A new episode has disrupted the launch of the RTX 50 series. It has been reported that some RTX 50s (mainly 5090s) did not have the initial number of ROPs activated on the GB202 GPU (168 instead of 176). If you’re in any doubt about your version, simply run a program like GPU-Z to find out. This difference is not insignificant, since we can see a drop in performance of up to 8.4% compared with a customised model overclocked at the factory. If you compare it to a ‘valid’ FE, the performance loss is 5.6%.
One less ROP, one more worry
A few hours ago, Nvidia not only confirmed the problem, but also revealed that the issue did not only affect RTX 5090s. In its communication, NVIDIA claims that 0.5% of RTX 5090 and 5070 Ti graphics cards are affected by missing ROPs.
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 ROPs below those specified. The average impact on graphics performance is 4%, with no impact on AI and compute workloads. Affected consumers can contact the GPU manufacturer for a replacement. The production anomaly has been corrected.
We won’t go on for 10 years about this problem, which is now known and recognised. However, what is slightly disturbing is the speed with which Nvidia has responded. The Greens are not in the habit of bouncing back from rumours, leaks and other statements. By providing a precise response, as well as an estimate of the volume impacted. Even if your mind isn’t in the right place, you’d be forgiven for thinking that the Greens have known about this problem for quite some time…
In the case of impacted cards, it is also important to realise that there is no real short-term solution. The only solution for customers who have been affected is to replace their cards…But there are no cards on the market at the moment. However, Nvidia has confirmed that people wishing to have their cards replaced will be able to do so when the situation is more favourable to them for an immediate exchange…