Vulkan And Ray Trave accelerator Card

I’m just thinking and wanted to talk, but wouldn’t it make sense that there be a ray traceing accelerator card to makebray tracing better. That way some of these older cards can stay relevant. For instance: Could an Inyel Ceon Phi Co Processors 61 cores be used as ray tracing cores to help say an AMD Card or even a Nvidia Card.

To me it would make sense that we could have ray tracing co processors for games.

What would make sense about that? In order to do ray tracing, you need to have the scene geometry live on the GPU that’s actually doing the ray tracing. And such scene geometry is typically going to be fairly large in size. Lower-end cards typically have less memory along with fewer computational units, so it’s not clear how such an “accelerator” would help.

Not only that, you’d then need to transmit the results of the ray intersection tests to the GPU doing the real work, along with the material parameters for the surface being intersected with. And then, that operation may itself need to do further ray tests, which means it would have to shove more data back to the “accelerator” GPU and wait for the results.

This is not an effective way to gain performance.

1 Like

Well it would send data the same way it sends data to ray tracing core. The distance would be greater because it would be on another chip though, but electricity as we all know is pretty fast. The speed of light. I know it is not the spped of light, but is still pretty fast… Like to get better performance it could do ray trace core processing for peripherals that way for fps’s and games that take focus you can stay on target.

Not that fast. Cross-GPU communication is not anything we would call “fast”. Not relative to a 16ms or less per-frame budget. It’s not about the physical speed of the electrons, but the structure of the electronics serving up those electrons.

This is also why solid state drives have a maximum speed that’s slower than memory, and main memory is slower to access than cache memory, and caches are slower to access than registers. They’re all done through electrons, but electrons and electronics aren’t the same thing.

And that brings me to the second phase of my post. Why not develop a pci card with ray tracing cores that is designed and engineered for this; that is easy to access. Nothing in the way.

Because PC’s don’t work that way. The access time for a PCI card is limited by the PCIe standard built into the motherboard and the card’s hardware. It’s not something the GPU gets to control.

Things like this are why modern, high-performance SSDs use M.2 sockets instead of SATA: they can access data faster than SATA can send it. So they use a different connection to avoid having to be limited by SATA access speeds.

So to make this work, you’d have to come up with a new motherboard connector so that other GPUs can more efficiently access its data. And that connection will have to be incredibly high-bandwidth (at least on the order of memory access times, if not faster) to make this work. So not only would you have to develop a special card socket, you’d have to do substantial motherboard work; it wouldn’t work for existing motherboards.

Oh, and the target GPU that’s going to read this data also has to have a higher performance interface to it. So it wouldn’t even work with existing GPUs; you’d have to buy a new one.

So, you could by a raytracing accelerator card + a new motherboard that has this special socket + a new GPU that can work with it. Or… just by a GPU that can do raytracing. Because those already exist.

The particular confluence of effects that made 3D accelerator cards viable in the mid-90s simply no longer exists today.

I thought most motherboards had a Pci Matrix these days… For instance HP z820 v z620… For Nvidia Optimus. I don’t know the specific on any of thAt stuff, but i thought matrix, Oh maybe they have direct lanes between pci slots?

And most of these modern gpu’s have a x8 pci slot on the top of the card. Maybe utilize this.

The whole purpose is to have better ray tracing as new tech comes out. That way we can attract new gamers to buy new hardware and the rest of us can do this. As far as I can see advances in games for next 5 years is better ray tracing. The cards from 2016 on can all play the games that are out just fine minus ray tracing.

What @Alfonse_Reinheart is trying to tell you is that the PCIe bus is unbearably slow for this. Consider why discrete GPUs have dedicated memory instead of using (part of) main system memory over the PCIe bus - that has the same problem: too much latency and too little bandwidth.

Just because an idea has a purpose doesn’t mean it’s workable. There was a purpose for physics accelerator cards too, but that didn’t make them viable in the market or useful to consumers.

And to be honest, even if such an accelerator were actually viable, any such card would still need to have 80-90% of all of the stuff a regular GPU has: lots of high-speed, low-latency memory; a bunch of compute units; the various complex memory controllers needed to feed data to all of those compute units; etc. The only stuff you’d be missing would be a rasterizer, ROPs, and the logic needed to output video. None of which are the expensive parts of a GPU.

yeah, but most modern gpu’s have a direct acces bus on the top to link straight into the gpu? Why is this not accessed. And to tell me a pci bus is slow is simply not true. you click and with in 1 ms it appears on screen. That goes through so many more channels than what it is hear that I am describing. In terms of physical hardware to link additional useable rt cores to an existing GPU. rt cores do is detirmine an integer. right? The ray hits an object and bounces to detirmine what color a pixel becomes.