Nvidia is preparing to launch new high-end GPUs, but the operative word may be “high-end.” In previous cycles, both AMD and Nvidia have typically led with a high-end card before debuting lower-end hardware within a matter of months. Exactly how quickly this rollout takes place varies from cycle to cycle — two months would be extremely quick, 12 months would be rather slow. We already knew Nvidia was doing things differently this time around, with a simultaneous GTX 2080 and 2080 Ti launch, but now there’s confirmation that the company is playing things differently with its lower-end stack as well.
According to Nvidia CFO Colette Kress, the RTX 2080 and 2080 Ti will exist alongside Pascal through the holiday season. At the recent Citi 2018 Global Technology Conference, Kress told the audience to expect Pascal in-market for a while yet. “We will be selling probably for the holiday season, both our Turing and our Pascal overall architecture. Remember, Turing is a leap forward in terms of overall capabilities. The performance improvement is much greater than the overall price. What that means is you are getting for your dollar spent tremendous more improvement.”
Whether the performance improvement will be commensurate with the price is very much an open question. Apart from some vague slides, Nvidia has released very little performance information and the new GPUs are significantly more expensive than previous generations. Nvidia’s current GTX 1080 is selling regularly for ~$ 469, while the new GTX 2080 will start at $ 700 ($ 800 for a Founders Edition). The 2080 Ti, meanwhile, is now a $ 1200 GPU. Performance data on ray tracing suggests that even top-end cards may struggle to hit 60 FPS at 1080p with ray-tracing enabled.
Now, granted, the question of whether someone would rather play at 4K with ray tracing off or 1080p with ray tracing enabled is a personal one. Both are valid options. But what we do know, at this point, is that Nvidia’s new GPUs are significantly more expensive than its previous, they set new records as far as high-end pricing across the GPU stack, and the performance gains these GPUs offer may or may not be in-line with those increases. As always, ET recommends a wait-and-see approach before committing to pre-orders. Any comparison of these cards will also need to take into account the pricing they actually command rather than MSRP. Hopefully, we won’t see a repeat of 2016, when Nvidia launched in May and cards were still difficult to find and selling at inflated prices four months later.
It’s not clear if Nvidia is planning to replace the 1080 and 1070 directly or if it will only keep Pascal cards in market in segments where it hasn’t introduced a Turing replacement yet. If we had to guess, we’d guess they’d keep the 1080 down — otherwise, Nvidia will leave a substantial gap in its own product space.
It’s not clear if Nvidia intends to do a complete refresh around Turing. It’s possible that Turing is a 12nm stopgap part, aligned as a high-end chip, while a new 7nm GPU is readied for next year. This, to be clear, is speculation on our part. But it might explain why Nvidia would launch primarily high-end GPUs while holding off on a complete refresh. That, and let’s face it — the company has little reason to aggressively push Turing down the product stack. With AMD not planning a refresh cycle this year, Turing isn’t going to have much company at the high end.