Remember when Nvidia “launched” a product and declared that while it was a great graphics card, it wasn’t named properly? …”Having two GPUs with the same RTX 4080 name is confusing.” That’s right, we’re talking about the “announcement” of the 12GB RTX 4080 about two months ago.
For those not keeping up, when Nvidia announced its latest GeForce 40-series, they used three offerings: the flagship RTX 4090, RTX 4080 16GB, and RTX 4080 12GB. The problem is that the two RTX 4080 variants are very different, and those differences aren’t limited to storage capacity, as the naming scheme suggests.
Because to reduce memory capacity without using low-density chips, the memory bus width must also be reduced. Using a different chip will result in wildly different performance, which is not a good move,
But this shady behavior is not new. Nvidia has done this before, such as with the DDR4 version of the GeForce GT 1030.
The biggest mistake was announcing the RTX 4080 12GB at the biggest event Nvidia has had in years and this particular anti-consumer movement in the world. Traditionally, they like to deal with dodgy product launches by sneaking them out at a later date with silent releases, meaning few notice and even fewer seem to care.
Well, that’s exactly what they did with the 8GB variant of the GeForce RTX 3060. Ironically, they launched it on October 27th, just 12 days after they introduced the 12GB 4080, because of the confusing name.
While it’s true that the 8GB and 12GB variants of the RTX 3060 are based on the GA106 chip — so they’re technically the same GPU — the changes still go well beyond memory capacity tweaks.
The memory subsystem has been significantly reduced on the 8GB model, and the L2 cache has been increased by 50% on the 12GB model, as has the bus width. Overall, the original RTX 3060 model saw a 61% increase in bandwidth to 360GB/s, while the “new” 8GB model has dropped to just 224GB/s – the same bandwidth as the RTX 3050.
Ultimately, this means the 8GB RTX 3060 will be slower than the 12GB model – spoiler alert, it’s 15% slower on average – but in some cases we’ve seen margins of over 30%, so that’s more than enough for us to draw your attention.
Additional note: Nvidia hasn’t made any official adjustments to MSRP, so the 8GB RTX 3060 is still $330. Looking at Newegg, we can see that the 12GB RTX 3060 starts at $350, with models on offer ranging from $350 to $380. On the other hand, the cheapest 8 GB version comes from Inno3D, starting at $385.
Though some of Asus’ $360 and $370 models lag behind. The point is, in terms of pricing, there’s no difference between the original 12GB model and these new gimped 8GB versions.
Here in Australia we bought a Galax model which we got at 10% off the original 12GB version, you might think “nice”, but it wasn’t. GeForce RTX 3060 has been released for almost 2 years, and everyone is already familiar with its performance. If you are not familiar with it, you can quickly check the rich benchmarks on the Internet. Meanwhile, this new 8GB model is nowhere to be found due to the lack of a review process. It is up to the inspectors whether to purchase the card and notify the consumer.
In fact, you may not even know there’s an 8GB version, or you may just assume the RTX 3060 will only come with 8GB of video memory, just like the 3060 Ti. The catch here is that unsuspecting consumers doing this could be tempted to pay RTX 3060 12GB money for something significantly slower or very close to it
We requested an official response from Nvidia regarding the 8GB version’s release, they simply stated that it was intended to provide more options to customers. They also said that the 12GB version would continue to be available in addition to the 16GB version. This is a sanctioned product from Nvidia, rather than partners going rogue, it’s possible that partners are obligated to produce it.
It appears that Nvidia is attempting to dispose of their lower-end Ampere GPUs in order to maximize profit, they are likely eager to do so. This also facilitates the sale of defective GA106 silicon that is more expensive than it would be if they sold it as a 3050.
Our concern is that we now have two products called “GeForce RTX 3060” and they don’t have the same level of performance, in many instances, they’re not even comparable. We’re not concerned with memory-intensive testing, such as exceeding the 8GB of VRAM, we’re concerned with performance when remaining within the 8GB limit.