Why does Sony create such wacky architectures?

V3 said:
Oh yeah you need to clarify, cheaper to buy or cheaper to manufacture

To buy, of course. If N2 is cheaper to manufacture, then Sega is really reaping me off :devilish:
IIRC, the silicon area of Naomi 2 was smaller than that of the PS2. There was no point in cost reducing it (i.e. putting everything into one chip) as it was for the arcade market where systems pay for themselves very very quickly. FWIW it would not have been much more expensive than the N1.

OTOH, (IMHO) I would think that software development costs are more important in the arcade market. In that respect, I believe N2 would be a far easier system to develop for.

BTW the N2 was completely backwardly compatible with N1.
 
OF course it was backward compatible Simon. It's Naomi1 on steoroids. Give the machines an extra PowerVR chip and the Elan T&L chip and you have Naomi2. It still bothers me why SEGA doesn't just update the board yet again, using a newer SH processor and two Kyro2's hand in hand. The Kyro could work with the same setup as the two PowerVR chips are doing in N2 right?

As far as cost to manufacture, try and think that N2 has parts that are cheaper to manufacture. The SH4 chip really costs little compared to the Emotion Engine even this late in the game, and GS still costs a bit more to make than one of the PowerVR chips, and I'm sure the 2nd one as well. Of course I'm guestimating on the graphics chips, but the CPU's are right I think. The overall cost to manufacture each arcade hardware is around the same, with the cheaper on going to Naomi2. But the cost to buy are greatly different or the same depending where you buy them from. Buy them from SEGA and well, they will cost ya some good ones. Namco is sometimes known for their high pricing of arcade machines as well. The boards themselves don't cost much when it comes to arcades, its the cabinets that you pay for.

Sonic
 
IIRC, the silicon area of Naomi 2 was smaller than that of the PS2.

Then Sega is really reaping me off :( Never mind, I guess :?

But its interesting you say that silicon area of N2 is smaller than that of PS2. Are you referring to EE+GS > SH4+ELAN+PVR+PVR ?

But still each of those chip has external buses to its own memory pool, which would add to cost.

FWIW it would not have been much more expensive than the N1.

Sega charged about double the price of N1 for N2. But not so bad as Model 3, which was really expensive.

OTOH, (IMHO) I would think that software development costs are more important in the arcade market. In that respect, I believe N2 would be a far easier system to develop for.

Tekken and SC2 is cheaper than VF4 during released. I think its more market driven though and don't really reflect development cost.

But do differing hardware really add that much to development cost ? Shouldn't the more powerful hardwares are the one that would have higher development cost due to more intensive art resources.

Anyway, if AM2 did what Namco did, by using System 246 from the start for VF4, they would save up on port cost as well ;)

BTW the N2 was completely backwardly compatible with N1.

Yeah I know. So where is N3 ? :p
 
Namco is sometimes known for their high pricing of arcade machines as well.

Yeah their System 22 and 23 are more expensive than Model 2.


The boards themselves don't cost much when it comes to arcades, its the cabinets that you pay for.

Yeah, and don't forget the game, they cost quite abit more than $50. ;)
 
PC-Engine:

the fact that N2 costs about the same as Sys246 to manufacture...maybe even cheaper

When? What time frame? In March 1999, in March 2000, today or somethime in 2005/6 when producution halts? Also how does the overal performance compete, what's the advantage/disadvantage of the given architectures? What potential does each hold? And don't give me that crap about Naomi 2 being better in every aspect, because I'd second that. Perhaps overal performance maybe better along it being far easier to tap, but in the end, considering production issues and costs, I doubt it would have been that much of a better choice. If you still want to persist otherwise, please fill me in on a few details of when the N2 was cheaper to manufacture and what the main advantages would be, compared to the PS2.

BTW; to my knowledge, the Sys246 board has more memory than the PS2 hardware. If this is correct, we should really stick to one platform, the PS2 and not the Sys246 board.
 
V3 said:
IIRC, the silicon area of Naomi 2 was smaller than that of the PS2.

Then Sega is really reaping me off :( Never mind, I guess :?
I think you want to use "ripping" not "reaping" :)
But its interesting you say that silicon area of N2is smaller than that of PS2. Are you referring to EE+GS > SH4+ELAN+PVR+PVR ?
IIRC EE+GS were more than twice the area of the SH4+CLX and that was, I believe, even with the former using a smaller process.
FWIW it would not have been much more expensive than the N1.

Sega charged about double the price of N1 for N2. But not so bad as Model 3, which was really expensive.
Well have you looked inside the model 3? It was enormous. Something like 3 large boards with umpteen processing chips.

OTOH, (IMHO) I would think that software development costs are more important in the arcade market. In that respect, I believe N2 would be a far easier system to develop for.
But do differing hardware really add that much to development cost ? Shouldn't the more powerful hardwares are the one that would have higher development cost due to more intensive art resources.
N2 was a very straightforward system - easy to get good performance. The libraries were also closely modelled on what Sega themselves used when using the SH4 as the T&L engine.

Anyway, if AM2 did what Namco did, by using System 246 from the start for VF4, they would save up on port cost as well ;)
I have it on good authority that the N2 was found to be far more powerful in real terms than other HW.

BTW the N2 was completely backwardly compatible with N1.

Yeah I know. So where is N3 ? :p
Nothing to do with me. I don't even know if there is or isn't such a beast.
 
Phil said:
PC-Engine:

the fact that N2 costs about the same as Sys246 to manufacture...maybe even cheaper

When? What time frame? In March 1999, in March 2000, today or somethime in 2005/6 when producution halts? Also how does the overal performance compete, what's the advantage/disadvantage of the given architectures? What potential does each hold? And don't give me that crap about Naomi 2 being better in every aspect, because I'd second that. Perhaps overal performance maybe better along it being far easier to tap, but in the end, considering production issues and costs, I doubt it would have been that much of a better choice. If you still want to persist otherwise, please fill me in on a few details of when the N2 was cheaper to manufacture and what the main advantages would be, compared to the PS2.

BTW; to my knowledge, the Sys246 board has more memory than the PS2 hardware. If this is correct, we should really stick to one platform, the PS2 and not the Sys246 board.

N2 had no production issues. The parts were mass produced most of them coming from the DC/NAOMI. Yields were high unlike the GS which used eDRAM. None of the parts were exotic including the memory which was good old SDRAM. NEC/Hitachi didn't need to build a new fab factory for the chips unlike the chips for Sys246, which AFAIK has the same amount of memory as PS2.
 
You still didn't answer what timeframe you're talking about, not to mention the whole performance issue...

as for the memory, I'm pretty confident the Sys246 has more (read it somewhere in a SC2 discussion). Maybe someone else can clear this up (I'm to lazy.. err.. busy to search for it myself). :)
 
Sonic said:
OF course it was backward compatible Simon. It's Naomi1 on steoroids. Give the machines an extra PowerVR chip and the Elan T&L chip and you have Naomi2.
No it did not automatically follow at all. Elan sits between the SH4 and the two CLXs. It had to have a compatibility mode specifically built in to make it 'transparent' so that the CPU thought it was talking directly to the first CLX when in emulation mode.

Sonic, explaining the N2 architecture to me is, (err, what's the expression) "like teaching your Grandmother to suck eggs". :D
 
Phil said:
You still didn't answer what timeframe you're talking about, not to mention the whole performance issue...

as for the memory, I'm pretty confident the Sys246 has more (read it somewhere in a SC2 discussion). Maybe someone else can clear this up (I'm to lazy.. err.. busy to search for it myself). :)

I'll let Simon F answer the former as I recall him stating the timeframe N2's design was complete. N2 is a lighting monster compared to Xbox, GCN, PS2. All the N2 games ran in progressive too. It's efficiency is demonstrated by it's sustained performance. Just go look at VF4 in the arcade and you'll see that it's very capable.

BTW, according to www.system16.com System 246 has 32MB of main RAM.
 
I'll let Simon F answer the former as I recall him stating the timeframe N2's design was complete. N2 is a lighting monster compared to Xbox, GCN, PS2. All the N2 games ran in progressive too.

Okay, you're probably right about the Sys246 being identical to the PS2 hardware. Anyways, judging by those specs, PS2 offers a few advantages over it, despite it probably being weaker in lightning and texturing. Also check out the difference in amount of RAM. I remember that at the time of PS2s launch and before, memory was quite expensive, leading me to believe that Sony may have gotten a cost advantage by using a different approach.
 
Phil said:
I remember that at the time of PS2s launch and before, memory was quite expensive, leading me to believe that Sony may have gotten a cost advant
age by using a different approach.
Well Sony did choose to use RDRAM (rather than say SDRAM) and so it would be rather expensive.
 
45 in 2005? I'm 99% sure that this will not happen. Do you have any infos I don't have?


No, unless they're really lucky. What I mean is that unlike other companies sony can design an architecture that might have lower than desired yield at 65nm because they know their product will be there for when they'll be able to scale down to 45nm... In other words since they don't have a six month cycle they can take the luxury of lossing 200+$ on initial h/w, and pushing a process to it's limits.


Something like CELL requires a totally different way of thinking to properly utilize, which doesn't come easy.

You do know that people have access to the project, don't you? The top brass at sony is as confident as ever, the only logical reason that could be is if.... EVERYTHING IS GOING ACCORDING TO PLAN, and they're getting what THEY EXPECTED TO GET...


PS2 graphics are a step back from even Dreamcast graphics, and that is inexcusable considering the PS2 is 50% more expensive and 15 months newer.

Wow, since when are Zoe2, and SH3 Less impressive than DC games... some of the models used for the ingame portions of many ps2 games feature more polys than ENTIRE DC games, even those in the top portion like Soul Calibur.... To my knowledge there is no xbox, gamecube game that uses what the ps2 uses for a game(200K+) in a single ingame model.

...Tw3o c5me....Ye3 mo2e s5rprises a5e co6ing....
Coupling that along with its incompatibility to superior displays like VGA and HDTV in all but a few games, I'm simply not able to display PS2 games as attractively as I can with all my Dreamcast and Xbox titles, and many GameCube games.

Oh, you mean the past ps2 titles, now a days most high profile titles feature progressive out of the box, you know the wants people normaly would buy.... and some have HIGHER THAN DC rez while in progressive....

The ps2 supports higher rez progressive than DC, and surround sound.

VGA/HDTV is a global effect that only comes from better output.

PS2 supports ALL HDTV standards in the specs, and even if games can't run at the highest of the standards... it supports many of them, many that the DC does not.

HDTV video has also been showcased.

something similar to CELL but soon realized it wasn't workable and moved on.

Investment in cell continues after years of research... so it is workable... mass manufacturing will begin in the near future...

Funny isn't it ? it couldn't be more true So I don't think that proves anything about the need of a tricky architecture because DC was an exemple of using of the shelves components in a standard architecture.

The ps2 was capped by manufacturing tech, it's release was also hastened, and it was released with poor documentation, as has been said. Next time it appears all of these factors will be addressed, and won't be a problem.

many games that look far better than anything on PS2

Far better for me is stuff like the fairy demo for the GFFX, even games like Doom 3, and Everquest 2, from what I've seen are not that far ahead... Although I do consider them significantly better looking than many of the cough current cough ps2 titles... Most current xbox games only feature a little better lighting here, and a little sharper texture there, or a little shader effect here that isn't saying much... even IQ doesn't appear to be better than the top stuff in ps2.

Amazing top games in the ps2 use several fold the geometry of dc, feature higher rez support, feature better textures, better IQ, better sound, multiple textures, more particles, significantly improved lighting and more animations per models... and it is considered inferior to the DC..

Yet the xbox pushes better textures, higher rez, better lighting, slightly higher polycounts in some games, some more shader effects, about the same IQ, and yet.... IT is considered Miles upon miles better than the ps2...

Hmmm, I smell double standard somewhere....


We do know that it was not the best design available at the time.

I'd say it is, I dunno about naomi2 specs, but I doubt it can top the EE when it comes to physics, The meat and bone of gameplay, if I wanted to do a 1k poly game and leave the rest to physics calculations, I'd tend to believe the ps2 would be more faithfull... Not to mention the fact that IIRC the first GSs were manufactured in .25m, not the intended .18m, If they had had a good .18process manufacturing abilitiy both EE and GS would have likely been more buffed as the designers probably intended. Yes, and let's not forget the secret untapped power hype it caused even when newer h/w was on the way... not to mention the manufacturing cost benefits later on, as previously mentioned here. The ps2 was also hastened in response to the DC, that affected the look of the final specs, let's not forget that won't likely happen again.

I believe the theoretical potential that was there was worth the risk, even if they didn't achieve it...

Just go look at VF4 in the arcade and you'll see that it's very capable.

Again had the design fullfilled it's possibilities it would certainly be superior to naomi2... As for VF4 I don't think many would consider it above many of the top upcomming games like halo 2...

IF it was only going to be such a small difference... it was not worth going the standard route...and missing the possibility the custom design could've brought...
 
I think you want to use "ripping" not "reaping"

Yes :oops: :) (I'll blame it on the flu medicine)

IIRC EE+GS were more than twice the area of the SH4+CLX and that was, I believe, even with the former using a smaller process.

Yeah I remember, how big EE and GS was for something that was going to cost $300.

Well have you looked inside the model 3? It was enormous. Something like 3 large boards with umpteen processing chips.

Yes, Model 3 is like a mini tower fill with chips. Was quite ahead for its time. You can find it for cheap now days, cheaper than N2 for sure :)

N2 was a very straightforward system - easy to get good performance. The libraries were also closely modelled on what Sega themselves used when using the SH4 as the T&L engine.

SH-4 wasn't slouch either for T&L. But that wasn't my point. Early on many developers talk about how developing for PS2 is more expensive compare to PSX, I am pretty sure they are referring to art resources.

I have it on good authority that the N2 was found to be far more powerful in real terms than other HW.

Why didn't MS adopt N2 for Xbox ? I always wonder about that.

Nothing to do with me. I don't even know if there is or isn't such a beast.

Sega said they are working on N3. No follow up announcement yet though.
 
I'd say it is, I dunno about naomi2 specs, but I doubt it can top the EE when it comes to physics..

Again had the design fullfilled it's possibilities it would certainly be superior to naomi2... As for VF4 I don't think many would consider it above many of the top upcomming games like halo 2...

I don't think physics would be a problem for the SH-4 as it isn't doing any TnL. IIRC the P3 in the Xbox and the G3 in the GCN doesn't have any problems with physics. You're acting as though the EE is dedicated to doing physics entirely and nothing else. Also I don't see why N2 wouldn't be able to do Halo 2 especially at 30 fps and considering the N2 technology is several years older than Xbox ;)


Why didn't MS adopt N2 for Xbox ? I always wonder about that.

Probably because they wanted something more PC port friendly not to mention up to date in terms of technology.
 
You're acting as though the EE is dedicated to doing physics entirely and nothing else. Also I don't see why N2 wouldn't be able to do Halo 2 especially at 30 fps

No, I'm acting as if it CAN be used mostly for physics, in which case it should fare better, I'd guess...

I personally don't like the look of the arcade VF4, and prefer that of DOA3, and I consider VF4 to be a small jump, even if it were a tech heavy jump, it doesn't look that much better than other games outhere...

The possibilities a custom manufactured at home design could have achieved are worth the risk.

Especially, now as it appears many are having trouble scaling down, and many of the outside manufacturers are planning on slowing down.

It is funny seeing how people attack the ps2 design, which as far as I know is the first HEAVY $$$ and R&D time console made by sce. Their first attempt could've turned better... but now as the industry's costs are rising, and sony/toshiba/ibm appear to moving on schedule.... I'd say keeping it a homebrewed solution appears to have been the right choice... Rather than relying on outside designers, manufacturers that might or MIGHT not deliver(cough fx cough.) neither in quantity, nor in performance, nor in price, nor in time... To keep as close to schedule as possible, and to have R&D going inside yrs prior to release is the SMARTEST CHOICE... If something better comes along, you can always pick it up, if not u can use what u have, and it could stand to be even better than what's outside...
 
I doubt in 2001 you could have fit Naomi 2, with all those RAM pools ( texture, models, main RAM ), two GPUs in parallel under $300...


Problem I see with Naomi 2 is the fill-rate of each GPU... yes I know about the "deferred rendering issue" ( still HW like Xbox can just do one pass to fill the Z-buffer using no texture or fragment program or lighting and then use early-Z checks to reject the occluded pixels... and PS2 can do something similar thanks to its 2.4 GPixels/s drawing speed when rendering untextured polygons )...

It would have been an improovement over DC, a big one... a dedicated T&L chip ( DX7 class right ? like the GC Naomi 2 would have to offload dynamic T&L to the host CPU... 200 MHz is not that much, but it should even out with the Gekko CPU in Vector FP code [8 FP ops/cycle] as the SH-4 has a 128 bits VU while gekko can only push 4 FP ops/cycle... Gekko big fat L2, OOOe and better branch prediction would have helped plenty in Integer code where I think Gekko smokes the competition... yes the bigger L2 should push Gekko in the same ballpark as the 733 MHz PIII with 128 KB of L2, maybe not faster, but in a very close range )
 
I doubt in 2001 you could have fit Naomi 2, with all those RAM pools ( texture, models, main RAM ), two GPUs in parallel under $300...

The PS2 costs more than $300 to manufature at launch ;)

Regarding the PS2, when multitexturing the fillrate drops not to mention the geometry has to be resent.
 
The cost of a PS2 in March 2000 to manufacture (launch date) was over $300. I believe in 2001 it was just around $300. Also, bear in mind PS2 first specs were released in March 1999. ;)
 
Look... it would have been more expensive to manufacture than PS2 IMHO if you wanted to come out in that time-frame... just because Namco is ripping people off by selling basically a stock PS2 as an Arcade machine, do not start telling me how easy Naomi 2 would have fit in a console...

Total RAM, 2 GPUs alone would have pushed it a little bit in the "not very suitable for the home market" segment...

Of course Intel could produce and sell an Itanium 2 chip at the very low cost of $120-200 per chip, but I would not expect it in a console coming out next year ( just to make an example )...
 
Back
Top