Graphics performance of AMD Navi graphics processor "Radeon RX" graphics apparently infiltrated

The AMD Navi GPUs are launched this year, but we have no information about their performance or the specifications they would give so far. At CompuBench, we could have our first list of a "Radeon RX" part of AMD Navi that gives interesting results.

AMD Navi "Radeon RX 66AF: F1": Unparalleled performance of GPU performance – Faster graphics, slowdowns in compute workloads

The list on CompuBench was revealed by Dylan522p in the r / hardware subdirectory. The site lists the specifications and performance measures of a device with the code name "66AF: F1". From the first impression, this seems to be a part of the next generation based on the Navi GPU, as this specific device ID has not yet been released. It should now be noted that this unit may be based on Navi's GPU architecture or something quite different. That's why I treat this information as a rumor, but the results are very interesting, so let's take a look.

Related NVIDIA discrete GPU market share increased to 81.2%, AMD to 18.8% in fourth quarter of 2018 – fourth quarter has been worse for AIBs for more than 10 years

The so-called AMD Navi graphics processor comes with 20 CPUs, which equates to 1280 stream processors. This would be the case if AMD retains the same CPU design with 64 shaders per CPU and the GPU has 20 CPUs, because CompuBench has the habit of incorrectly reporting the number of compute units.

When it comes to performance, we first need to determine if we are comparing different devices in the same operating system and API, because CompuBench can produce variable results when users run an operating system / Different API. A device running Windows / DirectX API would be much faster than a device running OSX and using OpenCL APIs or Vulkan. Regarding the performance results, the first tests are pure graphics in which the well-known AMD Navi GPU displays results and performance far superior to those of the Radeon RX 580. This almost corresponds to AMD Radeon RX territory. Vega 56.

However, in computational tests, the map is almost similar to the Radeon RX 580, which we also expected since AMD Navi GPUs were custom-designed for maximum graphics and game performance, rather than power. Calculation. The agreements made with Microsoft and Sony for their next-generation consoles have been very clear in this regard. So it's easy to say that desktop and laptop components will receive similar treatment.

Related Exclusive: AMD's flagship product Ryzen 7 3750H will be available in April

Now, if we compare the device to the Radeon RX Vega 64 graphics card, the performance performance of the Vega part is almost doubled, but the AMD Navi graphics processor manages to get close to it in some greedy tests. Graphics resources and bandwidth, which indicates that AMD would rely on HBM2 VRAM, even on their mainstream cards. Another possible reason could be the use of GDDR6 which also provides excellent results, since a broad 256-bit interface can reach 448 GB / s, which is close to the bandwidth of 512 GB / s Vega 56 and Vega 64 pieces and perhaps cheaper to implement.

That said, I always want to mention that it is only pure speculation on the part based on a Navi GPU. The only credible information available at the moment is that AMD Navi GPUs will feature the 7nm process node and a redesigned architecture designed for games first. The GPUs will be used in both next-generation consoles and will target the launch of 2H 2019, with consumer games being the priority target market. We'll keep you posted as soon as we have more information on AMD Next-gen Navi graphics cards.

Share
Tweet
Submit