Predict: The Next Generation Console Tech

[corduroygt]

Your delusional if you think the PS3 could compete with the next xbox in terms of pure hardware power...

Umm..Last I checked, PS3 doesn't have Fermi and 2GB RAM. I was talking about the next PS.

 

[corduroygt]

Last time I checked Fermi wasn't in PS3, or was it?!

I guess I wasn't clear, I was talking about the next PS.

 

[Schaden]

I tried to private message you, but looks like that is disabled for you. Enable it and I'll shoot you a reply.

I looked at the FAQ and in my user cp under messaging & notification, there are no options for PMs. I don't think my 'registered' account has the privilege yet.

Feel free to email me though.

 

[DuckThor Evil]

I tried to private message you, but looks like that is disabled for you. Enable it and I'll shoot you a reply.

Hey others want to hear that too

I guess I wasn't clear, I was talking about the next PS.

I know, but you replied to a speculation whether PS3 with 2GB could fight next Box. RobertR1 said it would get smoked, then you came in and said I'm not too sure about that, before you started talking about Fermi.

 

[Blazkowicz]

Actually no, Elpida has 1 Gb XDR memory in production since spring.

They also have it in a 32 bit configuration, which is a perfect fit for the PS3 which lets it cut the number of XDR circuits in half. I expect them to turn up in a cost reduced PS3 during next year, together with a shrinked RSX on 45 nm.

I also keep an open mind to the possibility Sony will replace the GDDR3 interface with an XDR-interface on the RSX, it would help reducing heat, number of circuits and help keeping up the volumes of XDR memory. GDDR3 is slowly dying anyway, being replaced by GDDR5 produced in high volumes.

DDR3 can probably replace GDDR3, but your idea sure is interesting.

 

[corduroygt]

Hey others want to hear that too



I know, but you replied to a speculation whether PS3 with 2GB could fight next Box. RobertR1 said it would get smoked, then you came in and said I'm not too sure about that, before you started talking about Fermi.

It's all in the same post though
I am not delusional, obviously the next PS will need a massive increase in GPU capability and Fermi might just be the ticket if it works as well as nvidia claims.

 

[Crossbar]

DDR3 can probably replace GDDR3, but your idea sure is interesting.

Yeah, I´ve been playing around with that idea as well, but I don´t think there are DDR3 memory with 32 bit data interface being produced, just 8 and 16 bits. Even if there were DDR3 memory with 32 bit data interface it would not help to reduce the number of memory components on the board. Going the XDR path may even let them cut the number of memory components in half once more when high speed (6.4 GHz, 7.2 GHz) XDR memory is entering production.

 

[eastmen]

I don't know where you're coming from ? RSX was a spin off from among the last of an older generation of discrete dx9 gpus and probably a bit disappointing as a future proof choice for a next gen console, but I think it was a matter of bad timing.

I don't believe the situation was as nefarious as you make it sound, "long since shelved", "never meant to be sold to anyone" ? Other than broken scaling, Nvidia delivered what they could in the time frame Sony required. PS3 was supposed to launch in 1st half of 2006. In between the launch of geforce 7 and 8.
I think the GPU Nvidia delivered looked especially weak since PS3 didn't launch until G80 was already on the market.



This article from July 2004 announces a collaboration between Nvidia and Sony for the next playstation. Furthermore, it claims the development work has already been ongoing for two years.

NVIDIA partners with Sony on PlayStation 3 graphics hardware

http://www.gamesindustry.biz/articles/nvidia-partners-with-sony-on-playstation-3-graphics-hardware

The problem with the 7x00s were the shaders . There were 2alus per texture pipeline capable of vec3 + scalar or vec2 + vec2 .

The r580 was a huge increase in shader power and moving foward all the next gen designs from ati and nvidia were dynamic including the xenos in the 360.

Its really that they went with a fillrate heavy design again as with the ps2 think it would help them keep up with newer designs again.

The g70 is a monster with 24t/16p/32z . The r580 on the other hand is 16t/16p/16z .

Mabye they were justl ooking at the wrong thing


The g80 was 32t/24p/192z with a 1280 textures per pass with a 128x scaler madd+ mull design .

Even a cut down verison of this for the cosnole would have utterly destroyed what they ended up with.

 

[Dr. Nick]

The problem with the 7x00s were the shaders . There were 2alus per texture pipeline capable of vec3 + scalar or vec2 + vec2 .

The r580 was a huge increase in shader power and moving foward all the next gen designs from ati and nvidia were dynamic including the xenos in the 360.

Its really that they went with a fillrate heavy design again as with the ps2 think it would help them keep up with newer designs again.

The g70 is a monster with 24t/16p/32z . The r580 on the other hand is 16t/16p/16z .

Mabye they were justl ooking at the wrong thing


The g80 was 32t/24p/192z with a 1280 textures per pass with a 128x scaler madd+ mull design .

Even a cut down verison of this for the cosnole would have utterly destroyed what they ended up with.

NV, ATI/now AMD and Microsoft already knew what was likely coming at the time, Sony on the other hand likely didn't. Look at the hardware they had before the PS3. It's a good thing that the Cell was powerful enough to make up for the RSX shortcomings but imagine what things would be like if the Cell wasn't needed to help the PS3 to keep up with the Xenos in the 360. If they had a scaled down G80. Something that would have been fully DX10 compliant. Something like that would not only have benefited Sony but PC gaming as well. In other words it would have been much closer to the vision laid out before developers and the public when the system was first announced.

 

[AmFreak]

This article from July 2004 announces a collaboration between Nvidia and Sony for the next playstation. Furthermore, it claims the development work has already been ongoing for two years.

NVIDIA partners with Sony on PlayStation 3 graphics hardware

http://www.gamesindustry.biz/articles/nvidia-partners-with-sony-on-playstation-3-graphics-hardware

Seems GI made a mistake - the date is wrong the month and day is flipped - its acutally the december 7th as this for example shows

http://www.nvidia.com/object/IO_17342.html

but i found this article by cnn money from as early as sept 2nd 2003 stating that Sony and nVidia are in talks so who knows when they started working togehter ...

http://money.cnn.com/2003/08/27/commentary/game_over/column_gaming/index.htm

 

[grandmaster]

I'm not sure what we can actually draw from statements and press releases. Surely based on the final product, the agreement would've essentially boiled down to "you can have our best-fit GPU for the launch period". It doesn't take two years to reformulate an existing GPU to fit into a console, surely?

 

[Shifty Geezer]

NV, ATI/now AMD and Microsoft already knew what was likely coming at the time, Sony on the other hand likely didn't. Look at the hardware they had before the PS3.

PS2 was an effective design. It was conceived before shaders and scaled effectively, giving a capable, versatile console that was cost effective and made Sony lots of money. What other choices were there while PS2 was being designed?

 

[function]

A multi-pipeline evolution of what went into the DC could have worked out well, offering several times the performance of the DC chip as well as excellent image quality and high resolution textures as standard. "Free" MSAA was just around the corner for IT's chips too. Paired with the EE and a decent amount of memory I think such a system could have produced some great looking games.

Or they could have used a ... er ... Voodoo 3?

 

[fehu]

considering the news about ibm abandoning the development of the cell, what are the more realistic options for sony?

something based on power7 like rumored for the nextbox?

 

[SEGA R&D]

Fujitsu Venus.

 

[Crossbar]

considering the news about ibm abandoning the development of the cell, what are the more realistic options for sony?

I don´t fully understand this question. Is there suddenly something hindering Sony from buying a custom CPU design based on Cell from IBM, just like the other console manufacturers has bought custom CPU designs from IBM?

Did people expect that the next playstation would use somekind of bog standard all purpose Cell CPU designed by IBM? Isn´t that a pretty far-fetched idea, given that the first thing IBM did after completing the STI Cell was to redesign it for the HPC market?

 

[Shifty Geezer]

considering the news about ibm abandoning the development of the cell...

That's not accurate. The news is they aren't producing PowerXCell32i. This is the original German article news, misreported in English. The future of Cell is unknown.

 

[fehu]

ah ok...

my assumption was: exotic cell redesign = many many money
and sony hasn't so many to fund ibm for this

on the other hand a simply customization of an existing chip is far less expensive


---


In this exact moment i've discovered the differential gddr5 standard that in 2012 with a 128bit bus, will allow for 200+GB/s and 4GB at acceptable prices
nice :smile:

 

[Crossbar]

my assumption was: exotic cell redesign = many many money
and sony hasn't so many to fund ibm for this

on the other hand a simply customization of an existing chip is far less expensive

OK, adding tonns of SPUs to some random CPU core shouldn´t in my opinion be more advanced than any other custom design after all the orignal Cell will probably exist already on whatever process they decide to base the next design on. That said they may swap out the PPC core if Toshiba offers some cheaper CPU IP and just cut a deal with Toshiba.

In this exact moment i've discovered the differential gddr5 standard that in 2012 with a 128bit bus, will allow for 200+GB/s and 4GB at acceptable prices
nice :smile:

Sounds awesome. Have they found some way to work around the Rambus patents or will they license their technology?

 

[Shifty Geezer]

OK, adding tonns of SPUs to some random CPU core shouldn´t in my opinion be more advanced than any other custom design after all the orignaö Cell will probably exist already on whatever process they decide to base next design on. That said they may swap out the PPC core if Toshiba offer some cheaper CPU IP and just cut a deal with Toshiba

do Toshiba have a CPU IP to offer? What about development tools? I think a Power/PPC core is still wanted, only a more potent one.