Nvidia’s Turing refresh cycle has been continuing apace these past few months, with new lower-end GPUs replacing the Pascal holdovers from earlier in the cycle. GPUs like the GTX 1660 Ti lack the Turing-specific features like ray tracing and DLSS support, the 1660 Ti was also the only GPU of the family that comes with a significant improvement in performance per dollar as well as raw performance. As such, there’s interest in seeing if the 1660 and 1650 will replicate this trend.
Both the 1660 and 1650 Ti are expected to use GDDR5, not the faster, more expensive GDDR6 that debuted with upper-end GPUs in the product stack. The GeForce 1660 should be competition for the RX 590, given that it increases core counts by ~10 percent over the GTX 1060. Overall power consumption will undoubtedly favor Nvidia in this comparison, whether the RX 590 or RX 580 is the reference target. We’ll reserve thoughts on absolute performance until we see shipping data, though we’d expect broad performance parity. Power on the 1660 looks to be provided by a single 8-pin PCIe port.
The 1650 is a bit more interesting. If the rumors of its configuration are accurate, the part looks positively underwhelming in comparison with the GPUs it will compete against. The bump from 768 GPU cores on the 1050 Ti to 896 on the 1650 will be welcome, but the RX 570 is an average of 1.42x faster than the 1050 Ti, and moving from 768 GPU cores to 896 is a gain of just 1.16x. That’s to say nothing of the RX 580, which has moved down to the <$200 price point and offers even more cores, memory bandwidth, and RAM to compete.
Assuming these are the configurations Nvidia brings to market, it doesn’t look as though the company intends to particularly challenge AMD below the $200 price point. To be fair, there isn’t much sign it needs to do so. The latest figures from Jon Peddie Research show AMD losing significant market share in Q4 2018 compared with Q3, and performing even more poorly against Q4 2017.
Clearly, holding on to the <$200 price points, even if AMD does it well, is no substitute for a competitive GPU stack top-to-bottom. Right now, that’s the one thing AMD doesn’t have. The price for a new Vega 56 has returned to $400 after a short discount, while a new Vega 64 runs $500. That’s simply not competitive against the current crop of RTX GPUs from Nvidia. Until AMD slashes its prices or brings Navi to market, it’s going to be effectively locked out of the higher band price segments.