Slashdot Mirror
Cloud Gaming With Ray Tracing
An anonymous reader writes "Since real-time ray tracing on a single desktop machine is still too slow for gaming, Intel has worked on a research project that puts the heavy workload up into the cloud, relying on multiple 32-core chips working together. That enables new special effects for games, like realistic reflections (e.g. on cars or the scope of a sniper rifle), glass simulations and multi-screen surveillance stations. The research paper also takes a closer look at various sources of latencies that become relevant in a cloud-based gaming approach."
First Line of the Article:
"A new technology from Intel called ray tracing could bring lifelike images and improved 3D effects to games on tablets and other mobile devices."
GAH!
My postings are informational and does not constitute legal advice. Act on it at your risk.
I was unaware Whitted worked for Intel. </sarcasm>
Can't say I've ever heard of him though. I use to play against someone called Polly but shes gone now.
"A new technology from Intel called ray tracing could bring lifelike images and improved 3D effects to games on tablets and other mobile devices." Ray tracing has been around a long time. Even ray tracing in the cloud isn't that new. NVidia has the RealityServer.
I wouldn't want my games up in Intel's cloud somewhere where I don't have any control and where I have to rely on my ISP to provide good latency. But it might be interesting to me to have a single powerful home server and then a couple laptops and a couple tablets that are basically IO devices and little else. Granted, you couldn't have everyone doing demanding, graphically intense games at all times, but a reasonably powerful desktop server should be more than capable of rendering 2x1080p laptop screens and 2x720p tablet screens, which is all most people have these days anyway (damn HDTV to hell for making people think 1080p is 'high resolution' for a PC monitor!). Of course, I'm not sure a wireless N network would keep up with the bandwidth requirements, but then you'd have the same exact problem (only much worse) trying to put those services on Intel servers in a datacenter somewhere.
It's not that impressive, either.
On the topic of raytracing - one thing that still stands out to me from the images in the paper are the lack of proper occlusion and shadows.
Take a look at the shot of the close up of the car side - look under the front wheel and it just looks .... artificial.
Unless there's some magic sauce that can be sprinkled on this without added a frame rate hit this isn't really all that wow at all.
That the article thinks "ray tracing" is a new Intel technology, or that it thinks "cloud" rendering is something that hasn't been around for 50 years?
I rtfa, and its confusing. It started with talk of cloud computing on mobile devices (with no mention how the constant speedbump of network lag were to be overcome) and then droned on about a new chip architecture.
Nothing to see here, moving along...
Hi, I Boris. Hear fix bear, yes?
This is garbage. Mobile gaming, cloud computing, eh rewriting wolfenstein to add ray tracing in the cloud??? I can see why that might make a POC, but Wolfenstein's not even 3D! "We have a red car sitting at a courtyard, which has a very shiny reflective surface. That can be rendered very good." OK, not speaking Inglish isn't a crime. But the editors should catch this kind of thing. UNworth reading.
"A new technology from Intel called ray tracing "
I stopped here.
Then every year its another ray tracing demo usually jammed awkwardly into an ID tech engine.
What other engine do you recommend? What other major label releases engines of its five-year-old games under the GNU GPL?
I can create photoreal images on my netbook with a GPU that has 24 stream processors.
At what resolution and frame rate?
Isn't this what Sony promised us the Playstation 3 would do, and the supposed reason why they went with the "Cell" processor? Because everything Sony that had to do heavy graphics lifting would have one, and they would all cooperate to make your games better? And of course, this never came to pass, and Sony never really used Cells for anything else (ISTR there might have been a Cell-based Blu-Ray player that wasn't the PS3, but maybe that was just a rumor.)
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
Intel wants to find a way to remain relevant, if they can come up with some clustering secret sauce then there is a reason to continue to buy products based on their technology. The PC Gaming market is in decline, the only space in computing that is growing rapidly is mobile computing, and corporations work on the model of boundless expansion. They need to conquer new markets to continue to exist. If Intel can get you to buy an intel-based phone, desktop, and tablet because they will cluster then it's a good strategy. Many technologies for clustering have been around for a very long time and it's long past time we saw some offerings on the consumer market. Next-generation wireless should offer pretty good throughput at close range, so if you put the AP where you spend most of your time it might be adequate for that kind of use.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
I was unaware Whitted worked for Intel. </sarcasm>
Actually, someone else may have beaten him to it .. but now I can't find the paper that cites the earlier reference.. so this post is a bit pointless :-(
Oh that's typical... I just found the paper. It's "Interactive Rendering with Coherent Ray Tracing" by Ingo Wald, Philipp Slusallek, Carsten Benthin, and Markus Wagner, published at Eurographics 2001.
Ah cloud computing... pauses to laugh... Ok, earth to the idiots at Intel - your network latency kills any benefits that could ever be imagined for this system. An average video card nowdays can push 80-100 Gbps, higher end cards are exceeding 150 Gbps and more. Let's look a video cable speeds - HDMI pushes 10.2 Gbps, VGA 10.8 Gbps, DVI 9.9 Gbps . Now lets look at the typical home internet connection today, it's avg 1.5-3 Mbps. Ok, let's do a thought experiment about how this stupid system would work. I need one frame rendered and let's pretend I can send out a request to all the computing power I need to render it. Let's say the time it takes to ray trace the frame takes almost no (zero) time since I can use the "cloud". Super!!! Now I need to put the frame together and send it back to my computer from the cloud. Uh ho!!! It's going to a HUGE amount of time get the final rendered frame back. Here's the math, let's pick a resolution of 1024 x 768 (yes, I know most of us run much higher than that) and we need 24 bits minimum of color. That's 18,874,368 bits. Using the typical internet connection, it will take 6 - 12 seconds to get each frame assuming it takes no time to render it. Most games run a minimum of 30 frames per second (50-60 frames per second is preferred). Now we have a system in which we get 1 frame every 6-12 seconds. That's a HUGE improvement Intel. Thank you for that.
To me it sounds more like the Intel Larrabee division has moved to the 'cloud'. Apart from that it's just a repeat from 2006, 2007, 2008, 2009...etc.
No sig today...
When tablet/mobile data plans won't be insanely expensive and when broadband will have no upload/download limits and decent speed, then you can start talking to me about rendering graphics in the "cloud".
~Syberz
how great would it be if /. automatically filtered stories which are about imagery but do not in fact have images in them.
Is this an actual example of a good usage of the term "cloud"? In the sense of some computers out there somewhere doing stuff for you and you getting the results? Not long ago I heard about the company OnLive and their cloud-based gaming, where all the computing and rendering is done on their servers, you send your control inputs across the net to them and they send you back sound and video.
Played it not long ago myself and expected the lag to be bad, but it turned out it wasn't bad after all. You can sense it, especially doing certain things, but it doesn't get in the way. And I hear they have more latency cutting measures in store. Pretty neat stuff.
Cloud gaming opens up the possibility of leveraging more computing power per player, so I can see fancy effects like ray tracing being incorporated into cloud games.
h264 doesn't work, you need a low latency codec. Computing the motion compensation between N keyframes means you're introducing N frames of latency.
So you need to transfer still images, encoded in MJPEG or something similar but more advanced. Is it possible?
of course it is! One solution was introduced recently with the windows SP1, the other one is open source and has been available for some years.
doing it from the cloud (i.e. fancy word for the internet) isn't so interesting, the technology sounds so much desirable on the company's lan, then on the home network. But it still is workable over the internet within conditions of bandwith and latency, i.e. you need a home connexion that both qualifies for HDTV over DSL and a good game of counterstrike. Good DSL may do, fiber would be much better. That's why it already exists again, and sold under the name of "OnLive".
I think the notion that games have to be any one thing is preposterous.
+0 Meh
If you read the paper I mentioned, you will see it, in turn, cites "A. Appel. Some techniques for shading machine renderings of
solids. SJCC, pages 27–45, 1968."
At 60Hz, one screen refresh is every 16ms, so the rendering takes either 8 to 14 images with 5 images caused by the network RTT..
Interesting.
What other engine do you recommend? What other major label releases engines of its five-year-old games under the GNU GPL?
How about something new that they have written themselves? Whats the point of demonstrating the supposedly next big thing in computer games with some obsolete five year old game? How are they ever going to get games written for raytracing if they can't even find somebody to put together a solid tech demo?
How about something new that they have written themselves?
Intel is in the hardware business and possibly the driver business. Making parts of a video game other than code needs a different skill set; otherwise, you will likely end up with the phenomenon called "programmer art". It's far cheaper to start with a 5-year-old Id game than to hire a producer and competent artists to come up with an original setting.