Rumors have been flying around for awhile now that Nvidia would be launching an RTX 2060 GPU with 12GB of memory, and today Nvidia confirmed it does exist. Instead of just launching the GPU via press release or with hardware-sampled reviews, it simply noted in its latest GeForce graphics driver 497.09 release notes that the RTX 2060 12GB is supported by the new driver. This “leak” means the GPU should be launching soon, most likely just in time for holiday shoppers to not be able to purchase one.
The GPU is a “Turing refresh,” and will be a slightly upgraded version of the card that launched in January 2019. In addition to having double the memory, going from 6GB to 12GB of GDDR6 14Gb/s, the card will also reportedly offer slightly more CUDA cores as well, going from 1,920 in the OG card to 2,176. These additional cores effectively make the card an RTX 2060 Super, which was, and still is, a good midrange GPU, as long as you’re not trying to play Cyberpunk with ray tracing at 4K. One big change though is that it will supposedly have a narrower memory bus, going from 256-bit in the Super version to 192-bit in the new GPU. This reduction in bus width should have a significant impact on the card’s ray tracing performance and ability to run rasterized games at higher resolutions such as 1440p. This will likely make it mostly a 1080p rasterizing card with little chance of running RT in AAA games. The card will boast higher TDP than the Super too, sucking down 184w compared to just 175w.
Regarding the question of why Nvidia would choose to launch an “old” GPU now, the reasoning seems to be twofold: it thinks it will be a card that gamers can actually purchase (imagine that!), and it will be easy to produce as well. Hitting up TSMC’s 12nm node means there’s not much competition for wafers at this time, and according to Tom’s Hardware the card’s lowly 33.06 MH/S Ethereum hash rate is so low that miners won’t even bother. That’s a third lower than an RTX 3060’s hash rate, so it’s possible miners just might pass this one by. However, if the past has taught us anything it’s that even if miners skip it (they probably won’t) there’s still bots to contend with, so we’re skeptical of the notion that any GPU could have widespread availability here in 2021.
The card is rumored to be priced at $300, which would aim it directly at the underwhelming Radeon RX 6600, which has an MSRP of $329. Gigabyte already accidentally leaked the existence of four models in an EEC posting previously, so it’s possible this new GPU will only be made by Nvidia’s partners, instead of the company producing its own Founder’s Edition boards. The GPU is supposed to launch on December 7th, so watch this space.
AMD Will Bring Smart Access Memory Support to Intel, Nvidia Hardware
AMD is reportedly working with Nvidia and Intel to bring hardware support for Smart Access Memory to other GPU and CPU platforms.
Nvidia Will Mimic AMD’s Smart Access Memory on Ampere: Report
AMD's Smart Access Memory hasn't even shipped yet, but Nvidia claims it can duplicate the feature.
Nvidia Unveils Ampere A100 80GB GPU With 2TB/s of Memory Bandwidth
Nvidia announced an 80GB Ampere A100 GPU this week, for AI software developers who really need some room to stretch their legs.
Tesla Ordered to Recall 150K+ Vehicles to Repair Memory Failures
Tesla has been asked — or "asked" — to recall some 159,000 vehicles to repair a NAND memory issue that will eventually cause failures on every affected vehicle.