Dell Pro Max AI with Blackwell GB300, 20 pflops and 784GB unified memory. Can it run Crysis? | Computer hardware notes | List of hardware components with examples | Cpu vs gpu vs ram | Turtles AI

Dell has unveiled the Pro Max with GB300, a desktop that delivers 20 petaFLOPS of processing power, powered by Nvidia’s GB300 superchip. This system is designed to handle large-scale AI workloads, giving developers the ability to prototype and scale complex models right on the desktop. The machine includes a Blackwell Ultra GPU, 784GB of unified memory, and supports network speeds of up to 800Gbps.

Key Points:

• Processing Power: 20 petaFLOPS thanks to the Nvidia GB300 superchip.
• Enhanced Memory: 784GB of unified memory, with 288GB HBM3e and 496GB LPDDR5X.
• High-Speed ​​Connectivity: Supports up to 800Gbps with Nvidia ConnectX-8 SuperNIC.
• OS Versatility: Available with Ubuntu or Nvidia’s base operating system.

At the recent GPU Technology Conference (GTC) in Silicon Valley, Dell unveiled its new desktop, the Pro Max with GB300, capable of reaching 20 petaFLOPS of processing power. This system is based on Nvidia’s Grace Blackwell Ultra GB300 superchip, which integrates a Blackwell Ultra GPU with next-generation Tensor Cores and FP4 precision. The configuration includes 784 GB of unified system memory, divided into 288 GB of HBM3e memory in the GPU and 496 GB of LPDDR5X for the CPU, optimized to handle large-scale training and inference workloads. The CPU and GPU are connected via NVLink-C2C, ensuring high-speed and consistent memory access.

The workstation is equipped with an Nvidia ConnectX-8 SuperNIC, optimized for AI workloads, and supports network speeds of up to 800 Gbps, enabling ultrafast interconnects for distributed AI training and data transfer. Kevin Terwilliger, vice president and general manager of commercial, consumer and gaming PCs at Dell, said the machine will make it easier for developers to prototype, test and scale their models to production environments. Terwilliger also expects developers to use the Pro Max with GB300 to run large-scale training and inference workloads that are typically deployed on servers.

Interestingly, the Pro Max with GB300 is based on the DGX Station that Nvidia unveiled at GTC yesterday, bringing the power of an AI supercomputer to the desktop. The machine will be available with Ubuntu or Nvidia’s base operating system, providing flexibility for users.

In addition to the Pro Max with GB300, Dell is introducing a mini-workstation powered by the Grace-Blackwell GB10 desktop processor, capable of reaching a petaFLOPS of performance. On the server side, Dell has updated its PowerEdge lineup to support the B300 NVL16, GB300 NVL72, and RTX PRO 6000 Blackwell Server Edition, all based on Nvidia’s new Blackwell Ultra silicon. Specifically, the PowerEdge XE8712 can accommodate up to 144 GB200 NVL4 accelerators, with up to 264 kW of power consumption per rack, cooled via Direct Liquid Cooling (DLC).

HPE and Lenovo are also integrating Nvidia’s latest hardware into their solutions. HPE’s ProLiant servers will support the HGX B300, while the DL384b Gen 12 machines will run the GB200 Grace Blackwell NVL4 Superchip. Lenovo, for its part, has announced solutions that can support Nvidia’s new chips, focusing on workloads such as AI reasoning, agent AI, and real-time video generation.

The new hardware solutions will be available in Q2 2025 or later, while related services are already active.

The introduction of the Dell Pro Max with GB300 represents a significant step forward in bringing high-performance computing power directly to the desktop, giving developers advanced tools to handle complex AI workloads.