The NVIDIA GeForce RTX 5080 16GB is now available in retail stores, though stock is limited and prices are inflated. Officially priced at $999, it matches the launch price of its predecessor, the RTX 4080 Super. However, the similarities don’t end there. Both GPUs share:
- A 16 GB memory buffer
- A 256-bit memory bus
- 64 MB of L2 cache
- TSMC’s N5 node
- Near-identical die sizes.
| Model | Die | Die Size | Process | Cores | Boost Clock | L2 Cache | VRAM | MemoryBus | Bandwidth | TBP | MSRP |
|---|---|---|---|---|---|---|---|---|---|---|---|
| GeForce RTX 5090 | GB202 | 750 mm² | TSMC N5 | 21,760 | 2407 MHz | 96 MB | 32 GB GDDR7 | 512-bit | 1.79 TB/s | 575W | $1,999 |
| GeForce RTX 5080 | GB203 | 378 mm² | TSMC N5 | 10,752 | 2617 MHz | 64 MB | 16 GB GDDR7 | 256-bit | 960 GB/s | 360W | $999 |
| GeForce RTX 4080 Super | AD103 | 379 mm² | TSMC N5 | 10,240 | 2550 MHz | 64 MB | 16 GB GDDR6X | 256-bit | 736 GB/s | 320W | $999 |
| GeForce RTX 4090 | AD102 | 609 mm² | TSMC N5 | 16,384 | 2520 MHz | 72 MB | 24 GB GDDR6X | 384-bit | 1.01 TB/s | 450W | $1,599 |
The RTX 5080 packs 500 more cores, but overall, the two GPUs are very similar on paper. The differences come up on the memory side. The RTX 5080 leverages GDDR7 memory clocked at 30 Gbps, while the 4080 Super sticks with 23 Gbps GDDR6X. This boosts the former’s memory bandwidth by 30%.
The RTX 5080 is also clocked slightly higher (2,617 MHz vs. 2550 MHz), contributing to its elevated 360W TBP. The RTX 4080 Super is limited to 320W in comparison.
NVIDIA RTX 5080: DLSS 4 & Multi Frame Generation Benchmarks
Multi-Frame Generation is the main reason for the existence of the RTX 50 series, and that’s where we’ll start. It works similarly to frame generation technology (frame interpolation) introduced with Ada Lovelace, but inserts multiple frames instead of one. To smoothen out the process, and reduce potential latency, the job us offloaded to Blackwell’s display engine which is twice as capable as Lovelace. Read more on the topic here.
Alan Wake 2
Alan Wake 2 is among the few games to showcase path traced lighting. Unfortunately, you can only experience it on high-end GeForce RTX graphics cards. Single frame generation and quality mode upscaling produce an average of ~60 FPS on the RTX 5080, versus 68 FPS on the RTX 4090. Enabling Multi Frame Generation buoys the former to 85 FPS and 110 FPS, with 1% lows of 72 FPS and 92 FPS.
Switching to performance mode upscaling boosts the RTX 5080 to 150 FPS and 119 FPS using frame generation 3x and 4x. Meanwhile, the RTX 4090 finishes with an average of 95 FPS which is still about 14% faster than the 5080 with single frame generation.
When used without upscaling, frame generation grants an 80%, 155%, and 224% boost using 2x, 3x, and 4x modes, respectively. At 80%, most of the uplift comes from the first frame. The second frame grants a 40% increase, while the third interpolated frame offers less than 30%.
As seen below, multi frame generation produces the most scaling at low framerates. However, even then the first frame is responsible for much of the gain (+92%), showing near-perfect scaling. Like the previous case, the third and fourth frames add 30-40% more performance over the first.
Cyberpunk 2077
Cyberpunk 2077 features one of the best implementations of ray and path tracing. The RTX 5080 gains a ~50% bump with frame generation 3x (over 2x), leading the GeForce RTX 4090 by 20%. Switching to frame generation 4x extends the lead to over 50%.
Combining quality mode upscaling with multi frame generation pushes the RTX 5080 to 125 FPS at 4K “Path Traced” settings, making it 50% faster than the RTX 4090 (with FG 2x). It’s worth noting that the 1% lows are only ~20% higher, unlike the last case where they were higher by more than 50%.
Switching to performance mode upscaling (with FG 4x) produces an average of 182 FPS on the RTX 5080, once again a lead of 50% over the RTX 4090 running FG 2x. In this case, the third frame adds only 25% more performance, improving the lows by a mere 10%. The overall scaling drops from 86% with DLAA (native) to 75% with half-resolution upscaling.
The “Psycho” ray-tracing preset extends the RTX 5080’s lead over the RTX 4090 (FG 2x) to 60% using frame generation 4x. Once again, low input framerates produce superior results with 89% scaling, albeit a limited 30% from the third frame.
Pairing rasterization with frame generation produces favorable results for the RTX 5080. It actully manages to beat the RTX 4090, with and without frame generation. Frame generation 4x grants it a whopping 90% lead over the 4090 (with FG 2x) at 4K “Ultra.”
Regardless, multi frame generation scaling past 2x remains limited to roughly 80%, with the bulk of the gains coming from the first two frames. It’s worth noting that regardless of the overall gains, the third frame tends to produce very similar gains across a wide range of tests.
Star Wars: Outlaws
Star Wars: Outlaws features ray-traced lighting and RTXDI. We run the latter separately in the next comparison. Running the highest native ray-traced settings grants the RTX 5080 (FG 4x) a 50% bump over the RTX 4090 (FG 2x) at 4K. As previously seen with low framerate DLAA inputs, frame generation 4x grants a 91% boost over 2x.
RTXDI shows even higher frame generation scaling with an average of 106 FPS. That’s again a near-50% higher than the RTX 4090 (FG 2x). Multi frame generation 4x grants an improvement of 98% over 2x, with 3x gaining 56% over 2x. The fourth frame bump remains under 30%.
Switching to performance mode upscaling elevates the RTX 5080 to nearly 150 FPS, maintaining a 50% lead over the RTX 4090 (FG 2x), an increase of 80% over single frame generation.
Marvel Rivals
Marvel Rivals shows spectacular scaling with multi frame generation 4x, averaging 181 FPS at 4K “Ultra” with DLAA. It’s 77% faster than the RTX 4090 (with FG 2x), yielding twice the framerate as single frame generation. Even the third frame grants a near-40% improvement over the second.
Running quality mode with FG 4x upscaling produces an average of 246 FPS on the RTX 5080, a sizable 73% higher than the RTX 4090 (with FG 2x). Despite the high input framerate, we see considerable scaling (85%) over FG 2x, though the third frame’s contribution remains minimal (20%).
Switching to performance mode upscaling doesn’t grant any significant gains over quality. The RTX 5080 holds a 63% lead over the RTX 5090 with FG 4x. However, it’s only 70% faster than single frame and 20% faster than 3x frame generation.
Hogwarts Legacy
Hogwarts Legacy exhibits strong MFG scaling on the RTX 5080, averaging 133 FPS at the maximum quality setting with DLAA. It leads the RTX 4090 (FG 2x) by 75% with FG 4x and 35% with FG 3x. We see a near-doubling of frames over single frame generation.
NVIDIA RTX 5080: Native & Upscaling Benchmarks
The GeForce RTX 5080 falls behind the RTX 4090 without its magic weapon. At 4K “Ultra” the Lovelace flagship is 10% faster than the 5080 in Alan Wake 2. Interestingly, this deficit increases to over 20% with ray-tracing.
The RTX 4090 is 20-25% faster than the 5080 in path traced benchmarks. Luckily, performance mode upscaling reduces this delta to under 20%.
Cyberpunk 2077 exhibits similar behavior. The RTX 5080 manages to one-up the RTX 4090 at 4K “Ultra” without ray-tracing, but falls behind by 20% as soon as it’s enabled.
The RTX 4090 remains 10-20% ahead of the RTX 5080 in path tracing, with performance upscaling reducing to nearly 10% at 4K.
The GeForce RTX 4090 is over 20% faster than the RTX 5080 in Star Wars: Outlaws across all three scenarios included below.
Marvel Rivals which uses Lumen GI maintains a 10-20% deficit between the RTX 4090 and the RTX 5080 at 4K “Ultra.”
At 4K native, the GeForce RTX 5080 and 4090 perform within 5% of one another in Hogwarts Legacy, both with and without ray-tracing.
Warhammer 40K: Space Marine is a bit of an anomalie. The RTX 4090 holds a 16% lead over the RTX 5080 at 4K “Ultra” with DLAA which shrinks to low single digit percentages at quality upscaling mode. Performance upscaling places the 5080 in the lead and that doesn’t change with frame generation thrown into the mix.
A Plague Tale: Requiem behaves similarly. The RTX 4080 leads by 12-13% at 4K Ultra (native), but this lead shrinks to 7% with quality and 1% with performance upscaling.
The RTX 5080 manages to beat the 4090 with single frame generation and performance upscaling enabled.
It’s the same story with Jedi Survivor, except here the RTX 4090 continues to hold a larger lead with frame generation enabled.
Black Myth: Wukong sees the RTX 4090 lead by 5-10% at 4K with Path Tracing maxed out. The deficit remains consistent across different upscaling and frame generation modes.
Power Consumption and Thermals
The GeForce RTX 5080 sipped between 320-340W of power in most titles. The power draw increased to 350-360W in path traced games, especially when tested without frame generation. Older and Lumen-based titles like Marvel Rivals used less than 300W, even at native 4K “Ultra.” The thermals were likewise very manageable. The GPU core recorded an average temperature of 68-72C, with peaks of 77-79C in intensive scenes.
Conclusion and Verdict
The GeForce RTX 5080 is not what we expected. In terms of sheer rasterization, it performs quite close to the RTX 4090, making it 15-20% faster than the RTX 4080 Super. Ray traced workloads see a larger deficit, often times higher than 20%, making it just over 10% faster than the 4080 Super. Then there’s Multi Frame Generation which can be a game-changer if implemented well.
At the time of review, we were limited to half a dozen MFG titles. The latest driver update enables MFG across 75 games, including Cyberpunk 2077, Stalker 2, Alan Wake 2, Flight Sim 2024, Final Fantasy XVI, Hellblade 2, Silent Hill 2, etc. However, there are numerious titles that still lack the technology, and unless adoption becomes more widespread, it will remain in an early access of sorts.
There’s also the matter of compeititive gaming where frame interpolation can’t be used, and of course, personal preference. There are quite a few people who can’t tolerate the input lag induced by MFG, especially folks on monitors with sub-1 ms response times.
Overall, the RTX 5080 is more of a refresh rather than a generational upgrade. You get slightly more cores, and faster memory which make it marginally faster than the RTX 4080 Super. Luckily, it’s also priced the same. As more games with MFG are release, we might see a larger interest in the RTX 5080 but for now, it’s only a valid upgrade path for RTX 20 and 30 series owners.
