Introduction

Today, we have for review the Gainward GeForce GTX 1630 Ghost. This card is a real “ghost” from the past, as it’s the first GeForce GTX product we’ve reviewed in almost two years. The GTX 1630 is, surprisingly, a brand-new SKU being launched today. Based on the older “Turing” graphics architecture and positioned in the 16-series which lacks real-time ray tracing or DLSS support, the GTX 1630 is expected to become NVIDIA’s new entry-level product.

The GTX 1630 is based on not just the older “Turing” architecture, but its silicon is also built on the older 12 nm FinFET node. Why NVIDIA decided to launch this now is really anyone’s guess. Perhaps the company was sitting on a mountain of 12 nm TU117 chips bound for GeForce MX mobile SKUs and decided to rope in the desktop segment to consume some inventory, but we’ll never know for sure. Based on its positioning, the GeForce GTX 1630 will compete with the AMD Radeon RX 6400 and Intel Arc A380.

The 12 nm TU117 silicon physically features 16 “Turing” SMs, or 1,024 CUDA cores, 64 TMUs, 16 ROPs, and a 128-bit wide memory interface. Even the GTX 1650 doesn’t max this out, featuring just 896 CUDA cores. The new GeForce GTX 1630 in today’s review has just half of the CUDA cores physically present on the silicon—512. That’s not all as NVIDIA even halved the memory bus width to just 64-bit while keeping the memory amount the same at 4 GB. GPU clock speed is significantly higher than the GTX 1650, at 1785 MHz compared to 1665 MHz on the GTX 1650.

The Gainward GTX 1650 Ghost features a simple monoblock fan-heatsink with a pair of fans. Although the typical board power of the GTX 1630 is rated at 75 W, Gainward has given it a 6-pin PCIe power input to ensure the card stays far away from the PCIe slot electrical design limits. We haven’t received any price information from Gainward or NVIDIA, but based on the rumors floating around, these cards are expected to sell for $150. I expect the Gainward GTX 1630 Ghost to sell at the baseline MSRP price point, too.