Why you can trust Tom's Hardware Our expert reviewers spend hours testing and comparing products and services so you can choose the best for you. Find out more about how we test.
It's hard to find things to like about the GTX 1630. If it arrived last year, at this same price, it might have had a few takers in the midst of the global GPU shortages. Today? It's too slow and too expensive. It's frankly absurd that some places like EVGA have GTX 1650 and GTX 1650 Super cards that cost less than the GTX 1630.
Maybe the pricing will correct, but realistically it needs to be cut in half for the cards to make any sort of sense. As a replacement in the product stack for the aging GTX 1050 — which hasn't been manufactured for a few years now, as far as we can tell — a GTX 1630 with a similar $110 price point would have been okay. You get double the VRAM at least, which means some of the games that choke on 2GB cards could still run. Alternatively, Nvidia could have made this a GT 1030 replacement, priced below $100. It still wouldn't be fast, but at least it would have the ultra-budget pricing angle going for it.
With AMD's Radeon RX 6400 delivering significantly better performance for $50 less, the only real selling point for the GTX 1630 would be its media encoding capabilities. If that's your primary consideration, waiting for Intel's Arc A380 to ship in the US would be a better choice. Gaming performance might be questionable, but the AV1 media encoding capabilities look quite promising.
The Turing TU117 GPU was never a potent chip, but disabling half of the cores and memory channels makes a relatively weak GPU into a solution that can barely hope to compete with integrated graphics. Yes, it's faster than the GTX 1050, but it's also slower than the GTX 1050 Ti, and both of those can readily be picked up used for about $100 these days.
You don't even get extra features compared to the 1050 Ti that would warrant the upgrade — it's the same NVENC hardware, more or less, with the same amount of VRAM, same video outputs, and worse performance. Power efficiency isn't an advantage either, as the GTX 1050 Ti only used 58W in our gaming test, and both are nominally 75W cards. So what's the point?
The point is to make something out of nothing, as far as I can tell. Nvidia probably had some TU117 chips that had functional NVENC hardware but with some of the memory controllers and SMs that couldn't hit the GTX 1650 or MX450 requirements. Just like there are GTX 1650 cards made using TU106 silicon, but with all the ray tracing and tensor core hardware disabled, and with only 14 of the potential 36 SMs enabled, now we have a GTX 1630 made with chips where 50% of the cores and memory controllers apparently failed.
If Nvidia or its partners priced this appropriately, some people would inevitably find a use for it. That means that it needs to cost significantly less than the pre-existing GTX 1650 cards, since those are universally faster. Even Intel's Arc A380 will likely deliver superior performance, though it's a gamble on whether or not Intel will ever get the drivers to fully functional status. We'll hopefully have a card soon that we can put through its paces.
- MORE: Best Graphics Cards
- MORE: GPU Benchmarks and Hierarchy
- MORE: All Graphics Content
Unreal Engine supervisor at ModelFarm blasts 50% failure rate with Intel chips — company switching to AMD's Ryzen 9 9950X, praises single-threaded performance
Nvidia reportedly preparing a Blackwell-based Titan GPU — maybe that 4-slot prototype cooler will show up again
AMD introduces a free Frame Latency Meter tool to measure mouse response time — works with all GPUs, no high-speed camera or manual frame-counting needed
