Chinese GPU manufacturer Innosilicon has launched their first standalone chip Fantasy One, which will be used in a series of graphics cards. The GPU was designed in collaboration with GPU designer Imagination Technologies.

GPU and its corresponding products will be displayed in the "Fantasy One GPU" product press conference. Fantasy One GPU is based on the BXT IP of its GPU partner Imagination Technologies. The company launched a total of four products, including solutions based on single GPU and dual GPU.

Huawei 7nm Kunpeng 920 CPU test is suitable for Chinese desktop computers

It is reported that the "Dream 1" GPU is the industry's first rendering GPU that adopts cutting-edge technologies such as the world's top GDDR6X high-bandwidth video memory technology and Innolink chiplets. China's patented physical unclonable encryption PUF technology, and self-developed advanced technologies such as PCIe 4.0, HDMI2.1, DP/eDP 1.4, etc. are also converged in one core.

"Dream 1" is the first domestic compatible 4K high-definition desktop and high-performance server-level GPU, supports domestic Linux, Android, Windows operating systems; supports OpenGL, OpenGLES, OpenCL, Vulkan, DX and other mainstream graphics frameworks; supports VR/AR/AI ; Smooth support for data center-level multi-user application scenarios such as multi-channel cloud games, cloud mobile phones, cloud office, and cloud desktops. The successful tape-out of "Fenghua One" demonstrates Inspur's strong strength and low-key style in the development of advanced technology IP and GPU customization.

InnoSilicon Fantasy One Type A graphics card

The first product to talk about is the Type A graphics card driven by Fantasy One GPU. Type A is the name of the graphics card, not to be confused with the USB Type-A port. The Type A card adopts a dual-slot and dual-fan form factor, and is said to provide up to 5 TFLOPs of FP32 and 160 GPixel/s fill rate. This seems to be a slightly reduced version of Fantasy One GPU, because the BXT GPU IP it is based on provides up to 6 TFLOPs of FP32, 24 TOPs AI and 192 GPixel/s fill rate. GPU also relies on multi-chip design, so we are talking about small chips here, but we don't have more details for now.

In terms of memory, a single Fantasy One GPU provides up to 16 GB of GDDR6X memory, running on a 128-bit bus interface. The memory chip runs at a transfer speed of 19 Gbps, providing up to 304 GB/s of bandwidth for the GPU. In addition, the card also provides DisplayPort 1.4, HDMI 2.1 and VGA output. The card is powered by a single 6-pin connector. It is said that the typical TDP for 4K rendering is about 20W, while the cloud rendering power is rated at 50W. There is also a specific model of the Type A server that uses a passive radiator.

InnoSilicon Fantasy One Type A SFF graphics card

There is also an entry-level/small solution based on Type A Fantasy One GPU. Considering its limited size and power, this graphics card will definitely lower the specifications and clock in the full version. The display card PCB adopts a 4+1 phase design and is equipped with a small fan heat sink. It has HDMI 2.1 and DisplayPort 1.4 output. There is no power connector on the card, and considering the size of the PCB, your memory capacity seems to be reduced.

The CPU independently developed by Zhaoxin provides support for a full range of x86 PCs in China-16nm chips, up to 8 cores, suitable for Chinese consumer desktops, notebooks and AIOs

InnoSilicon Fantasy One TypeB graphics card

On the high-end side, Innosilicon Type-B graphics cards have dual Fantasy One GPUs on the same PCB. The card uses an interconnect called Innolink, which may be a PLX chip that interconnects two GPUs on the same board. The graphics card provides up to 10 TFLOPs FP32 and up to 320 GPixel/s fill rate. It can also run up to 32 1080p/60 FPS or 64 720/30 FPS streams. Type B supports 32 GB GDDR6X memory through a 128-bit bus interface. Power is provided through a dual 8-pin connector. All cards are compliant with PCIe Gen 4.0.

Fantasy One Type A graphics cards target the consumer, workstation, and data center markets, while Type B graphics cards target data center, cloud computing, and AI use cases. These graphics cards will support a series of APIs including DirectX and have been demonstrated in Unigine Heaven, which is a fairly old benchmark of today's standards, but shows that the expected DirectX technology is indeed applicable to Fantasy One chips.

Innosilicon has not stayed on Fantasy One, and is already talking about Fantasy 2 and Fantasy 3 GPUs. It is said that they will be launched in 2022 and will use 5nm process technology.

News source: Videocardz, Ithome, Mydrivers

The latest technology news in your inbox, once a day!