Nvidia RTX 5000 Ada: Professional graphics cards with AD102 chip and 32GB GDDR6 now available

Nvidia’s RTX 5000 Ada, a professional graphics card for visualization applications, is now available. The card is equipped with Nvidia’s flagship GPU AD102, but in a much reduced configuration to lower the power consumption to 250 W. However, retailers are selling the card at inflated prices, if you can get it at all. These inflated prices are an annoyance for professional users who depend on the RTX 5000 Ada’s performance. They could cause businesses to opt for other options that are cheaper but don’t offer the same performance.

The Nvidia RTX 5000 Ada Generation graphics card is based on the AD102 GPU, which has 12,800 CUDA cores spread across 100 SMs. It offers an impressive processing power of 65.3 FP32 TFLOPS and is connected to 32 GB of GDDR6 memory via a 256-bit interface. In comparison, Nvidia’s top-of-the-line RTX 6000 Ada, also powered by the AD102 GPU, has 18,176 CUDA cores and delivers an impressive 91.6 FP32 TFLOPS. Furthermore, the RTX 6000 is equipped with 48 GB of GDDR6 memory, which is connected to the processor via a 384-bit interface.

The decision to reduce the GPU by 30% might seem unusual at first glance. However, it may have been the only way to create a product that is positioned below the RTX 6000 Ada while still maintaining a noticeable, if not dramatic, performance gap with the premium ProViz graphics card. Nvidia’s full AD103 has a maximum of 10,240 CUDA cores spread across 80 SMs, which would result in potentially lower performance compared to a scaled-down AD102. Interestingly, Nvidia has maintained a fairly high power rating of 250 W despite the significant reduction in GPU configuration. In terms of the cooling system, the board features a dual-slot cooler with a fan. In addition, the board uses a 12VHPWR PCIe auxiliary power connector.

It remains to be seen how Nvidia will use the AD103 chip in its professional offerings. The RTX 6000 Ada uses nearly all of the AD102 with 142 (out of 144) SMs and the full 384-bit memory interface. The RTX 5000 Ada, on the other hand, uses the AD102 with 100 SMs and only a 256-bit interface. Then a big jump occurs to the RTX 4500 Ada with a full AD104 implementation of 60 SMs and a 192-bit interface. The RTX 4000 Ada and RTX 4000 SFF Ada close the spectrum with a partial AD104 chip with 48 SMs and a 160-bit interface.

So there is a possibility for an RTX Ada card between the 4500 Ada and the 5000 Ada that uses the AD103. So far, however, Nvidia has not made such an announcement, and there doesn’t seem to be room for it within the standard naming structure (i.e. RTX 5500 Ada). It remains to be seen whether Nvidia will use the AD103 for professional cards at all.

Source: NVIDIA