Nvidia’s new next-gen GPU benchmarks cause concern among PC gamers, particularly with the RTX 5080 – but don’t panic yet

3 hours ago 2

A set of Nvidia RTX graphics cards next to a compact PC case.

(Image credit: Nvidia)

Nvidia has provided some new benchmarks for RTX 5000 graphics cards
Two of them don’t involve DLSS 4 and Multi Frame Generation
However, the gen-on-gen uplifts shown here are modest – but we still shouldn’t get carried away

Nvidia has released some more game benchmarks for its next-gen Blackwell GPUs, and we’ve caught a couple of results that don’t use DLSS 4 and its Multi Frame Generation (MFG) feature.

The trouble with the in-game benchmarks that use MFG – which is a big upgrade on Nvidia’s original frame generation, inserting more artificial frames to up the frames per second count – is that they aren’t a fair apples-to-apples comparison with RTX 4000 graphics cards using DLSS 3 frame generation (the latter can’t use DLSS 4 MFG, as it’s exclusive to RTX 5000). And that’s the case for most of the benchmarks aired thus far.

So, PC gamers are hungry to see generational comparisons that don’t use DLSS 4, avoiding this skewing of the results, and we’ve got two games where this has happened in this fresh benchmarking – reported by ComputerBase (via VideoCardz) – namely Horizon Forbidden West and Resident Evil 4.

Resident Evil 4 doesn’t use DLSS at all – but does have ray tracing turned on – and Horizon Forbidden West gives us a glimpse of rasterized (non-ray tracing) gen-on-gen performance, but with DLSS on (with no frame generation, though, crucially, so MFG is removed from the equation).

Going by the bars in the bar chart provided – estimating their relative lengths, as Nvidia doesn’t provide hard figures – it looks like the RTX 5090 is about a third (33%) faster than the RTX 4090 in these two games. However, there’s a much leaner 15% or thereabouts jump with the RTX 5080 versus the RTX 4080.

With the RTX 5070 and its 5070 Ti sibling, we’re looking at more like a 20% jump compared to their respective predecessors, again just in those two games.

The Nvidia GeForce 5090 GPU on display at CES 2025

Analysis: Fake frames outcry part umpteen

“See. Nvidia’s new graphics cards are a big con – without ‘fake frames’ they are going to be rubbish!”

Fake frames meaning frame generation, and this is the kind of vitriolic comment that’s popping up a fair bit following this revelation (and indeed before it, to be fair). But we do have to bear in mind that this is just a couple of games, in certain specific configurations.

Still, I concede the general point. On the one hand, Nvidia will obviously want to show off DLSS 4 and MFG as it’s a big leap forward (well, in theory at this point for all of us outside Team Green) for its GPUs. But on the other hand, it doesn’t feel great that most of the benchmarks shown thus far use MFG, and as noted, aren’t fair or direct comparisons with RTX 4000 graphics cards. These benchmarks show at least 30% to 40% gains (in previous airings), or a doubling of frames rates (as seen here with some games, and indeed a 2.9x gain with Indiana Jones and the Great Circle). Of course, this isn’t what you’ll see outside of games that support DLSS 4 with MFG.

Gamers would like to see a broader range of benchmarks, including pure rasterized performance without any DLSS – which we haven’t seen at all, as Resident Evil 4 above, the only game not to have DLSS turned on in Nvidia’s testing, is ray-traced performance.

Nvidia’s job is, at the prerelease stage, to stoke excitement for its graphics cards, obviously enough, but the slant towards that aim feels too skewed for gamers (and myself, I should add) in the way the RTX 5000 GPUs have been shown off so far.

Even so, the shakier-looking gen-on-gen performance of Horizon Forbidden West and Resident Evil 4 should not be used as a springboard to reach a conclusion along the lines of the (fake) fake frames rant I introduced this section with – that’s unfair, and going too far in the other direction

That said, to some extent, a lesser generational uplift is expected with Blackwell compared to Lovelace (RTX 4000), outside of the software plus AI tricks (neural texture compression) and the new trump card of MFG. After all, RTX 5000 is made on the same process as RTX 4000 (TSMC 4N, albeit an improved version, 4NP, for Blackwell), and so there’s no process drop to facilitate beefy generational gains there – that side of the equation relies purely on architectural enhancements.

Before we get too bogged down in the details here, one thing is clear enough – we need to wait for reviews before we get anything approaching a fully rounded picture of RTX 5000 performance. Which, of course, is always the case.

Still, there remains an inescapable feeling Nvidia is hiding something with the heavy slant towards MFG in this generation’s prerelease buildup – a lesson for Team Green to be more even-handed with its marketing efforts next time, perhaps. And of course, we don’t know how MFG is going to pan out in its execution and smoothness yet, either, for those PC games that do use the shiny new tech.

There are still a lot of unknowns, although all these questions will be answered soon enough. In theory the RTX 5090 review is coming next week, if the rumors prove to be correct.

Darren is a freelancer writing news and features for TechRadar (and occasionally T3) across a broad range of computing topics including CPUs, GPUs, various other hardware, VPNs, antivirus and more. He has written about tech for the best part of three decades, and writes books in his spare time (his debut novel - 'I Know What You Did Last Supper' - was published by Hachette UK in 2013).

Read Entire Article