1UP's Valve coverage concludes today with a two part (one, two) Gabe Newell interview in which he talks about his company, the industry and the next generation of systems. Yesterday they had a story about the history of Valve.
Interesting interviews. I especially like his insight into Longhorn, 360 and PS3. From his point of view Longhorn does nothing to help game developers or consumers playing their video games. This is disapointing. Obviously though WGF 2.0 will be nice for future graphic tech but that's similiar to the same experience end users have now with DX9.
I agree with him how nVidia/ATI/MS need some sort of auto update display driver for avg joe bob consumer who never checks nvidia.com and ati.com. I'm sure support gets flooded by this issue and it's sad this issue hasn't been addressed yet.
His views on 360 & PS3 are interesting also. I think it's obvious multicore CPU approach isn't ideal for programmers either especially the PS3. He mentions how one line of code on just one of the 7 SPE's and your whole code slows to a crawl by a factor of 80. He also speculates Sony went this route so game devs code is so specialized for the PS3 that it would be a waste of time & money to port games to other platforms. I'm not sure I believe that's that the reason. I tend to think Sony/KK are just really obsessed with theoritical floating point in hardware and not necessarily realistic hardware & software combination. PS2 is similiar to the PS3.
This isn't a problem for just the 360 & PS3 either though. PC cpu's are all going dual core right now and in 2006 virtually all cpu's will be dual core, even on laptops.
From all the info i've read from Carmack, Gabe and then the anonymous source article on Anand awhile back it really does sound like most software won't be using more then 1 CPU or SPE for quite awhile. Obviously games will still look awesome in the next generation but it sounds like the future of hardware really isn't ideal for game developers at all, no matter what platform you code for which hurts both the developers and the consumers.
..I tend to think Sony/KK are just really obsessed with theoretical floating point in hardware and not necessarily realistic hardware & software combination. PS2 is similar to the PS3.
I think Gabe has a point, Sony has other interests apart from a gaming machine, they want the cell to be everywhere in the future, they have made this clear on several occasions. So using it in a high volume, high profile and vibrant market makes a lot of business sense. You secure a lot of people trying to develop code for your technology, its then much easier to use it in other applications/products because the initial bump is already grind-ed down. Also, as Gabe notes, code made for the cell has practically no use whatsoever on any other architecture..clever..
..it sounds like the future of hardware really isn't ideal for game developers at all
Well apart from Sony (which I think have ulterior motives ) Microsoft had very few options to provide a console with high theoretical performance while having a low price. A multicore risc CPU is basically the only option. For desktops its more that we are approaching a point there we cannot increase the performance by speeding up or adding silicone. Also for desktops, you still have the O/S that can handle the multiple CPU's and the application can choose to ignore it (not necessarily valid for games, but in general).
From all the info i've read from Carmack, Gabe and then the anonymous source article on Anand awhile back it really does sound like most software won't be using more then 1 CPU or SPE for quite awhile.
I think people exaggerate this a bit. UE3 is multithreaded, What Crytek is working on is multithreaded, and many of the first generation Xbox 360 games are multithreaded and make use of multiple cores (the PGR team claim to be using all 3). So it's not impossible, it's just a lot of work.