Sony: 2M PS3s shipped

[Arwin]

He was also talking about "better" devkits - i.e. the Cell + G7x. Those configurations were far closer to the final system than the G5 + X800 kits for 360.

From a PC developers point of view, yes.

For the most part the main difference would be doing all the texturing from the 256-bit bus of the graphics board in the early dev kits, but then moving some of that to main RAM on the final kits if need be.

Again, if you develop as you do on PC, then yes. But if you were to play to the whole flexio's strengths, were going to use the Cell to be integrated into both ends of the graphics pipeline, then I'm not so sure. Ultimately the 360's architecture didn't get incredibly more complex from the early devkits, other than in terms of having to deal with tiling, which is why most just chose not to and sought solace in the arms of Ana instead. The 360 ultimately is about two things - optimising use of the 3 cores, and optimising the use of the Xenos' unique tiling/EDRAM architecture.

I hold my point that the PS3's architecture in many ways is far more foreign, but it all comes together in a way that simply wasn't available in the early devkits, and therefore early development resembled developing for comparatively 'weak' PC hardware, with very immature tools.

The reality is quite simply that PS3 development is behind on the 360's development. You can claim whatever you like, but Sony's platform is behind on the 360's in terms of development by a considerable margin, especially when taking into account the different technologies embedded in the two. The PS3 has the sixaxis, the Cell, the split memory architecture, Blu-Ray, and was slow in making APIs available for the sixaxis and online stuff. The 360 has mostly the tiling to worry about, but that problem was easy to mask through Ana, but did not otherwise introduce a lot of new stuff, and what was introduced was well-supported (again, except tiling). The 360 has made the lives of developers easier than the PS3, and is being handsomely rewarded for it. There is absolutely no need for arguing that PS3 developers had a head-start, except perhaps when it came to programming the shaders in the basic GPU side pipeline.

 

[TurnDragoZeroV2G]

Again, if you develop as you do on PC, then yes. But if you were to play to the whole flexio's strengths, were going to use the Cell to be integrated into both ends of the graphics pipeline, then I'm not so sure. Ultimately the 360's architecture didn't get incredibly more complex from the early devkits, other than in terms of having to deal with tiling, which is why most just chose not to and sought solace in the arms of Ana instead. The 360 ultimately is about two things - optimising use of the 3 cores, and optimising the use of the Xenos' unique tiling/EDRAM architecture.

What? Tell me I'm missing something here. FlexIO is a radical change, but a switch from limited framebuffer bandwidth to high framebuffer bandwidth, standard CPU-GPU bandwidth to 21GB/s, A move beyond just SM2.0, inclusion of Cache-enslavement from the GPU, MEMEXPORT and a Tesselation unit versus none of those.... those aren't significant in any way? That's totally unreasonable. If you're literally talking about becoming "incredibly more complex," then you're technically right. But that is also completely besides the point, because the PS3 didn't become incredibly more complex either. For that matter, the hardware changes from dev kits released back in early '05 to final kits were significantly smaller than the ones between alpha 360 kits and final kits--so if that is what you were trying to say with "incredibly more complex," then I just can't agree and can't see how you believe this to be true.

PS3 has to deal with split memory in final kits, but 360 had other trade-offs itself, such as the CPU having to go through the GPU to access main memory, a CPU that was significantly different from the alpha one with quite different strengths and weaknesses, and trading off higher bandwidth to main memory for very high framebuffer BW and OK main memory bandwidth, etc.

Of all the points you can make here, I don't believe alpha vs. final hardware is among them. You're downplaying the things 360 kits introduced that developers may have wanted to take advantage of, and downplaying the negatives that were introduced in moving to final kits.

 

[one]

For that matter, the hardware changes from dev kits released back in early '05 to final kits were significantly smaller than the ones between alpha 360 kits and final kits

There might be significantly bigger software changes, though!

 

[Arwin]

What? Tell me I'm missing something here. FlexIO is a radical change, but a switch from limited framebuffer bandwidth to high framebuffer bandwidth, standard CPU-GPU bandwidth to 21GB/s, A move beyond just SM2.0, inclusion of Cache-enslavement from the GPU, MEMEXPORT and a Tesselation unit versus none of those.... those aren't significant in any way? That's totally unreasonable. If you're literally talking about becoming "incredibly more complex," then you're technically right. But that is also completely besides the point, because the PS3 didn't become incredibly more complex either. For that matter, the hardware changes from dev kits released back in early '05 to final kits were significantly smaller than the ones between alpha 360 kits and final kits--so if that is what you were trying to say with "incredibly more complex," then I just can't agree and can't see how you believe this to be true.

PS3 has to deal with split memory in final kits, but 360 had other trade-offs itself, such as the CPU having to go through the GPU to access main memory

If this has significant implications on coding other than taking into account the shared bandwidth (which might be a pain by itself, mind), then I stand corrected. I didn't see this as causing a fundamental change in the pipeline or in the way you address main memory in your actual game-code, but maybe I am wrong.

, a CPU that was significantly different from the alpha one with quite different strengths and weaknesses, and trading off higher bandwidth to main memory for very high framebuffer BW and OK main memory bandwidth, etc.

Problems that were nevertheless finalised a year before they were on the PS3.

Of all the points you can make here, I don't believe alpha vs. final hardware is among them. You're downplaying the things 360 kits introduced that developers may have wanted to take advantage of, and downplaying the negatives that were introduced in moving to final kits.

That's not the point. People have been arguing that the PS3 developers have had less time to work with the final hardware to make full use of its architecture. The argument against this was made by stating that the devkits for the PS3 were closer to the final hardware earlier in time than the 360's were. I've just been trying to put that argument in the right perspective - i.e., the devkits that Sony developers got early, didn't put the 360s at a significant disadvantage at that stage of development because in terms of overall complexity of the pipeline the PS3 (not just on the graphics level) is just higher, and the 360 got final devkits out a year before the PS3 did.

This is all nothing shocking - you've got all these PC devs ranting on how the PS3 is a pain to code for, and you've got the whole strategy of MS based on having the 360 out there a year before the PS3 as well as give them great development tools and support (again, something praised especially by PC devs). I don't understand why now people find it suddenly necessary to claim that PS3 developers weren't at a disadvantage and use the 'devkit' argument.

 

[Laa-Yosh]

Once again the debate gets lost in irrelevant details.
The original point is that PS3 game releases are lagging behind the X360 games by a lot more than what the technical side could cause. There is a very obvious business related reason in effect here, which is the small size of the PS3's installed base.

 

[blakjedi]

and the 360 got final devkits out a year before the PS3 did.

But if the PS3 kits from the period when the 360's kits were finalizing were STILL closer to the PS3 final devkit than 360's were until 360's Finals were released... I just find your argument strange.

Once PS3 devs got a Cell, any GF6800 or better GPU would suffice (and did.) It seems that it has been easier to simulate PS3's final environment for a longer time than trying the 360's environment has been finalized.

 

[Kryton]

Once again the debate gets lost in irrelevant details.
The original point is that PS3 game releases are lagging behind the X360 games by a lot more than what the technical side could cause. There is a very obvious business related reason in effect here, which is the small size of the PS3's installed base.

Yes, that business argument is clear and not heard enough. Only issue with it is, what do you expect at launch? 360 does have more consoles 'out' and so you can expect developers targetted it first (assuming they were informed about the launch dates, I'm sure most were/had to be).

Once PS3 devs got a Cell, any GF6800 or better GPU would suffice (and did.) It seems that it has been easier to simulate PS3's final environment for a longer time than trying the 360's environment has been finalized.

I think the bus limitations and re-structuring of main memory would never have been emulated on a PC. Likewise, for the 360, you wouldn't get MEMEXPORT, some hybrid DX9-DX10 GPU, or the GPU controlling the I/O, but these seem redundant to getting games up and running with half-decent performance. The PS3 issues really hit hard at the common assumptions people would have made. Also, as was voiced many times, the lack of a solid SDK was a frequent complaint for the PS3 and not for the 360?

PS: Is the thread title still appropriate? This thread is going like a run-away train...

 

[Dave Baumann]

Errrm, its just bandwidth. Yes, much of what could be achieved on PS3 now could have been achieved on the earlier dev kits - some ops may just run slower (but still a damned sight better than emulation); but then much of that can be allievaed by the fact that the GPU in those early dev boxes had the bandwidth that wasn't present on FlexIO.

 

[Laa-Yosh]

Only issue with it is, what do you expect at launch? 360 does have more consoles 'out' and so you can expect developers targetted it first (assuming they were informed about the launch dates, I'm sure most were/had to be).

This is exactly the same argument that I've brought up a few days and pages ago - that launching a year earlier has indeed been a very good tactic for the X360 (I think Mythos was the one arguing with this). All we've had since about exclusives and other stuff was me trying to support that argument and then it kinda derailed the thread.
But now, the circle is complete

 

[nelg]

I think the bus limitations and re-structuring of main memory would never have been emulated on a PC. Likewise, for the 360, you wouldn't get MEMEXPORT, some hybrid DX9-DX10 GPU, or the GPU controlling the I/O, but these seem redundant to getting games up and running with half-decent performance. The PS3 issues really hit hard at the common assumptions people would have made. Also, as was voiced many times, the lack of a solid SDK was a frequent complaint for the PS3 and not for the 360?

What about Sony themselves? They had Cell in their hands before MS had Xenon, so why was the OS so long in production? This is something I thought might be telling in itself. If the designers of the CPU needed so much time themselves to finalize the OS for it, that may signal some difficulties. Furthermore, if they faced such difficulties, how hard would it be for others (devs) to achieve good utilization without the leg up that the creators themselves had?

 

[inefficient]

What about Sony themselves? They had Cell in their hands before MS had Xenon, so why was the OS so long in production? This is something I thought might be telling in itself. If the designers of the CPU needed so much time themselves to finalize the OS for it, that may signal some difficulties. Furthermore, if they faced such difficulties, how hard would it be for others (devs) to achieve good utilization without the leg up that the creators themselves had?

Cell is a radical design paradigm shift + Writing an OS and APIs and SDKs not the same as creating a game = Your answer.

 

[Shifty Geezer]

From a PC developers point of view, yes.

From any point of view! XB360's dev kits had an alien CPU and alien GPU for far too long. PS3 kits had the real CPU and similar GPU for yonks. Maybe on PS3 had the concern of different BW architecture in the Beta kits, but at least you could work on CPU code that'd drop straight into the final dev kits. On XB360, half your work had to be totally re-engineered!

 

[Acert93]

From any point of view! XB360's dev kits had an alien CPU and alien GPU for far too long. PS3 kits had the real CPU and similar GPU for yonks. Maybe on PS3 had the concern of different BW architecture in the Beta kits, but at least you could work on CPU code that'd drop straight into the final dev kits. On XB360, half your work had to be totally re-engineered!

I agree. I have been argueing since summer 2005 that I think Sony had a much better dev kit hardware approach. Sure, it wasn't final hardware--if it were, they could have shipped in 2005! But considering they had a real Cell chip in the kit and a NV4x class GPU (RSX is a NV47 derivative) that allows for a lot of testing and sampling code. In most circumstances your code is going to run faster on the final PS3 hardware.

The 360 went a reverse route. Xenon was late being finished, so the kits had G5s into the summer of '05. My guess is the kits had G5s because the original goal was OOOe PPC processors at 3.2GHz. The transition from G5 PPCs to Xenon is pretty harsh in many ways. And transitioning from a 9800 and X800 class GPU (DX9, SM2.0, 24bit percision) to Xenos (Hybrid featureset, SM3.0+, 32bit percision, eDRAM and tiled rendering, and unified shading pipeline) was probably a learning experience. As far as I know most of the other unique features of the GPU (tesselation, cache datastreaming, MEMEXPORT, and to a degree vertex texturing) haven't been used much because before 18 months ago there was no way to test them and see if there were useful implimentations of these technologies. They may all be bullet points without performance to back them up, so we may never see them (cough cough especially if MS keeps leveraging middleware and multiplatform titles cough cough) but I think, from a hardware standpoint, MS devs are just as far from tapping the hardware as Sony devs. Their may be higher ceilings on either end depending on what you are doing, but I think both platforms have a lot left in them just due to the fact of the dev kit situation and how long it takes to get a new title out the door (2 years or more).

Even a game like Heavenly Sword has a substantial amount of work done before it moved to the PS3. Imagine what DeanoC, nAo, and company will be able to do with a ground up effort from day 1 with HS2. Knowing the platform inside and out and exploiting the elements that give the best return is a valuable asset

 

[Johnny Awesome]

I would just like to point out that manufacturing costs alone don't dictate at what point one of the vendors can sell a console and still be profitable overall. Unless Sony significantly closes the gap in content sales (encompassing disc-based as well as download-able content) MS can seriously undercut the PS3's price, taking a loss on the hardware if necessary, and still make money.

Well, I would expect PS3 attach rates to be similar to X360 attach rates over the long term. There just aren't a lot of games for PS3 right now and the high price is preventing a lot of early adopters from getting the usual stuff even like Madden etc...

 

[Johnny Awesome]

I would think this just means that eastern publishers are going to have to follow Capcom's lead and look for a broader market. There certainly seems to be a strong worldwide demand for their games. If they continue to have a domestic approach to publishing then they are obviously going to have difficulties with a changing industry.

While there will still be a market for domestic only games, I think the publishers that think globally sooner rather than later are going to do better and not just for 360 games.

Very true. The evidence is the DQ move from PS3 to DS. That's a pretty significant change. It's looking like Japanese devs may be forced between choosing between handhelds in Japan and consoles everywhere else. If this is the case, then they have to at least consider the X360 for their console releases.

 

[Johnny Awesome]

quite a few errors.

DC was at 8/9M worldwide when the ps2 launched in the US and died around 11M. They sold 2M alone in Japan, something MS will probably never do.

In Fall 2000, Sony was 6 months after its Japanese launch and had ironed all its manufacturing problems, then was not supply limited.

You are right that the Dc died over night unlike the X360

Well the first thing to note is that the DC was $149 when the PS2 launched and X360 is $299/$399. That's a pretty big difference. The other thing is that outside of Japan, the X360 is doing quite a bit better than the DC. Just two points in a sea of differences IMO.

I imagine that if Sega was selling the DC for $299 when the PS2 launched and had 6 million of them in NA like the X360, then they would probably have stuck it out with the DC.

 

[Mintmaster]

Well, I would expect PS3 attach rates to be similar to X360 attach rates over the long term. There just aren't a lot of games for PS3 right now and the high price is preventing a lot of early adopters from getting the usual stuff even like Madden etc...

I think the other thing is that the hardcore gamers that are big spenders were probably itching for a new system and went nuts when the 360 came out by buying lots of games. Unfortunately for Sony, Microsoft nabbed most of them before the PS3 was even released.

Early 360 adopters had an unusually high attach rate, and it's probably just a blip due to HD and coming out first with reasonable software selection. I don't think it's a general trend for either Sony or Microsoft, so attach rates will probaby return to the norm now.

 

[Dizzy]

People forget that the DC didn't have EA support. As much as everybody hates EA a console can't be successful without them. I remember reading an article in Next magazine (I am not sure of the name is right they are no longer in publication) and they interviewed the head of EA at the time and he stated the reason they are not supporting the DC was that they didn't want to prolong the console war between Sega and Sony.

 

[mrcorbo]

Well, I would expect PS3 attach rates to be similar to X360 attach rates over the long term. There just aren't a lot of games for PS3 right now and the high price is preventing a lot of early adopters from getting the usual stuff even like Madden etc...

Attach rate is only really interesting from an analyst's viewpoint (both professional and amateur). It doesn't really figure into this unless a significantly higher attach rate is achieved for the PS3 over the 360 to make up for the difference in hardware units sold and put the two systems at parity when it comes to total software units sold. I would think that to developers, while attach rate is something you look at, the total unit sales of hardware and software are the most important statistics.

 

[Diamond.G]

I have a very difficult time taking anyone seriously who throws around "Dreamcast! Dreamcast I tell you!" without doing significant legwork to show that the parallel even remotely applies. It is about as good as saying...

Sony meet Nintendo. (i.e. the former market leader who got arrogent and foolhearted in regards to controlling developers/publishers through many methods, including media formats and architecture)

Of course such statements fly in the face of looking at the big picture and is unfairly picking and choosing points of contact.

How about some points of disconnect?

- MS has full/broad Publisher support. EA, in the corner, please say hello.
- MS's Xbox 360 has shown amazing customer support. Publishers who bet the 360 early are happy they did. Say hello Capcom, Bethesda, Ubisoft, and Epic. You can bet your bottom dollar they won't drop support anytime soon.
- MS has sold as many Xbox 360's in 12 months as the Dreamcast did... ever.
- MS was able to create a machine with the same physical memory footprint as the PS3 and bring to market a competitive graphics solution for the immediate and longterm. DC definately came up short on memory.
- MS has created a slow ebb toward (slowly) equalizing the exclusive market with key gains like GTAIV, VF5, AC, etc. How many PS2 exclusives hopped ship to be multiplatform on the DC?

And there are a ton more points of disconnect (MS is much richer; MS doesn't have a trackrecord of leaving pubs high and dry on platforms like 32X, Sega CD, SS, etc; MS has the three biggest IPs in NA on board in Madden/Halo/GTA; and so forth).



Using the Dreamcast as a 360 archetype is about as logical as using the N64 as the archetype to the PS3.

You know, console launched by the previous 2 gens market leader, a year late, more power, proprietary format, betting people will buy their platform the exclusives and so forth.

Of course that parallel is fraught with holes, so no one even bothers bringing it up.

Which makes me scratch my head why some of you cannot apply the same logic and see that the Dreamcast parallel just doesn't fit. It is a lazy way to avoid more involving points of discussion and treating each platform as unique. We of course should consider the impact and results of past decisions and consoles, but looking at them as absolutes on limited data (like 1 year head start! ) as some indication of the way the market "could" go, without seriously examining ALL the points of contact and disconnect and playing them out in the modern market, well, doesn't solve much.

Post of the Year!!!! This should be made into a sticky unto itself..