The optical read speed is at least offset by the universal presence of the harddrive, enabling virtual memory caching on a game-by-game basis. This effectively enables a total data transfer rate of the hard drive + the BD drive.
That's true although initially you can only load off the blu-ray when you first run the game so there is a speed hit there, unless you went with the mandatory install route. Many go with the mandatory install route because we can't hit Sony's required initial load time requirement with just the blu-ray drive.
More significantly though is that most of us will treat the hdd as optional and assume it's not there. This isn't due to any favoritism or conspiracy theories, but just business realities. Basically for many of us to stay alive, we desperately need to be able to work on the $199 console. Without that, the layoffs and company purges you are seeing happen would have been far more widespread. So as cool as the standard hdd may be, it will be used mostly in an optional manner.
I've heard that this OS overhead has been toned down and optimized significantly in the past year or so. Any truth to that?
Yeah they have steadily been improving it, default memory consumption is much better now than a year or so ago but they still lag behind. Likewise as Shifty says, there are 'optional' memory requirements as well, and those are still harsh.
TimothyFarrar said:
Resistance 2 is doing this 4xMSAA trick as well as a second pass for even lower resolution particles.
Hmm...you sure R2 is doing it that way? I had to do something on our PS3 title for particles as well as they were taking upwards of 8+ms sometimes (!!!), so I went with a 1/16th sized buffer and a single post process 'cleanup' step (~0.5ms) on that buffer before blending it back. It looks mostly ok but the artifacts I see on what I did look suspiciously similar to what I see on R2's particles as well...
TimothyFarrar said:
30% for a 60 Hz game might be a good upper bound. I think there was an Insomniac presentation which quoted a high number like this. Cannot remember if it was for early R2 (possibly a 60Hz target then?), might not make given (I think) the shipped R2 is a 30Hz game.
Ah ok, fair enough. I was defaulting to 30hz in my mind. For 60hz I guess 20% or even 30% may be possible, although I'm probably drop the hammer down closer to 20-25% if I could.
TimothyFarrar said:
As long as CPU/GPU memory access isn't a problem (ie your GPU can write out to CPU memory, or has unified memory, or the CPU can quickly access GPU memory), and the CPU has a serious vector unit (which the SPUs qualify for) with enough bandwidth, then doing post processing CPU side can make sense if you have free CPU time. Sure wouldn't try this on the 360 however. If you go way back to the old SGIs (O2, etc) which had unified memory, you might be doing some post processing CPU side. Likely in the future as GPUs get more CPUified, everything will stay GPU. However if PS4 ends up with both a DX11 class GPU and a lot of new SPUs, you might very well want post SPU side just to find enough to keep them busy.
You know I'm still not 100% sold on splitting post processing between the cpu and gpu. I guess there are cases where it can be done and/or makes sense, but it seems like the bang for the buck you get from the cpu in that regard isn't ideal on this generation. PS3 does have monster vector units, but you would munch up so many of them to do what the gpu would handle easier, not to mention dma bandwidth use, etc... Plus keeping post processing all gpu makes sharing and reuse of post process steps so much easier. Quick example of that, I was bored between projects so I implemented the 'sun shafts' shown in the ShaderX6 book. I think it was 4 or 5 post process steps for that. Easy enough and it looked nice. Then I went back and optimize it. Since all our post process is done in gpu it was easy to reuse some of the data that the sun shafts needed from other steps, and better yet I was able to piggy back some of it on other steps. In the end, the entire sun shafts post process effect ended up only costing me ~0.45ms or so once I reused/reduced/recycled it into the exiting gpu post process pipeline. I was very pleased with that
It still may very well make sense for Killzone 2, but games are always adding more post processing in subsequent versions. So I can't help but wonder if at some point if someone went to the trouble of cpu-i-fying post process that they may run out of cpu and have to start shifting them slowly back to gpu anyways. Maybe I'd be more sold on a 16+ spu PS4, but not sure about this gen.