With the launch of Nvidia’s RTX 4080 GPUs, plural, on the horizon, the company has taken an unprecedented step: it’s “unlaunching” the RTX 4080 12GB. As you might recall, this GPU was widely derided for being the RTX 4080 when given the same specs as the previous x70 GPU. These include a smaller AD104 die, far fewer CUDA cores, and a narrower 192-bit memory bus. Nobody was happy about this situation, as Nvidia had never before offered two GPUs with the same name and different specs. Why it chose this time will remain a mystery for ages. Regardless, it has now responded by removing the GPU from its stable. It’s unclear whether it will be rebadged as an RTX 4070, or disappear entirely. As of press time, the GPU has been removed from Nvidia’s website.
Nvidia announced the sudden U-turn in a blog post. We’ve double-checked to make sure this isn’t a parody site, because it’s amazing that Nvidia listens to its customers. The company holds the lion’s share of the discrete GPU market, so it can do whatever it wants. This time, however, seems to have gone too far.
In its post, the company clearly states: “The RTX 4080 12GB is a fantastic graphics card, but it’s not exactly named. Having two GPUs with the 4080 designation is confusing.” Um, you’re just figuring this out now, Nvidia? We knew it was confusing on launch day and said as much. The rest of the blog again mentions that the 16GB version is still coming. Nvidia also gave it a launch date: November 16. The company says that if the long lines at Microcenter for the 4090 launch are any indication, the 16GB version will undoubtedly sell out as well. It’s probably about right. RTX 4090s are already fetching $1,000 worth of dirt on eBay. It seems mining is dead, but scalping is. Still alive and well.
RTX 40-series spec comparison.
As you can see above, the two GPUs are radically different. In all previous generations, the 12GB card would be labeled as the RTX 4070. Nvidia hasn’t started talking about that specific GPU yet, so it may yet become one. However, it will probably be completely wasted. We hate to beat a dead horse, but this is the stuff that probably helped EVGA decide it was time to quit making Nvidia GPUs. It looks like the partners have already built some of these GPUs, as the launch was only a few weeks away. Also, the 12GB version was a partner-only card, while Nvidia refrained from making the FE version. This probably made the decision easier because it had no skin in the game.
The timing of the announcement makes us think that Nvidia is reading the 4080 benchmarks posted on its website. Every site that covers these things leads with a headline about how much slower the 12GB version is than the 16GB model. This reinforced the idea in everyone’s mind that the 12GB model was a cash grab. Would any gamer really pay $899 for a GPU with a 192-bit memory bus? You may think that this is a special nuance that no one pays attention to, but a person who spends so much money on a GPU will know. Nvidia’s announcement sparked an online uproar over those benchmarks, which leads us to believe the two events are connected.
F1 22 DLSS 3 benchmark, click to expand. (Photo: Nvidia)
There’s another reason for Nvidia’s backpedaling. Instead of doing the right thing to reduce consumer confusion, the company is likely trying to avoid launching a dud. AMD is set to unveil its RDNA3 GPUs on November 3 According to Igor’s Lab, the company won’t touch the RTX 4090 However, its flagship GPU might end up in the middle of both the Nvidia RTX 4080s Recall that the 16GB version cost $1,199 and the 12GB model was $899. Now imagine AMD putting a card between these two that performs near high-end for $700 or more. This will immediately make the 12GB card DOA.
Regardless of its motivation, Nvidia did the right thing here. Introducing a $900 card with x70 specs has led to accusations of nvidia being greedy. It was also an indictment against the company during the pandemic, as it relaunched the RTX 3080 with a $500 price premium along with various TI card launches. We know companies need to make a profit, but offering “new” GPUs with a 5 to 10 percent performance boost over an existing card seems pointless. Nvidia even bragged to investors during that period how its customers were paying $300 more than usual to upgrade. Back then, it sold every GPU it made, so it tried to make as many cards as possible. This has helped lead to the current oversupply of GPUs.
For now, we’ll have to wait and see what happens with the RTX 4070. Nvidia will likely sit back and see what AMD announces on November 3 and then adjust accordingly. It already leaves a big gap between its 40-series GPUs. There’s a huge delta between the RTX 4090 and 4080, so a 4080 Ti is probably on the cards. We already know that a 4090 Ti will arrive at some point. Where AMD’s chiplet-based GPUs will fit into this matrix remains to be seen.