Nvidia GeForce GTX 1630 Review: Lobotomized Turing

The Nvidia GeForce GTX 1630 started popping up in rumors in leaks a few months back. Part of me thought, “Surely Nvidia won’t release a new and pathetically slow Turing variant this late in the game.” But the realist in me knew it was only a matter of time — the GT 1030 and GT 730 cards that started shipping again last year was the only evidence we needed.

Let’s be blunt: The GTX 1630 isn’t anywhere close to being one of the best graphics cards, and in fact it lands near the very bottom of our GPU benchmarks hierarchy. The only slower GPUs that we’ve tested are the GTX 1050, RX 560, RX 550, and the aforementioned GT 1030. None of those are worth your time or money either, but at least they’re not being released in mid-2022.

The only real contender for the GTX 1630 is AMD’s recently launched Radeon RX 6400, but this new Nvidia card actually makes the lackluster 6400 look good. Actually, the real competition — and the reason no one should give the GTX 1630 the time of day — comes from the existing GTX 1650 and GTX 1650 Super. The latter basically doubles the specs of the 1630, and pretty much doubles performance as well.

Colorful sent us this sample for review, but there doesn’t seem to be an official price from Nvidia. What we can find online suggests that the soft MSRP has been set at $199, which is just silly. EVGA lists its own GTX 1630 for $199, or you can also buy the far superior GTX 1650 Super for $199. The 1630 also nominally replaces the GT 1030, which had a launch price of $79 — $70 for the faster GDDR5 variant. It feels as though the GTX 1630 was priced according to mid-2021, but it’s now laughably expensive.

Here’s how the specifications for Nvidia’s old-timer Turing TU117 GPUs stack up, with the RX 6400 and RX 6500 XT for comparison. 

GPU Specifications
Graphics Card GTX 1630 RX 6500 XT RX 6400 GTX 1650 Super GTX 1650
Architecture TU117 Navi 24 Navi 24 TU116 TU117
Process Technology TSMC 12FFN TSMC N6 TSMC N6 TSMC 12FFN TSMC 12FFN
Transistors (Billion) 4.7 5.4 5.4 6.6 4.7
Die size (mm^2) 200 107 107 284 200
SMs / CUs 10 16 12 20 14
GPU Cores 512 1024 768 1280 896
RT/RA Cores N/A 16 12 N/A N/A
Boost Clock (MHz) 1775 2815 2815 1725 1665
VRAM Speed (Gbps) 12 18 16 12 8
VRAM (GB) 4 4 4 4 6
VRAM Bus Width 64 64 64 128 128
ROPs 32 32 32 48 32
TMUs 32 64 48 80 56
TFLOPS FP32 (Boost) 1.8 5.8 4.3 4.4 3
Bandwidth (GBps) 96 144 128 192 128
PCIe Link Gen3 x16 Gen4 x4 Gen4 x4 Gen3 x16 Gen3 x16
TDP (watts) 75 107 53 100 75
Launch Date Jun-22 Jan-22 Jan-22 Nov-19 Apr-19
Launch Price $199 $199 $159 $159 $149
Online Price $199 $168 (opens in new tab) $149 (opens in new tab) ~$125 used (opens in new tab) ~$90 used (opens in new tab)

AMD’s RX 6400 basically tied the GTX 1650, so there’s little question it will easily beat the lower spec GTX 1630. About the only advantage that the GTX 1630 has is the presence of video encoding hardware — the TU117 has Pascal-era hardware, not the improved Turing encoder, but it’s still better than nothing.

What’s particularly odd with the GTX 1630 is that Nvidia has been shipping the same TU117 GPU in laptops as the MX450 (and more recently MX550) for a couple of years, though granted NVENC is disabled on those parts. Apparently, there were enough chips that couldn’t reach the required 14 SMs for the MX450 or GTX 1650, and Nvidia and its partners figured a cut-down 10 SM variant might still sell on desktops — to the uninformed, anyway.

You basically get all the same features as a GTX 1650, just with less performance. You don’t even necessarily get a lower power card, as the GTX 1630 models we’ve seen still come with a 6-pin power connector. That might be because these were less desirable chips with defects, or maybe it’s because Nvidia tried to make up for the lack of GPU cores with slightly higher clocks. At least the RX 6400 can be found in half-height models and doesn’t require additional power.

The 4GB GDDR6 comes clocked at 12Gbps, with a 64-bit interface. As noted already, that’s exactly half of what the GTX 1650 Super provides. Also note that, unlike AMD’s RDNA 2 GPUs, there’s no Infinity Cache to make up the difference, though the 512 GPU cores are already going to be a limiting factor.

With most modern graphics card selling at close to MSRP, the GTX 1630 feels like far too little, far too late. Last year we saw GTX 1650 cards going for $300 or more, and a $200 GTX 1630 might have made some kind of warped sense. Today, you can get the RX 6500 XT, RX 6400, GTX 1650, and even GTX 1650 Super for $200 or less. Maybe this was supposed to be for big OEMs, so they could toss in a weak GPU and claim to still offer dedicated graphics, but that’s ultimately just going to lead to disappointed customers.