The upcoming Nvidia GeForce RTX 2060 series of graphics cards will come in an assortment of different memory configurations according to a leak on the EEC (Eurasian Economic Commission) website.
The leak appears to be comprised of forty (not a typo) different Gigabyte SKUs based on the RTX 2060 line of graphics cards. While not all of the specs were listed, there are a few very interesting take homes here.
The first is that we have 3 different amounts of memory: 3GB, 4GB and finally 6GB of RAM. We also see GDDR6 memory and also GDDR5(X). It’s unclear whether the memory is just ‘vanilla’ GDDR5 or if it is the faster GDDR5X variant.
What’s interesting is that these leaks also appear almost simultaneously to another leak from a Laptop vendor in China. The Chinese OEM CJScope have a series of laptop devices (in both 15 and 17-inch sizes) which will feature a variant of Nvidia RTX cards.
The lower end SKU is an RTX 2060 Mobile variant, featuring 1536 CUDA cores, 6GB GDDR6 memory and with a base clock of 1320 MHz, and boosts to at least 1620.
The 1536 CUDA cores are significantly lower than the 1920 CUDA Cores we see for the 6GB desktop models, and I do wonder if the lower end RTX 2060 desktop variants will have a similar cut number of CUDA cores.
I have doubts that this wouldn’t be the case – Nvidia reduced the number of cores for the GeForce GTX 1060 on the lower RAM variants (such as the GeForce GTX 1060 3GB), but furthermore – if the 1920 CUDA cores are correct, the card will be faster than the GTX 1080. It would be very odd to equip a GTX 1080-ish card with only 3GB of GDDR5(X) memory, so I suspect the GeForce RTX 2060 will follow a similar trend.
In other news, the existence and specs of either the RTX 2050 or GTX 1150 have leaked. It’s very much more likely that it is the GTX 1150, as supposedly the lower end cards in the Turing lineup won’t be labeled as ‘RTX’, but this hasn’t been officially confirmed so I will hedge my bets.
According to the Geekbench entry, the 1150 / 2050 will feature 896 CUDA cores, and the memory appears to be GDDR5 and not GDDR6, in a 4GB configuration.
The RTX 2080 and RTX 2070 mobile cards sit as virtual mirrors of their desktop counterparts. With 2944 CUDA cores for both the RTX 2080 mobile and desktop (using the same memory bus width and 14gbps memory clock speed).
The RTX 2070 retains the same 2304 CUDA cores, the same 8GB GDDR6 at 14gbps as the desktop too.
What’s interesting is that the clock speeds found here are a little faster than the desktop variants. Perhaps they are cherry picked ‘A’ silicon variants, and capable of maintaining higher boost clocks compared to the typical chips.
The CJScope laptops will also sport the Intel 9th generation processor architecture and supposedly will see a launch of mid-January, 2019 (that’s apparently the same date as we’ll see the Desktop GeForce RTX 2060 launch too).
According to the benchmark results we see here, the RTX 2070 will significantly outperform the previous generation of Pascal based laptop graphics cards, but they are using only 3d Mark 11, so it’s hardly a comprehensive set of results.
The launch of the desktop RTX 2060 isn’t the only card Nvidia intends to release for the value-focused customer. The GTX 1160 (yes, GTX) will also see the light of day, which is supposedly the Turing architecture, but absent the Ray Tracing cores (Tensor cores haven’t been confirmed either way).
For gamers and power users on the go, the Turing architecture will be a fantastic option, because Turing is considerably faster in applications such as vRay and 3DS Max (as we recently tested).
Sources:
