NVIDIA has announced another member of its Blackwell PRO family of workstation GPUs: the RTX PRO 5500 Blackwell with 84GB of GDDR7 memory. The new card becomes the second-fastest model in the lineup, slotting between the RTX PRO 5000 and PRO 6000 to fill the sizable specifications gap between the two. Details on pricing and availability have not yet been disclosed.
The GPU initially surfaced on NVIDIA's regional websites, but its product pages have since gone global, alongside a now-removed but archived listing from PNY. NVIDIA now lists the full specifications on its U.S. website, and the GPU is listed as "Coming Soon".
Popular Now: Linus Tech Tips leaks Steam Frame first impressions, and Valve may be working on native Half-Life: Alyx support
The RTX PRO 5500 is based on a cut-down GB202 GPU with 21,760 CUDA cores across 170 SMs, leaving roughly 88% of the full chip's 192 SMs enabled. This also gives it the same CUDA core count as the GeForce RTX 5090. The GPU is paired with a massive 84GB of ECC GDDR7 memory, delivering nearly 1.4 TB/s of bandwidth, making it an attractive choice for those running local LLMs, if the pricing is right.
PNY's now-removed spec sheet listed a 416-bit memory interface, which would be possible by turning off three of GB202's 16 32-bit memory controllers. However, using 24Gb GDDR7 modules in clamshell mode would yield 78GB of VRAM rather than the advertised 84GB.
A 448-bit interface with 14 active memory controllers appears more plausible, as 28 24Gb GDDR7 modules in clamshell mode neatly produce the stated 84GB capacity. With NVIDIA still labelling the specifications as preliminary, the final memory configuration may still change.
The RTX PRO 5500 sports three NVENC encoders and three NVDEC decoders, giving it one additional decoder over the consumer-oriented RTX 5090 and bringing it in line with other Blackwell PRO GPUs in its class. The GPU also features four DisplayPort 2.1b outputs and a PCIe 5.0 x16 interface. NVIDIA is offering active air-cooled configurations, alongside support for liquid cooling through the RXM form factor.
The RTX PRO 5500 also supports Multi-Instance GPU (MIG), allowing its 84GB of VRAM and GPU resources to be partitioned into two isolated 42GB instances. The RTX PRO 5500 is rated for up to 600W of power, matching the flagship PRO 6000 Workstation Edition, while every other desktop Blackwell PRO GPU is rated at 300W or lower.
Frequently Asked Questions
Open a question for an answer from TweakTown's coverage of this news, or ask your own below.
What workloads are the RTX PRO 5500 Blackwell’s 84GB of GDDR7 memory and 1.4TB/s bandwidth aimed at?
What does MIG support let the RTX PRO 5500 do in workstation or AI setups?
What cooling options are available for the RTX PRO 5500, and how does the RXM form factor differ from air-cooled versions?
What power supply and system requirements should buyers expect for a 600W Blackwell PRO GPU like the RTX PRO 5500?
Have a question that isn't listed here? Ask below, and TweakBot will answer it.
NVIDIA has yet to disclose an official MSRP, but in the current market, we don't expect that price to hold for long. The RTX PRO 6000, which debuted at around $8,000, has doubled in price in recent weeks to $16,000. Given that the PRO 5500 is considerably closer to the PRO 6000 than the PRO 5000 in terms of specifications, we estimate the PRO 5500 could land around the $10,000 mark.
