Slashdot Mirror
Crytek Shows 4K 30 FPS Ray Tracing On Non-RTX AMD and NVIDIA GPUs (techspot.com)
dryriver writes: Crytek has published a video showing an ordinary AMD Vega 56 GPU -- which has no raytracing specific circuitry and only costs around $450 -- real-time ray tracing a complex 3D city environment at 4K 30 FPS. Crytek says that the technology demo runs fine on most normal NVIDIA and AMD gaming GPUs. As if this wasn't impressive already, the software real-time ray tracing technology is still in development and not even final. The framerates achieved may thus go up further, raising the question of precisely what the benefits of owning a super-expensive NVIDIA RTX 20xx series GPU are. Nvidia has claimed over and over again that without its amazing new RTX cores and AI denoiser, GPUs will choke on real-time ray tracing tasks in games. Crytek appears to have proven already that with some intelligently written code, bog ordinary GPU cores can handle real-time ray tracing just fine -- no RTX cores, AI denoiser or anything else NVIDIA touts as necessary.
Wait a fucking minute. "Only" costs around $450? If I tried to spend $450 on a video card for gaming, my wallet would jump up and slap me on the head. Visa would call me and ask if my credit card had been stolen.
You are welcome on my lawn.
which sounds like a lot, but I paid $300 for a Voodoo Rush in 95. That's $500 in today's money and it wasn't a big deal back then.
I'm not sure if it's the $229 GTX 1060 6gbs or just plain the worse economy (it is a lot shittier, You could make $12/hr starting at a call center in my dirt poor town back then, which is $20/hr now for a job a high school dropout could get, now the same job pays $9.50/hr, or about $5.70/hr in today's money) but these prices didn't used to seem all that nuts.
Hi! I make Firefox Plug-ins. Check 'em out @ https://addons.mozilla.org/en-US/firefox/addon/youtube-mp3-podcaster/
Turing (Nvidia) has exactly ZERO ray tracing units. All 'ray' algorithm maths calculations are done on the standard shader cores- same as with this demo on the AMD Vega 56. So what gives?
An ex-Nvidia engineer post on Beyond3d gave the game away. This engineer was partially responsible for the so-called 'ray tracing' enhancement on Turing. Put simply, this is what Turing does:
a = b + c * d
say the above is a ray maths calc (obviously it is not). The '+' and '*' are the maths operations done on the usual shaders. What every tech site misses is that there is another issue- the process that gets the VARIABLES to the shader ALU blocks.
What Nvidia did with Turing was to add tiny ASIC circuits that allow the VARIABLES that represent triangles and rays to be more efficiently moved to the STANDARD shader units from the other units (like geometry). No RT 'cores' (there are no such thing in Turing) but a tiny logic hack that allowes the GPU to be reconfigured to move certain kinds of data much faster to the shader cores.
However the significance of this is that if one arranges for ray/triangle data to be held in a more efficient form on a 'normal' GPU, the same ray 'acceleration' can be achieved.
Not that even on Turing does REAL-TIME ray tracing happen. Real ray tracing needs far too many 'bouncy' rays per screen pixel to ever be possible on any ordinary GPU- and the problem with real ray tracing is MEMORY COHERENCE, not the maths of the ray/triangle collision.
Turing 'ray tracing' is actually simple ray algorithms applied to real-time reflection maps (NOT true reflection) and shadows. Metro Exodus tried a very very basic form of ray averaging for lighting, which was no better than simply using more traditional light sources.
And the tensor cores on Turing? Well unlike the non-existent ray tracing cores, the Tensor cores are real and use vast numbers of transistors. Why are the Tensor cores real- and the main reason Turing exists? Because Nvidia spent hundreds of millions of dollars developing new crypto currency mining algorithms to run exclusively on Tensor. However, crypto currency collapsed between Turing's design and release. Nvidia's future GAMING GPUs will not have tensor cores.
Turing was 100% designed to displace AMD in the PC crypto currency mining space. Nvidia lost an absolute fortune with Turing cos of the collapse (and Nvidia actually reported this fact at its investor conferences in 2018).
PS all practical 'ray' algorithms can be far better done (faster, less energy) using traditional raster algorithms. Light probe lighting methods with voxel data sets do the real time lighting more than good enough. Real time reflection maps do not need 'ray tracing' to deploy in reflection enhancement. Same applies to shadows- where good enough is better than extreme GPU power/processing requirements.
Indeed, with shadows, the minor improvements to near shadows are not the issue- the issue is shadows being disabled beyond a certain z-distance- something ray methods actually make worse. Better shadows = MORE shadows and shadows across more of the scene.
Nvidia has always been full of it. To see them exposed as the frauds they are _this_ quickly is very nice.
Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
The biggest lie they ever told is that their stock value is worth ten times annual revenue, which explains why they were willing to outbid Intel for Mellanox, because they were basically paying half price.
When all you have is a hammer, every problem starts to look like a thumb.
What nvidia means by "ray tracing" with their RTX thing and the AI denoiser is actually path tracing, which uses incoherent rays and actually does simulate light bounces in a physically accurate way. Effects like depth of field, soft shadows, caustics, ambient occlusion, and diffuse interreflection are a natural result of the path tracing algorithm, but have to be specially accounted for in other algorithms like ray tracing. A good reference for this is Physically-Based Rendering, by Matt Pharr. Because the rays in a path tracer are incoherent, it's an inherently noisy algorithm that requires many samples to reduce variance to acceptable levels. That's where the AI denoiser comes in - it's able to take a noisy image made with fewer path-traced samples and reduce variance to an acceptable level in realtime.
The guys over at brigade also have an actual realtime path tracer, and while the work is world-class and draw-droppingly impressive, you can see how noisy it still is.
Why is it the case that objects in videos that show off graphics capabilities always look shiny, brand-new, crisp, all the time? Even when they are supposed to be old and dusty, they manage to look very shiny, brand-new and crisp.