Card brings cost and performance that differs little from the RTX 2060
THE Nvidia brought the GeForce RTX 3050, graphics chip originally launched for notebooks, for the market of dedicated video cards for desktop computers. Yes, it’s literally been a week since we’ve seen AMD do just that and go terribly wrong. Will it go this time? Or will this be another bad day for folks desperate for an entry-level graphics card, but still decently performing?
As is the rule in the entry-level hardware market, this product uses Nvidia’s current generation technologies, but with several restrictions. Despite that, this is where the differences show up when compared to the RX 6500 XT released by AMD last week. The reductions don’t go beyond the limit, especially in specifications crucial to the functioning of the card. So a total of 8GB of GDDR5 memory was kept – unlike the notebook versions, which have models with only 4GB. Also no cutting PCIe tracks, having a total of x16 communication channels, which means that even in a PCI Express 3.0 system it has enough bandwidth to keep the graphics chip in high performance.
Where are the cuts then? Comparing to an RTX 3060, we have slight reductions like CUDA cores (from 3,584 to 2,560) and a decrease of about 30% in available structures. On the other hand, there is also a price reduction, with the RTX 3060 being launched for US$ 329 while the RTX 3050 was announced for US$ 249. Lately, suggested and practiced prices are two very different realities, but we will comment on this more for front.
Starting with the performance, with FullHD, this chip is able to face any game in Ultra configuration or at least in Alto with high frame rates.
Even with ray tracing, it is able to run, but for these scenarios it needs to be combined with techniques such as AMD FidelityFX Super Resolution or Nvidia DLSS.
– Continues after advertising –
In all these scenarios, we can see a pattern where the card is 10 to 20% behind the RTX 2060, and some 25 to 30% below the RTX 3060. Versus AMD’s entry-level competitor, the Radeon RX 6600, we have two scenarios. In rasterization, the AMD card has a wide advantage, opening from 15 to 30%. But it’s turning on some feature related to Ray Tracing and Nvidia’s greater efficiency in this area makes both cards touch.
But there is one element in favor of the RTX 3050 that can make it attractive to the GTX 1050 and GTX 1060 crowd, at this point desperate for an upgrade to continue running releases in high quality: it is the cheapest RTX introduced to date. This makes it bring technologies like Ray Tracing acceleration and especially DLSS.
Upscaling techniques, especially DLSS, have brought impressive results, but they have limitations. And one of the biggest is that if you don’t give enough pixels of information, there’s no way an algorithm or even machine learning can reconstruct a final high-quality image. That’s why FullHD DLSS is much stricter than 4K DLSS – and the RTX 3050 doesn’t perform for 4K or DLSS.
In our comparisons, DLSS in quality mode still maintains a good level of graphics and definition at 1080p, but reducing it to Balanced mode will already have a negative impact on definition. This is not a surprise, since DLSS in Quality mode on FullHD already uses only 720p rendering as a base, and bringing it to balanced reduces this to just 1114 x 626. That is, we started to have very little information for the machine learning to work.
– Continues after advertising –
But there is a good in-between: 1440p. With the good performance of the RTX 3050 in 1080p, you can use DLSS in quality mode to scale from FullHD to QuadHD, with very positive results both in performance and in the final image quality. By forgoing Ultra for High, and using DLSS in quality, it is possible to achieve a high enough frame rate in virtually all titles.
And how much more advantageous is it to have DLSS instead of AMD FSR? In direct comparison to the Qualities and 1440p modes, DLSS clearly delivers more definition in the image and better resolves jagged edges on some edges. Here you need to give the FSR a helping hand, let it render more of the image using Ultra Quality mode, and still DLSS does better.
Looking at these adjustments in a performance graph, we have the following scenario:
Here we have a head start for the RTX 3050 (versus the RX 6600) for those with an eye on QuadHD. In games that have DLSS, you can look for a quality very close to the native one with a good level of performance, while the Radeon model will need to either render natively, or bring these losses in the final image quality to enable FSR. The result is still good, but less than what DLSS offers.
So, this time it makes sense to buy a notebook chip brought to PCs? Here we have a completely different situation from the RX 6500 XT. This chip maintains all the elements to deliver a good experience, such as enough video memory, PCIe lines not being a bottleneck even in 3.0 configuration, possibility of streaming or recordings accelerated by NVENC and, above all, a very consistent performance, ready to face FullHD in high quality and even QuadHD with the help of DLSS. But it all depends on the price.
With value close to the RTX 2060 and with less performance in many of the tests, this model is a portrait of the stagnation of the entry-level graphics card market. It arrives struggling to show a clear improvement over the card released three years ago, with only a few improvements in RT and more memory – and well behind the Radeon RX 6600 in many scenarios, which is also very close in cost.
Again a board that is barely relevant versus models that have been on the market for years
In theory, it’s a card with a good level of performance, but, in practice, for those looking for a new video card, it’s bringing practically nothing new. Nor will it help anything in the sad situation of looking for an entry-level graphics card, with an acceptable balance between cost and performance.