GIGABYTE, a supplier of high-performance computing (HPC) systems, today disclosed four NVIDIA HGX A100 platforms under development. These platforms will be available with NVIDIA A100 Tensor Core GPUs. NVIDIA A100 is the first elastic, multi-instance GPU that unifies training, inference, HPC, and analytics. These four products include G262 series servers that can hold four NVIDIA A100 GPUs and G492 series that can provide eight A100 GPUs. Each series also distinguishes between two models, which support the 3rd generation Intel Xeon Scalable processor and the 2nd generation AMD EPYC processor. The NVIDIA HGX A100 platform is a key element in the NVIDIA accelerated data center concept that brings huge parallel computing power to customers, thereby helping customers accelerate their digital transformation.

With GPU acceleration becoming the mainstream technology in today's data center. Scientists, researchers and engineers are committed to using GPU-accelerated HPC and artificial intelligence (AI) to meet the important challenges of the current world. The NVIDIA accelerated data center concept, including GIGABYTE high-performance servers with NVIDIA NVSwitch, NVIDIA NVLink, and NVIDIA A100 GPUs, will provide GPU computing power required for different computing scales. The NVIDIA accelerated data center also features NVIDIA Mellanox HDR InfiniBand high-speed networking and NVIDIA Magnum IO software that supports GPUDirect RDMA and GPUDirect Storage.

Here's the first picture of a Tesla A100 "Ampere" AIC (add-in card) form-factor board, hot on the heals of the morning big A100 reveal. The AIC card is a bare PCB, which workstation builders will add compatible cooling solutions on. The PCB features the gigantic GA100 processor with its six HBM2E stacks, in the center, surrounded by VRM components, and I/O on three sides. On the bottom side, you will find a conventional PCI-Express 4.0 x16 host interface. Above it, are NVLink fingers. The rear I/O has high-bandwidth network interfaces (likely 200 Gbps InfiniBand), by Mellanox. The tail end has hard points for 12 V power input. Find juicy details of the GA100 in our older article.

NVIDIA is invested in the development of multi-GPU, specifically SLI over NVLink, and has developed a new multi-GPU rendering technique that appears to be inspired by tile-based rendering. Implemented at a single-GPU level, tile-based rendering has been one of NVIDIA's many secret sauces that improved performance since its "Maxwell" family of GPUs. 3DCenter.org discovered that NVIDIA is working on its multi-GPU avatar, called CFR, which could be short for "checkerboard frame rendering," or "checkered frame rendering." The method is already secretly deployed on current NVIDIA drivers, although not documented for developers to implement.

In CFR, the frame is divided into tiny square tiles, like a checkerboard. Odd-numbered tiles are rendered by one GPU, and even-numbered ones by the other. Unlike AFR (alternate frame rendering), in which each GPU's dedicated memory has a copy of all of the resources needed to render the frame, methods like CFR and SFR (split frame rendering) optimize resource allocation. CFR also purportedly offers lesser micro-stutter than AFR. 3DCenter also detailed the features and requirements of CFR. To begin with, the method is only compatible with DirectX (including DirectX 12, 11, and 10), and not OpenGL or Vulkan. For now it's "Turing" exclusive, since NVLink is required (probably its bandwidth is needed to virtualize the tile buffer). Tools like NVIDIA Profile Inspector allow you to force CFR on provided the other hardware and API requirements are met. It still has many compatibility problems, and remains practically undocumented by NVIDIA.

ZOTAC GAMING is excited to introduce the new GeForce RTX SUPER series of graphics cards pushing more CUDA Cores, more GDDR6 memory, more memory bandwidth, and more power. Continuing the push for next-gen gaming, each SUPER Series enable extraordinary performance with real-time ray tracing with ray tracing cores and tensor cores. Each ZOTAC GAMING GeForce RTX SUPER series is equipped with the most powerful cooling hardware, IceStorm 2.0 cooling, and beautifully lit with SPECTRA lighting.

NVIDIA today announced its support for Arm CPUs, providing the high performance computing industry a new path to build extremely energy-efficient, AI-enabled exascale supercomputers. NVIDIA is making available to the Arm ecosystem its full stack of AI and HPC software - which accelerates more than 600 HPC applications and all AI frameworks - by year's end. The stack includes all NVIDIA CUDA-X AI and HPC libraries, GPU-accelerated AI frameworks and software development tools such as PGI compilers with OpenACC support and profilers. Once stack optimization is complete, NVIDIA will accelerate all major CPU architectures, including x86, POWER and Arm.

"Supercomputers are the essential instruments of scientific discovery, and achieving exascale supercomputing will dramatically expand the frontier of human knowledge," said Jensen Huang, founder and CEO of NVIDIA. "As traditional compute scaling ends, power will limit all supercomputers. The combination of NVIDIA's CUDA-accelerated computing and Arm's energy-efficient CPU architecture will give the HPC community a boost to exascale."

Motherboard vendors are betting big on the success of AMD's "Valhalla" desktop platform that combines a Ryzen 3000-series Zen 2 processor with an AMD X570 chipset motherboard, and have responded with some mighty premium board designs. GIGABYTE deployed its full spectrum of Aorus branding, including Ultra, Elite, ITX Pro, Master, and Xtreme. The X570 I Aorus Pro WiFi mini-ITX motherboard is an impressive feat of engineering despite its designers having to wrestle with the feisty new PCIe gen 4 chipset. It draws power from a combination of 24-pin and 8-pin connectors, and conditions power for the SoC with an impressive 8-phase VRM that uses high-grade PowIRstage components. A rather tall fan-heatsink cools the X570 chipset, with a 30 mm fan.

Connectivity options on the X570 I Aorus Pro WiFi are surprisingly aplenty. The sole expansion slot is a PCI-Express 4.0 x16, but the storage connectivity includes not one, but two M.2-2280 slots (reverse side of the PCB), each with PCI-Express 4.0 x4 and SATA 6 Gbps wiring. Four SATA 6 Gbps ports make for the rest of the storage connectivity. Networking options include 2.4 Gbps 802.11ax WLAN, Bluetooth 5.0 (Intel , and 1 GbE, all pulled by Intel-made controllers. USB connectivity includes six 5 Gbps USB 3.2 gen 1, and two 10 Gbps USB 3.2 gen 2 ports (of which one is type-C), and two 5 Gbps ports by headers. The onboard audio solution has 6-channel analog output, but is backed by a premium Realtek ALC1220VB Enhance CODEC (114 dBA SNR).

Intel has been working on CXL, short for Compute Express Link gen 1, for over four years new. This new interconnect protocol was donated to a new consortium of tech companies for release as a the CXL 1.0 standard. Its protocol layer will pave the way for PCI-Express gen 5.0 to sustain its bandwidth growth target of being twice as fast as PCIe gen 4.0. CXL 1.0 is out to compete with other established PCIe-alternative slot standards such as NVLink from NVIDIA, and InfinityFabric from AMD. It has one killer advantage, though: the CXL 1.0 is pin-compatible and backwards-compatible with PCI-Express, and uses PCIe physical-layer and electrical interface.

This reduces hardware upgrade costs for data-centers. CXL maintains memory coherency between the CPU's memory-space and memory on installed devices. The CXL Consortium, or SIG, includes data-center and cloud-computing giants, including Alibaba, Cisco, DellEMC, Facebook, Google, HPE, Huawei, Microsoft, and of course Intel. CXL will be used bot as a socketed/slotted interface for add-on cards and GPU boards, and as an embedded interface. We estimate bandwidth of CXL to be 32 Gbps per lane, or four times that of PCIe gen 3.0, keeping in line with PCIe gen 5.0 bandwidth growth estimates.

Late January it was reported that Intel is looking to buy out Israeli networking hardware maker Mellanox Technology, in what looked like a cakewalk USD $6 billion deal at the time, which was a 35 percent premium over the valuation of Mellanox. Turns out, Intel hasn't closed the deal, and there are other big tech players in the foray for Mellanox, the most notable being NVIDIA. The GPU giant has reportedly offered Mellanox a competitive bid of $7 billion.

NVIDIA eyes a slice of the data-center networking hardware pie since the company has invested heavily in GPU-based AI accelerators and its own high-bandwidth interconnect dubbed NVLink, and now needs to complete its hardware ecosystem with NICs and switches under its own brand. Founded in 1999 in Yoqneam, Israel, Mellanox designs high performance network processors and fully-built NICs in a wide range of data-center relevant interconnects. Intel is by far the biggest tech company operating in Israel, with not just R&D centers, but also manufacturing sites, in stark contrast to NVIDIA, which opened its first R&D office in 2017 with a few hundred employees.

Update: NVIDIA's bid for Mellanox stands at $7 billion.

Last week was quite busy with news, and one thing that slipped through the cracks then was the announcement of ZOTAC's new NVLink bridging solutions. Marketed under the ZOTAC Gaming brand, these follow in line with other major NVIDIA add-in card (AIC) partners in providing an alternative to the first-party OEM offering, just in case you preferred a different aesthetic. These are available in either a 3-slot (74.5 x 75.5 x 25 mm) or a 4-slot (74.5 x 96 x 25 mm) spacing option, with backlighting support for the ZOTAC Gaming logo in the center that can be configured by their SPECTRA software lighting suite.

These NVLink bridges have not yet been made available for purchase in most regions, with expected retail availability ranging from December 14-20, but those in East Asia can already find them available for comparable prices as with other such options from MSI, ASUS, and GIGABYTE. ZOTAC also put out an unboxing video which gives a better look at the product relative to the renders below, for those interested.

NVIDIA today introduced NVIDIA TITAN RTX , the world's most powerful desktop GPU, providing massive performance for AI research, data science and creative applications. Driven by the new NVIDIA Turing architecture, TITAN RTX - dubbed T-Rex - delivers 130 teraflops of deep learning performance and 11 GigaRays of ray-tracing performance.

"Turing is NVIDIA's biggest advance in a decade - fusing shaders, ray tracing, and deep learning to reinvent the GPU," said Jensen Huang, founder and CEO of NVIDIA. "The introduction of T-Rex puts Turing within reach of millions of the most demanding PC users - developers, scientists and content creators."

It's been a busy past 48 hours at EVGA, with the launch of the B360 Micro motherboard, unveiling of the NVLink bridges, and now a teaser of what could very well be the company's Z390 DARK motherboard, targeted at professional overclockers. K|ngp|n shared this teaser image of the board on social media, revealing a socket LGA1151 motherboard that's laid out like an LN2 overclocker's dream - memory slots north of the CPU sockets, CPU VRM to its west and south, and power drawn from a combination of 24-pin ATX and two 8-pin EPS connectors angled away toward the east. A cluster of 7-segment LED displays put out diagnostic codes. The designers seem to have opted for an expensive 8~10-layer PCB that's rich in copper. We'll hear more about this beauty as the Z390 platform launches later this month.

EVGA unveiled the design of its NVLink bridges, closely following ASUS and GIGABYTE. Its design has the most pronounced C-shape among all aftermarket bridges, and four nearly independent RGB LED diffusers along the top shroud, which somewhat has the design language as its recent graphics cards. Like the others, the EVGA NVLink bridge could come in 3-slot (2 slots between cards) and 4-slot (3 slots between cards) sizes, and could be priced at $80, if not more.

After NVIDIA's own $80 NVLink bridge, and ASUS' slightly more functional ROG NVLink Bridge, it's becoming clear that at $80, or higher price than entry-level graphics cards and motherboards, NVIDIA is guiding its partners to make a business of selling $5 pieces of hardware at $80. The latest entry to this contempt toward gamers is GIGABYTE, with its Aorus-branded NVLink Bridge, available in 3-slot (2 slots between cards) and 4-slot (3 slots between cards) variants. The bridge is characterized by a less pronounced C-shape than NVIDIA's bridge, through not completely trapezoid like ASUS'. The Aorus logo is capable of 16.7 million color RGB LED illumination, which you control via GIGABYTE RGB Fusion software. Like every other 2-way NVLink bridge, this one will go for $80.