New computer GPUs have launched at a furious pace the past few months, mostly in the $500-and-up sector. This week, we finally see a 2020 GPU arriving at a lower price than a brand-new gaming console: the Nvidia RTX 3060 Ti, priced at $399 and launching tomorrow, December 2.

But once again this year, Nvidia is leaving people in the dark about how many of these cards we can expect to reach stores. We know the company manufactured at least one of them, at any rate, because my review hardware arrived last week. The usual gamut of benchmarks confirms performance on par with last year’s RTX 2080 Super, at nearly half the cost.

Like other RTX-branded GPUs, the Nvidia RTX 3060 Ti features proprietary processing cores on its silicon—namely, its Tensor cores (for AI computation) and its RT cores (to manage all things ray tracing). To get to a $399 price, the 3060 Ti drops specs compared to its higher-ranked siblings in the usual categories, particularly CUDA cores, but it also severely drops its Tensor and RT core counts. Nvidia’s trick here is that those core types have been updated since last year’s model to do more work per core.

If this were a market where you could easily snap up a $499 RTX 3070, some of these RTX 3060 test results would be hard to swallow, considering the price-per-dollar comparison. But there’s no getting around this new card’s ability to match the RTX 2080 Super (original MSRP: $699) in every category that counts.

Like most Ars Technica GPU reviews, we limit our benchmarks to 4K tests, owing to the fact that lower-resolution benchmarks typically become CPU-limited and thus don’t tell the full story of how a GPU will turn out in your particular PC. (If you’re wondering, my testing rig sports an i7-8700K CPU, overclocked to 4.7GHz, plus 32GB DDR4-3000 RAM, an 850W PSU, and a PCI-e 3.0 SSD.)

Last month, we saw an exception to this testing standard thanks to the 128MB of L3 cache in AMD’s Radeon RX 6000 series, which drives improved 1440p performance. But AMD’s cheapest card as of press time, the $579 RX 6800, isn’t a fair comparison with the RTX 3060 Ti. (In other words: yes, its extra $180 delivers more power in 4K and 1440p modes.) Should AMD ever launch a lower-priced 6000-series card, we’ll be sure to go back and test 1440p modes accordingly.

Otherwise, there’s not a ton to say about RTX 3060 Ti that hasn’t been spelled out with its costlier siblings. DLSS still impresses as a proprietary upsampling and anti-aliasing system, and that, combined with solid ray-tracing tech, continues to make Nvidia cards a tantalizing option—especially when clock speeds and CUDA cores have been reduced to hit the $399 sweet spot while still otherwise looking quite performative.

Meanwhile, if your favorite games don’t tap into DLSS, you should expect to tinker with their settings to maximize their 1440p or 1080p performance levels—and I can’t help but imagine AMD has a response to this exact use case with any future lower-priced RX 6000-series GPUs. But nothing of the sort has been announced yet, so for the time being, Nvidia takes the lead at this price point.

Listing image by Nvidia

https://arstechnica.com/?p=1726395