Ubisoft Claims CPU Specs a Limiting Factor In Assassin's Creed Unity On Consoles
MojoKid (1002251) writes A new interview with Assassin's Creed Unity senior producer Vincent Pontbriand has some gamers seeing red and others crying "told you so," after the developer revealed that the game's 900p framerate and 30 fps target on consoles is a result of weak CPU performance rather than GPU compute. "Technically we're CPU-bound," Pontbriand said. "The GPUs are really powerful, obviously the graphics look pretty good, but it's the CPU that has to process the AI, the number of NPCs we have on screen, all these systems running in parallel. We were quickly bottlenecked by that and it was a bit frustrating, because we thought that this was going to be a tenfold improvement over everything AI-wise..." This has been read by many as a rather damning referendum on the capabilities of AMD's APU that's under the hood of Sony's and Microsoft's new consoles. To some extent, that's justified; the Jaguar CPU inside both the Sony PS4 and Xbox One is a modest chip with a relatively low clock speed. Both consoles may offer eight CPU threads on paper, but games can't access all that headroom. One thread is reserved for the OS and a few more cores will be used for processing the 3D pipeline. Between the two, Ubisoft may have only had 4-5 cores for AI and other calculations — scarcely more than last gen, and the Xbox 360 and PS3 CPUs were clocked much faster than the 1.6 / 1.73GHz frequencies of their replacements.
"We could be running at 100fps if it was just graphics, but because of AI, we're still limited to 30 frames per second."
Uh, have you guys tried running the AI calculations less frequently than graphics redraws? You don't have to keep them in sync, you know.
I really hate that Sony dropped their cell processors going from the PS3 to the PS4 in favour of an x86 based system. We didn't see a lot of devices using cell and because of that, a lot of cell super-computer clusters were even made using actual PS3s. Even the prior MIPS processers of the earlier PlayStations are used in computer architecture texts books to this day (albeit overly-simplified versions of MIPS's pipling systems).
I really want to see more architecture options, not less. Intel bought Alpha, killed it, screwed up with their own VLIW attempt with the Itaniums (which use EPIC) and I haven't heard anything about Transmeta in years. Today everything is ARM or x86_64 (with MIPS still seen in some embedded systems, mostly home routers). IBM still produces new POWER systems, but they're limited to a specific server niches.
TFA just mentions the interview without a clear reference to it. Looking for it I found two other articles that suggest that the 900p resolution and 30fps targets came from other factors. http://www.gamespot.com/articl... says that 30fps is "more cinematic" and 60fps "looked really wierd." http://www.gamespot.com/articl... suggests that some non-graphic computation is going on the GPU, but also has a quote that mentions "technically CPU bound."
What we don't know from these articles is why some or more of the AI computation can't be done in the GPU.
In the gaming community we've had to call the big publishers out on their marketing garbage quite a lot lately. Shitty PC ports. Framerate locking. Terrible, buggy releases, low resolutions. (And don't even get me started on cynical euphemisms like "cinematic experience" and other jabs at framerates above 30. I've never wanted to punch something so much)
If there's one thing the new consoles have right is a mind numbing amount of CPU power. Plenty of of modern cores. These things would have been supercomputers a decade a go. What they don't have is extremely fast single threaded performance. But we've known that for years now. Many cores has always been the path forward for consumer electronics. We've even seen it in smart phones and tablets.
Ubisoft's marketing shitcocks are covering for inept programming (Or rather a shit management style that encourages/rewards inept programming) because it's much more difficult to take advantage of many slower cores than one, fast single thread. You aren't going to get fast single threaded performance out of an integrated AMD SoC. (If you want fast single thread, you go Intel. End of story.)
That said, the above excuse is still complete bullshit. It's well known that the chips in the new consoles have GPUs that have the effective computing power of what you'd find in a mid-grade gaming PC from 2010, with the PS4 coming out ahead of the Xbone by a healthy margin. What happens when you try to make a cutting edge pretty 2014 game run 60FPS at 1080p on a four year old graphics card?
Exactly what you think fucking happens! The graphics on the new consoles don't "Look great" they look like garbage because they're rendered at barely-above-720p at pitiful frame rates.
Ubisoft and company: Stop making excuses for your slam-down-your-customer's-throat-marketed AAA console shiftests. Make real ports for the real serious gaming platform, the PC. If your game is so ineptly developed that you can't decouple the framerate and game logic then fire your managers and let the programmers do their jobs.
You pay more, and you get more. You want a game machine that's capable of being a full-blown music recording studio? Or a video editing suite? Or can run Matlab, Mathematica and fluid dynamics simulations? That's a PC.
Consoles became less interesting to me as I grew up. Games didn't become less interesting to me, but the notion that I would sit in my family's living room with a controller in my hand just became an artifact of childhood.
Of course, some dedicated game box is going to be less powerful than a PC. The only reason they exist is so that game companies can manage licenses. They're not meant to be for your benefit. They're consumption machines, designed to tie you into a corporate "ecosystem". If I was 13, I would love one. Now, they just seem like evidence of the failure of the gaming market to mature.
You are welcome on my lawn.
You're missing the fact that AI code is typically branch-heavy, which kills pipelining and makes all of what you say about "instructions per clock" moot. In branch-heavy code, a high clock speed is very important. Metrics of "N instructions per clock" only apply to the theoretical best-case of non-branching code.
Oh, yea?
An Intel 1.7Ghz i7 is TWENTY FIVE times faster than an Intel Pentium 4 4.0Ghz
http://www.cpubenchmark.net/co...
That settles it. Ignore clock speed across generations.
My God can beat up your God. Just kidding...don't take offense. I know there's no God.