Manli introduced its newest high-end graphics card, the Nebula GeForce RTX 5090 powered by the NVIDIA Blackwell architecture. This card is Manli's fourth RTX 5090 version and the first time a flagship GPU model has joined Manli's Nebula lineup alongside the Gallardo and Stellar RTX 5090 series. The Nebula RTX 5090 runs at base and boost clock speeds of 2017 MHz and 2407 MHz respectively. The card has 32 GB of GDDR7 memory with a 512-bit interface providing memory speeds of 28.0 Gbps and bandwidth up to 1,792 GB/s.

The graphics card needs 575 W of power and keeps its GPU temperature at a maximum of 90°C. In terms of cooling it features a large heatsink with a triple cooler setup. This includes three 100 mm fans (which stop spinning during idle states) and nine 6 mm heat pipes. For enhanced aesthetics and structural rigidity Manli's Nebula GeForce RTX 5090 features a die-cast cover and a metal backplate. Connectivity-wise, the graphics card offers three DisplayPort outputs and one HDMI port. It measures 345 x 150 x 65 mm and takes up 3.5 slots. It's worth noting that this model is intended for international markets, unlike the 5090 D variants that Manli offers specifically to the Chinese market. Exact pricing and availability are yet to be known.

Reuters reports that according to two sources with knowledge of the situation, NVIDIA has placed orders for 300,000 H20 Blackwell chipsets to TSMC. The order shows high Chinese demand that pushed the U.S. chip giant to go beyond its current stock. The Trump administration gave NVIDIA permission to start selling H20 graphics processing units to China again changing an April ban meant to stop advanced AI chips from reaching Chinese markets because of national security worries. NVIDIA made the H20 chip just for the Chinese market after U.S. export limits targeted its other AI chipsets in late 2023. The H20 has less computing power than NVIDIA's H100 or newer Blackwell series sold in other places.

The new TSMC orders would add to NVIDIA's current H20 stock of 600,000 to 700,000 chips, sources said. To put this in perspective, NVIDIA sold about one million H20 chips in 2024, based on research from SemiAnalysis. After April's sales ban, NVIDIA warned of potential $5.5 billion inventory write-offs and $15 billion in foregone sales. During a trip to Beijing this month, NVIDIA CEO Jensen Huang said H20 order numbers would decide if production would start again pointing out that getting the supply chain going would take nine months. After his visit, reports suggested NVIDIA told customers it had limited H20 stocks with no plans to start wafer production again soon. The company has requested that Chinese buyers submit new documentation including order volume forecasts from clients. Despite competition from Huawei's less powerful alternatives, NVIDIA products remain popular in China, evidenced by increased repair demand for banned GPUs, many smuggled into the country.

NVIDIA is looking to hit the PC processor market with shock and awe, suggests the CPU core counts and iGPU CU counts. Having seen the success of Qualcomm in this market, NVIDIA is designing its new N1X PC processor in anticipation of Microsoft opening up the Windows 11 Arm Copilot+ AI PC ecosystem, letting in new players such as NVIDIA and MediaTek. A new Geekbench online database submission sheds light on what is under the hood.

Geekbench is able to detect the number of CPU cores and GPU OpenCL compute units it is able to address. A user with access to an N1X test machine ran Geekbench 6.4.0, and the benchmark yielded a score of 46361 points in the OpenCL test. The chip is detected having 20 CPU cores, and exposes 48 compute units (OpenCL terminology) to the GPU compute benchmark. The N1X is expected to implement a heterogenous multicore CPU complex, and while it has 20 cores, these are not all the same. The processor is expected to pack a powerful iGPU based on the Blackwell graphics architecture, with 48 SM (streaming multiprocessors), The idea behind such a chip would be to compute with Apple's M-Pro and M-Max series SoCs powering its latest MacBook Pros, as well as x86-based PC chips such as the Ryzen AI Max "Strix Halo." 48 Blackwell streaming multiprocessors is identical to that of the desktop GeForce RTX 5070, but the iGPU, making it possibly the fastest iGPU at launch.

INNO3D, a leading manufacturer of high-end multimedia components and innovations proudly announces the new additions to the GeForce RTX 50 Series, the GeForce RTX 5090 & RTX 5080 Frostbite Pro, featuring Alphacool's 1-Slot-Design with backplate for high-performance liquid cooling. Founded in 1998 with the vision of developing pioneering computer hardware products on a global scale. Fast-forward to the present day, INNO3D is now well-established in the gaming community known for our innovative and daring approach to design and technology. We are Brutal by Nature in everything we do and are 201% committed to you for the best gaming experience in the world.

Designed for enthusiast gamers, overclockers, and workstation professionals, the Frostbite Pro takes unparalleled thermal efficiency to the next level with Alphacool's enterprise-grade 1-slot water cooler. Its chrome-plated copper cooling block ensures maximum heat dissipation, while carbon and brass components deliver durability and seamless integration into high-density environments.

NVIDIA has revealed last week its plan to roll out 600,000 to 800,000 SOCAMM memory modules in 2025. This move puts SOCAMM technology in a position to replace current high-bandwidth memory (HBM) options. NVIDIA aims to use these modules in its upcoming AI products such as the GB300 "Blackwell" platform and AI PC Digits system, which it showed off at GTC 2025. Industry reports indicate that NVIDIA has communicated projected order quantities to memory and substrate suppliers. NVIDIA engaged three major memory manufacturers: Samsung Electronics, SK Hynix, and Micron Technology for SOCAMM co-development. As Digitimes Asia reports citing ET News and Wccftech, Micron has secured approval for volume production ahead of its competitors.

The System on Chip Advanced Memory Module (SOCAMM) technology uses LPDDR DRAM to boost performance for low-power high-bandwidth AI computing tasks. Micron says SOCAMM gives 2.5 times more bandwidth while reducing size and power use by a third compared to standard server RDIMM modules. The planned SOCAMM rollout though smaller than NVIDIA's expected 9 million HBM unit buy for 2025, marks a key market shift. Industry experts see this as a possible game-changer for memory and substrate fields. SOCAMM's rollout covers business and consumer markets first aiming at AI servers and workstations before moving to personal computers. The technology links cost-effective growth with high-performance AI workload needs.

Team OGS has demonstrated an impressive feat of extreme overclocking by extracting every last drop of performance from GALAX GeForce RTX 5090D HOF edition. Armed with a custom Extreme OC BIOS, authorized for up to 2,000 W of power, and two 12V-2x6 connectors delivering 1,200 W, the group cooled the card with liquid nitrogen and fine-tuned voltages to push clock speeds far beyond factory limits. In the GPUPI 32B benchmark, the GPU achieved a record-breaking core frequency of 3,650 MHz and a completion time of just 39.434 seconds, securing a new world record for GPU-accelerated Pi calculation.

The team's success extended to graphics tests as well. In 3DMark Port Royal, the GPU operated at 3,570 MHz on its core and 2,250 MHz on its GDDR7 memory, achieving 47,469 points. Under Unigine Superposition's 1080p Extreme preset, they maintained 3,540 MHz and 2,250 MHz, respectively, reaching 38,237 points, an increase of several thousand over previous records. Pushing memory speeds to an effective 36 Gbps translates into 2.304 TB/s of bandwidth, nearly 29% above the standard specifications. These results show the untapped potential of NVIDIA's Blackwell architecture and suggest that, with similar power, cooling, and VBIOS modifications, professional cards like the RTX Pro 6000 series could achieve even more significant gains, given the higher CUDA core count.

NVIDIA's freshly completed GeForce RTX 50 family, powered by the new Blackwell architecture, has begun to register meaningful numbers in Steam's June 2025 Hardware Survey. Since first appearing in May, cards from this lineup, except for the as-yet-unavailable RTX 5050, now account for 3.69% of surveyed systems. Leading the pack among the newcomers, the RTX 5070 grabs nearly 1% of overall share, up substantially from its debut, while the RTX 5080 and RTX 5070 Ti follow closely behind. The more budget‑oriented RTX 5060 Ti and RTX 5060 have also made their mark, and even the top‑end RTX 5090 has registered on enough machines to appear in the survey.

These figures show a swift uptake by desktop gamers eager for improved performance and AI-driven features, even as the tried-and-true RTX 4060 Laptop GPU holds onto its position as the most prevalent NVIDIA part, with just under 5% of the installed base. Meanwhile, AMD's latest Radeon RX 9000 series and Intel's Arc B‑series remain absent from the survey results, suggesting that shipment volumes for those cards have not yet reached the critical mass needed to register with Valve's monthly sampling of millions of Steam users. This demonstrates NVIDIA's continued dominance in add-in board sales, where it has consistently captured over 90% of the market.

Initial benchmarks shared on China's Weibo platform by INNO3D hinted that NVIDIA's upcoming GeForce RTX 5050 may not quite catch up to the GeForce RTX 4060 in raw performance. The leaked figures, which focused on a handful of 3DMark tests and a selection of game runs, showed the RTX 5050 lagging just a few percent behind its established last-gen mid-range sibling. While those early results did not cover a comprehensive range of titles, they suggested that the new card would fall slightly short of the RTX 4060's average frame rate and 1% low metrics when gaming at 1080p settings.

A more complete picture emerged from South Korea's Quasar Zone review of the Colorful iGame GeForce RTX 5050 Ultra OC White edition. In tests across fifteen popular 1080p games, the RTX 5050 delivered roughly 7% lower average frame rates and similar 1% low results compared to the RTX 4060. Even Intel's Arc B580 managed to edge it out by a small margin in those same trials. However, NVIDIA's Multi‑Frame Generation options dramatically change the story by reconstructing additional frames on the fly. When these technologies are enabled, the RTX 5050 not only leaves older-generation cards well behind but also overtakes the higher‑priced RTX 4060 Ti in several benchmarks. As we await more reviews to emerge, it will be interesting to see if the newest DLSS and MFG are sufficient for gamers to purchase the entry-level Blackwell GPU or opt for an older mid-range Ada Lovelace.

NVIDIA has quietly engineered a third variant of its flagship GPU in response to new US export regulations, aiming to maintain performance while avoiding restrictions. A few weeks ago, we got leaks about the GeForce RTX 5090DD, which was supposed to be a new version based on the modified RTX 5090D. The initially released GeForce RTX 5090D was designed to comply with the Department of Commerce's updated rules while delivering virtually identical results to its unrestricted counterpart in both gaming and synthetic benchmarks. Now, the newly named RTX 5090D v2 features the same 21,760 CUDA cores and a 575 W power draw, ensuring that the core Blackwell architecture performs as users would expect, albeit with a reduced 24 GB VRAM buffer.

NVIDIA plans to introduce an updated RTX 5090D v2 model in August. According to reliable industry sources, this refresh will feature 24 GB of GDDR7 memory on a 384-bit bus, down from the original 512-bit interface, while preserving overall thermal and computational characteristics. For most gamers, the reduced VRAM capacity is unlikely to be noticeable, as only a handful of titles require more than 24 GB of VRAM. Ultimately, the success of the v2 will depend on whether it reaches the market at a price point that aligns with the expectations of its flagship. Given that the target market is China, we have to wait for sales and performance figures first. According to a well-known leaker, kopite7kimi, they noted that there is some surprise, indicating that NVIDIA has made a significant change to comply with export regulations, so it could be a firmware update preventing these cards from running AI workloads.

Manli Technology Group Limited is proud to announce the Manli GeForce RTX 5050 graphics card. Step up to NVIDIA Blackwell with the GeForce RTX 5050, featuring 4th-gen ray tracing and 5th-gen Tensor Cores for game-changing AI capabilities and performance in top games and apps.

Manli Design
There are 2,560 CUDA cores onboard powering the RTX 5050. It also features 8 GB of memory, and GDDR6 memory speeds of up to 20 Gbps, with 5th Gen Tensor Cores, delivering up to 421 AI TOPS. Manli offers RTX 5050 in Polar Fox & Nebula series. We sincerely invite you to experience NVIDIA Blackwell with our innovative IP characters - Polar. Plus, there's also classical and durable version - Nebula. Expect you to have a whole new adventure with Manli's graphic cards. Stay with Manli, you always have the choice!!!

MSI proudly unveils its latest lineup of graphics cards powered by the newly launched NVIDIA GeForce RTX 5050 GPUs, featuring the GAMING, VENTUS, and SHADOW series. Engineered with cutting-edge graphics technology and robust thermal solutions, these new graphics cards are built to deliver consistent, reliable performance across a wide range of gaming and creative applications. Advanced cooling technologies ensure efficient, quiet operation—even under sustained workloads—making them a trusted choice for users who prioritize stability, efficiency, and long-term value.

Powered by NVIDIA Blackwell, GeForce RTX 50 Series GPUs bring game-changing capabilities to gamers and creators. Equipped with a massive level of AI horsepower, the RTX 50 Series enables new experiences and next-level graphics fidelity. Multiply performance with NVIDIA DLSS 4, generate images at unprecedented speed, and unleash creativity with NVIDIA Studio. Additionally, users can harness NVIDIA NIM microservices, a suite of advanced AI models designed to help developers build AI assistants, agents, and workflows—optimized for NIM-ready systems.

NVIDIA's DLSS Transformer officially graduates from beta today and is rolling out in its full, stable form across supported games. Following its debut as part of the DLSS 4 update, which features multi-frame generation and significant improvements in image quality, the technology has proven its worth. By replacing traditional convolutional neural networks with Transformer models, NVIDIA has doubled the model's parameters, boosted compute throughput fourfold, and delivered a 30-50% uplift in ray-traced effects quality, according to internal benchmarks. Even more impressive, each AI-enhanced frame now processes in just 1 ms on an RTX 5090, compared with the 3.25 ms required by DLSS 3.

Under the hood, DLSS 4 can tap into Blackwell-exclusive hardware, from FP8 tensor cores to fused CUDA kernels, and leans on vertical layer fusion and memory optimizations to keep overhead in check even with models twice the size of their CNN predecessors. To fine-tune performance and eliminate glitches such as ghosting, flicker, or blurring, NVIDIA has quietly run this network through a dedicated supercomputer for the past six years, continuously iterating on its findings. The result is a real-time AI upscaling solution that pairs a higher-performance AI architecture with rigorously validated quality.

INNO3D, a leading manufacturer of high-end multimedia components and innovations is proud to unveil its latest additions to the GeForce RTX 50 Series lineup: the INNO3D GeForce RTX 5050 GPUs. Powered by NVIDIA's highly efficient Blackwell architecture, these state-of-the-art graphics cards deliver exceptional performance and a suite of industry-leading features.

INNO3D has enhanced both the design and performance of its standout cooler series, featuring the sleek TWIN X2 and TWIN X2 OC models. Meanwhile, the RTX 5050 receives a small form factor upgrade with the single-fan COMPACT edition. These strategic enhancements underscore INNO3D's dedication to delivering extraordinary visual experiences, continuing to push the limits of graphics innovation for gamers and creators alike.

NVIDIA is rumored to have advanced the launch date of its GeForce RTX 5050 mid-range GPU to July 1, 2025. It was earlier reported that the RTX 5050 would launch by the end of July. MEGAsizeGPU, a reliable source with NVIDIA leaks, says that NVIDIA informed AICs of the advancement, and that no cards will be ready to ship on that date. We are not quite sure what NVIDIA's strategy with the RTX 5050 is. The RTX 5060 launch was eclipsed by Computex, back in May. There are no major press events planned in July. One idea would be to announce the card and have AICs make it available to purchase as soon as they can, so the company could cover some ground over the mid-Summer.

The GeForce RTX 5050 is rumored to be based on the "GB207" silicon, the company's smallest chip implementing the "Blackwell" graphics architecture. This chip is rumored to physically has 20 SM, which the RTX 5050 maxes out, for 2,560 CUDA cores, 80 Tensor cores, 20 RT cores, 80 TMUs, and an unknown number of ROPs. It's expected to offer 8 GB of GDDR6 memory (either 18 Gbps or 20 Gbps), across a 128-bit wide memory bus. NVIDIA is looking to target a sub-$250 price-point with the RTX 5050, to compete with the Intel Arc B580.Update 14:44 UTC: Turns out the RTX 5050 was announced today, with availability "in the second half of July".

Instructions per clock is a metric used to define and compare CPU architecture performance usually. However, enthusiast colleagues at ComputerBase had an idea to test the IPC improvement in GPUs, comparing it across current and past generations. NVIDIA's Blackwell-based GeForce RTX 50 series faces off against the Ada Lovelace-based RTX 40 generation, while AMD's RDNA 4-powered Radeon RX 9000 lineup challenges the RDNA 3-based RX 7000 series. For NVIDIA, the test used RTX 5070 Ti and 4070 Ti SUPER, aligning ALU counts and clock speeds and treating memory bandwidth differences as negligible. For AMD, the test matched the RX 9060 XT to the RX 7600 XT, both featuring identical ALUs and GDDR6 memory. By closely matching shader counts and normalizing for clock variations, ComputerBase isolates IPC improvements from other hardware enhancements. In rasterized rendering tests across 19 popular titles, NVIDIA's Blackwell architecture delivered an average IPC advantage of just 1% over the older Ada Lovelace.

This difference could easily be attributed to normal benchmark variance. Ray tracing and path tracing benchmarks showed no significant IPC uplift, leaving the latest generation essentially on par with its predecessor when normalized for clock and unit count. AMD's RDNA 4, by contrast, exhibited a substantial IPC leap. Rasterized performance improved by around 20% compared to RDNA 3, while ray-traced workloads enjoyed a roughly 31% gain. Path tracing results were even more extreme, with RDNA 4 delivering nearly twice the FPS, a 100% increase over its predecessor. These findings suggest that NVIDIA's performance improvements primarily stem from higher clock speeds, increased execution unit counts, and enhanced features. AMD's RDNA 4 represents a significant architectural advance, marking its most notable IPC gain since the original RDNA launch.

NVIDIA's upcoming GB300 NVL72 "Blackwell Ultra" rack-scale systems are reportedly going to get a humanoid robot assembly, according to sources close to Reuters. As readers are aware, most of the traditional manufacturing processes in silicon manufacturing, PCB manufacturing, and server manufacturing are automated, requiring little to no human intervention. However, rack-scale systems required humans for final assembly up until now. It appears that Foxconn and NVIDIA have made plans to open up the first AI-powered humanoid robot assembly plant in Houston, Texas. The central plan is that, in the coming months as the plant is completed, humanoid robots will take over the final assembly process entirely removing humans from the manufacturing loop.

And this is not a bad thing. Since server assembly typically requires lifting heavy server racks throughout the day, the humanoid robot system will aid humans by doing the hard work, thereby saving workers from excessive labor. Initially, humans will oversee these robots in their operations, with fully autonomous factories expected later on. The human element here will primarily involve inspecting the work. NVIDIA has been laying the groundwork for humanoid robots for some time, as the company has developed NVIDIA Isaac, a comprehensive CUDA-accelerated platform designed for humanoid robots. As models from Agility Robotics, Boston Dynamics, Fourier, Foxlink, Galbot, Mentee Robotics, NEURA Robotics, General Robotics, Skild AI, and XPENG require models that are aware of their surroundings, NVIDIA created Isaac GR00T N1, the world's first open humanoid robot foundation model, available for anyone to use and finetune.

Sycom is a Japan-based OEM that specializes in build-to-order (BTO) gaming PCs and components such as graphics cards. Last year, the company released the Silent Master GeForce RTX 40-series "Ada" graphics cards that come with a custom cooling solution that uses an aluminium fin-stack heatsink, and a pair of large 120 mm Noctua case fans for ventilation; and this week, they launched the Silent Master RTX 50-series "Blackwell" graphics card series.

The series includes three models, the RTX 5070 Ti, the RTX 5070, and the RTX 5060 Ti 16 GB. At this point, it's not known if Sycom designed the heatsink or PCB—there's a possibility that these are sourced from a graphics card OEM without the cooler shroud and fans, and topped up with a custom cooler shroud, backplate, and a pair of Noctua-sourced high-end fans. The RTX 5070 Ti and RTX 5070 cards come with larger coolers that use a pair of 120 mm NF-A12x25 (25 mm-thick), while the RTX 5060 Ti 16 GB card comes with a pair of 92 mm NF-A9x14. All three cards lack any form of factory overclock, their main claim to fame is their low noise. Unlike the ASUS Noctua Edition graphics cards, the three cards have relatively well managed card thickness of 3.5-slot or under. All three cards are being sold under the BTO (build to order) model, exclusively in Japan.

When we heard the news that NVIDIA will be halting shipments to Chinese customers for its GeForce RTX 5090D GPU in Q2, there were some speculations that the company is preparing to make a comeback. Leaving out the Chinese market is a tough decision from a business standpoint, but NVIDIA complies with export regulations imposed by the US administration under Donald Trump. However, the company now plans to keep its presence in China with the release of the export-abiding GeForce "RTX 5090DD" GPU. Carrying a GB202-240 die, it will be a cut-down version with reportedly only 14,080 CUDA cores present, paired with 24 GB of GDDR7 memory.

Chinese gamers will notice a significant difference, but so will anyone trying to acquire these GPUs for non-gaming purposes. As the US administration has banned shipments of high-end chips to China to stop them from training military-grade AI, gaming GPUs have been a target of AI firms. Now, a modified GeForce RTX 5090 DD, which not only cuts with raw compute but also firmware, should be sufficient for gaming. Carrying a PG145 SKU 40 board, the new RTX 5090DD GPU will also feature a modified 384-bit bus for its 24 GB of memory, down from the original RTX 5090D's 512-bit bus.

Update 09:45 UTC: According to one of the most accurate NVIDIA leakers, kopite7kimi, the core count for this GPU is supposed to be 21,760 CUDA cores, with 575 W TDP. The leaker did note that there is some surprise, so NVIDIA has definitely changed something big under the hood to comply with export regulations.

ASUS Republic of Gamers (ROG) today announced the ROG Astral GeForce RTX 5080 Dhahab CORE OC Edition graphics card, built to take style and performance to new frontiers. With the latest NVIDIA GPU architecture, cutting-edge thermal design and a premium aesthetic, the ROG Astral GeForce RTX 5080 Dhahab CORE OC is built for gamers who want a PC that plays well and looks incredible doing it.

The gold standard of GeForce RTX 5080 performance
The ROG Astral GeForce RTX 5080 Dhahab CORE OC Edition graphics card stands ready to let users reap the benefits of the new Blackwell architecture at the heart of the NVIDIA GeForce RTX 50 Series. This delivers fourth-generation ray tracing cores for incredible performance. Users also get NVIDIA DLSS 4 Super Resolution, Multi-Frame Generation and Ray Reconstruction, which help games run smoothly with graphics cranked up.

The June 2025 Update of Adobe Premiere Pro, the industry standard video editing application, added support for GPU-accelerated 4:2:2 color format video encoding and decoding on NVIDIA "Blackwell" GPUs. It implements Video Codec SDK 13.0, which NVIDIA released in February. This adds 4:2:2 color format support for H.264 and HEVC. The 4:2:2 color format provides twice the amount of color detail as 4:2:0, but with a nominal increase in stream bitrate or file-size. Adobe says that the added color detail of 4:2:2 is ideal for "precise color grading, cleaner chroma keying, and crisper text rendering."

Generative AI has reshaped how people create, imagine and interact with digital content. As AI models continue to grow in capability and complexity, they require more VRAM, or video random access memory. The base Stable Diffusion 3.5 Large model, for example, uses over 18 GB of VRAM - limiting the number of systems that can run it well. By applying quantization to the model, noncritical layers can be removed or run with lower precision. NVIDIA GeForce RTX 40 Series and the Ada Lovelace generation of NVIDIA RTX PRO GPUs support FP8 quantization to help run these quantized models, and the latest-generation NVIDIA Blackwell GPUs also add support for FP4.

NVIDIA collaborated with Stability AI to quantize its latest model, Stable Diffusion (SD) 3.5 Large, to FP8 - reducing VRAM consumption by 40%. Further optimizations to SD3.5 Large and Medium with the NVIDIA TensorRT software development kit (SDK) double performance. In addition, TensorRT has been reimagined for RTX AI PCs, combining its industry-leading performance with just-in-time (JIT), on-device engine building and an 8x smaller package size for seamless AI deployment to more than 100 million RTX AI PCs. TensorRT for RTX is now available as a standalone SDK for developers.

Today Kubuntu Focus announces the availability of the Sixth-Generation M2. This model includes substantial upgrades in CPU, GPU, display, and NVMe. The Focus Team claims the M2 GEN 6 laptop is a superb choice for anyone looking for the best out-of-the-box Linux experience with optimized and validated hardware. Focus systems are used by developers, DevOps, ML researchers, creators, and others who value their time and the unmatched Linux-first support.

Improvements from the prior generation include:

• Faster, cooler, and quieter CPU performance with the Intel Core Ultra 9 275HX. With 24 cores, it gains 5-15% performance yet is up to 40% more efficient.
• Between 30-44% faster graphics benchmark performance and 50% more VRAM with the NVIDIA GeForce Blackwell 5070 Ti GPU. This features 12 GB GDDR7 VRAM and fourth-gen ray tracing and fifth-gen tensor cores.
• A larger and taller 525-nit 16.0" display with WQXGA LCD, 2560×1600 resolution, G-Sync, and a 16:10 aspect ratio.

NVIDIA today announced it is working with European nations, and technology and industry leaders, to build NVIDIA Blackwell AI infrastructure that will strengthen digital sovereignty, support economic growth and position the continent as a leader in the AI industrial revolution. France, Italy, Spain and the U.K. are among the nations building domestic AI infrastructure with an ecosystem of technology and cloud providers, including Domyn, Mistral AI, Nebius and Nscale, and telecommunications providers, including Orange, Swisscom, Telefónica and Telenor.

These deployments will deliver more than 3,000 exaflops of NVIDIA Blackwell compute resources for sovereign AI, enabling European enterprises, startups and public sector organizations to securely develop, train and deploy agentic and physical AI applications. NVIDIA is establishing and expanding AI technology centers in Germany, Sweden, Italy, Spain, the U.K. and Finland. These centers build on NVIDIA's history of collaborating with academic institutions and industry through the NVIDIA AI Technology Center program and NVIDIA Deep Learning Institute to develop the AI workforce and scientific discovery throughout the regions.

PEGATRON, a globally recognized Design, Manufacturing, and Service (DMS) provider, is showcasing its latest AI server solutions at GTC Paris 2025. Built on NVIDIA Blackwell architecture, PEGATRON's cutting-edge systems are tailored for AI training, reasoning, and enterprise-scale deployment.

NVIDIA GB300 NVL72
At the forefront is the RA4802-72N2, built on the NVIDIA GB300 NVL72 rack system, featuring 72 NVIDIA Blackwell Ultra GPUs and 36 NVIDIA Grace CPUs. Designed for AI factories, it boosts output by up to 50X. PEGATRON's in-house developed Coolant Distribution Unit (CDU) delivers 310 kW of cooling capacity with redundant hot-swappable pumps, ensuring performance and reliability for mission-critical workloads.

Supermicro, Inc., a Total IT Solution Provider for AI, Cloud, Storage, and 5G/Edge, is announcing an expansion of the industry's broadest portfolio of solutions designed for NVIDIA Blackwell Architecture to the European market. The introduction of more than 30 solutions reinforces Supermicro's industry leadership by providing the most comprehensive and efficient solution stack for NVIDIA HGX B200, GB200 NVL72, and RTX PRO 6000 Blackwell Server Edition deployments, enabling rapid time-to-online for European enterprise AI factories across any environment. Through close collaboration with NVIDIA, Supermicro's solution stack enables the deployment of NVIDIA Enterprise AI Factory validated design and supports the upcoming introduction of NVIDIA Blackwell Ultra solutions later this year, including NVIDIA GB300 NVL72 and HGX B300.

"With our first-to-market advantage and broad portfolio of NVIDIA Blackwell solutions, Supermicro is uniquely positioned to meet the accelerating demand for enterprise AI infrastructure across Europe," said Charles Liang, president and CEO of Supermicro. "Our collaboration with NVIDIA, combined with our global manufacturing capabilities and advanced liquid cooling technologies, enables European organizations to deploy AI factories with significantly improved efficiency and reduced implementation timelines. We're committed to providing the complete solution stack enterprises need to successfully scale their AI initiatives."