Speculation from the inquirer about what might be giong on:
http://www.theinquirer.net/?article=12523
http://www.theinquirer.net/?article=12523
That does bring new light to the subject. But I believe someone pointed out that the Cell processors are week and need to be run with many cores running together. Who knows how many will be needed to run a neXtBox efficiently... but that's not really why I replied... I read somewhere... maybe Sony's own website... not sure... but anyway...
The way Cell is designed it could take advantage of other Cell chips in your home... it'd be funny to see both the PS3 and the Xbox come out using this technology. The PS3 would be a leech on your neXtBox and the other way around. lol... I'll try to find something on that.
The way Cell is designed it could take advantage of other Cell chips in your home... it'd be funny to see both the PS3 and the Xbox come out using this technology. The PS3 would be a leech on your neXtBox and the other way around. lol... I'll try to find something on that.
MadRat
Lifer
- Oct 14, 1999
- 11,999
- 308
- 126
MS only needs to port their functions using a HAL as someone else has already described. IBM could use AMD's K8 core coupled to cell sub-processors. The PowerPC processor is probably more expensive to produce for a gaming console purpose than the K8. Plus MS intends to keep it on a tight schedule so they aren't likely to introduce too many technology hurdles into their way. Correct me if I am wrong but doesn't cell technology consist of multiple cores on a chip? Only so many cores can fit on a chip and then they link chips together to scale to their needs. IBM could use the HDT links of the K8 to tie into a parent Cell-based chip and then tie two Cell-based child cores to the parent. That would put the power well within reach of PS3's design.
The second or more cores would be to offload integer and FPU functions to run them in parrallel across multiple cores, processes that IBM does well without resorting to traditional Intel/AMD technology. Existing XBOX programs using MMX/SSE/SSE2 would still be suported by the K8. 3DNow! would also theoretically be supported.
I was going to say $600, but that seemed a little high for today's market. If its billed as the "Media Center" concept then $600 is a bargain.
I mean that MS is gearing for functionality on par with what they had in mind with their "Media Center" concept. This is not simply a game console that doubles as a DVD player, but also one that can interface to control stereos and record analog video. Inclusion of XBOX compatibility just adds awesome value to the system. Even at $600 its a bargain for what MS is been working on.
Well for my sake and everyone elses... I hope they do go with Cell technology. As someone pointed out... it surely would suck for MS to unleash it's beast only to have Sony stomp all over it. Consumers knowing that the PS3 would be several times as powerful as the neXtBox would surely lessen the amount of MS consoles sold... even if they had to wait another year to have the PS3. I did some reading on this Cell technology... yeah it's something to be talked about. just one core is supposed to be 100 times more powerful than a 2.4GHz P4. I'd like to see just one of these chips in a desktop. lol. The thing is it seems these chips are designed to work in 4s. And can work their way up to 16. 4 chips running at the strength of 100 P4s. That's bragging rights for sure. And no way could a PowerPC stand up to the likes of Cell. With IBMs .1 micron process (barely bigger than Intels planned .09{{which isn't even out yet}}) we could see these chips running pretty cool for the punch they have.
Here's some news if you want to learn... Seems the President of Sony came up with the idea for Cell and brought it to IBM's attention. Here's the link. How Toshiba got into it I don't know... I have yet to read all that.
Here's some news if you want to learn... Seems the President of Sony came up with the idea for Cell and brought it to IBM's attention. Here's the link. How Toshiba got into it I don't know... I have yet to read all that.
Eug
Lifer
- Mar 11, 2000
- 24,165
- 1,809
- 126
1) I'd be surprised if they went with cell, unless MS pays everyone a LOT of money. This is Sony's baby, and MS is one of their biggest competitors.
2) Cell will take a LONG time, and likely too long for Xbox's launch. Indeed, there is info out there that suggests Sony might release an interim solution considering the time needed for Cell, and the potential competition.
3) If the hype is to be believed, then Cell wouldn't even need ATI's GPUs.
The mistake here is comparing today's PPC (or x86) technology with the Cell of many years down the line. Cell is a great idea, but it's not necessarily going to blow the doors off everything. Things change. Remember Rambus? It's fast technology, but it certainly didn't revolutionize the industry. And where are we today with memory?
2) Cell will take a LONG time, and likely too long for Xbox's launch. Indeed, there is info out there that suggests Sony might release an interim solution considering the time needed for Cell, and the potential competition.
3) If the hype is to be believed, then Cell wouldn't even need ATI's GPUs.
Can you say marketing hype?
Well, the GameCube runs PPC. That chip ain't the fastest, but it is supposedly cheap as borscht.
As long as they don't run something like that terrible Windows XP Media Center Edition.
And there's no guarantee that old Xbox programs will even be supported. IMO this is NOT a deal-breaker feature.
It's likely that at the launch of Cell, similar generation GPUs will be faster. The benefit of Cell is scalability, but you'll need blazingly fast interconnects and memory, etc. This relates to one of the above points. It's likely that memory that supports Cell's requirements won't be available as soon as Microsoft or even Sony wants.
The mistake here is comparing today's PPC (or x86) technology with the Cell of many years down the line. Cell is a great idea, but it's not necessarily going to blow the doors off everything. Things change. Remember Rambus? It's fast technology, but it certainly didn't revolutionize the industry. And where are we today with memory?
IBM should be on 0.09 by 2004.
If Sony goes with Cell and MS doesn't I'd be surprised. It would put MS in a bad position up against the PS3. But who knows. Granted it was Sony's Idea... the whole Cell deal that is... but without IBM Sony wouldn't be able to do it I don't think. Who knows who actually has control over Cell. If it is Sony which would make sence... then we'd never see Microsoft use the technology. But MS does have a lot of money... and I read a while ago that between Sony, Toshiba and IBM they put in just over 400 million dollars on the Cell Technology. Microsoft does have a way of getting its nose into a lot of stuff... 400 million is nothing for them... They could prolly drop that in on the deal and get involved doubling what they have to work with on the technology... but who knows. A PowerPC does seem a bit odd considering the whole X86 format of Xbox the 1st... Cell seems a bit weird just cause it's "Sony's baby" lol. It was their idea afterall. And Intel definetly seems out... No word on AMD or Via. And I know it's been said before... I even posted it once somewhere... ATI wouldn't really need to play a part in neXtBox if Cell was incorporated and truely is as good as they say. But we do know that ATI will play a role and we do know that IBM will play a role. And we know EUG is smart so why can't he read minds and just tell us what's up. lol
wow...this is crazy
ATI has its own little deptartmnts rivaling to produce a better GPU for the GC2 and XBOX2
And now the Xbox2 will be powered by a IBM cpu - just like how the Cell is a joint venture between sony and IBM
ATI has its own little deptartmnts rivaling to produce a better GPU for the GC2 and XBOX2
And now the Xbox2 will be powered by a IBM cpu - just like how the Cell is a joint venture between sony and IBM
IncredibleHutch-
Sony's design theory has been placing all of the strain on the processor and relying on software based rasterization of effects that we are used to seeing in hardware. Take the PS2 versus the XBox. In terms of processor power, the PS2 crushes the XBox and yet the XBox still whips the PS2 on the visual front and handily. The reason for that is quite simple, the PS2 has a graphics chip that has a feature set comparable to the Voodoo1 to go along with their very powerful EE.
The PS3 from everything we have seen is going to amplify this. The want to use Cell to hit around 1TFLOP for their processor while MS and Nintendo will both likely go with a far more mainstream processing solution. Sony is likely to pair the Cell with a very basic rasterizer core(shouldn't be called a GPU) that may be able to match up feature wise with the TNT this time(much better then the V1 class chip they are running now) so then we have Cell trying to go toe to toe with the R500 which isn't a situation that is going to look to great for Sony.
If we look at the core functionality of the R500 we should be looking at complete FP32 VS/PS units that will be handling all of the shader duties for the XB2. We can already see pixel shader performance now of current hardware versus processors, a reasonably complex shader that can be rendered at 60FPS on something like a R350 usually runs around 1 frame per minute on a current top of the line processor. 3600 times faster give or take. Obviously they need to come up with a better shading implementation for processor, which is why Sony is investing the Stanford HLSL, but then again, we are talking about a graphics core that debuted in 2002 not the R500 which will be significantly faster. Perhaps the Stanford HLSL will run fifty times faster then current processor based shader languages, but how much faster will the R500 be then the R9800?
Another issue is the complexity of development. Cell isn't even done yet, compiler and development tools for the platform are going to be extremely limited at best and will likely take several years for them to start to get a grip on. On the other side of the fence, we have DirectX dev environment simply ported over to use a new processor. Cell is a monster processor no doubt, but without pairing it with a monster GPU they are not going to be showing off nearly what the chip would do given the proper kind of support.
Of course, most of this is moot as with the next gen consoles the main limiting factor is going to be artists, all of the consoles are shaping up to be significantly more powerful then the current gen which still is holding up pretty d@mn well on the visual front. Dev costs are going to start being increasingly limited by content creation instead of coding, with the studios who have the best/fastest artists taking away a good deal of the spotlight on the visual front from the Carmack's and Sweeney's of the game world(although their next gen engines are the types of ones needed to exploit this).
Sony's design theory has been placing all of the strain on the processor and relying on software based rasterization of effects that we are used to seeing in hardware. Take the PS2 versus the XBox. In terms of processor power, the PS2 crushes the XBox and yet the XBox still whips the PS2 on the visual front and handily. The reason for that is quite simple, the PS2 has a graphics chip that has a feature set comparable to the Voodoo1 to go along with their very powerful EE.
The PS3 from everything we have seen is going to amplify this. The want to use Cell to hit around 1TFLOP for their processor while MS and Nintendo will both likely go with a far more mainstream processing solution. Sony is likely to pair the Cell with a very basic rasterizer core(shouldn't be called a GPU) that may be able to match up feature wise with the TNT this time(much better then the V1 class chip they are running now) so then we have Cell trying to go toe to toe with the R500 which isn't a situation that is going to look to great for Sony.
If we look at the core functionality of the R500 we should be looking at complete FP32 VS/PS units that will be handling all of the shader duties for the XB2. We can already see pixel shader performance now of current hardware versus processors, a reasonably complex shader that can be rendered at 60FPS on something like a R350 usually runs around 1 frame per minute on a current top of the line processor. 3600 times faster give or take. Obviously they need to come up with a better shading implementation for processor, which is why Sony is investing the Stanford HLSL, but then again, we are talking about a graphics core that debuted in 2002 not the R500 which will be significantly faster. Perhaps the Stanford HLSL will run fifty times faster then current processor based shader languages, but how much faster will the R500 be then the R9800?
Another issue is the complexity of development. Cell isn't even done yet, compiler and development tools for the platform are going to be extremely limited at best and will likely take several years for them to start to get a grip on. On the other side of the fence, we have DirectX dev environment simply ported over to use a new processor. Cell is a monster processor no doubt, but without pairing it with a monster GPU they are not going to be showing off nearly what the chip would do given the proper kind of support.
Of course, most of this is moot as with the next gen consoles the main limiting factor is going to be artists, all of the consoles are shaping up to be significantly more powerful then the current gen which still is holding up pretty d@mn well on the visual front. Dev costs are going to start being increasingly limited by content creation instead of coding, with the studios who have the best/fastest artists taking away a good deal of the spotlight on the visual front from the Carmack's and Sweeney's of the game world(although their next gen engines are the types of ones needed to exploit this).
What's to say that Microsoft will even write a Windows like OS for the neXtBox. They can't just port it over without rewriting it to fit the instruction set of the PowerPC CPUs... and if they did rewrite any portion of windows for the PowerPC that same code would run on any PowerPC... what's to say that directX support doesn't come to the Mac... and really you can't say because of the OS because the OS is built around the instruction set of the CPU as are the rest of the programs. How hard would it be to make a piece of code pertaining to DirectX run with the Mac OS... I mean they make and add stuff to Windows all the time... In a sence IBM could rewrite one file that pertains to the way graphics are handled and tell it to point to the code that IBM and Microsoft will write up for DirectX to work on the PowerPC CPU that will power neXtBox.
xbox runs a stripped down windows nt derived os, iirc. besides, windows ce and windows nt can be run on powerpc. ms would definitely not write a whole new os just to run on a game machine. as for direct x, windows nt did have directx version 3. ms will port directx over to the mac when hell freezes over.
xbox runs a stripped down windows nt derived os, iirc. besides, windows ce and windows nt can be run on powerpc. ms would definitely not write a whole new os just to run on a game machine. as for direct x, windows nt did have directx version 3. ms will port directx over to the mac when hell freezes over.
Hey that may be... Mac and Windows are two different sides to the war. If DirectX were to go to Mac though... MS could make more $$$. lol You know how back in the day there was something saying that every PC sold had to have a windows OS on it... I'm not sure what other stipulations there were but it helped Microsoft get to where it is now big time. Imagine MS and IBM signing a contract of some sort for DirectX were MS gets o say 10 bucks per copy of the Mac OS. Now I know I'm just throwing this out there and may never happen but MS really is all about getting their hands into a little bit of everything. No one thought they'd even get into gaming consoles until they pulled Xbox out of their rears (not literally... go figure).
I'm glad someone brought to my attention the fact that the EE (the whole deal really) on the PS2 isn't so great... I've read up on all that and found a few things I didn't much care for. I don't know how far along ATI will be with their GPUs by the time Xbox the 2nd does come out but R500ish sounds about right. To tell you the truth I'd be plenty glad with a 9800XT in there. It would be a set piece of hardware so games would slowly start to utilize every drop of power that it had to offer instead of trying to write code meant to run on a variety of GPUs. I still really want to know what kind of deal with IBM that Microsoft has exactly. Seems obvious that it would be the main CPU... but... PowerPC, Cell, something entirely new?
I'm glad someone brought to my attention the fact that the EE (the whole deal really) on the PS2 isn't so great... I've read up on all that and found a few things I didn't much care for. I don't know how far along ATI will be with their GPUs by the time Xbox the 2nd does come out but R500ish sounds about right. To tell you the truth I'd be plenty glad with a 9800XT in there. It would be a set piece of hardware so games would slowly start to utilize every drop of power that it had to offer instead of trying to write code meant to run on a variety of GPUs. I still really want to know what kind of deal with IBM that Microsoft has exactly. Seems obvious that it would be the main CPU... but... PowerPC, Cell, something entirely new?
The EE is the ultimate jack of all trades - master of none. It's been used to calculate game physics, to T&L, shaderish effects, and even do dolby digital 5.1! The thing with sony's philosophy is that if you create a jack of all trades, once programing efficency is up and you no longer need all of the power to go to graphics, you can use it to do other things. Or if you want to do something besides graphics, you can dedicate the all purpose CPU to it.
On the R500 on the other hand, there is no way in hell it would be able to do physics calculations or dolby digital or all of the funky things the PS2 programers have used the EE for.
I think the next generation of game consoles should have dedicated physics processers, though. Due to the proprietary nature of consoles, integrating one would be alot less painful than integrating one on the PC.
On the R500 on the other hand, there is no way in hell it would be able to do physics calculations or dolby digital or all of the funky things the PS2 programers have used the EE for.
I think the next generation of game consoles should have dedicated physics processers, though. Due to the proprietary nature of consoles, integrating one would be alot less painful than integrating one on the PC.
Eug
Lifer
- Mar 11, 2000
- 24,165
- 1,809
- 126
Actually, nVidia has been dropping hints that they intend to do this for some time(integrating physics on to the GPU). With the announcement of them moving SoundStorm to an add in board, I'm wondering if they are perhaps thinking about moving it there, but either way this is something that nVidia has been at least looking at for some time. It is possible that MS may provide some sort of DX features that would allow such a thing to be done easily in DX10. If nVidia is considering it, ATi will be too I'm sure. This move could have enormous potential for the gaming market, take Havok and run it at its highest possible settings in game currently and you will kill the most powerful CPU, but with a dedicated physics processor the picture changes and leaves a lot of CPU cycles for AI.
I'm not saying that the R500 will have this feature, or any other future part in particular, but it is something that has been looked at and mentioned several times by at least one of the big two.
Maybe it was said, but I don't have time to read entire post. Is sony really aiming to create a teraflop processor in the ps3... that's what I read about a month ago? And does anyone know how that's going?
I'll try to elaborate... lol. Cell is designed to be the 'be all chip'. That in mind my guess is there will be a lot of coding needed to be done to get it to do just what you want it to do. Running Graphics code all the while running physics code and AI code, Dolby Digital, and whatever else you could imagine running on the chip... will. Doing this efficiently may be a problem... The instruction set would be probably really big... which would all have to stay in cache for it to work at top speed. If the instruction set isn't all that big it would definitely mean a lot more code in the actual software to take and use the minimal instruction set code to get the chip to do what you wanted. Now take all this with a grain of salt because the only programming knowledge I have is 2 years of Turbo Pascal... But this should help give an idea...
I was referring to the Intel project a while back. Where the theory was, to basically let the CPU act as the Sound
Card, Video card etc etc.
No. Intel canned a project to integrate north bridge, south bridge, video and sound into the CPU itself.
The only reason why it never took off is because it took RDRAM, back in the day of the PIII.
I'm sure you could take a PIII tualatin, integrate an i815E (modified for DDR400) with a south bridge into the chip, have it run DDR 400, and it'd run great. I'm just waiting for a company try to do this again.
On a second note, this is probably why Intel is very reluctant to do the memory contoller integration thing, again.
The only reason why it never took off is because it took RDRAM, back in the day of the PIII.
I'm sure you could take a PIII tualatin, integrate an i815E (modified for DDR400) with a south bridge into the chip, have it run DDR 400, and it'd run great. I'm just waiting for a company try to do this again.
On a second note, this is probably why Intel is very reluctant to do the memory contoller integration thing, again.
zephyrprime
Diamond Member
- Feb 18, 2001
- 7,512
- 2
- 81
The instruction set can't be really big because they have so many cores on 1 chip. That means little real estate for each core. The whole idea of large # of cores procs is that each core is kinda weak but there are just so many cores you can make up for it.
In order to keep all the cores fed, the thing will have to have a huge memory bus. At least 512bits.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 23K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter