QNAP has introduced a NAS system that pairs a 2019 EPYC processor with NVIDIA’s RTX PRO 6000 Blackwell GPU, blending legacy compute power with the latest in AI acceleration. The combination suggests a deliberate engineering tradeoff: leveraging mature, cost-effective server-grade CPUs while tapping into Blackwell’s massive 96 GB HBM2e memory and third-generation Tensor Cores to handle demanding data workloads.
- Processor: AMD EPYC 7002 series (Zen 2, up to 32 cores)
- GPU: NVIDIA RTX PRO 6000 Blackwell (96 GB HBM2e, 18,432 CUDA cores)
- Memory: Up to 512 GB DDR4
- Storage: 24 drive bays (SAS/SATA/NVMe), up to 2 PB raw capacity
- Cooling: Liquid cooling with dual fans for GPU, air cooling for CPU
The system is designed for edge AI applications where operational cost matters as much as performance. The EPYC 7002 series, though six years old, remains a workhorse in data centers due to its efficiency and scalability. Meanwhile, the Blackwell GPU brings NVIDIA’s latest advancements—including 96 GB of ultra-fast HBM2e memory and third-gen Tensor Cores—to tasks like large-scale inference or training without the power draw of more recent CPU architectures.
One notable tradeoff is thermal management: the RTX PRO 6000 Blackwell requires liquid cooling, while the EPYC remains air-cooled. This dual-approach design reflects a balance between legacy hardware longevity and cutting-edge acceleration. Whether this setup will outperform newer x86-based NAS systems in real-world AI workloads remains to be seen, but it underscores QNAP’s focus on cost-effective edge solutions.
The system is expected to launch with pricing aligned with enterprise-grade storage, though exact details have not been confirmed. It marks a rare instance where a manufacturer pairs such divergent generations of hardware, hinting at both innovation and pragmatism in edge AI infrastructure.
