It honestly feels far too soon to be typing the letters CES in sequence, but the show is not that far away anymore and there is some news worth sharing. Yesterday it was revealed that AMD's CEO, Dr. Lisa Su, will be the keynote speaker at CES. This has lead to rumours that during her speech she will reveal Navi, AMD's next generation GPU. This was later confirmed, so we will be seeing at least some information about AMD's 7nm GPU, with or without ray tracing, during her keynote and at the AMD booth. With NVIDIA's successful launch of some rather expensive GPUs, some of which are even available, we can hope that AMD's new GPUs give them a run for the money; in one way or another.
"The GPU outfit announced that its CEO, Dr Lisa Su, will deliver a keynote address at next year's CES and spill the beans on the new chips and GPUs. Yes, it's an announcement of an announcement."
Here is some more Tech News from around the web:
- Nvidia's RTX 2080 Ti GPU will finally start shipping this week @ The Inquirer
- Linux, HCI and more, all from the new release of Windows Server 2019 @ The Register
- The KillMii is an entire Wii console in a tin of mints @ The Inquirer
- Apple forgot to lock Intel Management Engine in laptops, so get patching @ The Register
- TiVo Bolt OTA review: For would-be cord-cutters with great antenna reception @ Ars Technica
“AMD’s 7nm GPU, with or
“AMD’s 7nm GPU, with or without ray tracing,”
You can Ray Trace till the cows some home on Pascal/previous 3 generations(CUDA) and Vega previous 3 generations(OpenCL), and also on CPUs(That were orginially used for ray tracing and still are).
Vega and Navi may not have any on GPU dedicated Ray Tracing hardware and the first to use any dediceted Ray Tracing IP on the GPU was Imagination Technologies on their PowerVR Wizard GPUs for the mobile market.
Ray Tracing is a compute workload so CPUs were where Ray Tracing happened first and then CUDA(Nvidia) and OpenCL(Developed by Apple that now Khronos-Maintains and AMD, Intel, and others use for GPU compute) languages were developed for doing compute workloads on the GPU’s shader cores.
Maybe you can say GPU Hardware Based Ray Tracing for Nvidia and Imagination Technologies while AMD/Others use software methods. AMD is certianly adding some AI/Tensor related Vega micro-arch extentions to Vega 20 for that Pro AI market’s needs but AMD can probably work that into Navi also and it’s the AI/Tensor core based denoising that makes Nvidia’s very Grainy Hardware Ray Tracing output of use for gaming!
Nvidia is still using more raster oriented methods on its Turing/RTX SKUs and very limited Ray Tracing for only specific effects because it’s not GigaRays per second it’s more like MegaRays per 33.33 millisecond frame times(30 FPS) or even less MegaRays available per 16.67 milleseconds Frame Times(60 FPS). So that’s not very many Rays per Frame Time to give each pixel at 1440p enough Rays Per Pixel to the entire scene or even any extra Rays for the Ambient Occlusion(AO), Shadow, other Ray Traced render passes!
So that more Hybrid rendering for Nvidia and Raster Workloads mostly still with Limited Ray Tracing mixed in for better reflections, refractions, AO, other Ray traced Effects on a limited amount of a scene’s Mesh-Objects/materials values.
I think that probably in 2021 that AMD could bring something similar to Discrete AMD GPUs and most likely Microsoft and Sony are funding AMD’s semi-custom developments for at least AI based Upscaling(AI based similar to Nvidia’s “DLSS” IP). The console makers are big upscaling IP users. Nvidia upscales 1440P to 4K as the DLSS targets that to allow for all that Limited RTX technology to work more efficiently conisdering even the top end Turing SKUs still lack enough GigaRays/s for even useable limited/Hybrid full native 4K rendering in gameing with sufficient Rays available per linited amounts of pixels.
But Lisa is sure going to be one Happy CEO come 2019 looking at to the pervious 2 business quarters’ Growth Figures. And AMD’s going all 7nm TSMC process node is something to be damn proud of for AMD! AMD will be making more revenues selling Epyc SKUs than Nvidia makes selling GPUs sometimes around 2020-21 the way things are heading with Epyc. Vega 20 and any other deals between AMD and Xilinx for more Epyc CPUs paired with Xilinx AI accelerators also.
AMD and Xilinx now that would make for some interesting children! Can anyone say FPGAs on the HBM#’s DRAM stacks and “DLSS” like AI sorts of processing done right in the VRAM. That would be a possible method that maybe the Console Customers would look at as a quick to certify method while AMD works on some later GPU micro-arch that fully includes the same sorts of dedicated on GPU hardware solutions as Nvidia’s RTX IP.
thanks for your insight
thanks for your insight
New hardware generally isn’t
New hardware generally isn’t mainstream for a generation or two after it is released, so they have time. It is al marketing at this point.
AMD might actually have a
AMD might actually have a chance to catch up this time. That would be cool, not that I am in the market for a new GPU.