Just like the leaked information, Nvidia has officially launched the new RTX Super graphics card series with noticeable upgrades in performance while the price is almost unchanged.
Super series currently has 3 versions including RTX 2060 Super, RTX 2070 Super and RTX 2080 Super. I already have two versions of Founders Edition of RTX 2060 Super and RTX 2070 Super, open the box for you to see and share a little bit about its differences compared to the original versions.
In the Founders Edtion series alone, the design of RTX Super and RTX 20 series is no different, still the familiar "gas stove" style with 2 fans, still a very high-end finish with a finely cut aluminum cover. But on the black plastic panel between the two radiator fans, next to the RTX 2060 or RTX 2070 series, there will be Super letters on Nvidia green background. This is also the only factor that helps distinguish Super generations from the original generation.
In addition, I discovered that there are some minor changes in the design that the metal ledge catches the card into the case where the screw holes are bent down to make it easier to catch the snail, increasing the capacity. load bearing for the card without the need for a stand. The overall design is still very beautiful and many brothers will still love the Founders Edition series, although it is difficult to buy.
So how is RTX 2060 Super RTX 2060 and RTX 2070 Super different from RTX 2070?
RTX 2070 Super uses a GPU TU104 variant of the same as RTX 2080 but with the Stream Multiprocessor layout of 40 SM, its CUDA multiplier will be 2560 and 320 multipliers, the RT multiplier corresponding to each SM cluster also It is 40. According to Nvidia's design of each texture texture (TPC) cluster, it will include 2 SM + 1 PolyMorph engines, so it can be deduced that the number of texture processing units (TU) will be 160 TU and 20 engine PolyMorph. In addition, the GPU clock is also increased slightly by Nvidia with the basic pulse now being 1605 MHz and Boost 1770 MHz.
So RTX 2070 Super will lie between RTX 2080 and RTX 2070. RTX 2080 also uses TU104 with 46 SM, 2944 CUDA cores, 368 Tensor cores, 46 RT cores, clock frequencies 1515 – 1800 MHz. Meanwhile RTX 2070 uses TU106 GPU with 36 SM, 2394 CUDA cores, 288 Tensor, 38 RT cores, 1410 – 1710 MHz clock. As for memory and bandwidth settings, it is unchanged with 8 GB GDDR6, 1750 MHz memory pulse, 256-bit bus width, and bandwidth of 448 GB / s.
The last time I evaluated RTX 2080 FE, it showed the same power as the GTX 1080 Ti, not to mention the hardware Ray Tracing with RT cores. So with this upgrade, the RTX 2070 Super will be even more powerful than the original RTX 2080, Nvidia said, stronger than the GTX 1080 Ti. Having 320 Tensor multipliers will help this card handle better machine learning applications and of course you can experience games with better Ray Tracing than RTX 2070 thanks to 40 accelerated RT multipliers for processing. Another notable point is that it has a TDP of only 175 W, 10 W lower than the RTX 2070 and 50 W lower than the RTX 2080.
Another big difference between the RTX 2070 Super and RTX 2070 is that on the Super generation it has NVLink bridges – something that RTX 2070 doesn't have. Nvidia corrected it with the RTX 2070 Super because GTX 1070 or GTX 1070 Ti previously supported SLI multi-GPU.
More interesting is the RTX 2060 Super, which is easy to buy when its price is below 10 million VND. RTX 2060 starts at 349 USD, RTX 2060 Super version is 50 USD more expensive but if based on upgrades in this version, 50 USD is worth it. Basically the RTX 2060 Super still uses the TU106 GPU as the RTX 2060 or RTX 2070 but it is worth noting that Nvidia has increased the number of SM to 34 SM, from which the CUDA multiplier increases to 2176, along with 272 multipliers. Tensor, 34 personnel RT and number of TU also increased to 136 TU. GPU clock is slightly increased with basic pulse level of 1470 MHz and Boost 1650 MHz. Improved memory with capacity up to 8 GB GDDR6, memory pulse on par with RTX 2070 is 1750 MHz, bus width is now 256-bit instead of 192-bit to memory bandwidth of RTX 2060 Super peer with more advanced lines, ie 448 GB / s.
Compared to the RTX 2060, this card also uses the TU106 GPU with 30 SM, 1920 CUDA cores, 240 Tensor cores, 30 cores RT, 120 TU and goes with only 6 GB GDDR6 memory and lower bandwidth. The RTX 2060 Super approached the RTX 2070 and last time in the RTX 2060 review, I said it was as strong as GTX 1080, surpassing the GTX 1070 Ti, this time, the RTX 2060 Super will be a rank higher than GTX 1080.
The card's TDP is unchanged at 160 W and the card still uses 8 PCIe power sources like the RTX 2060 version.
Equipped with output port on RTX 2060 Super (left) and RTX 2070 Super. Both support USB-C port output as well as to support new generation of VR glasses using only one USB-C cable. The rest of the RTX 2070 Super has an additional DisplayPort port (Nvidia comes with a adapter to transfer DisplayPort to DVI) and RTX 2060 Super is DVI.
The performance of these cards, I will soon send you a detailed review next week.