Quote:

The battle of the weedy GPUs

Benchmarks featuring an AMD Radeon RX 6400 graphics card have been spotted on Geekbench. Benchleaks has shared a quartet of Geekbench browser results for what appears to be a singular PC system going through multiple Geekbench runs, perhaps checking the performance changes after system tweaks. As Geekbench isn't really a good GPU benchmark, take these results with a grain of salt.

The Radeon RX 6400 was launched at the same time as the Radeon RX 6500 XT, but it's an OEM-only part (for now). While that should mean the scores come from a pre-built PC, it's entirely possible to remove the card and test it in a different system — there's nothing locking the card to only certain motherboards, for example. The particular system specs consist of an AMD Ryzen 5 5600X, installed on an ASRock X570 Taichi, with 32GB of dual-channel DDR4 RAM. The system had Windows 11 installed and ran v 5.4.1 of Geekbench.

It's worth noting that ASRock has a history of working a bit outside the lines, so it's possible these results are from ASRock testing the GPU for a potential retail release.

If we take the above as typical results for an RX 6400, we can make some rough performance comparisons. First of all, the RX 6400 is going to obviously live in the shadow of its bigger brother, the RX 6500 XT. The biggest differences between the siblings are that the RX 6400 has 25% fewer cores, has a nominal 53W TGP (Total Graphics Power), and runs nearly 500MHz slower than the RX 6500 XT that boosts up to 2.815 GHz (reference).  So, how much slower do these changes make the RX 6400?

According to the Vulkan score comparisons, having picked through the Geekbench database, it looks like the RX 6400 is just 7% slower than the RX 6500 XT, which isn't too terrible considering how much has been chopped from the silicon. But again, this is Geekbench. Some other comparisons might help show the kind of company the RX 6400 keeps. Its score of 34,749 is easily beaten by the GeForce GTX 1060 6GB (37,185), but is almost level with the GTX 1650 (34,764), while the higher of the two results puts it ahead of the typical RX 6500 XT scores.

Moving along to OpenCL scores, let us say the RX 6400 can achieve 40,272 in this benchmark (half way between the above leaked OpenCL results pairing). Doing the same three comparisons, the RX 6500 XT scores 52,287, which is nearly 30% faster than the unreleased smaller sibling. The RX 6400 performance in this test (40,272), which usually shows AMD GPUs in a good light, takes revenge on the GTX 1060 6GB (36,405), and beats the GTX 1650 (38,111) by a small margin.

The AMD Radeon RX 6500 XT and the Radeon RX 6400 both went official in early January. Together they represented AMD's first assault on the sub-$200 graphics card market with the RDNA/2 architecture, but so far we have only been able to see examples of the former released. Large OEMs often have longer lead times when it comes to adopting new hardware, and it looks like the Radeon RX 6400 is now out in the wild. Whether we'll ever see a retail launch, and what the price might be, remain to be seen.

Given the significantly lower TGP, the only real draw for the RX 6400 is that it doesn't require any form of external power connector. That means it can work in virtually any PC built within the past decade. That's the good news. The bad news is that with any older PC that doesn't support PCIe 4.0, performance will suffer — we measured an 8% drop on average when testing the card on a PCIe 3.0 i9-9900K system compared to our i9-12900K baseline. It could be much worse on older PCs that also lack resizable BAR support.

The Radeon RX 6500 XT hasn't been selling all that well, based on the GPU prices we're tracking on eBay, so not even scalpers are able to sell the cards. Cutting performance another 20–30% shouldn't win any additional buyers.
...