ASUS has quietly entered the pre-order phase for its MATRIX GeForce RTX 5090 D V2, a specialized variant of NVIDIA’s flagship Blackwell-based GPU designed exclusively for China’s domestic market. Unlike its predecessor—the 32GB RTX 5090 D V1 (a limited-edition, $3,999 monstrosity with 3oz copper PCB layers and a Bosch IMU)—this iteration scales back VRAM to 24GB while retaining the same core architecture. The price, however, remains steep: 29,999 yuan (~$4,341 USD), placing it among the most expensive consumer GPUs ever released.

What sets the V2 apart isn’t just its cost but its raw performance under load. ASUS demonstrated the card hitting 28,638 points in 3DMark Time Spy Extreme, a benchmark typically reserved for professional workstations. This figure surpasses even NVIDIA’s RTX PRO 6000—a card built on a slightly larger GPU die (GB202) with 24,576 cores—by leveraging an 800W power envelope. The implication is clear: ASUS has unlocked near-maximal performance from Blackwell’s 21,760-core configuration, a feat that suggests NVIDIA’s consumer-grade silicon isn’t as far behind its professional counterpart as previously assumed.

The V1’s Legacy and China’s GPU Dilemma

The RTX 5090 D V1, launched last November as a 30th-anniversary limited edition, was a statement piece—both in design and ambition. Its $3,999 price tag, dual 16-pin 12VHPWR connectors, and inclusion of a Bosch BMI323 IMU (a sensor typically found in smartphones) made it a curiosity. Yet, its production was cut short, likely due to geopolitical tensions restricting NVIDIA’s supply chain into China. The RTX 5090 D V2 appears to be ASUS’s workaround, offering a 24GB alternative that complies with local regulations while still catering to AI workloads and high-end gaming.

Curiously, the V2’s pricing aligns with leaks from late 2025 predicting $5,000+ RTX 5090 models driven by AI demand. While the V2 stops short of that threshold, its $4,341 USD asking price reflects the same market forces: enterprises and enthusiasts willing to pay a premium for Blackwell’s compute prowess. For context, an MSI RTX 5090 D Gaming Trio (32GB)—a non-MATRIX variant—was briefly listed at 25,999 yuan (~$3,762 USD) earlier this year, underscoring how ASUS’s custom design justifies its markup.

Who Needs This?

The RTX 5090 D V2 isn’t for casual gamers. Its 800W power draw requires a robust PSU and cooling solution, and its 24GB VRAM—while generous for gaming—is a downgrade from the V1’s 32GB, which was better suited for AI inference tasks. That said, the V2’s performance suggests it could still handle 8K rendering, multi-monitor setups, and even light AI workloads with ease. The target audience is likely a mix of

  • Chinese AI startups needing Blackwell’s Tensor cores without the hassle of professional GPUs.
  • Hardcore enthusiasts willing to push the limits of gaming and content creation.
  • Corporate buyers in regulated markets where NVIDIA’s professional cards are restricted.

The card’s launch also raises questions about NVIDIA’s broader strategy. With reports of up to 40% production cuts for the RTX 50 series in early 2026 and rumors of a SUPER lineup at CES 2026, the company may be tightening supply to prioritize high-margin professional and AI-focused hardware. ASUS’s V2, in this light, could be a stopgap—offering Blackwell’s performance to a market starved for options.

What’s Next?

Availability remains unconfirmed, but pre-orders are open via JD.com. Whether this card becomes a fleeting novelty or a blueprint for future China-specific GPUs depends on how NVIDIA and partners navigate export controls. One thing is certain: the RTX 5090 D V2 isn’t just a graphics card—it’s a test case for how high-end PC hardware evolves in a fragmented global market.