Graphical parity, yes. But at what cost? I believe Sony's financial losses from selling the hardware far, far exceed those of its direct competitor.
Was Cell any good? *spawn
- Thread starter Shifty Geezer
- Start date
- Status
Do you think is it caused to the cell cost?
I think the problem was more to the bluray drivers cost... cell it's right & damning impressing to be a cpu of 2006. I can't to imagine what the ps3 could have does with a gpu at the same level of the cell...Hey Tuna, i could be wrong, but.. i actually think in this case.. i am not
Without the Cell the PS3 would be a disaster all thanks to a obvious GPU screwup/disaster that left them with a RSX that was not in anyway what i think they intended from the start. It made Nvidia alot of money though
The CELL is basically the reason the PS3 can hold it´s own against a strong competition where the obvious weak points are much more rare.
It would be alot more fun to have this discussion if the Cell had been married with a strong GPU, then there might have been a point about wasted potential.
i dont normally put my 2p in but, most of the advanced cg techniques this gen have either been on cell or p.c. also it helped kick start the "job/task manager" style of thread management.
we cant say if it has made or lost money for them as this cycle is not over (the end is not even in sight yet we don't have any detail on what the next home playstation will look like hardware wise, so the $400m may end up being returned 10 fold.).
it was the first mainstream multi core cpu (before intels "q" class core 2's i think) so it will have its place in history
I would still like to see what a ps3 powered by a cell + cell of cells (if i remember the old ps3 rumours right) or cell + massive fill rate gpu (ps2 style) would be able to display! I also still want a dev ps3 (and manuals) one day...
i have a question, how difficult is it for cell to talk to memory used by rsx?
we cant say if it has made or lost money for them as this cycle is not over (the end is not even in sight yet we don't have any detail on what the next home playstation will look like hardware wise, so the $400m may end up being returned 10 fold.).
it was the first mainstream multi core cpu (before intels "q" class core 2's i think) so it will have its place in history
I would still like to see what a ps3 powered by a cell + cell of cells (if i remember the old ps3 rumours right) or cell + massive fill rate gpu (ps2 style) would be able to display! I also still want a dev ps3 (and manuals) one day...
i have a question, how difficult is it for cell to talk to memory used by rsx?
I don't see how one can use RSX as a validation for Cell's use in the PS3. MS had Xenos in 2005. Sony should have been able to have something similar by 2006 if a strong GPU was part of their design focus. After all, Nvidia launched the Geforce 8 series the same month that Sony launched the PS3 and actually ended up first to market with a unified shader architecture in the PC space. If Sony hadn't put so much money and deign focus into Cell maybe they could have used those resources to include a better RSX.
Also, since IBM's work on Cell's PPE seems to have formed that basis of what became Xenon, it seems worth wondering what both the 360's and PS3's CPUs would have looked like if Cell never existed. Would IBM have still been able to produce something like Xenon for both?
Also, since IBM's work on Cell's PPE seems to have formed that basis of what became Xenon, it seems worth wondering what both the 360's and PS3's CPUs would have looked like if Cell never existed. Would IBM have still been able to produce something like Xenon for both?
From memory, the three companies were willing to spent $400 million developing Cell. Before STI sat on the table, IBM already has some form of the PPU and Toshiba already have the early SPU. I guess Sony brought in the vision of putting PPUs and SPUs on a single chip with ringbus and DMA engine as well as north and south bridge. I think Sony only went to IBM for the manufacturing process. If not it would just be Sony and Toshiba like the PS2 Emotion Engine. Maybe with ARM or MIPS for the PPU.
So I don't think it was that expensive relatively in developing Cell. NV spent far more money in developing the architecture that becomes RSX and their 6000 and 7000 series line up.
In a way Cell is using off the shelf components unlike GPUs from NV or ATI.
I think more interesting question is why Sony went with NV RSX ? That must be desperation.
My theory went like this, Sony followed the same development timeline for PS3 as they did with PS1 and PS2. That is they had developed a really powerful rasteriser with no processing units that can make it into a functioning GPU. So like PS2 they need a CPU that can provide the FLOPS. So again like Emotion Engine they went to Toshiba for some DSPs and later to IBM. At this point Sony is investing in Cell development and the manufacturing of their GS successor.
At some point Xbox 360 announced and launched with 512 MB of memory, first signed of trouble, Sony only intended PS3 with 256 MB, since that's what they had ordered. Later IBM only managed to put in a single PPU and 8 SPUs into Broadband Engine, far less than what Sony needed to power their beastly rasteriser. I think at that point they had Toshiba GPU and NV GPU for consideration. They picked NV to meet their FLOPS requirement for PS3 and they can attach another 256 MB to match Xbox 360, problem solved. Only later, they discovered RSX can't emulate PS2 and it went downhill from there.
The early 360 ports look bad on PS3, it took years before it's near parity. I think if Sony had dropped Cell as well at that point and went with some other CPU, PS3 would never catch up to 360.
Not too sure if PS4 will be PS3 done right, ie one 2 billion transistors chip dedicated mostly to SPUs for FLOPS and another 2 billion transistor processor dedicated to raster. Looking at Vita, it seems the current Sony is more concerned with ease of development than crazy architecture.
He hasn't said nothig new or surprising imho... but I'm pretty sure Rage isn't built with ps3 architecture in mind and I doubt Carmak is interested to push the sony hardware in some way. He has more interested to follow his 'path' & his ideas.
Last edited by a moderator:
D
Deleted member 13524
Guest
assurdum, are you using some kind of automated text translation? Some sentences in your posts make no sense at all...
Ok, I cut the post. My apologies.
Or maybe they were pushing the PS3 pretty hard, and it's just not as good as the 360 at the technology they wanted to put in the game? A multiplatform game can't afford to tiptoe around all of a particular systems weaknesses like a first party game can, and it can't hide inside the PR/perception bubble that comes from it never being seen running on the other guys machine.
RSX seems to have gifted the PS3 some significant weaknesses that multiplatform games can't afford to tiptoe around. Cell might have done that too, at least initially, but that certainly doesn't seem to be the case now. It appears to be carrying out the same stuff that the 360 CPU does, along with propping up RSX.
If you judge how good a CPU is by what it actually ends up doing then I don't think there's any question that Cell is "good", even if Sony's approach to designing the PS3 wasn't the best one (but that's not really Cell's fault is it?).
How about you address his specific points? He talks about IO and the rasterizer, RSX, etc. Do you have anything to dispute in the context of a multiplatform game or are we going to have to hear these same comments with almost every game where a developer makes a comment?
This is from 2 years ago when Rage was running in 20fps on PS3. Now he says that both consoles are around the same.
Also, Carmack opinons about both consoles changed with times:
2007 (or 2008?) - PS3 slighty stronger than 360
2009 - PS3 weaker than 360
2010 - PS3 slighty stronger than 360
2011 - both consoles are around the same, but he prefers trades on 360
Nope.
2005 post E3 - Geoff's interview, PS3 more peak performance
2006 - 2009 - Anything strictly graphical is slower on PS3 because rasterizer
2010 - Performances are pretty much the same, both consoles very close in actual performances.
2011 - PS3 IO troubles and memory limitations.
Pretty much spot on IMO as PS3 version tends to have the same performances as 360 version, dropping only resolution( but he could have gone to dropping frames instead, no body would even bother). Pop in is another problem which they seem couldn't deal with completely on PS3.
Thanks for corrections. As for 2010 (which is april 2011 actually, crappy memory ):
"You have to use a separate tool chain, the debugging is crappier, and all this. The upside of that is, there is more raw performance for computing there than there is on 360"
http://www.gamingunion.net/news/carmack-ps3-has-more-raw-power-than-the-xbox-360--4657.html
):"You have to use a separate tool chain, the debugging is crappier, and all this. The upside of that is, there is more raw performance for computing there than there is on 360"
http://www.gamingunion.net/news/carmack-ps3-has-more-raw-power-than-the-xbox-360--4657.html
Carmack is just talking about his game and the way he uses Cell. "Performancewise, the main CPU is about the same" doesn't quite tell the full story. ^_^
Also, Sony did not allow full install. In othe posts, his devs mentioned that unbuffered IO and more memory would help. Carmack's himself wasn't entirely satisfied with the streaming implementation too. And 360 version also has popins but less obvious ?
Cell is interesting not because it is used to overcome GPU bottlenecks. It helped to advance GPU work too. Those postproccessing quality is different from GPU based ones on the consoles. Moving the entire level independently while playing the game is interesting too. Hope the devs find more use for it !
Also, Sony did not allow full install. In othe posts, his devs mentioned that unbuffered IO and more memory would help. Carmack's himself wasn't entirely satisfied with the streaming implementation too. And 360 version also has popins but less obvious ?
Cell is interesting not because it is used to overcome GPU bottlenecks. It helped to advance GPU work too. Those postproccessing quality is different from GPU based ones on the consoles. Moving the entire level independently while playing the game is interesting too. Hope the devs find more use for it !
Are you talking about Uncharted? Because it's annoying when it happens in Gears 3 since it throws off your aim. Or do you mean some other game where cell is doing something impossible on anything else? Please clarify.
Last edited by a moderator:
What I am said of so blatant?
I only think ps3 & Rage are two different 'beasts' with different 'fates'; Carmak surely not interests to extrapolate the best of the platform without 'future' perspective but wisely follows his path. From a ps3 perspective, I don't know how good has been begin the works on 360 (I remember Bioware has to have to rewrite the whole mass effect engine just to catch the same quality in the multiplaform development, starting on the ps3), I only said to have some doubt about this choice but what has been done, has been done, so... Although I found Rage 'virtually' the same on the ps3, I beg to hypotize probably something more technically we can to achieve on the ps3 (not that much) but why if anyone would have noticed the difference at the end? The same Carmak claims more or less the same thing: we can outperform both platform working specifically in the single, but I trade 360...
Last edited by a moderator:
- Status
Similar threads
- Locked
