Nvidia’s RTX announcement was really exiting, it’s not every generation that we get to see big implementations like Ray Tracing. They did a pretty good job, explaining the technology at Gamescom and why it could be a game changer for the industry.
Adopting first generation of any new implementation can be a bit risky, as we did see the RTX 2080ti struggle to maintain 60 fps on the new Tomb Raider Game with Ray Tracing turned on, that too at 1080p. But people looking forward to the RTX cards wouldn’t need to worry if Nvidia got the basics right and deliver a adequate performance boost over previous generation, at affordable prices.
The RTX 2080Ti and RTX 2080 look great on paper and no doubt they will be great performers, but not every gamer can afford to spend 800$+ USD on a graphics card. This is where the RTX 2070 comes in. The 70 series cards from all generations have acted like a bridge between the mid-range and the high-range, it actually brings you very close to flagship performance at a decent price.
Now this recent development really worries me, according to VideoCardz, the RTX 2070 will be a TU106 GPU.
Over the years, the 80 and the 70 series cards have had similar architectures, like both the GTX 670 and GTX 680 were on the GK104. The GTX 970 and 980 were both on the GM204, barring the 700 series where we did see the architectural difference.
Yes, all the three cards are based on Turing but on a different die. The RTX 2080Ti will be on the TU102, the RTX 2080 on the TU104 and if VideoCardz is correct, the RTX 2070 will be on TU106.
Although this doesn’t mean the RTX 2070 will be a bad performer, but it might not be as efficient as the RTX 2080, bandwidth might be lower too. Because the Ti series cards have always been based on slightly different architectures than the 80 or 70 series cards, they always had considerable performance gains, fitting in a lot more transistors. For perspective, both the GTX 1080 and 1070 packed in 7.2 billion transistors, compared to the 1080ti’s 12 billion.
Cuda cores can be a good performance metric here, the 2070 will have 78% cores of 2080. That’s about the same difference between the GTX 1080 and GTX 1070. So even if performance doesn’t take a significant hit, efficiency might be affected. Ray Tracing performance might also take a hit, but that has to be tested.
There’s a 50% price increase in RTX cards across the board compared to previous generation, so I don’t really see why Nvidia wouldn’t keep the same architecture for both the RTX 2080 and RTX 2070, they could simply just pack less Cuda cores, like they have previously. But I still hope that these changes don’t pull back the RTX 2070’s performance. Benchmarks will be out on September 19th, when the NDA on reviews end.