Nvidia has announced the release of the RTX 4000, what they proudly tout as “the most powerful single slot GPU on the planet” It has 20GB of GDDR6, 6,144 CUDA cores, and a 130W TDP. This remarkable innovation, however, comes with a catch – it’s primarily targeted at the workstation market and carries a premium price tag of $1,250.
Equipped with an Ada Lovelace AD104 chip, the RTX 4000 shares the graphical architecture of the entire RTX 40-series GeForce lineup. Positioned between the RTX 4070 and RTX 4070 Ti in the GeForce hierarchy, it boasts additional VRAM but doesn’t align precisely due to this memory variation.
Notably, this GPU offers expanded VRAM but exhibits reduced bandwidth, utilizing a 160-bit memory bus for an aggregate of 360GB/s. Although possessing a leaner core count compared to the RTX 4000 mobile chip in workstation laptops, Nvidia has previously optimized desktop card versions similarly. Its characteristics align with the RTX 4000 SFF, albeit with significantly lower power requirements.
In contrast to the RTX 4070’s 200W, this card operates at a modest 130W. This discrepancy influences its performance figures, with a rating of 26.7 TFLOPs, slightly below the RTX 4070’s 29.15 TFLOPs. Remarkably, the RTX 4000 boasts more cores and a greater memory capacity despite this reduction.
Trade-offs are evident due to the compact form factor, emphasizing Nvidia’s intent for multiple-card setups to compensate for individual performance. For instance, envision four cards at $1,250 each—totaling $5,000.
In principle, these RTX cards could serve for gaming, considering familiar drivers. Yet, their steep prices guarantee their exclusivity to enterprise applications.
Nvidia introduces three new enterprise-grade GPUs: the most budget-friendly being the RTX 4000, followed by the $2,250 RTX 4500 in October, and the present RTX 5000, distinct from potential next-gen GeForce 50-series cards, available for $4,000.