Nvidia GeForce RTX 4080 Super review: the 4K GPU shoot-out

In performance terms, there's a law of diminishing returns to the excitement level surrounding the arrival of each new Super, from the generally positive reaction to the RTX 4070 Super to the more muted response to the RTX 4070 Ti Super - and now, ambivalence with the 4080 Super improvement. The truth is, it's a pricing adjustment dressed up in a marketing announcement and this is generally fine overall - but how much of a value enhancement are we actually getting?

I've noticed that 'dollars per frame' based on a multi-game average is a comparative measure used by some of the PC press and I think it's a fantastic 'at-a-glance' look at value with much to commend it. Obviously it will be defined a great deal by myriad factors including the choice of game and how many RT benches are included on one end and how many esports on the other. Still, I thought I would give it a go and the table with my own results is seen below (with a tip of the hat to Will Judd for breaking out the Excel for the numbers!).

I have one reservation, however. Averaging a bunch of averages based on a non-linear scale (frames per second) doesn't make sense to me. So I asked developers for their opinion and this comment in particular stood out to me: "If you average two games at 30fps and 300fps, the first game will become insignificant while in real life, 0-30fps will be far more noticeable than 270-300fps. Or, the other way around, 10 games at 30fps will have as much of an impact as one game at 300fps."

Their suggestion? Average the results from all GPUs for each game, then calculate variance of each card to that average. Then at the end, scale all normalised values by the average of all average FPS on a per-game basis to return values that are 'averageable', enabling a more consistent, normalised 'cost per frame'. In effect, we're 'porting over' a non-linear series of results to something that isn't distorted in the averaging process.

Based on the results of all of our RT and raster benches, here's how that looks - and I've included the more usual averaging of fps averages number too.

Whichever calculation method you choose, the evidence is pretty clear. When it launched at a massively misjudged $1200, the RTX 4080 offered no additional value than the RTX 4090. The notion of an exhorbitantly expensive flagship giving way to much better value further down the stack diminished to nothing - and we actually had examples where the RTX 4090 delivered better value than the RTX 4080. Astonishing.

The cost per frame values for the 4080's predecessor - RTX 3080 - are $12.52 ($12.47 normalised), so while relative value only slightly tips towards the RTX 4080 Super, at least we're looking at something fairer. In effect, this looks very much like the kind of top-end pricing that the RTX 4080 should have had at launch. If you're looking for proportionately better value, the 4070 Super is clearly worth taking a look at, based on the table above.

There's one more thing I'd like to point out though. We can't ignore that reaching a certain performance threshold makes more experiences viable, especially at 4K resolution, and that may be worth paying the premium. Path-traced Alan Wake and Cyberpunk 2077 are doing things that the RX 7900 XTX never will. It's that simple. DLSS spatial upscaling, DLSS 3 frame generation temporal upscaling, DLSS ray reconstruction - they're combining to make special things happen.

1 of 4

Caption

Attribution

Ultimately, I can't help but think that if you're 'dropping' a grand on a new GPU, you should have access to top-tier experiences at 4K resolution. AMD's value is clear and present, but I think this is all worth taking into consideration. In the gallery above, you'll note that even on a top-tier AMD sponsored game - Avatar: Frontiers of Pandora - the RTX 4080 Super is ahead of the RX 7900 XTX and DLSS quality mode is delivering improved image quality over its FSR2 equivalent.

In summary then, RTX 4080 Super does what it needs to do in delivering the kind of price/performance/features we didn't get at launch. Similar to the RTX 4070 Ti Super, it feels like the kind of package it should have been back in the day: expensive but not egregiously so. The price is improved then, but it arrives 14 months on from the 4080's launch and while there is a slightly better level of proportionate value compared to the rightly celebrated RTX 3080, it's still hard not to feel that this is still a pretty steep asking price for an 80-class product. At least now there is a gap for a 4080 Ti, but with the AI boom, it does feel less likely we'll ever get one.

Nvidia GeForce RTX 4080 Super Analysis

• Introduction
• RT benchmarks: Dying Light 2, Cyberpunk 2077, Control
• RT benchmarks: Metro Exodus Enhanced Edition, F1 22, Marvel's Spider-Man: Miles Morales
• RT/DLSS/FSR2/DLSS3 benchmarks: Cyberpunk 2077, Dying Light 2, Forza Horizon 5, Marvel's Spider-Man: Miles Morales
• Game benchmarks: Control, Cyberpunk 2077, F1 22, Forza Horizon 5
• Game benchmarks: Hitman 3, A Plague Tale: Requiem, Returnal, Marvel's Spider-Man: Miles Morales
• Conclusions, value and recommendations [This Page]