NVIDIA To Make Good On Rare GeForce RTX 50 GPU Defect With Missing ROPs

GeForce RTX 50 series card on a black background.
It turns out that a spattering of reports about some GeForce RTX 5090 graphics cards shipping with missing render output unit (or raster operations pipeline) hardware were the result of a "production anomaly" and not an issue with the BIOS. NVIDIA said as much in a statement confirming a "rare issue" with some cards, which also includes the GeForce RTX 5090D for the Chinese market and the GeForce RTX 5070 Ti.

Apparently, the GeForce RTX 5080 and upcoming GeForce RTX 5070 (non-Ti model) are not affected, or if they are, NVIDIA isn't saying so at the moment. To be clear, the only reports of missing ROPs that we've seen so far are tied to GeForce RTX 5090 and 5090D models.

In a statement provided to The Verge, NVIDIA's GeForce global PR director, Ben Barraondo, offered some context and guidance to owners of affected cards.

"We have identified a rare issue affecting less than 0.5% (half a percent) of GeForce RTX 5090 / 5090D and 5070 Ti GPUs which have one fewer ROP than specified. The average graphical performance impact is 4%, with no impact on AI and Compute workloads. Affected consumers can contact the board manufacturer for a replacement. The production anomaly has been corrected," Barraondo said.

The issue came to light when a TechPowerUp forum member reported seeing fewer ROPs listed in the GPU-Z utility for their Zotac GeForce RTX 5090 Solid (non-OC model) graphics card. It just so happened that TPU had the same custom model on hand, which also showed fewer ROPs—168 instead of the 176 it's supposed to have.

TPU reran some comparison benchmarks and found that the gimped Zotac model was the least performant of the stack of GeForce RTX 5090 models in its stable (it tested eight cards, by our count). Those benchmarks showed the affected model to be 5.6% slower than NVIDIA's own Founders Edition model, and 8.4% slower than a factory-overclocked ASUS ROG Astral GeForce RTX 5090 OC model.

Since running the story, other reports have come in of missing ROP hardware on cards from a wide variety of hardware partners, including ASUS, Gigabyte, MSI, Palit, and NVIDIA's own FE model (in addition to Zotac and potentially others).

NVIDIA's official statement doesn't get into specifics on the production anomaly at play, only that it results in one fewer ROP than what's supposed to be there. The company is talking about one hardware unit, each of which can process eight pixels per GPU clock cycle, the latter of which is how ROP counts are advertised. In the latter context, the GeForce RTX 5090 is supposed to have 176 ROPs.

Prior to NVIDIA issuing a statement and guidance, notable leaker MEGAsizeGPU (@Zed__Wang) posted on X that the "root cause is the chip," claiming that a "small batch of GB202 is defective, and the BIOS can not do anything with the issue." This was echoed by @harukaze5719, another prominent leaker who stated that missing ROPs are the result of a "chip-level defect."

How To Tell If Your GeForce RTX 50 Card Is Affected

GPU-Z screenshot of a GeForce RTX 5090 graphics card.

If you own a GeForce RTX 5090, GeForce RTX 5090D, or GeForce RTX 5070 Ti, you can check if yours is defective by running the GPU-Z utility. Several core specs are listed on the main 'Graphics Card' tab. You'll find ROPs/TMUs clumped together a few rows down. The first figure is the ROP count (and the second one is the number of TMUs).

The image above is a screenshot from a GeForce RTX 5090 FE model that we have in-house. It shows the correct number of ROPs. Both the 5090 and 5090D should show 176 ROPs. If you own a GeForce RTX 5070 Ti, the correct number is 96. If it says 88 (or any number other than 96), then it's defective.