Besides it’s not that the premise is unrealistic, it’s comming. It’s probably not comming in DX11, but it’s comming.

And yes Oscar Forth you’re the only one that thinks that. Since the intel chip in question (and it wasn’t originally done on one chip but I believe 4 machines using a PVM or equivalent, and 4 cores each) would be a 16 core behemoth that dwarfs the performance of any modern buyable intel CPU, while at the same time being dramatically easier to program than a PS3. One of my pet projects is helping (I make no pretext of being lead on this) is setting up a small PS3 cluster for some astrophysics types, writing software for it is a bit, different, clearly manageable (e.g. DMC4 and GTAIV), but both the 360 and PS3 trap themselves with a pitiful amount of main memory and the funky architecture of the memory, SPE’s & PPE probably hampers success a bit.

No, this is not usual for PCMech. I just got completely HAD here.

Why can’t we just give DirectX 11 a rest until maybe the second half of next year or two? DirectX 10.1 has already been out since the days of ATI Radeon HD 3xxx series, but maybe not quite…

For those that know anything about Ray-tracing … the problem isn’t ray tracing its the scene changing. Updating the scene optimisation trees is a nightmare.

Also .. am i the only one who really thinks that if you can render Quake 4 at 100FPS on an intel chip then ONE PS3 should be able to render that vehicle?

yeah, on April 1, lol ! http://www.techarp.com/showarticle.aspx?artno=526

Besides that, nearly every argument in this article is crap.

Not that the principle of adding ray tracing to DX is unreasonable. OpenRT aims to be ray tracing in the style of OpenGl, and presumably there will be a DirectRT in future.

However, the basic mathematics sort of fall apart. Ray tracing, while demonstrably practical for older engines on extremely fancy setups with 16 cpu cores is probably 2 or 3 hardware generations from being implementable at that level on the desktop and another generation or 2 away from being able to do that on (what will then be) modern engines. My research is on RT on GPU hardware, which faces essentially the same curve: theoretically possible now but don’t expect anything any one would want to actually use any time soon.

It is possible, however unlikely, that one of AMD, Intel or Nvidia will release hardware custom built for RT in the same way current GPU’s were for rasterization, but even there you’re looking at years for it to filter down much (AFAIK the university of I think it’s Saarland has something along these lines already but I have no idea how good it is or the costs). That would create a whole slew of questions with compatability etc… as well, which means we likely won’t see that for some time (though likely that will be the future).

Granted MS may decide that DX10.x is as far as they need to go with raster graphics and DX11 will include ray tracing at some point in the distant future, but don’t expect it to be anything other than an exercise for academics and researchers for a few more years. I would also say, given the relatively bland reception DX10 has recieved the MS marketting types will be happy to shove something out the door, no matter how minor an improvement they can call DX11 and hope for some better press.