AMD seems to be working on RDNA cards for cryptocurrencies too

The first signs that AMD might be building a specialized gas pedal for cryptocurrency mining appeared back in October 2020. A Linux kernel update added a new Navi 10-based gas pedal with ID 731E. This card, as clearly spelled out in the description, will not support DCN (Display Core Next) and VCN (Video Core Next), which means that it is a graphics processor without image output capability.

This graphics card can be used either to accelerate computing in clusters or for cryptocurrency mining. According to a new patch on March 3, AMD has created the Navi 12 GPU, which also lacks VCN, that is the ability to output an image. The latest GPU comes just after NVIDIA made an announcement about its mining solutions, which are expected to launch later this month. Interestingly, Navi 12 is not available in any free-market Radeon graphics card. It was designed specifically for the Apple Mac as the Radeon Pro 5600M. It is the only Navi 1X GPU with HBM2 memory. There are no mining tests available for it, so it's hard to predict how well it will perform in cryptocurrency mining.

In terms of specifications, both Navi 10 and Navi 12 offer 2,560 stream processors (40 processing units). Navi 10 is available with 8 GB of GDDR6 memory, while Navi 12 has 2 HBM2 memory stacks of 4 GB. The GDDR6 memory has a bandwidth of 448 Gbps, while the Navi based on HBM2 can offer 394 Gbps. If we believe the words of a rather reputable informant Komachi, AMD started preparing variants of the RX 5700 XTB, RX 5700 B and RX 5500 XTB specifically for cryptocurrency, no later than last November. These could be working names for the new cards, but it's unclear if AMD would want to use the Radeon brand for such gas pedals.

Updated Tesla Model S's on-board computer gets powerful AMD Navi graphics processor 23

Renowned whistleblower Patrick Schur posted on Twitter a flowchart explaining how the upgraded Model S electric car's onboard computer will be able to run such graphically demanding video games as «The Witcher 3», Cyberpunk 2077 and many others.

According to the diagram, the new Model S electric cars' entertainment system, which is likely to also act as a navigation system, will use the AMD Navi 23 GPU.

The diagram indicates that the graphics subsystem will be equipped with 16Gb (2GB) GDDR6 memory chips from Samsung (K4ZAF325XM). There will be four such memory modules. Thus, the total amount of video memory of the Model S on-board computer will be 8GB.

The chart also reports that the 14 Gb/s video memory will feature 128-bit bus support, for a maximum throughput of 224 Gb/s, similar to what's possible in Xbox Series S consoles. Additionally, Tesla will have two DisplayPort and one HDMI connector for connectivity to in-vehicle monitors, as well as a PCIe Gen4 x8 interface for connectivity to other cards.

According to Tesla itself, the onboard computer of the new Model S will be able to provide up to 10 Tflops of processing power, which is almost equivalent to the performance of a PlayStation 5 game console. To do that, the 32 GPU Navi 23 processing units with 2048 stream processors would have to run at a minimum of 2.44 GHz.

Samsung has promised AMD graphics as early as the next generation of Exynos processors

At the Exynos 2100 launch event today, Samsung announced that its collaboration with AMD to integrate RDNA graphics architecture into Exynos chips has borne fruit.

President of Samsung System LSI (mobile chips, image sensors, neuroprocessors, security chips and mobile displays) Dr Inyup Kang confirmed during the event that the company is building next-generation mobile graphics with AMD, which will be launched in its next flagship product. That's all that was said on the subject. What exactly was meant by the next flagship product remains a mystery. One would hope that a new single-chip system with AMD graphics would come out no later than the Galaxy Note21 (although there are rumours of the Galaxy Note series being discontinued). Perhaps the flagship with AMD GPU will be the Galaxy Z Fold 3 or another bendable smartphone?

All in all, it's very likely that a new Exynos chip with AMD integrated graphics will be available before the end of this year. The actual products will probably not appear until early 2022 (though, I'd like to be wrong and see the fruits of AMD and Samsung's collaboration as early as possible). It's not yet clear whether we're talking about using RDNA architecture or RDNA 2 already (perhaps it will be their mobile hybrid). The latest Exynos 2100 is the 5nm chip that will underpin the Galaxy S21 in most markets. It uses standard ARM cores in a bundle of one most powerful Cortex-X1 @2.9GHz, three performance Cortex-A78 and three power-efficient Cortex-A55. LPDDR5 RAM and UFS 3.1 storage are supported. ARM Mali G78 is used as graphics, there is a powerful neural processor, a new 5G modem and an improved signal processor. Samsung has also released a promotional video for this single-chip system:

Patent suggests AMD is going to apply chipsets to graphics cards, despite difficulties

A new application filed with the US Patent and Trademark Office on the last day of last year describes AMD's approach to a potential chiplet-based graphics accelerator design. The manufacturer pointed out the problems in such a design and explained how they could be avoided in the future.

According to AMD, graphics processors are still monolithic chips for a reason: there are plenty of problems with the interaction of multiple chips. The current graphics pipeline is not particularly efficient for multiple GPUs (the same goes for configurations like Crossfire or SLI), as it is sometimes difficult to distribute tasks between multiple active chips in a system. AMD also describes synchronisation of memory content between multiple GPU chips as being complex and costly from a design perspective.

AMD believes that such problems could be avoided by introducing «high bandwidth passive crosslinks». According to AMD, the first GPU chipset should be directly connected to the CPU, while each of the graphics chipsets in the array & ; are connected to the first GPU via passive crosslinks. Such links must be placed in a silicon interposer layer, which is an intermediate wafer with conductors in one or more layers. A group of interconnected chips would work, essentially, like a system on a chip.

Conventional GPUs have a common last-level cache (LLC), but to avoid timing issues AMD believes that each of the chipsets should get their own LLC, with all of them being unified and consistent across all chipsets.

AMD has not publicly confirmed that it is working on a chiplet-based graphics accelerator design. However, there have been rumours that the RDNA 3+ design could be chiplet-based. AMD has a lot of experience with multi-chip designs, especially in its Ryzen processors and various APUs, including gaming consoles.

It is expected that both competitors, NVIDIA and Intel, could also go down this path, allowing them to produce accelerators with higher performance. Intel has already confirmed that it is working on Xe-HP graphics cards with a multi-chip design & ; they are expected to debut this year. NVIDIA is rumoured to unveil its first such accelerator along with the Hopper architecture. Back in 2019, NVIDIA's chief science advisor Bill Dally  said that the company has developed technology to build a graphics accelerator with a multi-chip layout.