End of Cell for IBM

[Mobius1aic]

Its a good point, though i dont see BC as the best solution. Id rather they alowed me to trade/gift/sell my content on a built-in marketplace so that DD isnt always at a disadvantage in this respect. If im stuck with something i dont want anymore, and would have given it away or sold it, BC doesnt help at all.

If a completely DD system was implemented, then demos both MP and SP would be very important considering I wouldn't be able to try things out first, though it's a very similar predicament with me and my PC gaming. However, most of what I play had a demo in the first place for me to try and it's not an issue. I very much like Capcom for always creating some kind of PC benchmark for PC versions. While playability would be nice, the word on gameplay is already out considering these PC versions are out months after their console brethren, so we just need to know about performance. Now if only Capcom wasn't so Nvidia-centric......... Even still, these days, most major PC releases get a demo, huge holiday titles like Left 4 Dead 2, to less publicized stuff like Tropico 3. So far, MS has been pretty good about this on the 360, hell, all the platforms have. It's just that going with a DD focused model places extra pressure on demo availability and publishers can't hope too much for some misinformed holiday shopper (IE grandma) buying some random crap as a present.

 

[specwarGP2]

Its a good point, though i dont see BC as the best solution. Id rather they alowed me to trade/gift/sell my content on a built-in marketplace so that DD isnt always at a disadvantage in this respect. If im stuck with something i dont want anymore, and would have given it away or sold it, BC doesnt help at all.

They're not gonna let us trade/gift/sell DD... cmon man, we're not in the matrix!!!

 

[ShadowRunner]

They're not gonna let us trade/gift/sell DD... cmon man, we're not in the matrix!!!

Yea i know they wont, but they should.

DD will always come with pro's and cons like this. The fact that one day something you download may not work anymore is a potential con that people should take into account when purchasing like all the other issues.

 

[patsu]

http://www.andriasang.com/e/blog/2009/12/25/ps4_cell/

According to Goto, Sony, IBM and Toshiba recognized some of the problems programmers were having with Cell and, early on, came up with a couple of plans for fixing the issues. Included in these was a plan for something that was at one point called "SPU2." This new version of the SPU would shift the 256 kilobyte local store space that's included on each SPU chip into the role of a hardware management cache, allowing the SPUs direct access to main memory and allowing programmers to program for a single memory space, similar to a standard PC CPU (for Japanese readers, the original article explains this in much greater detail).

Goto says there are signs that some time this summer, Sony was looking into using this updated SPU design in the core of the PS4. At the very least, he says, the design was a strong candidate.

However, he's recently started hearing about other plans. While he's unable to get into the specifics, at a broad level, the new plans call for a PC-like multicore setup.

...

It's work in progress, so plans may not be cast in stone yet. Might want to treat it like a regular rumor for now.

 

[JardeL]

Thanks for the news patsu.

I'd like to know the benefits of this kind of [ direct access to memory / SPU vs. SPU2 ] enhancement.

 

[Shifty Geezer]

Thanks for the news patsu.

I'd like to know the benefits of this kind of [ direct access to memory / SPU vs. SPU2 ] enhancement.

Ease of programming, but also a drop in memory performance and data throughput in some cases. I think this change (as I read it) is about compromising the design of Cell, dropping the pursuit of ideal performance and accepting that developers aren't that interested and would much rather have good, easy performance, than hard, mind-boggling performance. After all, if half your processing power goes unused, there's no point having it!

 

[obonicus]

Would there still be any advantage in sticking with the asymmetry?

 

[semitope]

I am a bit afraid it will become a situation like what the 360 has (yes the 360 has a situation). Games on the ps3 continue to get noticeably better and better as its life continues and thats partly due to its learning curve. Also exclusives on it tend to be of good quality and worth buying.

I kind of agree with

"We don't provide the 'easy to program for' console that (developers) want, because 'easy to program for' means that anybody will be able to take advantage of pretty much what the hardware can do, so then the question is, what do you do for the rest of the nine-and-a-half years?" explained Hirai.
Huh? But his explanation didn't end there.
"So it's a kind of--I wouldn't say a double-edged sword--but it's hard to program for," Hirai continued, "and a lot of people see the negatives of it, but if you flip that around, it means the hardware has a lot more to offer.

as long as the system is functional , at launch the ps3 was a mess, and has a lot to offer from the start, which is likely with the current knowledge of ps3 programming already around, it should fair well with a setup that presents some challenge to fully exploit.

On the issue of asymmetry, if sharing a pool of memory will slow down either the gpu or cpu access then it should be separate. The RSX already can get to the xdr the cell uses so separate pools with a wide enough bandwidth to minimize the effect sounds ok.

 

[Rangers]

9.5 years? perhaps somone should show Hirai that PS2 had 2% software market share last months NPD..PS3 will be lucky to even have that at 9.5 years, since it has sold much worse than PS2 so far.

I suppose ease of programming is a double edge sword, but 360 games have shown vast improvement as well, possibly even more than PS3 has shown imo. Gears of War once wowed me, it was a vast step up on any game before it on 360 in 2006. Now there are many 360 games that make it look average (RE5, Alan Wake, Gears 2, MW2, Battlefield BC2, Mass Effect 2, Final Fantasy 13, etc). The 360 launch lineup graphical star imo was PGR3..which says a lot. I bought Perfect Dark Zero at launch and it was barely next gen. I also well remember GRAW knocking everybodies socks off in March 2006, another downright ugly game today. PS3 on the other hand had Motorstorm at launch and Heavenly Sword soon after (IIIRC) which were great looking games from the start. I dont know if you can argue that the PS3 with Cell has truly advanced any more than 360 with traditional tri-core, comparing launch to today on both. Even allowing that those in the know on B3D know the 360 was hampered early by weak dev kits.

The bottom line is that if it's great looking games you want, then focus on the GPU not the CPU. Every dollar spent on the GPU will make better looking games than a dollar spent on the CPU, to make it do GPU things.

 

[patsu]

PS3 on the other hand had Motorstorm at launch and Heavenly Sword soon after (IIIRC) which were great looking games from the start. I dont know if you can argue that the PS3 with Cell has truly advanced any more than 360 with traditional tri-core, comparing launch to today on both.

It'd be subjective anyway. Besides technology, art direction will affect the visuals as well -- especially when you're comparing completely different games. Early PS3 dev kits were buggy and incomplete too. But if you simply compare launch titles to today's titles, the 360 has one more year of development. May be that's why you see more improvement there since launch.

On the PS3, I'd say there is a visible difference between MotorStorm and MotorStorm 2, or between Uncharted 1 and Uncharted 2. It's more than pretty pictures. The scale, scope and amount of activities have improved. The AA in Saboteur looks stellar too compared to any launch or new titles. The SPUs are great for advanced audio systems too.

I do agree that the GPU will make the most difference visually. However, I'm also curious to see if the SPUs can really handle natural interface with complex graphics. e.g., adding head tracking to Gran Turismo 5.


As for PS4 (rumors)... we really need to look at the overall system architecture to see how issues are solved. Using a PC multicore architecture would mean general purpose code runs faster. They may delegate specialized tasks to the new GPU and other natural interface hardware (SpursEngine ? FPGA ?).

 

[Lazy8s]

Rather than require workloads to conform to such arbitrary and narrow limits in order to be processed efficiently, a CPU should be good at a flexible range of workloads -- able to switch contexts with little overhead (or no overhead, like Meta), -- even for processing gameplay as opposed to running a PC OS.

How this wasn't obvious to the Kutaragi regime is incredible, yet decisions like that are why they got themselves displaced at Sony. At least now, the company's hardware engineers are free to select the most competitve solutions from the open market -- in the handheld sector, anyway, where the legacy won't be as difficult to handle -- and not forced to use inferior, homegrown tech.

 

[Lucid_Dreamer]

I just love the use of "regime", "forced" and "inferior" in that post. Someone is making use of all the negative power words as a means to an end.

Inferior compared to what exactly? To me, the top end results on the PS3 are very much superior to any other consoles' games. Apparently, the PS3's design benefits gamers. As a gamer, should I care about anything else besides a more robust gaming experience? As a developer, isn't the fun suppose to be in the mental challenge of taking advantage of the hardware's STRENGTHS in order to bring the robust gaming experience into reality?

It should be interesting to see the design they choose for the next-gen. Sony's decision to bow to some of the developers or stay dedicated to the best hardware performance for gamers (and the product lifecycle which is also a benefit to gamers) will be reflected in their architectural choice. That's for sure.

 

[semitope]

I just love the use of "regime", "forced" and "inferior" in that post. Someone is making use of all the negative power words as a means to an end.

Inferior compared to what exactly? To me, the top end results on the PS3 are very much superior to any other consoles' games. Apparently, the PS3's design benefits gamers. As a gamer, should I care about anything else besides a more robust gaming experience? As a developer, isn't the fun suppose to be in the mental challenge of taking advantage of the hardware's STRENGTHS in order to bring the robust gaming experience into reality?

It should be interesting to see the design they choose for the next-gen. Sony's decision to bow to some of the developers or stay dedicated to the best hardware performance for gamers (and the product lifecycle which is also a benefit to gamers) will be reflected in their architectural choice. That's for sure.

Exactly.

To add, the cell is a great console processor IMO and anything replacing it better be able to do similarly. The processor in desktop PCs today and the one in the 360 cannot do what I think is necessary. The CPU should have GPU capabilities to supplement whatever GPU is included. It would most likely excel at whatever other tasks are programmed for it if it can do this and it would be able to bump graphics beyond any system using a separate (non graphics/or weak graphics capable CPU) with an equally powerful/slightly more powerful GPU. IMO if the RSX was equal to the xenos (is it?) the ps3 would easily be capable of better looking games simply because it has that extra bit in the processor. That applies to the next generation as well, if MS uses a quad core similar to i7 for example and use a GPU that ends up just as powerful as the one in the PS4 then a cell based processor would give the advantage. Should take the example from the PC realm where u can have GPU/CPU bottlenecks just because their tasks don't overlap enough (Dx11 addresses this a bit?).

Its also amazing to hear about devs doing graphics work on the cell and STILL doing amazing physics amongst other things. General purpose cores sure, but add some spus or the equivalent in another architecture. This is not exactly "general purpose" work the system using the processor is intended for. In super computing environments and in gaming and similar implementations its not general purpose.

 

[DuckThor Evil]

I kind of agree with

http://news.cnet.com/sony-ps3-is-hard-to-develop-for-on-purpose

I don't, I think that's something that sounds "kinda cool" to some people, but in reality offers no benefits whatsoever. In no way would it have been worse if Sony had games the quality KZ2 or Uncharted 2 at launch. Every system will always show improvements as the system, tools and the experience of the developers matures, but there is absolutely no sense in purposefully hindering the ease of development.

edit: and to be honest, I think Kaz was just saying something that was supposed to sound good, given the situation they were in. I wouldn't be surpised if instead of planning a hard to develop chip, he'd rather have something else in there and maybe not just in hindsight either.

 

[patsu]

Every system will always show improvements as the system, tools and the experience of the developers matures, but there is absolutely no sense in purposefully hindering the ease of development.

Yap ! I don't think the purpose of Cell is to hinder ease of development. It's supposed to achieve higher performance for developers who're willing to go an extra mile -- regretably, at the expense of ease of development in the general case.

 

[obonicus]

In no way would it have been worse if Sony had games the quality KZ2 or Uncharted 2 at launch.

It would have been worse for your man-baby gamer, who would be all 'sigh, this generation has peaked already, where are my end-of-generation surprises?' But then anything that annoys those guys is a net positive.

 

[NRP]

This generation has shown me how effective marketing and PR can be. I mean, just look around. After all these multiplat games, post mortem analyses, dev accounts, etc, some people are still drinking the Kool-Aid.

 

[patsu]

We have both positive and negative dev accounts though. The truth is usually somewhere in between. And the developer's effort will still count the most. e.g. We have a cross-platform middleware provider who tries to maximize extra resources on the Cell.

Comparing SingStar at launch and now, they added SPU voice recognition and more may come with renewed interest in natural interfaces. There are also interesting games like EyePet and Eye of Judgment today. I am still very very interested in Saboteur's technique. Unfortunately, these are isolated cases and do not represent any general trend.

In any case, the WatchImpress article seems to report on what Goto heard over the last few months. We don't know what's the real situation yet. Personally, I think enhancing/getting rid of the PPU is a good thing. Not sure what they intend to do with the SPU portion yet.

 

[Shifty Geezer]

(and the product lifecycle which is also a benefit to gamers)

I don't understand this thinking. Deliberately making the hardware hard to use so it shows better improvements over the product lifecycle? You could achieve the same with a simple many-core processor and just disable half the processors and clock it at 2/3rds clock speed for launch, then gradually unlock potential over the years. Ooooo, look how the developers increasingly harness the power of the console then! I don't particularly care to have my first 2 years of games 'rubbish' and look forward to a following two years of good titles, if my first two years could be good also. Developers will learn to do things differently. eg. SSAO. Let's leave it to software engineering and improved design and a more open market to provide new and wonderful experiences, rather than gimped and awkward hard just for the sake of being gimped/awkward artifically limiting what developers can achieve. I'm sure the developers would prefer easy systems too.

 

[jonabbey]

I don't understand this thinking. Deliberately making the hardware hard to use so it shows better improvements over the product lifecycle? You could achieve the same with a simple many-core processor and just disable half the processors and clock it at 2/3rds clock speed for launch, then gradually unlock potential over the years. Ooooo, look how the developers increasingly harness the power of the console then! I don't particularly care to have my first 2 years of games 'rubbish' and look forward to a following two years of good titles, if my first two years could be good also. Developers will learn to do things differently. eg. SSAO. Let's leave it to software engineering and improved design and a more open market to provide new and wonderful experiences, rather than gimped and awkward hard just for the sake of being gimped/awkward artifically limiting what developers can achieve. I'm sure the developers would prefer easy systems too.

The PS3 was not made difficult to program to preserve a ramp up on the learning curve, it was made difficult to program so that it would have a better chance at remaining price/power performant as the rest of the industry continues along the Moore's law curve.