Many people this weekend will be unwrapping a copy of Criterion’s latest racer, Need For Speed: Hot Pursuit. However, some people may still be sitting on the sidelines wondering how a game which runs at 30 frames per second can compete with one that runs at the industry’s holy grail of 60fps, like Burnout Paradise did and of course Gran Turismo 5 aims for.
Criterion themselves are champions of 60fps but the Need For Speed franchise has been firmly entrenched in the 30fps camp, aiming for a cinematic rather than arcade feel. For Hot Pursuit, Criterion have adopted the series’ 30fps. This is where a lot of people tuned out, fearing that the halving of the frame rate will lead to a near doubling of controller response.
Firstly, there is the common perception that twice the framerate must automatically mean it’s better – which is fair enough, but there is another spin you can put on it by saying that you have double the amount of time available to  actually render the frame. This doubling of the render time has allowed Criterion to introduce a new range of technologies, like a Killzone 2 style deferred rendering which has allowed them to add an amazing level of detail and a gob-smacking 15km draw distance. It has also allowed for some fantastic looking lighting effects, which don’t have to wait for a replay mode to make an appearance.
The downside of this extra time is of course a potential near doubling of controller response, meaning that despite the graphical advantages of the doubling of the time available to render a frame, the trade-off of a laggy controller response just isn’t worth it.
Criterion have cracked this problem by incorporating some highly technical code-type voodoo, somewhere amongst the game’s code. Digital Foundry, as usual, are on hand to unearth what’s happening, speaking to Criterion’s Alex Fry who had this to say.
“The way the architecture works is to run the game simulation internally at 60FPS, and it’s polling the controller once for every simulation step so you get as up-to-date inputs as possible,”.
“The render code (building display lists for the GPU to consume the next frame) immediately follows the two controller poll/simulate loops, and then it waits for v-sync. Thus on CPU we get two 60FPS updates and one 30Hz render in a total of 33ms.
“When the CPU is done and v-sync is hit, the GPU kicks off and renders the scene while the next simulate/render frame happens in parallel on CPU. Once done the GPU also waits for v-sync (which also syncs it to the CPU), thus adding another 33ms. Then TV scan-out happens to get the final image to screen, which adds a final 16.6ms.”
The total of all this is 83ms which makes it just only one measly extra frame of response time over the industry-leading Burnout Paradise. Pretty astonishing given the graphical advantages of having twice as long to render the frame. This is a new record for a game running at 30fps and probably represents the theoretical limit of controller input, PC owners are also in for a record-breaking treat with a response time of 50ms, which is faster than any other previously tested game.
By running the game internally at 60fps but outputting at 30fps gamers get the best of both worlds. A highly responsive control system but also extra graphical effects, lighting and the incredible 15km draw distance that probably wouldn’t be possible if the rendering time was halved in order to push 60fps.
teflon
This kind of approach isn’t exactly new, I’ve seen and heard of it being used on a few other games (Of the top of my head, I can only think of Supreme Commander, which detached the simulation refresh and the frame rate from one another, simply as a compromise to allow it’s huge scale to work).
Still, a good solution to a common problem, and good to see they’re giving the very best for their games.
Kennykazey
Sounds clever. Looking forward to try this game out for real hopefully within next week.
TSBonyman
Interesting article – and thankfully nothing to be too concerned about before i buy this game.
besidavi
The title of this article is a bit confusing – I saw it and thought the worst but it’s actually a positive story!
cam the man
Same hear, the newspapers have been doing that for years.
beeje13
‘running the game internally at 60fps but outputting at 30fps’
hmm not sure I agrre with that, a 60fps image feed makes the game easier to look at as well as it being more fluid and realistic.
Call of duty games have a very low resolution (I think BO on PS3 renders at 540p) but the 60fps makes them look so good.
ScottW-1976
Having played BO for a few hours today for the first time on PS3 I would argue that it doens’t particularly “look good” at all. Very dissapointed with the whole experience…
beeje13
yep thats what I though, the first mission was almost boring because of the lack of detail, but i think you’ll probably get used to it.
Saying that, if I went back to mw2, I would probably notice the upgrade in detail
DrNate86
I didn’t understand a word of this but it all sounds very nice! I didn’t notice any lag playing the demo, so I don’t know why people were worried.
Topov81
That’s sounds very interesting, but I’d rather they gave us something to do in free roam :(
bunimomike
Wouldn’t that contradict the very nature of free roam?
freezebug2
Free roam is for taking fancy car pictures in awesome settings really…period.
topgear
Am I the only person that thinks that the graphics are a turd on pc?
ScottW-1976
With regards to NFS:HP or in general?
beeje13
‘ on PC ‘ Yes, if its under £400 or under £600 for a laptop, but consoles cannot come near a £600 shop bought rig, let alone if you make one yourself.
fps_d0minat0r
if GT5 can get 60fps, i dont see why burnout cant.
surely its not demanding more technically from the ps3.
bunimomike
It’s probably nowhere as near as efficiently coded for the PS3 hardware (what with it being multi-platform). Also, keep in mind that GT5 is almost medical looking with its uber-anally-retentive look and Hot Pursuit is trying to look far more real-worldly. The latter takes far more effort (ignoring the whole car count).
DJ Judas
Someone should tell Criterion to test UT’99. Their PC record wouldn’t last long.
DJ Judas
I’m not sure it would actually after having a read around. I’d be interested to see the setup they tested the 50ms response on.