proving our theory regarding FSB dependency in games like Q3A).

And thats only 100Mhz and not 133Mhz.

 

[nismotigerwvu]

From you link when being mainly CPU limited:


And thats only 100Mhz and not 133Mhz.

The Celeron 800 in both of those articles was a 100mhz fsb chip, not a 66. Even jumping ahead, if you look at a Tulatin (which the xbox used the older Coppermine core) the P6 based cores were slower clock for clock than the K7 ones by and large (with a few exceptions). I know you're going to keep parroting in vain but it's absolutely true. http://www.anandtech.com/show/804/10 It just might be hard to see this as having been the case considering how different the competitive landscape is today.

 

[Soulkeeper]

Intel overnerfed Celeron. 66Mhz bus and L2 cache that was both smaller and only 4 way associative only were bad combo for general performance.

Xbox had what was P3 with half of cache and was in fact very competitive for its day.

And I am not sure if anyone was buying those Celerons in these forums? The rage was 100Mhz bus coppermines running 140-150 bus speed? Or others turned to dark side and used Durons/K7 Athlons.

I remember those times well here on the forums.
I'd say those of us on the "dark side" here were atleast half, if not more
AMD was much more competitive back then.

 

[ShintaiDK]

The Celeron 800 in both of those articles was a 100mhz fsb chip, not a 66. Even jumping ahead, if you look at a Tulatin (which the xbox used the older Coppermine core) the P6 based cores were slower clock for clock than the K7 ones by and large (with a few exceptions). I know you're going to keep parroting in vain but it's absolutely true. http://www.anandtech.com/show/804/10 It just might be hard to see this as having been the case considering how different the competitive landscape is today.

What relevance does your second link have?

You admit you compare with lower FSB models. And the Anandtech article specificly saying the FSB is the limit. A limit that is offset in the Xbox CPU using a 133Mhz FSB.

 

[BigDaveX]

Integer performance would probably have been about the same between the Xbox's CPU and a similarly-clocked Duron, but the former would likely have been able to do better when SSE code was used, while the latter would have been way ahead in x87 code.

As an aside, this revelation got me thinking about whether the XBOX HUEG case design was originally designed with the Duron's higher heat dissipation requirements in mind. But then I remembered the original controller design and realized, nope, Microsoft were just working by the "bigger is better" maxim. :biggrin:

 

[Sweepr]

Integer performance would probably have been about the same between the Xbox's CPU and a similarly-clocked Duron, but the former would likely have been able to do better when SSE code was used, while the latter would have been way ahead in x87 code.

As an aside, this revelation got me thinking about whether the XBOX HUEG case design was originally designed with the Duron's higher heat dissipation requirements in mind. But then I remembered the original controller design and realized, nope, Microsoft were just working by the "bigger is better" maxim. :biggrin:

Almost as huge as Dreamcast's controller.
They quickly released an improved Controller S which was much better IMO.

Xbox could put up some neat graphics back in early 2000s.
Rallisport Challenge 2 (silky smooth 60 FPS), Ninja Gaiden, Doom 3, Half Life 2, Halo 2, Conker Live and Reloaded, Splinter Cell (series), Panzer Dragon Orta just to name a few.

 

[Insert_Nickname]

(some Durons had SSE later, but I don't know if it would have made it into the Xbox)

That would be the "Morgan" and "Applebred" cores. Ah, those where the days... :biggrin:

Had a Morgan Duron 1200MHz coupled with a ECS K7S5A board. At the time it was an unbeatable value combination.