What will Sony add to their RSX GPU development?

[Shinjisan]

As i said, reached a certain level of performance, any extras are just unwanted expenses Sony will try to avoid at all costs, considering how expensive PS3 is already.

I'm not saying "they'll put a GPU that's good enough" as if to say they could put a NV40 and be done with it.

I'm saying that reached a certain level - say G70 level, just an example - anything extra is just unwanted cost.

Just like they could push for Cell to have all 8 SPE's enabled, but a 7 SPE's Cell is "good enough", after taking into consideration the level of performance they need.

By your reasoning, Sony would just keep adding features and more advanced hardware just to reach this utopian "best ever hardware" status, when it really isn't needed - after a certain level as i said.

1GB Ram would be perfect and make PS3 much more powerful, but will they do it? Of course not.

Sony have performance targets for RSX and will stick to them. Anything extra will be scrapped.

Consoles are all about compromise, remember. The platform will be very powerful, but not unneccessarily powerful, or we'd end up with a NeoGeo kind of scenario, with a vastly more powerful hardware than the competition that's just too expensive for anyone to care. Sony need to sell PS3 at a certain price point and will try to cut their losses any way they can.

I totally agree that consoles are all about compromise.
But it they have to buy a 90nm GPU and they plan to have the product out in late spring 2006 they don't buy a GPU developed around a 110nm process.
That's not compromise,it's dumb.
A good compromise is to keep the frequency at 550MHZ which for a 90nm GPU is definetly low,PC counterparts will reach 650-700MHZ.

 

[Jawed]

The 128-bit GDDR3 bus indicates a pretty constrained design for RSX.

Bearing in mind that "G71" has the appearance of being designed to run with memory at nearly three times the bandwidth of RSX (~60GB/s versus 22.4GB/s - G71 may be designed to work with GDDR4, 80GB/s+), it seems like there'd be an awful mismatch if RSX was actually "G71" (whatever that is). By "mismatch" really what I mean is "disproportionately expensive part" rather than "won't work properly".

RSX having 60% of the bandwidth of G70 is not something to be proud of.

Jawed

 

[Dave Baumann]

I would still suggest half the number of ROPs as well. Removing half the memory interface and still having 16 ROPs doesn't really fit - even at 8 ROPs that would still be twice the number of ROPs for any current 128-bit graphics chip being built by the graphics vendors (but RSX's bandwidth is more favourable, so 8 makes more sense).

 

[silhouette]

I totally agree that consoles are all about compromise.
But it they have to buy a 90nm GPU and they plan to have the product out in late spring 2006 they don't buy a GPU developed around a 110nm process.
That's not compromise,it's dumb.
A good compromise is to keep the frequency at 550MHZ which for a 90nm GPU is definetly low,PC counterparts will reach 650-700MHZ.

Shinjisan, designing a chip for 90nm is not a simple scaling down from 130/110nm. Because of increased leakage currents, they need to re-design some parts of the chips to cope with that problem.. otherwise they end up with a chip which consumes more power than its 130/110 nm counterpart and also produces more heat (which will be a bigger problem when put in a small form-factors like consoles). On the top of that, the design rules for Sony's foundaries might be slightly different that those for TSMC (G70s are manufactured there). So, NV also may need to change the design of the chip for that. Add the new memory interface + removal of purevideo, it may take them this much to re-design/validate and test the real chip.

Btw, going from 550 to 700 Mhz is not free as well. They need to increase voltage which will result in more heat, which will become a bigger problem again for small spaces such as in PS3. I am not suprised that 550 Mhz would be close to the limit that will satisfy their power/heat requirements.

 

[leechan25]

Sony not worried about cost

look guys,

Sony is saying it wants PS3 lifecycle to last about ten years. Why in the world would they simply put a year old GPU with minor mods when their wanting this thing to last that long and still be competltive with future GPUs or concoles? For this reason I believe Sony is willing to spread their development cost over a ten year timeframe. That why we're are getting a blueray drive, cell microprocessor, true HD features, and a GPU that will pretty much outshine any PCish counterpart for some time in the future. So i'm asking you to be a little more forward thinking in your responds to my questions.

Cell is forward thinking

Blueray is forward thinking

So why would RSX be simply based only last years GPU.

There's more to GPU then meets the eye.

 

[Carl B]

10 years, yes... but not as their flagship console.

PS4 will come along roughly five/six years from now and pick up from where PS3 left off. This isn't saying anything about the GPU in PS3 mind you, but I don't think anyone's worried about putting a chip in there that will be competetive with the leading-edge eight years down the line. Because that's impossible.

 

[pc999]

I would still suggest half the number of ROPs as well. Removing half the memory interface and still having 16 ROPs doesn't really fit - even at 8 ROPs that would still be twice the number of ROPs for any current 128-bit graphics chip being built by the graphics vendors (but RSX's bandwidth is more favourable, so 8 makes more sense).

In that case they would save lot from those 300M+ transistores they said in E3 (as g70 already as 302M),right? if so this mean they should be doing some serios moddification, I guess.

 

[Dave Baumann]

302M transistors is the number for G70. I very much doubt that once they've added stuff and taken stuff away they will end at 302M again.

 

[Bobbler]

302M transistors is the number for G70. I very much doubt that once they've added stuff and taken stuff away they will end at 302M again.

The number given at E3 was just 300m+, as pc999 just said -- I'm not even sure they knew the exact number at that point (or maybe they didn't want competition to know -- tee hee).

 

[leechan25]

I see your point but I don't think these new consoles lifecycle should realistically be so short because developers have less time to overcome the technical learning curve. Additionally, covering develop cost of content has not evolve properly enough to fund these software programs that are be coming more like developing movies. It will take time for the gaming industry to make this transition.

Cost of PS3 game - about 5 to whatever millions of dollars

Cost of PS4 game - 15 million and up

I mean developers like carrmack are already crying because of Xbox 360 and PS3 tech curve . So a ten year lifecycle is more for the industry to development than simply just PS3's user base.

 

[Carl B]

Well he wasn't saying it was 302 though, he was using that number for G70, while relating that RSX is supposedly over 300 if E3 presentations are to be believed. So indeed PC999, good catch should bandwidth concerns possibly have NVidia and Sony looking to a chip with fewer ROPs, as per Dave's theory. Certainly that would imply some substantial logic coming to sub in to keep in excess of the 300 million figure.

Still though, I think it's best just to wait for info without putting too much effort into theorizing. It's almost just too open-ended at this point. All we know for certain is that it's origins are rooted somehwere in the G70 architecture.

 

[Carl B]

I see your point but I don't think these new consoles lifecycle should realistically be so short because developers have less time to overcome the technical learning curve. Additionally, covering develop cost of content has not evolve properly enough to fund these software programs that are be coming more like developing movies. It will take time for the gaming industry to make this transition.

Cost of PS3 game - about 5 to whatever millions of dollars

Cost of PS4 game - 15 million and up

I mean developers like carrmack are already crying because of Xbox 360 and PS3 tech curve . So a ten year lifecycle is more for the industry to development than simply just PS3's user base.

Well, be that as it may I honestly could see Microsoft pursuing a continued release cycle more aggressive than the five-year average that has become the de facto standard. I truly believe there is no way it will take 10 years for PS4 to be released, Sony was simply refering to the fact that they will support the system for ten years, as well as continue selling it as PSThree down the line. So it *is* a ten year product lifecycle, but not to say it breaks the five-year console new release cycle though.

 

[Fox5]

I see your point but I don't think these new consoles lifecycle should realistically be so short because developers have less time to overcome the technical learning curve. Additionally, covering develop cost of content has not evolve properly enough to fund these software programs that are be coming more like developing movies. It will take time for the gaming industry to make this transition.

Cost of PS3 game - about 5 to whatever millions of dollars

Cost of PS4 game - 15 million and up

I mean developers like carrmack are already crying because of Xbox 360 and PS3 tech curve . So a ten year lifecycle is more for the industry to development than simply just PS3's user base.

A 10 year life cycle with no upgrade is almost suicidal. 7 years in it would be easy for someone to come out with a vastly superior hardware at not much more cost, and even without developers taking advantage of it it could look much better.

 

[pc999]

If the PS3 demos are for real I doubt that anythingh could look much better, as some of them are (almost) photo-reallistic, and if we consider what good dev can make from a console...

More the added cost woulnd worth it IMO that I think is shared by many.

Plus the less the console life cicle is the less proffit they made, I really doubt they like to spend Billions in R&D unless they really need to do, and IMO there will pass a lot of time before they need.

 

[Carl B]

I don't understand what we're arguing here anymore. PC999 are you saying you're thinking a ten year lifecycle (before PS4) as well? That's craziness.

Listen, no one spent billions on R&D, they spent billions on fab roll-out. Now, that has a very real effect on the PS3 to be certain - but it's costs (and ultimate utility) exist outside of the Playstation roof. Even if PS3 were to have a one year life cycle and PS4 come out immediately, those billions on fabs, and the hundreds of millions on Cell R&D, would still be put to use.

I agree that longer gens equal greater profits, but no one's so unrealistic as to think that someones not going to push the button of a new gen *way* before the ten year mark.

PSOne lasted ten years, PSTwo will last ten years, so will PSThree...

That's what they're talking about, not ten years between now and PS4. People would have to really be drinking the kool-aid to consider PS3 to be packing the power to take on the XBox 720 and the Nintendo Exodus.

 

[pc999]

I don't understand what we're arguing here anymore. PC999 are you saying you're thinking a ten year lifecycle (before PS4) as well? That's craziness.

Not exact number like 10 but just longer than PS3, anyway if you dont like Sonys (as they will use it for others thinghs and yet we dont know which amout of is directely related to PS3) eg take the MS how spent 4B in R&D (from memory, sorry but no link) for XB360 plus they lost ?M to sell their firsts XB360 and we dont know till when they will sell bellow cost(many of them with problems that will cost even more), plus ads/PR, they had lost a lot of money they will only make this again if they need, plus the rest of the industry dont like much too (15Mx? in games cost, new tools, new programesetc...).

Do you really think that if we got consoles 10X more powerfull racing games would look (perceptible) better than GTV, MS, F1 demos on PS3, I really doubt it.

I am one how think that next gen should have waited a bit longer as current gen still blow everythingh that we expected from them see Black, FCI,SC4..., they should have waited for 65nm so they can have both better specs/features (just to make sure ), cooler/thiner and more affordable consoles with good next gen disks.

Then the biggest reason, constant new HW releases does not make the industry as mainstream as others (do you think that if people needed o buy a new VCR/DVD player 2-1 years after they get affordable in order to see new movies they would be as mainstream as they are , I am very sure that they wouldnt be, and once that their main proffit come from games the more consoles they have sold the more games they sold.

On the other side the lossing company may try to make the next next? gen to came sonner (like MS did) but I think that it will be much less efficient than ever before.

Dont curse me but I have my doubts if there will be a PS4, but I am almost sure that if there is a PS4 as a much better PS3 (spec) with a few improvements (like PS3) I am sure that will not exist a PS5,I think I am sure I will only think that a real next next gen is worth if they bring a real big new thing (VR, holographic scenario that evolves around you etc...) then PS4 (for me, probably for you is 5) will worth but after all it is only my opinion.

 

[Alpha_Spartan]

Expect the RSX to be a relatively mature, relatively inexpensive solution that can be passively cooled and produced by the millions. Neither the G70 or the G71 fit all of these criteria at the moment.

Once again, this thread is allowed to live because certain people relish the thought that just because Sony called it the "RSX" that it's going to be something new and revolutionary instead of a heavily modified version of a current GPU.

 

[scooby_dooby]

4B in R&D (from memory, sorry but no link) for XB360 .

are you kidding? 4B lost on XBOX1 according to a forbes 'estimate', not to mention 2billion of that was directly spent on XBOX Live. X360 r&d? 4 Billion?

My guess is MS will launch the new xbox after 5 years, which will be 4 years into the sony lifecycle. Sony will have a tough call at that point since the new 360 will be many times more powerful than the PS3. This whole 10year cycle thing is nonsense, keep dreaming. 5-6 year lifecyle until the new console launches tops.

 

[Guilty Bystander]

Hey guys I just thought about something here that could show why the RSX can't be the G71.

The G71 can do 36 shaderops per cycle.
As shown by nVidia in the Sony E3 conference the RSX can do 136 shaderops per cycle.

This would mean both GPU's could process the following amount of shaderops/s.

G71 (thought as a 550MHz GPU):
550MHz x 36 : 1000 = 19,8 billion shaderops/s

RSX:
550MHz x 136 : 1000 = 74,8 billion shaderops/s

This would really shed some light onto why nVidia said by the time the PS3 is launched worldwide our PC GPU's might have closed the gap in performance (on paper).

Also the Xenos can do 96 shaderops per cycle which would make it able of the following:

Xenos/R6xx Hybrid:
500MHz x 96 : 1000 = 48 billion shaderops/s

This 48 billion shaderops/s is precisely what M$ says on their Xbox.com website at 360 specs.
Check it out here.
This all would certainly show why we are seeing so much visual differences between 360 and even the highest end PC's using multiple GPU's if the 360 is coded for.
Just look at 1st gen games like Kameo with hundreds maybe thousands of fully detailed trolls going at it without any hickup all with the same high framerate or Ninty Nine Nights with thousands of enemies fully detailed all at 60fps.

Yeah I know about CoD2 being slightly less detailed than on a PC with MAX settings or Quake 4 dropping all over the place but couldn't this have to do something with just porting or even emulating the PC versions.
Surely the 360 does have more than enough power to emulate the PC and still run this game at 60fps.

This should also explain why Infinity Ward is coming with a patch for the 360 version to fix save game corruption problems.

What do you guys think?

 

[Bobbler]

are you kidding? 4B lost on XBOX1 according to a forbes 'estimate', not to mention 2billion of that was directly spent on XBOX Live. X360 r&d? 4 Billion?

Live probably cost a few hundred million at most -- there is nothing about Live that costs 2 billion dollars, not even close. Honestly, it doesn't cost billions to set up a matchmaking service and a software client that is able to DL videos, and game demos and have a friends list of sorts -- the most expensive part is the servers that can handle a few million users "logged" in at a time, and that is likely to be done rather simplistically since it isn't a mission critical situation (especially given the closed nature of the client/server). 2 billion is a lot of money.