Welcome to PC Hardware Finder

Create an account to save your builds and preferences, or continue browsing as a guest

Hardware5 min read0 views

NVIDIA Blackwell GB200 Launches — Breaking Power and Performance Records

NVIDIA's Blackwell architecture is now in full production with the GB200 NVL72 superchip powering major AI clusters at Google, AWS, and Microsoft. It is a massive leap over Hopper, though its power draw of up to 1,200W per module is raising data centre concerns.

May 10, 2026
By PC Hardware Finder
NVIDIABlackwellGB200AIData CentreGPU

NVIDIA's Blackwell architecture has entered full production ramp, with the GB200 NVL72 superchip — a rack-scale system combining 72 Blackwell GPUs and 36 Grace CPUs — now powering AI training clusters at Google, AWS, and Microsoft Azure.

The Performance Leap

The improvement over the previous Hopper architecture is substantial. NVIDIA claims a 4–5x improvement in training throughput for large language models at comparable energy spend, driven by architectural changes to the Tensor Core design, an NVLink interconnect running at 1.8 TB/s, and a new memory subsystem using HBM3e at 192 GB per GPU.

The Power Problem

The consumption figures are significant. The full GB200 NVL72 rack-scale system draws approximately 120 kW. Individual GB200 modules approach 1,200W under sustained AI workloads. For data centre operators, this means building or retrofitting facilities with new power distribution and liquid cooling infrastructure — a capital project adding lead time and cost to every AI cluster deployment.

The power requirements are a genuine constraint on how quickly the industry can deploy Blackwell at scale. New data centre construction with the required electrical and cooling infrastructure takes 18–36 months from planning to operation.

Effect on Consumer Hardware

From a PC hardware perspective, the Blackwell ramp has direct market effects. HBM3e production capacity is near its ceiling, with AI accelerators claiming the majority of output from Samsung, SK Hynix, and Micron. This constrains supply for gaming GPUs that use GDDR7 and GDDR6X, which compete for the same advanced packaging facilities.

NVIDIA's consumer Blackwell products — the RTX 50-series — share the architecture in heavily scaled-down form. The gaming variants use GDDR7 rather than HBM and operate at power levels consistent with enthusiast desktop use.

About the Author

CD
Callum Duce
Founder & Hardware Editor, PC Hardware Finder

Callum Duce is an Australian PC builder and the founder of PC Hardware Finder. With years of hands-on experience building gaming and workstation PCs, he created this site to give Australians clear, unsponsored hardware advice based on real-world experience and current AUD pricing. He covers compatibility guides, component reviews, and buying recommendations to help readers build confidently without overspending.

Related Articles

Hardware

Microsoft Now Recommends 32 GB RAM for No-Worries Gaming

Microsoft now recommends 32 GB of RAM for a no-worries gaming PC, a significant step up from the 16 GB baseline. We explain why the threshold shifted, what 32 GB actually buys you, and whether upgrading is worth it right now.

5 min read
Hardware

AMD Readies RDNA 4 Reveal — RX 8000 Series Targets Midrange Value Over Raw Power

AMD is expected to showcase its Radeon RX 8000 series soon, focusing on energy efficiency and price-performance rather than competing head-on with NVIDIA at the high end. Enthusiasts expect limited high-end models and a midrange-first strategy.

5 min read
Hardware

Intel Battlemage Benchmarks Leak — Performance Between RTX 4070 and 4070 Ti

Intel's upcoming Battlemage GPU architecture has leaked benchmark results showing performance in RTX 4070–4070 Ti territory. With improved drivers and XeSS updates, it could be Intel's first real shot at competing in the midrange gaming GPU market.

5 min read