A Closer Look
So let’s quickly discuss the market, because it’s an odd one. NVIDIA’s lowest-priced card until now was the RTX 5060, an 8GB card that launched for $299 and, in all honesty, not just from ourselves, but it wasn’t what I’d call, best received. That really came down to one key thing: 8GB of memory on a 128-bit bus. But it came with a difference, because at least it utilised GDDR7 memory to leverage some higher bandwidth. Though with games like Indiana Jones and Doom Eternal, to name a few, showcasing the issues that 8GB of VRAM presents in 2025, the RTX 5050 is already on the back foot, especially given that NVIDIA have plumped for GDDR6 instead, leading to a pretty measly, at least in the grand scheme of things, 320 GB/s of bandwidth.
Comparing that to what AMD have, it’s very much the same. That features the same GDDR6 memory and uses the same 128-bit interface, so very similar in terms of bandwidth. While AMD and their partners didn’t want to send out the 8GB model for review, us and other reviewers made our own conclusions based on the 16GB model, and again we saw the 8GB variant dead on arrival.
So you’ll probably be starting to see a bit of a trend with these cards, and it all comes down to the memory. 8GB isn’t enough, but why? When the RTX 30 series launched, some cards had 8GB of VRAM, and no one complained. Yes, it was a different time, as we take a trip down memory lane, but no one batted an eyelid. So what’s changed? Well, in all honesty, a lot of it comes down to the consumer, and sorry, but I’m going to blame you guys to a degree. We as gamers have been a little spoiled, so now it’s almost expected or a given to be able to run the latest games on max settings, even if at a lower resolution. It’s almost unacceptable to have anything less, and with higher settings generally having larger textures, or raytracing baked in, much like we see on games like Indiana Jones, which is a culprit of this, it just doesn’t fit the bill, and that leads to issues.
Now, there is a caveat to all of this, and that comes down to price. If AMD, Intel, or even NVIDIA release a card that is below par in the court of public opinion, as long as the price stacks up, then there is a saving grace. And until now, we’ve not really had that, whereas the RTX 5050 coming in at $249 just scrapes into that margin that could see it considered as a worthy contender in the more budget-focused space.
Not only content with butchering the memory design on this card, NVIDIA have also plumped for using the PCIe 5.0 x8 interface, whereas cards like the RX 9060 XT, even in its 8GB form, uses x16 bandwidth and is only $50 more. So there are many angles to take this from, but our focus is on how this GB207-based GPU performs in all of the latest games, if we can get them running at all.
Now, while we did reach out to a lot of AIB partners to get a card, as mentioned, they were somewhat reluctant. Though Colorful came to the rescue with their Ultra W Duo OC model, and while we will have some more content specific to this model and its cooling capabilities, along with breaking down the card to look into the design and quality, today is all about the GB207-300-A1 core. A core that is only used on the 5050 series, including mobile and equivalent RTX PRO models.
This particular model does improve on the stock clock speeds by increasing the boost clock by around 3% from 2572MHz up to 2647MHz, which isn’t a huge amount. But this is done through the one-key overclock button on the rear, which again we’ll be looking at in a separate review, while we stuck with the stock clocks, so our results will be pretty typical of what most RTX 5050 models will be like.
