I took a look at the new RTX 2060 SUPER Founders Edition first and was impressed that Nvidia basically turned it into the RTX 2070. With that in mind, it only made me wonder even more where that put the also announced RTX 2070 SUPER. So today I’m going to dive into the changes, take a look at the Founders Edition card, and then run the RTX 2070 SUPER FE through our ever-expanding benchmark suite. After that, we can take a look at pricing and see where the new card fits into the market. I don’t think I will ever get used to the SUPER branding when Nvidia had a few throwback options like GT and Ultra that they have used before. But beyond that, I’m excited to see how the 2070 SUPER performs.
Product Name: Nvidia RTX 2070 SUPER Founders Edition
Review Sample Provided by: Nvidia
Written by: Wes Compton
Pictures by: Wes Compton
Amazon Affiliate Link: HERE
SUPER? What’s changed
Normally before diving into testing a new GPU, I would want to dig into the architecture changes and see what is new but the SUPER lineup isn’t an architecture change nor is it a die shrink. But that isn’t to say that nothing has changed. I went into detail on the changes with the RTX 2060 SUPER in that review so for this review I am focusing on what has changed with the RTX 2070 SUPER. To do that in the table below I have included the GTX 1070, RTX 2070, and the RTX 2080 alongside the new card. The idea is to get an idea of how far things have come from the GTX 1070, to compare the 2070 vs the 2070 SUPER, and then to see how the new card compares to the older RTX 2080. The RTX 2060 SUPER ended up being more like a baby RTX 2070, will that happen with the 2070 SUPER as well?
So like with the 2060 SUPER Nvidia didn’t just change clock speeds they did open up addition cores on the TU106 GPU for the new SUPER model. They went from 36 SMs up to 40 which translates to a jump from 2304 to 2560 CUDA cores. This, however, doesn’t get the 2070 SUPER as close to the RTX 2080 as they did with the 2060 SUPER and the RTX 2070. Part of that is because the 2080 is actually running on a different GPU, the TU104 not the TU106 that the 2060 and 2070 shared. To help compensate they did, however, bump the clock speeds up a lot more. In fact, the 2070 SUPER is clocked higher than the RTX 2080 as well with its 1605 MHz base clock and 1770 MHz boost clock speed. That’s well above the 1410 MHz core clock of the original 2070! The clock speed change and the additional SMs did bring the TFLOP rating up from 7.5 to 9.1 on both FP32 and INT32 workloads which is closer to the 10 TFLOPS of the RTX 2080 than the 7.5 of the 2070. Memory capacity and the controller didn’t change, the RTX 2070 SUPER has the same 8GB GDDR6 and 256-bit controller that the RTX 2070 and RTX 2080 have. The increase in clock speed and core count did, however, change the TGP needed. It went from 175 watts to 215 watts which matches the RTX 2080.
I mentioned all of this in the RTX 2060 SUPER review as well, but if you missed that Nvidia’s presentation focused on a few other areas as well and I wanted to touch on a few of those. Obviously, they did put some basic information about the three new cards. The 2080 Super is faster than the Titan XP which I thought was interesting and the 2070 Super is faster than the 1080 Ti. The 2080 Super is up ahead of the GTX 2080 as well.
They also focused a lot on all of the big ray tracing games of E3 this year, as I mentioned at the start ray tracing is starting to take off when just 9 months ago at the RTX launch there were only a few demos that had it. Big games like Cyberpunk, Watchdogs, Doom, and Wolfenstein all will have ray tracing. Even more importantly all of the big engines and APIs are getting it as well which should translate to a lot of the smaller titles starting to introduce it as well.
Nvidia’s other big announcement alongside the SUPER launch is a software introduction for Frame View. Frame View does a few different things. It is similar to FRAPS in that you can run benchmarks to track FPS but it allows looking at the FPS at the rendered FPS which is at the beginning at the pipeline and the displayed FPS which is at the end of the pipeline. You can also check out percentile FPS in a few different variations. You can also check out the video cards power usage with GPU only or TGP power usage. That right now is only supported by Nvidia cards. All of this can be outputted to excel files and also put on an on-screen overlay. The scan files can get really interesting when putting the frame rate information together with the power usage numbers. You can chart performance per watt graphs. I had a bit of a late start on my testing so I haven’t had the chance to play with Frame View too much but it could be very interesting for testing!
Before getting into testing I also took a look at GPUz to confirm that our test card was running at the same clock speed as listed in the specifications. This also lets me document what vBIOS our card is running and the driver used for testing (the pre-launch press driver).
Nvidia also released a hype video for the new SUPER lineup with a SUPER hero feel this morning. You can check that out here as well.