When people first asked me whether I had any plans for an NVIDIA RTX A2000 review, I was like "wut? why would I?" After taking a closer look at the product and what it offers for the non-professional crowd, my interested was piqued enough to buy one in retail. While it's not your typical "GeForce" card, the RTX A2000 has several unique capabilities. First and foremost, it's extremely compact. The market for small-form-factor graphics cards has gotten smaller and smaller in recent years, which usually means your options for discrete graphics are the RX 550, GT 1030, GTX 1050 Ti or GTX 1650—weak cards that are unable to handle Full HD gaming in 2022. The NVIDIA RTX A2000 is based on the modern GA106 GPU, which also powers the GeForce RTX 3050 and RTX 3060. Unlike the latter, the RTX A2000 runs with PCIe slot power only, so you don't have to worry about separate power cables, which often are not available on some SFF PSUs.

Averaged over our full test suite at 1080p Full HD, we got very good results from the RTX A2000 given its form factor and positioning. The card matches RTX 3050 almost exactly, which means it's a bit quicker than the GTX 1660 Super and GTX 1660 Ti, 25% faster than the Radeon RX 6500 XT, and roughly 30% faster than RX 580 and GTX 1060. 20 out of our 25 games ran at over 60 FPS—at maximum settings. I'd say that definitely makes the card 1080p Full HD capable; sacrificing one or the other setting for better FPS shouldn't be a big deal. If you've been thinking "but integrated graphics," consider our Ryzen 5 5700G IGP results. They show that the GeForce GTX 1060 is almost twice as fast as the 5700G's integrated graphics, which are some of the fastest available. This makes the RTX A2000 over 3x (!) as fast as "good" integrated graphics; i.e., 20 FPS vs. 60 FPS.

While the RTX A2000 does support ray tracing, and we tested it in this review, it's not something I'd consider realistic for this card. It simply does not enough horsepower—the card's fully loaded with RT off trying to achieve 60 FPS. Of course, it never hurts to have the capability, and it might help with resale value. Compared to the Radeon RX 6500 XT, RT performance is much better, more in line with the RTX 3050 and RX 6600. A more useful capability is full support for NVIDIA's DLSS upscaling technology, which gives you an additional dial to sacrifice a little bit of image quality for higher framerates.

In terms of visual design, I really like what NVIDIA has done with the RTX A2000; it's basically just a black box with a fan installed. The black, matte surface gives the card a much higher-end look and feel. What tarnishes the impression is the lack of a backplate, though. Imagine if NVIDIA added a nice backplate of the same material as the cooler shroud, which would come to $2, maybe $5? C'mon. Under the hood, the RTX A2000 uses a really simple cooling design. It's basically a slab of metal with fins cut into it sitting in the airflow of the single fan. This fan is of the radial blower type, which usually means noisy and high-pitched. Not so on the RTX A2000. With 27.5 dBA, the card is very quiet in idle, almost inaudible. Unfortunately, it does not support the idle fan stop capability included with all GeForce 30 Series cards, but it is quiet enough that you won't hear the fan in a closed SFF case when used as a media player, for example. Once the card is fully loaded while gaming, the fan spins up noticeably, and you will be able to make it out over other noise sources if you focus on it, but with 31 dBA, the card is reasonably quiet, especially for its tiny form factor. Temperatures are great and lower than expected with 71°C; there's plenty of headroom for a custom fan curve. The radial blower fan design is a good choice because hot air is exhausted through the slot, so the insides of your SFF case aren't heated up. Due to the way NVIDIA's fan control works, the lowest fan speed you can set manually is 30%—the idle fan speed. This gives you a way to quieten down the card under load, which runs at 41% fan speed. Idle noise can't be lowered further because the card already runs at 30% fan speed in idle, though.

Where the NVIDIA RTX A2000 really shines, and I double-checked my numbers, is power consumption and efficiency. The card is designed to operate with PCIe slot power alone, which of course simplifies cable routing because there is no cable to route, and PSU requirements because you don't need a 6-pin or 8-pin plug. On the other hand, this means the card will have to make do with the 75 W the PCIe slot is able to provide—quite a challenge given we've been seeing 130 W from the RTX 3050, and even AMD's brand-new 6 nanometer RX 6500 XT uses 89 W in gaming. So what's NVIDIA's secret sauce? It is "just" the same 8 nanometer GA106 GPU that powers the RTX 3050 and RTX 3060. The answer lies in the voltage-frequency curve. This curve defines what voltage the card operates at to achieve a certain clock speed. NVIDIA's Boost algorithm dynamically scales the frequency up and down depending on the power limit and other factors, like temperature. While we usually see the GA106 run at 1.08 V (RTX 3050) and 1.02–1.04 V (RTX 3060), NVIDIA runs the RTX A2000 at only 0.74 V on average in gaming. That's a huge difference to anything we've seen in recent years, and it has paid off. The energy efficiency of RTX A2000 is pretty much off the charts—twice (!!) as good as the 6 nanometer RX 6500 XT, twice that of RTX 2060, 80% better than RTX 3050, and 40% better than the RX 6800 non-XT, the most efficient graphics card we ever tested until this review. This is undervolting from the factory at its best and the reason NVIDIA managed to get this card to work with 75 W slot power only, and such a puny heatsink. Great job!

The undervolting and conservative frequency settings are probably also the reason why overclocking worked so well. We gained 16% in real-life performance, which is MUCH more than the overclocking potential we're usually seeing on other graphics cards. What makes the overclock even more impressive is that it was achieved with the card still respecting its 70 W power limit. Actually, it's not possible to increase the power limit manually, as the manual adjustment range is set to +0%. I feel like the option to increase the power limit could have been useful for even general operation. Any somewhat decent motherboard will be able to provide 80 or 90 W on the slot easily, which is a significant increase (+14% and +28%) that could easily yield additional performance as the card is held back a bit by its power limit right now.

NVIDIA announced an MSRP of $450 for the RTX A2000, which is surprisingly reasonable given this is a professional-grade graphics card. Of course, all MSRPs are a fantasy these days. I bought my card for €720 including 20% VAT, which roughly converts to US$700 without tax. At that price point, the card is definitely not cheap; the RTX 3050 at $500, RX 6500 XT at $300, and RX 6600 at $500 are definitely superior in terms of price/performance, but none of them offer the ultra-compact form factor of the RTX A2000, and they all require additional power inputs. This puts the RTX A2000 into an interesting niche for builders of high-performance, small-form-factor systems willing to pay the SFF tax. If your case does have more room for a full-height graphics card, maybe also consider the Palit RTX 3050/3060 StormX OC, which is full-height, but only 17 cm long. Overall, the RTX A2000 is an engineering miracle that's hard to recommend for a general audience—that said, it still deserves our Innovation award.