Slashdot

There are very few real differences left between engines, graphically speaking. In the end you've got endless oodles of parallax maps, specular maps, diffuse maps, and all that nonsense. These tools are available to anyone, and they are everywhere. In the end the deciding factor between a game looking damned sexy vs. load of crap comes down to target hardware and artistic talent.

There are some interesting and exciting ideas floating about, but as far as I know none of the truly impressive ones have made it into a production game that's been announced yet. In the end the line drawn between different engines are almost entirely dependent on lighting technique, and the vast majority of games simply use the standard set by Doom 3 (aka parallax/normal maps, direct lighting, with hacky ways to predict an ambient value which Doom 3 sorely lacked). GOW (and by extension UE3) uses a very interesting hybrid precomputed radiosity + dynamic lighting solution that is akin to the method used in HL2, though avoids some of the larger mistakes Valve made with their implementation. It looks great, and it is probably the state of the art for lighting technology at this point.

The next step in the holy grail of graphics will be real-time radiosity and global illumination. We're not there yet. I've seen some interesting papers on the subject, but AFAIK neither Heavenly Sword nor Killzone 2 are going to be using that kind of tech. I don't know of any production game that has that type of tech rolled in. Indeed, I don't know of any hardware that is capable of running that type of simulation at playable framerates!

In a roundabout way, what I'm saying is that in this shader age, the "graphical capabilities" of an engine are really measured with the shaders, and in that arena there are very few techniques being employed. Some devs make optimizations and changes that make their results look marginally better, but often times this is not a one-size-fits-all solution and will only look good with whatever it is they're working with, content-wise. This is the source of my "all engines are created equal" comment. The true difference between engines now come down to data organization - how large you can balloon your maps while remaining manageable, how good your netcode is (this varies greatly between engines, truly). As you can imagine, a lot of what sets each engine apart from another is incredibly niche, and that is also why I object to any labeling of any engine as inferior or superior. Those two terms are simply not valid for describing game engines. Depending on the effects you want to achieve, there are different engines that suit your needs. To evaluate engines graphically, however, is foolish. Often times my non-graphics-coding buddies would comment between two screenshots, claiming that one looks far better than another. In 99% of these cases what they're noticing is the quality of the art, not the capabilities of the code running underneath. I've taken incredibly "poor" looking engines and made them look on par with Quake 4 (which granted is no longer truly state of the art) within hours. Any engine that supports HLSL/GLSL in the end can look just like any other, and even fancy shadowing techniques are very homogeneous across the industry.

Last I checked, the PS3 came out in 2006, and the existence of Cell in the machine was only known, what, a year before that? Unless you're a time traveler from the year 2012, check those numbers.

Not to mention that there is still no threading support in the PS3 SDK. All those SPUs are not much good on their own, and Sony's massively advertised throughput for the processor assumes peak efficiency in all SP