Most games are 80% general purpose code and 20% FP...

[cthellis42]

In Apple's case it was not as much "fast enough" as "fast enough, cheap enough, and in the direction they want." IBM is perfectly content to experiment with their Power and PowerPC architectures (obviously), but if they're not going to move in the direction and towards the designs Apple needs, they'd have to switch architectures regardless. (And there was little chance they'd switch to a less-known, less-tested, and less-able-to-see their roadmap architecture over a source like Intel that already fit their needs, and would continue to grow with a PC focus, whereas IBM sold off their strictly-PC business end.)

...we now return you to your regularly-scheduled debate, already in progress.

 

[London Geezer]

But look at how things turned out last generation: Sony's super-custom EE was roundly out-performed by a low-end P3, mostly due to better caches (and out-of-order-execution) on the P3.

The EE wasn't "beaten" by the P3. It would be better to say that the EE+GS was "beaten" by the P3+NV2A. Hardly surprising, since it came out 18 months later, and since the NV2A alone has more transistors than both the EE and the GS put together. Had MS released at the same time as Sony, with PC parts available at the time, things would be very different.

It's quite possible that MS, the word's most successful software company, might know something about computer architecture design. And don't forget that IBM, who also knows something about computer architecture design, also proposed a four-PPC architecture to Sony, only to have Sony insist on the more risky Cell architecture.

Sony didn't insist on the Cell architecture any more than IBM did.

 

[Shifty Geezer]

By accounts, Sony brokered a peace between Toshiba and IBM to a unified format. Can't see any 'insisting' on Sony's part.

 

[jvd]

The EE wasn't "beaten" by the P3. It would be better to say that the EE+GS was "beaten" by the P3+NV2A. Hardly surprising, since it came out 18 months later, and since the NV2A alone has more transistors than both the EE and the GS put together. Had MS released at the same time as Sony, with PC parts available at the time, things would be very different.

a celron 500 with a kyro 2 gpu with an elan tnl processer ? No shader performance but stellar performance none the less

 

[darkblu]

I think if MS thought that Cell was the best design approach, then they would have gone with a 4-SPE-and-one-PPC design. Sure, it might not have been exactly an SPE, due to licensing and IP issues, but I'm sure that IBM could have legally come up with something similar enough that it wouldn't make much difference.

so why did MS equip each of their CPUs with some serious SIMD FP arms instead of using that transisor budget for something 'more useful' wrt general-purpose code?

But look at how things turned out last generation: Sony's super-custom EE was roundly out-performed by a low-end P3, mostly due to better caches (and out-of-order-execution) on the P3.

you're kidding, right? a 'super custom EE' and a 'low-end P3' ?!?

that 'low-end' P3 was a representative of an uber-general-purpose-cruncher architecture, and the only 'low-end' thing about it was in comparison to the top-of-the-line desktop parts of that same architecture, not to the humble mips. that P3 outgunned the poor mips in every single bloody aspect but one - flops. if anything, last generation showed how a mediocre general-purpose-int design but with some special flops powers actually stood head-on against a general-purpose champion with a top-line GPU, and did that well.

It's quite possible that MS, the word's most successful software company, might know something about computer architecture design.

it is. that's why i still expect them to actually show that.

And don't forget that IBM, who also knows something about computer architecture design, also proposed a four-PPC architecture to Sony, only to have Sony insist on the more risky Cell architecture.

quite understandable postion from sony, given what they learned from the previos gen - that two VUs can battle against a superiour general-purpose cpu and a top-line TnL GPU.

Another way of looking at things is that MS is betting on improving game code as it is now, while Sony is betting on improving game code as it might become. Sony is betting that if they provide 7 SPEs then developers will figure out how to use them, even if it means contorting their game code to use far more streaming floating point than they otherwise would.

i don't see how MS are betting on improving 'game code as it is now', as latter would underperform severely on the xbox360 cpus.

Historically, we saw that the Pentium's high-speed floating point gave it an advantage in some important games (specificly Quake) over other x86 processors. Perhaps history will repeat itself this time.

On the other hand, perhaps not.

It will be very interesting to see how things work out!

oh yeah.

 

[London Geezer]

The EE wasn't "beaten" by the P3. It would be better to say that the EE+GS was "beaten" by the P3+NV2A. Hardly surprising, since it came out 18 months later, and since the NV2A alone has more transistors than both the EE and the GS put together. Had MS released at the same time as Sony, with PC parts available at the time, things would be very different.

a celron 500 with a kyro 2 gpu with an elan tnl processer ? No shader performance but stellar performance none the less

As i said, things would be very different. Try doing even the first Jak&Daxter on that.

 

[Sis]

The XB360 isn't a great general purpose system, and an X2 or dualie Pentium would have worked alot better, so I find Microsoft's protestations on this issue pure FUD.

Compared to its competitors, the XB360 is arguably better. I have not heard one quote from MS where they say, "The XB360 is the greatest GP processor ever."

Maybe MS should run a stress test of Havok/Novodex/ODE, and count FP performance. And if MS is right, why did Ageia design a hardware physics accelerator that is not a general purpose CPU, but a highly parallel SIMD FP machine?

If you are building a phsyics card, don't use a general purpose cpu. Therefore, if you are building a cpu for a console, don't use a general purpose cpu? I'm not seeing the connection.

I don't know why this is considered FUD. It seems to be a loose rule, in past generations, and so they went with that design. You even admit this in your own post where you criticize them for spreading FUD:

It may be that existing last generation games exhibit these statistics...

Unless you are saying that the Xbox 360 competitors are all better at general purpose work?

.Sis

 

[jvd]

As i said, things would be very different. Try doing even the first Jak&Daxter on that.

prob a cake walk

kyro 2 was almost 3 times faster than the power vr chip in the dreamcast . THe elan cuold transform 10 million polygons with i believe 4 or 6 lights with no penalty . The celron would most likely be fine for ai and game data .

It prob would have cost less too since
the kyro 2 was only 15m transistors vs the nv2x which was 57m transistors (sorry don't know the nv20a counts but i'm sure its very close ) not to mention kyro 2 was on 180nm vs 150nm of nv20a . I believe that both the elan and the kyro 2 would be smaller than the nv20a So they could have added more ram to the console . Most likely the price would have dropped faster too . Oh and it would most likely easly be able to play dreamcast games . Could have made sega an exclusive party and have acess to a great library of games off the bat

 

[Fox5]

As i said, things would be very different. Try doing even the first Jak&Daxter on that.

prob a cake walk

kyro 2 was almost 3 times faster than the power vr chip in the dreamcast . THe elan cuold transform 10 million polygons with i believe 4 or 6 lights with no penalty . The celron would most likely be fine for ai and game data .

It prob would have cost less too since
the kyro 2 was only 15m transistors vs the nv2x which was 57m transistors (sorry don't know the nv20a counts but i'm sure its very close ) not to mention kyro 2 was on 180nm vs 150nm of nv20a . I believe that both the elan and the kyro 2 would be smaller than the nv20a So they could have added more ram to the console . Most likely the price would have dropped faster too . Oh and it would most likely easly be able to play dreamcast games . Could have made sega an exclusive party and have acess to a great library of games off the bat

Eh, didn't the Geforce 2 outperform the Kyro 2 under nearly every situation, sometimes substantially, such as in any space/flight sim?
And I'm not sure if the current xbox could do Jak and Daxter, it's built to the strengths of the PS2, a better looking xbox game could be made, but that doesn't mean Jak and Daxter would be a good fit for the xbox.
Most xbox games don't seem to go beyond the capability of a Geforce 2 either, though the best ones usually do.

 

[jvd]

the kyro 2 was able to keep up with the geforce 2 ultra in many games . Don't forget that ms could have had the kyro 2 made on the same process as the nv20a 150nm instead of the 180nm the kyro 2 was released on thus most likely clocking it higher and giving it more head room .

The major draw back of the kyro was its lack of tnl , but that is fixed with the elan chip

You'd be looking at naomi 2 + lvl graphics easly on that machine

 

[Fox5]

the kyro 2 was able to keep up with the geforce 2 ultra in many games . Don't forget that ms could have had the kyro 2 made on the same process as the nv20a 150nm instead of the 180nm the kyro 2 was released on thus most likely clocking it higher and giving it more head room .

The major draw back of the kyro was its lack of tnl , but that is fixed with the elan chip

You'd be looking at naomi 2 + lvl graphics easly on that machine

Hmm, I dunno, Tekken 4 and 5 look much better to me than Virtua Fighter 4, but perhaps having the easier system to develop for would have had xbox to have much better looking games out way before anything worthwhile came out on ps2.

 

[jvd]

Remember 15m transitor chip on 150nm would have been very cheap and clocked very high

it was a 2x2 at 175mhz at 180nm . Giving 350mp fillrate . That is 250mp fillrate more than the dreamcast had . Plus more usable features .

THe money saved would have most likely gone to more ram .

Its tough to say how good things would loko , but i'm sure they still would have looked great and would have ended up costing less .

They could have even gone with a 2 x kyro 2 set up on a chip at 150nm , that would only be 30m tranistors mabye 35m , the kyro was scalable .

Ms had many options at the time . I really don't know why they went nvidia . They might have been able to get the next gen of power vr

 

[Shifty Geezer]

Perhaps nVidia were a good choice for integrated chipset? Plus the namesake?? :?

 

[Lessard]

that 'low-end' P3 was a representative of an uber-general-purpose-cruncher architecture, and the only 'low-end' thing about it was in comparison to the top-of-the-line desktop parts of that same architecture, not to the humble mips. that P3 outgunned the poor mips in every single bloody aspect but one - flops.

I agree

if anything, last generation showed how a mediocre general-purpose-int design but with some special flops powers actually stood head-on against a general-purpose champion with a top-line GPU, and did that well.

Completely disagree ...
The VO is used at 2% and the V1 is used at 56% for T&L (almost) exclusively (these stats come from a Sony document)
And this is 5 years after the launch of the console ...

The teams which have converted Xbox games to PS2 should have met
a few pbs i think :?

 

[London Geezer]

Completely disagree ...
The VO is used at 2% and the V1 is used at 56% for T&L (almost) exclusively (these stats come from a Sony document)
And this is 5 years after the launch of the console ...

What Sony document?! :? And over what range of games?

Those figures seem a bit off.

 

[Johnny_Physics]

Completely disagree ...
The VO is used at 2% and the V1 is used at 56% for T&L (almost) exclusively (these stats come from a Sony document)
And this is 5 years after the launch of the console ...

What Sony document?! :? And over what range of games?

Those figures seem a bit off.

It's from that PA presentation called "How Far We Got" released by SCEE, you can find it on their site.
Or click on these links:
http://www.scee.sony.co.uk/sceesite/files/presentations/PSP/HowFarHaveWeGot.pdf

http://www.scee.sony.co.uk/sceesite/files/articles/develop/PerformanceAnalyser.pdf
http://www.scee.sony.co.uk/sceesite/files/presentations/agdc2002/PerformanceAnalyser.pdf


The document was released in 2003 and the author of it even posted a few times on these boards to clear up a thing or to.
PS. Lessard fudged the numbers, don't trust his lies!

 

[Lessard]

What Sony document?! :? And over what range of games?

"Reaching for the limits of PS2 Performance. How far have we got?"

An exceptional document done by SCEE, one of the rare documents
where there is 0% marketing inside ...

Those figures seem a bit off.

Not for me ... do you think SSE or MMX units either are massively used in PC games ? I think they are not used AT ALL (0%) in the majority of games and when they are their utilisation rate should be under 2% (speculation from my part though)
Marketing, marketing ...

 

[Lessard]

PS. Lessard fudged the numbers, don't trust his lies!

Huh ? Page 14 Vector Unit Analysis :

2% VUO usage
- most games don't use VUO
- Best performing games use up to 8% VU0

56% VU1 usage
...

PS: thx for the link

 

[Johnny_Physics]

PS. Lessard fudged the numbers, don't trust his lies!

Huh ? Page 14 Vector Unit Analysis :

2% VUO usage
- most games don't use VUO
- Best performing games use up to 8% VU0

56% VU1 usage
...

PS: thx for the link

Come on, that wasn't really what you said was it?

the V1 is used at 56% for T&L (almost) exclusively (these stats come from a Sony document)
And this is 5 years after the launch of the console ...

So the document really says that the VU1 is used at 56% for T&L (almost) exclusively?

And the document was released in 2003 and the PS2 was released in early 2000(japan) so this was not five years after the launch of the console.

The devil is in the details you know.



The "don't trust his lies" was just a Memento-joke by the way, except it was "don't belive his lies" IIRC, it suddenly flashed in my head when i read LBs post so I typed it out - I wasn't calling you a liar so I hope that you didn't get offended.

 

[Lessard]

So the document really says that the VU1 is used at 56% for T&L (almost) exclusively?

No the document didn't say that effectively but the T&L is calculated by the VU1 (remember the GPU of the PS2 has not T&L unit).
The other tasks are done by the VU0 (physiks ... )
We can conclude the VU1 is used exclusively for T&L as the VU0 is not used at all ...

And the document was released in 2003 and the PS2 was released in early 2000(japan) so this was not five years after the launch of the console.

Yep my fault i didn't see the date ... but i believe the stats were mostly taken from snd and maybe even from third generation games ...