News about Rambus and the PS3

Yeah... I was not thinking about it... I am sorry, I really made a mistake :(

You would only transfer the subdivision suurfaces or triangle strips from main RAM... Let's say though that there is a 25.6 GB/s Redwood connection between CPU and GPU... the number should still work there...

But thinking about what we just said there is even more bandwidth left for textures and other data ( to stream data to e-DRAM and keep both of those on chip pools fed ) coming from the main RAM...
 
If sony didn't care about cost we'd have a ps3 that had 4 cell chips in it each with 256 megs of on die ram and 2 gigs of system ram all for the low cost of 200 bucks.


I mean why not sony doens't care about cost they will just make it up on software sales . Grall i thought you were smarter than that. In busniess everything comes down to cost.
 
That still slow compared to a S-ATA based HD which have 266MB/s of bandwidth IIRC. Plus the access time is far better than any know optical solution. Just pointing out a couple of things here.

I dunno if that is true, but later gen. hdds should be faster than what we have today.

As for ram, look at nintendo they placed a good amount of cheap ram in the system... It looks GOOD ;) on the spec sheets... and from what I've heard the big N could've gotten significantly superior cheap ram for a similar price.

MS is gonna tout whatever it can as an advantage to the ps3... placing ram even useless additional cheap ram will only hinder the big MS efforts, and the big M will only use similar or even more ram of a not so cheap kind :devilish:
 
Actually if the PS3 CPU is configured as the Broadband Engine is in the patent, then it will have 4 Cells.

Each PE (Processing Element) is a Cell. the Broadband Engine is made from 4 PEs.

but anyway, yea, if Sony made PS3 with 4 full Cell-based CPUs, several gigabytes of RAM, and sold it for a $300, it would be...dreamy :)



The article from Mercury News awhile back, that said PS3 would have 72 sub-processors, suggested that the CPU would be made from 8 PEs. thus 8 PowerPC CPUs plus 64 APUs.

but regardless of the sub-processor configuration on PS3's CPU, I think they should not cut the eDRAM down from 64 MB to 32 MB.


64 MB for CPU is what is described in patent. 32 MB could be used on the GPU though. giving us 96 MB eDRAM plus several MB more of local SRAM storage, and at least 256 MB main memory.
 
minimum PS3 specs:


CPU: 1 TFLOPs

main memory bandwidth: 25+ GB/sec
main memory size: 256 MB

eDRAM bandwidth: 100+ GB/sec
eDRAM size: 32MB + 32MB (CPU+GPU)
 
megadrive0088 said:
Actually if the PS3 CPU is configured as the Broadband Engine is in the patent, then it will have 4 Cells.

Each PE (Processing Element) is a Cell. the Broadband Engine is made from 4 PEs.

but anyway, yea, if Sony made PS3 with 4 full Cell-based CPUs, several gigabytes of RAM, and sold it for a $300, it would be...dreamy :)



The article from Mercury News awhile back, that said PS3 would have 72 sub-processors, suggested that the CPU would be made from 8 PEs. thus 8 PowerPC CPUs plus 64 APUs.

but regardless of the sub-processor configuration on PS3's CPU, I think they should not cut the eDRAM down from 64 MB to 32 MB.


64 MB for CPU is what is described in patent. 32 MB could be used on the GPU though. giving us 96 MB eDRAM plus several MB more of local SRAM storage, and at least 256 MB main memory.


Personally I would like to see Sony go with a higher pice with the PS 3 to allow for additional ram, blu-ray, and maybe even a second processor. At $399 the PS 3 will sell just as well, if not better than a $299 priced console.
 
I'm not so sure about that. A lot of funny things happen to your mainstream consumer base once you cross over the $300 mark. If the competitor consoles come in at or below $300, then that would only spell more trouble for Sony, if it came in substantially above $300.
 
randycat99 said:
I'm not so sure about that. A lot of funny things happen to your mainstream consumer base once you cross over the $300 mark. If the competitor consoles come in at or below $300, then that would only spell more trouble for Sony, if it came in substantially above $300.

$399 isn't a bad price if it come with Blu-ray,Remember Sony will advertise the Cell as the most versitlie Chip out in the Market being compatible with many other Sony/Toshiba/IBM products.....maybe they can sell a stripped down version without Hdd and Tivo feathers a $100 less.
 
Panajev2001a said:
Incorrect, the DC had 16 MB of main RAM...

16+8=24

It's ~ 24 MB of RAM in a UMA. Since PS2, GCN, and Xbox all store some textures in main RAM it makes sense to add the 8 MB of VRAM to the 16 MB of main RAM on DC.
 
Panajev2001a said:
Fine then :)

:D Seriously though, I honestly don't know if PS3 will have 512 MB+ of UMA though. Looking at the previous and current console's memory capacity doesn't really give us enough information to make any informed conclusion especially when we don't know how much it will cost to manufacture XDR.
 
Psychogenics said:
$399 isn't a bad price if it come with Blu-ray,Remember Sony will advertise the Cell as the most versitlie Chip out in the Market being compatible with many other Sony/Toshiba/IBM products.....maybe they can sell a stripped down version without Hdd and Tivo feathers a $100 less.

You can argue any price is "reasonable" then, as long as you keep throwing on more fancy components. The average consumer isn't going to care if it has BR or not- just that it plays the game they want when they put the disc in, and the cost is within their reach. $300 is about where that threshold is between "cost within reach" and "cost out of reach". $180-200 is the jackpot price, though.
 
I dont see why we need to keep scaling bandwidth up at the same rate that fillrate and processing power goes up.

On a console, we are fixed at low resolution, and already with todays minimum spec bandwidth (like xbox) we can draw a scene 5 or 6 times over at 30hz.

It seems to me we are more or less set for framebuffer bandwidth, even with antialiasing. Texture BW might increase, but probably not at the same rate as computation requirements do.

The argument is that pixel and vertex shaders are going to keep increasing in size and scope, and that will necessitate larger and larger bandwidth reserves. The problem I see with that, is again the rate of growth relative to computation requirements but moreso development times. Writing shading programs with a lot of branching and loops like Cell might entail, is going to be hellish to debug and that by its own nature will limit the scope.

IMO, more memory is better than faster memory for the next generation.
 
Fred said:
I dont see why we need to keep scaling bandwidth up at the same rate that fillrate and processing power goes up.

I would never expect this comment... out of you.

The argument is that pixel and vertex shaders are going to keep increasing in size and scope, and that will necessitate larger and larger bandwidth reserves. The problem I see with that, is again the rate of growth relative to computation requirements but moreso development times. Writing shading programs with a lot of branching and loops like Cell might entail, is going to be hellish to debug and that by its own nature will limit the scope.

First off, why do you assume the current PC status-quo architecture will apply to PS3? Any achitecture built around the basic ideal of utilizing a mass of reductionist cores which can together surpass the common Von Neumann architecture in tranistor/preformance and even absolute preformance will see that the ultimate wall encountered is feeding the massivly parrallel array that emerges in the design.

Pixel Fusion (circa 2000) was this way, as is Cell. Any architecture designed for dynamic media processing as basically envisioned by Diefendorff has useable bandwith as an uttern necessity.

IMO, more memory is better than faster memory for the next generation.

Typical PC user.... ;) AFAIK the basic tenets of dynamic processing contradict this, especially on a closed box PC. You need to keep those sea of processing units fed.

PS. Your responce to the thread on Cg was ultimatly wrong, just never got a change to repsond. Compering the Cube and it's architecture which is hardly PC-centric as an example was fallicious. Rather look at what the commonality that Cg would bring by paralleling the XBox's easy compilation/porting to the PC and the influx of games that are nothing but PC ports....
 
randycat99 said:
Psychogenics said:
$399 isn't a bad price if it come with Blu-ray,Remember Sony will advertise the Cell as the most versitlie Chip out in the Market being compatible with many other Sony/Toshiba/IBM products.....maybe they can sell a stripped down version without Hdd and Tivo feathers a $100 less.

You can argue any price is "reasonable" then, as long as you keep throwing on more fancy components. The average consumer isn't going to care if it has BR or not- just that it plays the game they want when they put the disc in, and the cost is within their reach. $300 is about where that threshold is between "cost within reach" and "cost out of reach". $180-200 is the jackpot price, though.

Randy they could throw in a kitchen sink but if everything in there doesn't scale down in price fast enough they are screwed.


Say the xbox 2 and the ps3 come out . The xbox 2 is 300$ , the ps3 is 400$ Both sony and ms loose 200$ on each system. Ms is already in the advantage. They have a system that costs a 100 less to make and they are able to sell it retail for cheaper at launch. Now say the ram in the ps3 has alot of problems and doesn't drop in price So the ps3 only drops on average 10 dollars a year. But the xbox 2 drops 25. This puts ms at an even larger advantage. It be very hard for sony to compete against a 200$ xbox that isn't loosing money. Kinda like whats happening now to ms . Its esp harder if the xbox 2 launches first.

There is alot more for sony to think about than if something is cool to add to the ps3.
 
I would never expect this comment... out of you.

I think his comment is fair. With sufficient caching, which is what the PS3 should have with all the embedded RAM, more memory (in this case the lowest level in the storage hieracrchy before we hit disks) is a good thing.
 
Hey I like Cell, and im well aware of its advantages. However, quantitatively it strikes me that we are hitting a point in graphics where bandwidth is less and less important, and computation power more and more (one of the advantages of Cell IMO). I am not arguing against the fact that more bandwidth is better and that the sea of processors needs to be perpetually filled. But we also have to identify that it assumes defacto that the bottleneck is bandwidth, which need not be the case.

Ultimately, mathematically lets look at the problem we need to solve. We have a set amount of pixels to fill, a set amount of texels to fetch. A large number of tasks that were once the primary bandwidth hog, have hit a point where we simply do not need all that more to reach a satisfactory solution or at the very least the problem has moved to a different stage (eg framebuffer vs texture bandwidth).

The future trends in gfx seems to be doing a lot more math per pixel (that seems to be growing fast), rather than fetching and writing (which seems to have stayed roughly constant) Eg, the calculations are by and large growing in the shader stage (this irregardless of architecture).

So while I am well aware of the PS2 model vs the PC model, where we send small chunks of data over a very fast bus, ultimately this cant remain insensitive to the actual math that is required by the problem per se.

Now, you could argue for an increase in bandwidth due to things like physics and low lvl crosstalk, but from experience typically that is more sensitive to low latency memory moreso than a wide data path.
 
jvd said:
randycat99 said:
Psychogenics said:
$399 isn't a bad price if it come with Blu-ray,Remember Sony will advertise the Cell as the most versitlie Chip out in the Market being compatible with many other Sony/Toshiba/IBM products.....maybe they can sell a stripped down version without Hdd and Tivo feathers a $100 less.

You can argue any price is "reasonable" then, as long as you keep throwing on more fancy components. The average consumer isn't going to care if it has BR or not- just that it plays the game they want when they put the disc in, and the cost is within their reach. $300 is about where that threshold is between "cost within reach" and "cost out of reach". $180-200 is the jackpot price, though.

Randy they could throw in a kitchen sink but if everything in there doesn't scale down in price fast enough they are screwed.


Say the xbox 2 and the ps3 come out . The xbox 2 is 300$ , the ps3 is 400$ Both sony and ms loose 200$ on each system. Ms is already in the advantage. They have a system that costs a 100 less to make and they are able to sell it retail for cheaper at launch. Now say the ram in the ps3 has alot of problems and doesn't drop in price So the ps3 only drops on average 10 dollars a year. But the xbox 2 drops 25. This puts ms at an even larger advantage. It be very hard for sony to compete against a 200$ xbox that isn't loosing money. Kinda like whats happening now to ms . Its esp harder if the xbox 2 launches first.

There is alot more for sony to think about than if something is cool to add to the ps3.


The whole point of a console coming at a higher price is to increase the power. If your suggesting that Microsoft would match the PS 3 hardware specs with a lower priced console, I say no way. More powerful specs costs money, they would be losing even more money then they are now.

A $399 PS 3 is a nightmare situation for MS in my opinion. Sony has the hype and momentum to justify a machine at this price to the public. Microsoft would have a hard time convining people to plunk down $399 for its console after the inital release. What does MS do? Stick with $299 and a lower speced machine? Go higher with a $349 or $399 and try to match the PS 3?
 
Back
Top