A thought on X-Box 2 and Creative Labs.
- Thread starter Brimstone
- Start date
BenSkywalker
Regular
Given the typical life cycle of a console and the rate that RAM is advancing for the next gen we are likely going to be looking at ~50GB/sec for system memory, roughly in the 512MB range give or take. Figuring for the limitations of current HDTV standards which we are likely to see in place for a couple of decades at least running 1080i with 4x AA isn't going to present much of a bandwith issue, it won't be one at all if all players have moved to embedded buffers. Bandwith is a short term problem, display technology moves far too slow to make it otherwise. Moving to 8x AA from 4x at the higest level of resolution makes very little difference, particulalry when people will be moving from 480p without AA most of the time to 1080i with it in the next transition. Aliaisng will be all but a non issue, with significantly more bandwith left over then anything available today. Bandwith is relatively speaking a non issue for the next generation.
Sony's goal seems to be sending developmet costs into the stratosphere with their obscenely complex design, something that dev houses are already having problems with(coder's ideals for what they can do given a near limitless amount of time is obviously not going to fly with publishers). Using a programmable 'GPU' with an easy to utilize API can significantly reduce development costs which was a major goal for MS making the choice of a PC native graphics supplier, particularly one that works closely with DirectX, a logical one for the next generation. Simplicity of development for this gen has already made it much more difficult for Sony to keep key exclusives, something that is going to be increasingly clear as we progress through this generation.
Creative Labs has proven absolutely nothing. Not only have they failed to exhibit that they can maintain a graphics development house properly, they have also been shown up twice on the market that they hold a near monopoly on, once by Aureal and then again by nVidia. They don't have the type of developer relations that nVidia has, nor have they exhibited that they can produce anything revolutionary without a serious threat looming.
PowerVR has not exhibited that they can compete on the high end of the spectrum in a regular time frame. When the DC shipped there was a lot of criticism towards the fact that it wasn't as powerful as top tier gaming PCs, a first for a new set top console. The PS2, XBox and GC did not have to deal with that stigma. Their arcade part is either cheap or fast, not both(scaling the unit to its most powerful configuration is quite costly). PowerVR has proven that they can be innovative without a doubt, they just haven't proven that they can compete with the high end. Saying they could is all well and good, without proof it isn't going to help.
MS is not going to risk billions of dollars on a company that has not proven it can do it, this is why nVidia got the nod over Gigapixel.
As of right now it appears that MS has one of three companies to chose from- VIA, nV or ATi if they don't decide to dev the whole thing in house. VIA has to prove that they can compete in the high end of graphics chips, they already are proving themselves as a platform provider. nVidia has to work out an agreement with MS on the current problems, although as of this point it isn't looking very good. That leaves ATi which as of this point would seem to be the odds on favorite, although we are still years away. Another important factor is that Nintendo is almost certain to use one of the other companies that MS doesn't chose, it may be worth more to MS to go with one company over another, even if their is pricing conflicts, to stop Nintendo from being able to one up them.
Creative Labs isn't known for being cheap or groundbreaking. PowerVR is cheap and groundbreaking, but they aren't powerful. nVidia is powerful and groundbreaking but they aren't cheap. ATi is powerful and groundbreaking but they aren't too cheap either(cheaper then nV, but not as powerful). VIA is cheap and have proven they can compete a powerful platform solution, but not a graphics solution nor have they proven to be innovative. Of course, we are still about half a decade away from the launch of the next generation. It wouldn't surprise me if MS was in contact with multiple parties at this point to get as many different proposals as possible then they can have the luxury of waiting a bit on finalization as they see how each company is holding up. Five years from now BitBoys could be dominating the market(and they could well be the company that gets the contract if they prove they can deliver), just as five years ago noone could have seen nVidia and ATi dominating the market with 3dfx being dead and consumed.
Sony's goal seems to be sending developmet costs into the stratosphere with their obscenely complex design, something that dev houses are already having problems with(coder's ideals for what they can do given a near limitless amount of time is obviously not going to fly with publishers). Using a programmable 'GPU' with an easy to utilize API can significantly reduce development costs which was a major goal for MS making the choice of a PC native graphics supplier, particularly one that works closely with DirectX, a logical one for the next generation. Simplicity of development for this gen has already made it much more difficult for Sony to keep key exclusives, something that is going to be increasingly clear as we progress through this generation.
Creative Labs has proven absolutely nothing. Not only have they failed to exhibit that they can maintain a graphics development house properly, they have also been shown up twice on the market that they hold a near monopoly on, once by Aureal and then again by nVidia. They don't have the type of developer relations that nVidia has, nor have they exhibited that they can produce anything revolutionary without a serious threat looming.
PowerVR has not exhibited that they can compete on the high end of the spectrum in a regular time frame. When the DC shipped there was a lot of criticism towards the fact that it wasn't as powerful as top tier gaming PCs, a first for a new set top console. The PS2, XBox and GC did not have to deal with that stigma. Their arcade part is either cheap or fast, not both(scaling the unit to its most powerful configuration is quite costly). PowerVR has proven that they can be innovative without a doubt, they just haven't proven that they can compete with the high end. Saying they could is all well and good, without proof it isn't going to help.
MS is not going to risk billions of dollars on a company that has not proven it can do it, this is why nVidia got the nod over Gigapixel.
As of right now it appears that MS has one of three companies to chose from- VIA, nV or ATi if they don't decide to dev the whole thing in house. VIA has to prove that they can compete in the high end of graphics chips, they already are proving themselves as a platform provider. nVidia has to work out an agreement with MS on the current problems, although as of this point it isn't looking very good. That leaves ATi which as of this point would seem to be the odds on favorite, although we are still years away. Another important factor is that Nintendo is almost certain to use one of the other companies that MS doesn't chose, it may be worth more to MS to go with one company over another, even if their is pricing conflicts, to stop Nintendo from being able to one up them.
Creative Labs isn't known for being cheap or groundbreaking. PowerVR is cheap and groundbreaking, but they aren't powerful. nVidia is powerful and groundbreaking but they aren't cheap. ATi is powerful and groundbreaking but they aren't too cheap either(cheaper then nV, but not as powerful). VIA is cheap and have proven they can compete a powerful platform solution, but not a graphics solution nor have they proven to be innovative. Of course, we are still about half a decade away from the launch of the next generation. It wouldn't surprise me if MS was in contact with multiple parties at this point to get as many different proposals as possible then they can have the luxury of waiting a bit on finalization as they see how each company is holding up. Five years from now BitBoys could be dominating the market(and they could well be the company that gets the contract if they prove they can deliver), just as five years ago noone could have seen nVidia and ATi dominating the market with 3dfx being dead and consumed.
Dave B(TotalVR)
Regular
Like via?
At what cost though?.. if you could get away with the same performance with only 20GB/s or less and so far less cost why not do that?
What about with 8x AA
.. what about 64bit rendering?Cost is always quite a big issue with consoles though.
In raw specs maybe, but in reality DC could put out better visuals then any PC game at its time of release.
A top of the range system was still far faster in raw specs then a XBox or GameCube at release. Geforce 3 TI500 and 1.4ghz T-Bird on a DDR motherboard for instance.
No but they've proven that they can produce great technology for fixed units like consoles and arcades. Dreamcast is still on the level with PS2 in graphics and its now 4-5 years old and their Naomi arcade tech currently holds 70% of the arcade market. On their own or with STM I'd agree with you that they weren't a great option for MS with XBox2. But together with VIA they are a great option IMO (of course this all relys on IMGTEC and VIA being partners but I think that's only a matter of time now). XBox2 could be made completely by VIA with all parts coming from them and IMGTEC. Motherboards.. VIA can do that brilliantly, CPU's VIA can do that and with their new chips their starting to do it well, for the GPU they can call on PowerVR and for the sound they can call on ensigma (a division of IMGTEC). Not only can the cost be kept down due to a more efficient architecture but the cost can also be kept down by having all the parts coming from one company (VIA).
Now put VIA and IMGTEC together and you have cheap powerful platform, cheap innovative graphics. As for powerful graphics, put Kyro III (which has been finished now for a little while) in a console right now and it could easily outperform XBox (at 640x480x32 with FSAA) and be a cheaper solution to boot. So I'm sure that with MS throwing money at IMGTEC and VIA to develop this chip they could come up with a really nice chip with excellent performance and a cheap cost.
I think at this point assuming IMGTEC and VIA are together it would be a close fight between ATI, Nvidia and IMG/VIA for XBox2. But with Nvidia pissing on MS big time and ATI doing GameCube and so possibly sticking with Nintendo for next gen that leaves IMG/VIA in a great position. Of course maybe when MS starts to look for the companies to make XBox 2 the position may have changed.. who knows.
Even if IMG/VIA didn't get XBox 2 and instead ATI did.. then that leaves Nintendo to choose between Nvidia and IMG/VIA and considering each companies qualities and what Nintendo likes in a console I think Nintendo would chose IMG/VIA.. so I do see PowerVR tech in at least one consoles in the next gen (with a possible next gen handheld console being based on MBX).
Joe DeFuria
Legend
My thoughts....
I also think that PowerVR tech would be better "suited" for a Nintendo console vs. the next X-Box. Main reason: Direct X.
No doubt, MS's next console will be based on some future version of DirectX (DX 11?), which will be "mostly compatible" with the PC DirectX of the same time frame. It's also almost a given that it will be backward compatible with the original X-Box and DX 8.0. One thing that PowerVR has not been able to "prove" is that they can deliver a part based on their technology that MS feels comfrtable using as the "reference platform" for some DX version. For maximum compatibilty and minimal risk, I see Microsoft choosing a vendor that consistently delivers parts that are on the bleeding edge of DirectX technology, and that to me means ATI or nVidia.
Remember, the original X-Box chip was a battle between nVidia and Gigapixel. The same arguments for PowerVR can be used for Gigapixel: the part should have been a much better bang for the buck compared to NV2A. However, MS went with nVidia. Why? IMO, it's because nVidia has a DirectX track record, and that posed minimal risk, which outweighed the price/performance advantage of Gigapixel.
Nintendo, which will not be relying on the DirectX environment, sounds like a much better match for PowerVR tech than X-Box. A custom programming environment can be utilized without any regard to backwards compatibility.
I also think that PowerVR tech would be better "suited" for a Nintendo console vs. the next X-Box. Main reason: Direct X.
No doubt, MS's next console will be based on some future version of DirectX (DX 11?), which will be "mostly compatible" with the PC DirectX of the same time frame. It's also almost a given that it will be backward compatible with the original X-Box and DX 8.0. One thing that PowerVR has not been able to "prove" is that they can deliver a part based on their technology that MS feels comfrtable using as the "reference platform" for some DX version. For maximum compatibilty and minimal risk, I see Microsoft choosing a vendor that consistently delivers parts that are on the bleeding edge of DirectX technology, and that to me means ATI or nVidia.
Remember, the original X-Box chip was a battle between nVidia and Gigapixel. The same arguments for PowerVR can be used for Gigapixel: the part should have been a much better bang for the buck compared to NV2A. However, MS went with nVidia. Why? IMO, it's because nVidia has a DirectX track record, and that posed minimal risk, which outweighed the price/performance advantage of Gigapixel.
Nintendo, which will not be relying on the DirectX environment, sounds like a much better match for PowerVR tech than X-Box. A custom programming environment can be utilized without any regard to backwards compatibility.
Im more curious who will develop the processor for the XBOX2 ... the PS2's had its process to drag it down, but with Sony's new processor being codeveloped with IBM and a deal in place to give Toshiba access to IBM's fab technology it will IMO be plain impossible for m$ to compete with them in the same way they did now. With commodity desktop processors and graphics chips produced on a relatively low logic-speed process (compared to the IDM's). Technology might not sell consoles, but people do expect some improvement for the company who lags the market.
Intel or AMD will have to be tied in with the XBOX2 in some way IMO. Problem is that m$ would really have to start investing now, and I just dont see them having the balls to do what Sony is doing with Cell given the XBOX's lackluster performance.
Intel or AMD will have to be tied in with the XBOX2 in some way IMO. Problem is that m$ would really have to start investing now, and I just dont see them having the balls to do what Sony is doing with Cell given the XBOX's lackluster performance.
DX compatability of later PowerVR designs (such as KYRO) is fine, assuming developer stick to certian DX coding guidlines. It a lot easier to control how the developers code within a console environment as well.
Because Gigapixel had never produced a single chip based on their designs while Nvidia were a company with a proven history (all be it in the PC market not the console market). IMGTEC don't have the same problem as they have delivered chips to the PC market and have been very succesful in the console/arcade market.
I take your point on MS wanting to use Nvidia/ATI because of new DX API's being based on their top chip's features though.. it may have some bearing on the decision.
However for actual compatability I don't see any problem with a PowerVR chip, it runs DX games absolutely fine. The only problems that have occured with compatability in the past is due to lack of communication and some devs using some unusual techniques in their games without considering any chips like Kyro 1 and II. However this all seems to have been sorted out now, their are no games that have problems on Kyro II now and any problems that have occured in the past wouldn't have happen on a console anyway.. because obviously the dev would be making/porting the game with the PowerVR tech in mind.
I think the main reason why PowerVR could be a good choice for Nintendo is because with GameCube they've went with a cheap efficient design. So if they want to stay with that sort of console design (cheap but powerful) then PowerVR tech could be the perfect solution for their next console IMO.
BenSkywalker
Regular
Darren-
What's the price rift between PC66 and PC2100 at? About 3% if that? Dropping 60% or more bandwith to save a few percent doesn't make much sense.
Hmm, and how are tilers going to hold up with 128x AA and 512bit rendering When you hit 1080i with 4x AA you are closing in on human perception limits.
Saving $1.50 and dropping 60% or more? I don't think its worth it. Given the amount of time we have until the XB2 is ready to go 60GB/sec isn't looking at bleeding edge, that should be fairly mainstream.
A bit over one quarter the gometric throughput is faster?
If you honestly think that the DC is holding up to the PS2, you are delusional The launch titles for the PS2 were certainly nothing worth getting excited about, but I haven't seen anything on the DC that comes close to FFX.
VIA is going after third world nations with their CPUs, they are not going to be close to Intel or AMD for many years, let alone a highly customized CPU such as that being used in the PS3. Right now, over two years after the launch of the PS2 VIA still doesn't have a processor remotely close to the MIPS/EE.
I have to hear this, how do you figure? Also, you assume that it would be cheaper then the XBox, what about adding in the cost of the North and SouthBridge chipsets? Then tack on the cost of an audio chip. nV is doing a lot more then just providing the graphics chip.
Between the two of them you are still missing a sound chip, and then all of that combined needs to come in at the right price point and be competitive against what nV or ATi has to offer.
Right now IMGTec/VIA haven't proven that they can do it. nVidia gave them a platform, not just a graphics chip. ATi is working that way(chipsets at least), I see them as being the closest right now. As far as the current pricing issue goes, I can't blame nV for being a bit miffed at having to drop their price point within six months of launching the console. Sony has had their platform out for over two years to put things in perspective.
Kind of hopeful for IMT/VIA in that scenario particularly considering that ATi has the team that has handled the last two Nintnedo consoles.
What's the price rift between PC66 and PC2100 at? About 3% if that? Dropping 60% or more bandwith to save a few percent doesn't make much sense.
Hmm, and how are tilers going to hold up with 128x AA and 512bit rendering
When you hit 1080i with 4x AA you are closing in on human perception limits.Saving $1.50 and dropping 60% or more? I don't think its worth it. Given the amount of time we have until the XB2 is ready to go 60GB/sec isn't looking at bleeding edge, that should be fairly mainstream.
A bit over one quarter the gometric throughput is faster?
If you honestly think that the DC is holding up to the PS2, you are delusional
The launch titles for the PS2 were certainly nothing worth getting excited about, but I haven't seen anything on the DC that comes close to FFX.VIA is going after third world nations with their CPUs, they are not going to be close to Intel or AMD for many years, let alone a highly customized CPU such as that being used in the PS3. Right now, over two years after the launch of the PS2 VIA still doesn't have a processor remotely close to the MIPS/EE.
I have to hear this, how do you figure? Also, you assume that it would be cheaper then the XBox, what about adding in the cost of the North and SouthBridge chipsets? Then tack on the cost of an audio chip. nV is doing a lot more then just providing the graphics chip.
Between the two of them you are still missing a sound chip, and then all of that combined needs to come in at the right price point and be competitive against what nV or ATi has to offer.
Right now IMGTec/VIA haven't proven that they can do it. nVidia gave them a platform, not just a graphics chip. ATi is working that way(chipsets at least), I see them as being the closest right now. As far as the current pricing issue goes, I can't blame nV for being a bit miffed at having to drop their price point within six months of launching the console. Sony has had their platform out for over two years to put things in perspective.
Kind of hopeful for IMT/VIA in that scenario particularly considering that ATi has the team that has handled the last two Nintnedo consoles.
Ben, unlike Nvidia, IMG TEC has their own proprietary and kickass Meta DSP for audio. (And not necessarily only audio, but I won't go there.) Check www.Metagence.com
There's another discussion of just that over at, hmm, where was it, oh yes, Console Talk
There's another discussion of just that over at, hmm, where was it, oh yes, Console Talk
Who's talking about PC66 or PC2100?.. I'm talking about top of the range video ram vs lower end video ram, more like 350mhz 128bit DDR ram vs 150mhz 128bit DDR ram.
The answer is far better then an IMR
Anyway the 512bit comment is certainly out of place mate, 64bit is coming like it or not.Exactly what res is 1080!??.. I know nothing about HDTV's.
Once again we're not talking about system ram so old that its no longer even used (PC66) and new system ram that's used in most new PC's. We're talking about top of the line ram that's hardly used in any great number and is very expensive vs ram used in lower end cards that have been bought in their ten's of millions. The price gap is not $1.50 their, look at the TI4200 vs the TI4600, the biggest difference their is ram speed goes from 200mhz in 4200 to 325mhz in 4600 and the price of the card rockets.
Really? Xbox 2 will likely be released in under 5 years. They'll have to get it all ready in under 4 years, so in 4 years you think 60gb/s will be mainstream? (like 166mhz DDR is now?). 4 years ago top of the line video ram was around 2.7gb/s, now mainstream ram is around 5.18gb/s. So current mainstream video ram has just under double the amount of bandwidth that top of the line video ram had 4 years ago (and that's without taking into account the fact that DDR is less efficient then SDR). But you think that 4 years from now mainstream ram (the 166mhz 128k DDR ram of today) will have 6 times more bandwidth then the current top of the line ram (325mhz 128bit DDR)?
I'm talking about raw specs, I've heard many a person say "XBox isn't that powerful.. look at it, its only got a 750mhz CPU and a 200mhz graphics card with only 6gb/s of bandwidth... my PC has a 1.4+ GHZ CPU and has a graphics card with 10gb/s bandwidth.. its much more powerful then XBox.
DC might not be holding up to PS2 in stuff like polygon counts but allot of the time PS2 games don't have that high polygons counts and the DC make up for its slightly lower polygon counts with better texturing. Overall DC can still hang with the PS2 for visuals IMO. Of course PS2's best looking games are better looking then anything on DC but not by a long way, and most of the games on PS2 and DC look pretty similar in quality.
They may not have to, the XBox doesn't have a CPU that even comes close to the PS2's CPU.. but it makes up for that in other ways. Look at the XBox's CPU, its barely a mid end solution in the PC space.
Its pretty obvious isn't it? The Kyro III with 200mhz DDR ram (like XBox) would reach a far higher fillrate at 640x480x32 with FSAA then the NV2A could with 200mhz DDR ram. Because of the tiling the FSAA would not take extra ram or framebuffer bandwidth, no Z-buffer, no overdraw.. you do the maths.
Are Nvidia's south/north bridge so cheap that VIA could not match them in price?... VIA have a hell of a lot more knowledge and experience in motherboards then Nvidia do. As for sound, again is Nvidia's solution anything special?
Did you not read my comment on Ensigma?.. the sound division of IMGTEC.
Well that goes without saying.
And Nvidia will undoubtedly be the favourite for XBox 2 if they can get this little squable with MS sorted out and get back to being good friends again.
And you think ATI are going to make the chips for both XBox 2 and Nintendo's next console? I said if ATI go with XBox 2 then I believe Nintendo may choose VIA/IMGTEC because of Nintendo's current philosophy of cheap yet efficient consoles.
Once again, this is all assuming that Nvidia and MS don't fully recover their friendship after this bad fallout with chip costs. As I've said if Nvidia and MS become best friends again then Nvidia are undoubtedly the strong favourites for XBox 2. But if not then MS may have to choose someone else. If its ATI then I think IMG/VIA have a hell of a chance of being chosen ahead of Nvidia by Nintendo. But then if were talking dev support and drivers ect then ATI are hardly great at that.. infact IMGTEC are clearly better then ATI in this area IMO.
Dave B(TotalVR)
Regular
Its not just the cost of the memory chips, its the cost of the added traces on the board (pushing it to more layers), pins on the chip and complexity of the memory controller too. By reducing the memory speed you cut costs with the chips (and the savings would be higher than you make out here) the board complexity and more importantly the silicon complexity of your main GPU core.
Also, as Gunhead has beaten me to saying, IMGTEC have quite a kickass audio technology of their own.
Ugh, am I the only one that finds it extremely annoying when people split up their posts into tens of single line quotes?
In this thread Teasy leads with 15 quotes in a single post, tightly followed by BenSkywalker with 11 and Dave B(TotalVR) with 10.
This kind of extreme splitting makes you need to read the whole thing you replied on, you also loose all kind of context. Otherwise you can just quickly see what you replied on just by quickly eyeing on the first line of the quote and get the whole context without need to look back in the topic to find where it was cut from. I usually don't bother reading posts with more than 3 or 4 quotes in it.
In this thread Teasy leads with 15 quotes in a single post, tightly followed by BenSkywalker with 11 and Dave B(TotalVR) with 10.
This kind of extreme splitting makes you need to read the whole thing you replied on, you also loose all kind of context. Otherwise you can just quickly see what you replied on just by quickly eyeing on the first line of the quote and get the whole context without need to look back in the topic to find where it was cut from. I usually don't bother reading posts with more than 3 or 4 quotes in it.
History... threw the eyes of Ben
No, its breaking the mold and outputting designs that are years ahead. They made a CPU in '99, using the 0.25um process and drawbacks with it that still outpaces any today in FP intensive tasks. They're the only producer of a single chip 3D processor with 48GB/sec of sustainable bandwith, only entity with eDRAM... I mean, give respect where it's due - SCEI dev's kicks ass.
How silly is this - I mean common now. Developers, like Carmack - who much be the biggest proponent - are drewling for full programmability. As long as IBM/Sony can provide the libraries, dev relations, and tools; then it's no diffrent than anything else - just much, much faster and more powerful for those who have the talent to exploit.
Um... What rasterizer? Your not much of a dynamic thinker, huh? How is a 1 TFlop processor weak? Lets think this threw, interesting stuff. I truely believe there will be no hardwired rasterization in PS3, rather an extremely fast, sustainable, CMP processor with SMT ability. The indications comming from analysts following Sony, to Kutaragi's vision, to their recent usage of 'MPU' instead of 'CPU'. I guess it depends on how you see the future...
This is for all future graphics, not just Cell.
I see fragment shading and texturing going bye-bye and replaced by enormously dense polygon counts and high-level programmable vertex shaders. I just removed the main functions in a current 'GPU's 3D pipeline. All TCUs and related hard-wired logic (filtering, et al) are gone...
http://www.gamasutra.com/features/20020410/brickhill_01.htm
I mean, think of the Nv25 and it's peak of 1 TFlop... It's unilaterally agreed that the vast majority of this processing is fixed function (70%?) and the remaining shaders and programmable blocks constitute a small part of this chip. (Even the Avenger core/Voodoo3 could do 100GFlop) So, if the Nv25 can do so much with the little preformance backing it's shaders... just imagine over a TFlop of sustained preformance.
And because it's totally programmable, the develpers can suite the entire 3D pipe to the game, or perhaps to the individual scene.
Ease of development all comes down to the tools and dev support given in Cells's case. If they follow in nVdia's developer footsteps and provide developers with sample code and basic 3D pipe implimentatiosn to work off of, then it's easy as pie for those who just cut and paste and immensly powerful and flexible for those who are adventerous.
No, its breaking the mold and outputting designs that are years ahead. They made a CPU in '99, using the 0.25um process and drawbacks with it that still outpaces any today in FP intensive tasks. They're the only producer of a single chip 3D processor with 48GB/sec of sustainable bandwith, only entity with eDRAM... I mean, give respect where it's due - SCEI dev's kicks ass.
How silly is this - I mean common now. Developers, like Carmack - who much be the biggest proponent - are drewling for full programmability. As long as IBM/Sony can provide the libraries, dev relations, and tools; then it's no diffrent than anything else - just much, much faster and more powerful for those who have the talent to exploit.
Um... What rasterizer? Your not much of a dynamic thinker, huh? How is a 1 TFlop processor weak? Lets think this threw, interesting stuff. I truely believe there will be no hardwired rasterization in PS3, rather an extremely fast, sustainable, CMP processor with SMT ability. The indications comming from analysts following Sony, to Kutaragi's vision, to their recent usage of 'MPU' instead of 'CPU'. I guess it depends on how you see the future...
This is for all future graphics, not just Cell.
I see fragment shading and texturing going bye-bye and replaced by enormously dense polygon counts and high-level programmable vertex shaders. I just removed the main functions in a current 'GPU's 3D pipeline. All TCUs and related hard-wired logic (filtering, et al) are gone...
http://www.gamasutra.com/features/20020410/brickhill_01.htm
I mean, think of the Nv25 and it's peak of 1 TFlop... It's unilaterally agreed that the vast majority of this processing is fixed function (70%?) and the remaining shaders and programmable blocks constitute a small part of this chip. (Even the Avenger core/Voodoo3 could do 100GFlop) So, if the Nv25 can do so much with the little preformance backing it's shaders... just imagine over a TFlop of sustained preformance.
And because it's totally programmable, the develpers can suite the entire 3D pipe to the game, or perhaps to the individual scene.
Ease of development all comes down to the tools and dev support given in Cells's case. If they follow in nVdia's developer footsteps and provide developers with sample code and basic 3D pipe implimentatiosn to work off of, then it's easy as pie for those who just cut and paste and immensly powerful and flexible for those who are adventerous.
Similar threads
