Nvidia has successfully confirmed that the RTX 2070 goes to be the underside rung of the ray tracing ladder, with any GPU beneath it unlikely to get a good expertise with the {hardware} out there. We spoke with Nvidia’s Tom Petersen, who defined that the extent of efficiency you’re going to get from the RTX 2070 will basically be the minimal {hardware} for actual time ray tracing when it launches in October.
“It turns out five billion is about entry price for a good ray traced experience,” says Petersen. “So Pascal, at our highest level, could cast – if you software emulated – around a billion. So we’re going from a billion to 5 or 6 billion.”
The upcoming RTX 2070 goes to be able to 6 Giga Rays/sec efficiency, so something beneath that’s going to essentially wrestle – particularly because the RTX 2070 goes to must curtail its ray tracing ambitions in contrast with the higher-spec RTX 2080 Ti and RTX 2080 playing cards. Those playing cards are able to 10 and eight Giga Rays/sec respectively.
Ray tracing is a significantly high-end know-how even when it’s solely used for lighting results, so it’s not shocking that you just’re unlikely to get the total RTX items on the extra mainstream finish of Nvidia’s new 20-series GPUs. This would appear to place pay to any hopes of the RTX 2060, or GTX 2060, getting in on the ray tracing motion every time we see the extra price-focused variations of the brand new GeForce GPUs drop.
But it’s not altogether inconceivable. While 5 Giga Rays per Second appears to be shorthand for the minimal {hardware} necessities for actual time ray tracing, that does depart a small quantity of wiggle room beneath the RTX 2070. If Nvidia chooses to construct an RTX 2060 from the identical TU106 GPU because the 2070 then it’s potential if may simply sneak in above that efficiency restrict.
From some very tough calculations you’d most likely want round 30 SMs, and subsequently 30 RT Cores, to hit the mandatory 5 Giga Rays per Second mark. That would give us an RTX 2060 with 1,920 CUDA cores and possibly a really excessive value level too. It would then nonetheless be a 445mm2 GPU with round 11bn transistors inside it, in spite of everything, and that’s virtually the identical measurement because the top-end GP102 chip that powered the Titans and the GTX 1080 Ti.
Then there may be additionally the potential for various ranges of ray tracing in games to offset the decrease energy RTX silicon, with games doubtlessly having completely different settings relying on how exhausting you need to push your GPU.
“So I would say with 6 Giga Rays you can definitely imagine you’re not going to be able to ray trace the same way you do with a 2080 Ti,” Petersen tells me, “but kind of like level-of-detail, or like Ultra, Min, Max settings, you’re going to cast fewer rays. It’ll be a little less sharp in the reflection, a little less clear in the ambient occlusion, but it’s still going to look pretty good.”
Source
