Skip to main content

Nvidia GeForce RTX 5080 review: performance worthy of the name?

The pros and cons of DLSS 4 multi frame generation vs RTX 4090 and 4080 Super.

We're dedicating a whole page to DLSS 4 and multi frame generation as it's one of the few defining factors of the product over the existing RTX 4080 and RTX 4080 Super. It's also a strong point of Nvidia marketing to the point wherep CEO Jensen Huang even made the claim that the upcoming RTX 5070 offers RTX 4090 "performance". With the arrival of the RTX 5080, we can start to put those claims to test. After all, if the RTX 5080 can't reach that standard, that provocative 5070 claim lays in tatters.

First up, let's quickly recap on what DLSS 4 multi frame generation does and how it works. It's an offshoot of the existing technology found in 40-series cards. Essentially the current frame is rendered as per normal, and the following one too, which is then buffered. The original frame-gen implementation would then use an optical flow accelerator tied into the game engine to calculate an intermediate frame, boosting frame-rate. The thing is that that buffering the extra frame and calculating the middle image takes time, which means increased latency. Multi frame generation works in much the same way but with key differences: optical flow is now handled by the tensor cores and two, three and four interpolated frames are possible, while pacing hardware within the Blackwell architecture works to ensure smooth delivery of those extra frames.

So the bad news is that frame generation adds latency, but the good news - relatively - is that the amount of latency added by two or three generated images vs the standard two is more limited. There were not many games available for testing DLSS 4 during the review period, but Cyberpunk 2077 and Alan Wake 2 were supported and tie in nicely with our automated benchmarking technology, so let's dig in - the results are compelling.

Cyberpunk 2077, RT Overdrive, Streaming Test, DLSS 4

Frame Rate (FPS)20015010050
Frame Time (ms)5.010.015.020.025.0
Overall frame-rate (FPS)
Resolution:
0100200300400500600700
RTX 5090 Native 2160p
RTX 5090 DLSS Perf 2160p
RTX 5090 DLSS Perf 2x MFG 2160p
RTX 5090 DLSS Perf 3x MFG 2160p
RTX 5090 DLSS Perf 4x MFG 2160p
RTX 5080 Native 2160p
RTX 5080 DLSS Perf 2160p
RTX 5080 DLSS Perf 2x MFG 2160p
RTX 5080 DLSS Perf 3x MFG 2160p
RTX 5080 DLSS Perf 4x MFG 2160p
RTX 4090 Native 2160p
RTX 4090 DLSS Perf 2160p
RTX 4090 DLSS Perf 2x MFG 2160p
RTX 4080 Super Native 2160p
RTX 4080 Super DLSS Perf 2160p
RTX 4080 Super DLSS Perf 2x MFG 2160p
Hover to compare:
PC Latency RTX 5090 RTX 5080 RTX 4090 RTX 4080 Super
4K Native 57.25ms 74.92ms 87.48ms 112.65ms
4K DLSS Perf 27.39ms 36.52ms 31.66ms 39.08ms
4K DLSS Perf 2x MFG 33.33ms 46.04ms 38.91ms 48.83ms
4K DLSS Perf 3x MFG 35.50ms 48.97ms - -
4K DLSS Perf 4x MFG 36.78ms 50.93ms - -

Cyberpunk 2077 is an excellent standard bearer for frame generation technology and it's quite remarkable to see the results playing out on a 4K 240Hz panel. The reason why RTX 5090 works so well - and RTX 5080 works fairly well - is that the base frame-rate with DLSS Super Resolution in place is high, translating into low latency. 27.4ms and 36.5ms are low enough that in a game of this nature, it's hard detect the extra latency incurred by frame generation - up to a point.

However, a lower-spec card produces less impressive results than the higher-spec one - so the RTX 5090 with full frame-gen active actually delivers higher frame-rates and significantly lower lag than the RTX 5080. It's quite remarkable to see the 5090 result with 4x MFG effectively on par with RTX 5080 with no frame generation active at all.

RTX 4090 vs RTX 5080? Putting aside the definition of true performance, the RTX 5080 with 3x MFG has a 16 percent frame-rate advantage over the RTX 4090 - but this comes with a cost, an average 10ms deficit to latency across the fairly lengthy clip. The frame-rate boost provided by multi frame generation increases to a nigh-on 46 percent with 4x MFG, which clearly looks a lot smoother. The cost here in latency terms rises, however, to 12ms. Image quality holds up, so the trade is equitable - but a true performance boost should come with no downsides at all.

Alan Wake 2, High, Full RT, DLSS 4

Frame Rate (FPS)20018016014012010080604020
Frame Time (ms)10.015.020.025.030.0
Overall frame-rate (FPS)
Resolution:
050100150200250300350
RTX 5090 Native 2160p
RTX 5090 DLSS Perf 2160p
RTX 5090 DLSS Perf 2x MFG 2160p
RTX 5090 DLSS Perf 3x MFG 2160p
RTX 5090 DLSS Perf 4x MFG 2160p
RTX 5080 Native 2160p
RTX 5080 DLSS Perf 2160p
RTX 5080 DLSS Perf 2x MFG 2160p
RTX 5080 DLSS Perf 3x MFG 2160p
RTX 5080 DLSS Perf 4x MFG 2160p
RTX 4090 Native 2160p
RTX 4090 DLSS Perf 2160p
RTX 4090 DLSS Perf 2x MFG 2160p
RTX 4080 Super Native 2160p
RTX 4080 Super DLSS Perf 2160p
RTX 4080 Super DLSS Perf 2x MFG 2160p
Hover to compare:
PC Latency RTX 5090 RTX 5080 RTX 4090 RTX 4080 Super
4K Native 111.58ms 237.19ms 143.15ms 269.42ms
4K DLSS Perf 44.40ms 70.07ms 54.71ms 75.21ms
4K DLSS Perf 2x MFG 53.17ms 82.86ms 67.26ms 90.01ms
4K DLSS Perf 3x MFG 58.60ms 86.79ms - -
4K DLSS Perf 4x MFG 59.02ms 91.06ms - -

Alan Wake 2 shows similar results to Cyberpunk 2077 in some respects, highlighting the frame-rate boosting prowess of multi frame generation. However, base latency is significantly higher than Cyberpunk 2077 running at the same frame-rate - which is a profound piece of information. The contention has always been that running a game at 40-60fps should provide a good basis for frame generation. And I guess it probably is in a great many titles. However, Alan Wake's best latency results - as measured with DLSS performance super resolution active - is higher than Cyberpunk's with frame-gen active! There's a strong argument here that we should stop thinking about base frame-rate for frame-gen suitability and instead concentrate on latency first and foremost.

Looking at RTX 5080 in example, results are considerably worse than RTX 5090. Our best possible latency here is a remarkable 70ms, meaning that once we've tested through each frame-gen iteration, we're up to a staggering 91ms with 4x MFG. What you get is remarkably smooth gameplay, but there's a disconnect between how the game looks and how the game plays. This suggests to me we need to test many more titles to ascertain base frame-rate and how the multi frame gen results play out.

With all of this in mind, the RTX 4090 vs RTX 5080 comparisons flatter to deceive. Once again, 3x MFG on the new card offers a higher frame-rate than 2x frame-gen on RTX 4090 - it's a good 14 percent faster with 3x MFG and around 46 percent to the better with 4x MFG. However, the 4090 is offering an average 67ms against 87/91ms on RTX 5080. The gap is too wide for the RTX 5080 to provide what you might call a like-for-like experience. More testing is going to be needed on more titles, but it raises serious questions about Nvidia's claims that RTX 5070 will offer RTX 4090 performance bearing in mind that 5070 results can only be worse compared to 5080.

Nvidia GeForce RTX 5080 Analysis