AMD's Radeon PRO W9000: A Strategic Approach to Workstation GPUs for AI and Professional Tasks

Forget 96GB GPUs - AMD's Radeon PRO W9000 leak hints at smarter design, better price, and real-world power where it matters

With a 356mm² die size, the Radeon W9000 could still punch hard in real-world tasks AMD is set to expand its workstation GPU lineup with the imminent release of the new Radeon PRO W9000 series, built on the RDNA 4 architecture. This chip is aimed at professionals working in demanding fields such as video editing, 3D rendering, and AI development. A leak from Hoang Anh Phu claims the lineup will include a 32GB model based on the new Navi 48 XTW architecture. AMD is supposed to be chasing Nvidia, but the RTX PRO 6000 Blackwell boasts an enormous 96GB frame buffer and 24,064 CUDA cores. This is far beyond the 32GB configuration that AMD appears to be offering, suggesting the Radeon PRO W9000 may be targeting professionals who don't require large datasets and instead value performance per dollar, along with efficiency and affordability. As with all unconfirmed rumors, the details should be taken with a grain of salt - but if accurate, the leak suggests a shift in how AMD prioritizes performance relative to cost. The previous-generation Radeon PRO W7900 featured 48GB of memory, so the 32GB - likely GDDR6 - is a reduction. While the capacity is lower, the Navi 48 XTW die, measuring 356mm², is still expected to deliver performance improvements across a wide range of tasks. It is also likely to bring benefits to CAD workloads, CGI rendering, and real-time simulations, areas where the best laptops for video editing or Photoshop also show gains. The chip is expected to come in XL, XT, and XTX variants, each tuned for different professional needs. The XTW model is believed to feature hardware enhancements optimized for pro workflows, making it a strong option for users needing a GPU for production-grade tasks. As Phu, a known hardware leaker, noted, "It's not as beefy as the last generation, but pricing remains key." One remaining question is software support. RDNA 4 does not yet have full integration with AMD's ROCm platform, which is critical for AI and machine learning developers. Although final specs and an official name are still under wraps, all signs point to a strategic reveal aligned with Computex 2025 and AMD's "Advancing AI" event in June 2025.

AMD Radeon PRO W9000 Series: Navi 48 XTW Chip and Memory Configuration

AMD is preparing to launch the Radeon PRO W9000 series workstation GPUs, which are expected to be showcased at the 2025 Taipei International Computer Show. These new cards will feature the Navi 48 XTW chip, part of AMD's RDNA 4 architecture, and will come with 32GB of GDDR6 memory. The decision to stick with a 32GB configuration instead of opting for larger memory sizes, like in previous generations, suggests that AMD is taking a more measured approach for its high-end workstation GPUs. The Navi 48 chip used in the W9000 series has a process area of 356mm², and while it's not clear if the XTW version of the chip will differ in size, it is likely that the professional-grade models will incorporate enhancements designed to better handle professional workloads. The W9000 series is expected to feature three different versions: XL, XT, and XTX. The XTX variant will contain 64 compute units (CUs), similar to the Radeon RX 9070 XT, a consumer-grade card. However, the XTW version of the W9000 series might include special optimizations, particularly for professional users, making it more suited for complex tasks such as 3D rendering, AI workloads, and scientific simulations. In terms of memory, the 32GB configuration of the Radeon PRO W9000 is smaller than the 48GB offered by the previous generation W7900 and much smaller than Nvidia's RTX PRO 6000, which boasts a massive 96GB of memory. Despite this, the Radeon PRO W9000's 32GB is still expected to offer solid performance for many professional applications, and the move to a more conservative memory configuration may reflect AMD's strategy to offer a better price-to-performance ratio while still addressing the needs of high-end users. AMD has also been making strides with its ROCm platform, which aims to optimize RDNA 4 for AI and machine learning applications. This may be highlighted further during the "Advancing AI" event in June. As of now, the specific release date and pricing for the Radeon PRO W9000 series remain unconfirmed, but they are expected to launch in the coming months. Source: ithome

AMD rumored with 'Navi 48 XTW' based Radeon PRO workstation GPU with 32GB of VRAM

TL;DR: AMD is developing the RDNA 4-based Radeon PRO W9000 workstation GPU featuring 32GB of GDDR6 memory, targeting AI and professional workloads. This Navi 48 XTW GPU offers double the VRAM of current consumer models, enhancing performance for AI developers. Official details are expected by Computex 2025. AMD is working on a new RDNA 4-based workstation GPU that will feature 32GB of GDDR6 memory, which should be released as the Radeon PRO W9000 series. In a new post on X by leaker @AnhPhuH, we're hearing that the new "Navi 48 XTW" GPU will feature 32GB of VRAM, and will be part of a new PRO series of workstation GPUs. AMD launched its RDNA 4-powered Radeon RX 9070 series GPUs recently, with its flagship Radeon RX 9070 XT featuring 16GB of GDDR6 memory. The purported Navi 48 XTW GPU with 32GB of GDDR6 will be its highest-spec configuration in terms of VRAM, with consumer GPUs left with only 16GB. For AI developers, the additional 16GB (32GB total) will be a big win, but it's nothing we haven't seen before. Last year we saw GIGABYTE release its Radeon W7900 Dual Slot AI TOP 48G graphics card, with 48GB of GDDR6 memory with ECC (Error Checking and Correction). Not only that, but it was rumored that AMD could release a Radeon RX 9070 XTX graphics card with 32GB of GDDR6 for AI workloads, so a new Navi 48 XTW with 32GB of GDDR6 isn't too surprising. We have Computex 2025 right around the corner with less than a month to go, so we should expect to hear some more news of the Navi 48 XTW-based workstation GPU by then, or at AMD's upcoming Advancing AI event on June 12.