RTX 4080 vs RTX 5090
These two are relatives, not rivals. The RTX 4080 and the RTX 5090 share a lineup, so the question is which version fits you. If you only read one line, make it this: the RTX 5090 takes 2 of 3 compared areas, and the gpu & performance gap is the one to look at first. Price is part of the story too: the RTX 4080 launched at $800 less than the RTX 5090.
At a glance
- GPU & Performance RTX 5090 wins
- Memory RTX 5090 wins
- Power & Thermals RTX 4080 wins
- Price RTX 4080 wins
- GPU & Performance3.9
- Memory1.5
- Power & Thermals7.6
- GPU & Performance9.7
- Memory10.0
- Power & Thermals0.0
Spec scores (0 to 10) are calculated from manufacturer specs against other models in the same class. See how the formula works. They are not reviews or user ratings.
Key Differences
-
The RTX 5090 has 115% more theoretical FP32 compute
When it comes to fp32 compute, the RTX 5090 wins comfortably with 104.9 TFLOPS against 48.7 TFLOPS. FP32 compute is a theoretical peak that combines shader count and clock speed, a quick way to size up raw power.
-
The RTX 5090 has 16 GB more VRAM
It’s a lopsided result (32 GB for the RTX 5090, 16 GB for the RTX 4080) and one of the main reasons to choose between them. Newer games tend to ask for more video memory, so the RTX 5090’s extra VRAM helps the card age better.
-
The RTX 5090 has 124% more shader units
The spread here is large: 21,760 on the RTX 5090 versus 9,728 on the RTX 4080. Expect to notice it. Shader units do most of the heavy lifting in games, so more of them usually means more frames per second within the same architecture.
-
The RTX 4080 has 255 W lower board power
Of all the specs we compared, this is one where the RTX 4080 makes its strongest case: 320 W against 575 W. Lower board power means less heat, quieter cooling and a smaller impact on your electricity bill.
-
The RTX 5090 has 150% more memory bandwidth
The RTX 5090 clearly leads on memory bandwidth, and 1,792 GB/s versus 717 GB/s is a difference that reaches well beyond the spec sheet. Bandwidth needs vary by architecture, so compare it alongside VRAM and cache rather than on its own.
Design & Size
- RTX 4080 11.97 in × 3 slots
- RTX 5090 11.97 in × 2 slots
Their dimensions are close enough that size should be one of the last things you weigh.
Size & Connectivity
| Spec | RTX 4080 | RTX 5090 |
|---|---|---|
| Reference card length | 11.97 in | 11.97 in |
| Slot width | 3 slots | 2 slots |
| Display outputs | 3x DisplayPort 1.4a, 1x HDMI 2.1 | 3x DisplayPort 2.1b, 1x HDMI 2.1b |
GPU & Performance
RTX 5090 leads
In the gpu & performance category, the RTX 5090 isn’t just ahead; it’s clearly ahead. The RTX 5090 has the edge with 115% more theoretical FP32 compute. The RTX 5090 has more theoretical compute, which often, but not always, translates into higher frame rates. Choose the RTX 4080 and you get. If this section decides it for you, choose the RTX 5090.
FP32 compute
RTX 4080: 48.7 TFLOPS RTX 5090: 104.9 TFLOPS Architectures use their compute differently, so treat the gap as a guide and look at game benchmarks for the details.
Shader units
RTX 4080: 9,728 RTX 5090: 21,760 Treat shader count as a strong hint about performance tier, then check benchmarks for the games you actually play.
Memory
RTX 5090 leads
The memory numbers point firmly in one direction: toward the RTX 5090. The RTX 5090 comes with 16 GB more VRAM. At high resolutions or with texture packs, running out of VRAM can hurt more than a small speed gap. For anyone who cares most about memory, the RTX 5090 is the safer choice.
VRAM
RTX 4080: 16 GB RTX 5090: 32 GB More VRAM lets the card hold higher-resolution textures and bigger scenes without stutter.
Memory bandwidth
RTX 4080: 717 GB/s RTX 5090: 1,792 GB/s Memory bandwidth decides how quickly the GPU can feed itself with data, which matters most at 1440p and 4K.
Features
Nothing to tally in the features section, but plenty to read. The table lists what the RTX 4080 and the RTX 5090 each use.
Power & Thermals
RTX 4080 leads
Here the RTX 4080 opens up one of its biggest leads in the comparison. The RTX 4080 takes this one with 255 W lower board power. Higher board power often comes with higher performance, so the RTX 5090’s appetite may be part of the price of its speed. Anyone shopping mainly on power & thermals should start with the RTX 4080.
Total board power
RTX 4080: 320 W RTX 5090: 575 W The RTX 4080 draws less power, which makes it easier to cool in compact cases.
Recommended PSU
RTX 4080: 750 W RTX 5090: 1,000 W The recommended power supply tells you whether your current PSU can handle the card without an upgrade.
Full Specifications
GPU & Performance
| Spec | RTX 4080 | RTX 5090 |
|---|---|---|
| Architecture | Ada Lovelace | Blackwell |
| GPU die | AD103 | GB202 |
| Process | TSMC 4N | TSMC 4N |
| Shader units | 9,728 | 21,760 (winner) Most shader units |
| Ray tracing units | 76 | 170 (winner) Most ray tracing units |
| Base clock | 2,205 MHz (winner) | 2,010 MHz |
| Boost clock | 2,505 MHz (winner) | 2,410 MHz |
| FP32 computeTheoretical peak; AMD counts dual-issue, so compare within a brand only | 48.7 TFLOPS | 104.9 TFLOPS (winner) Most compute in class |
Memory
| Spec | RTX 4080 | RTX 5090 |
|---|---|---|
| VRAM | 16 GB | 32 GB (winner) Most VRAM in class |
| Memory type | GDDR6X | GDDR7 |
| Memory bus | 256-bit | 512-bit (winner) Widest memory bus |
| Memory bandwidth | 717 GB/s | 1,792 GB/s (winner) Most memory bandwidth |
Features
| Spec | RTX 4080 | RTX 5090 |
|---|---|---|
| Upscaling | DLSS Super Resolution | DLSS 4 Super Resolution |
| Frame generation | DLSS Frame Generation (2x) | DLSS 4 Multi Frame Generation (up to 4x) |
| PCIe interface | PCIe 4.0 x16 | PCIe 5.0 x16 |
Power & Thermals
| Spec | RTX 4080 | RTX 5090 |
|---|---|---|
| Total board power | 320 W (winner) | 575 W Most power-hungry in class |
| Recommended PSU | 750 W (winner) | 1,000 W Biggest PSU requirement |
| Power connector | 1x 16-pin (12VHPWR) | 1x 16-pin (12V-2x6) |
Size & Connectivity
| Spec | RTX 4080 | RTX 5090 |
|---|---|---|
| Reference card length | 11.97 in | 11.97 in |
| Slot width | 3 slots | 2 slots |
| Display outputs | 3x DisplayPort 1.4a, 1x HDMI 2.1 | 3x DisplayPort 2.1b, 1x HDMI 2.1b |
Price
| Spec | RTX 4080 | RTX 5090 |
|---|---|---|
| Launch price (MSRP) | $1,199 (winner) | $1,999 Priciest in class |
“N/A” means the manufacturer does not specify the value or we could not confirm it. Figures marked “manufacturer rating” are the maker’s own claims, not our measurements.
Price & Value
Compare the launch prices and they sit 67% apart, or $800 in real money. The RTX 5090 costs $800 more, and for that you get 115% more theoretical FP32 compute and 16 GB more VRAM.
Launch prices in the US as of January 2025; street prices move, so check before you buy.
Which One Should You Buy?
Choose the RTX 4080 if…
- you care about your power bill and a quieter case (320 W vs 575 W)
- you want to keep your current power supply (750 W vs 1,000 W recommended PSU)
- you want the higher guaranteed base clock (2,205 MHz vs 2,010 MHz)
- you care about peak clock speed (2,505 MHz vs 2,410 MHz boost)
- you don’t need the extras that cost $800 more
Choose the RTX 5090 if…
- you run compute-heavy work alongside gaming (104.9 TFLOPS vs 48.7 TFLOPS FP32)
- you play at high resolutions with high-quality textures (32 GB vs 16 GB VRAM)
- you want more raw GPU hardware (21,760 vs 9,728 shader units)
- you want more memory bandwidth (1,792 GB/s vs 717 GB/s)
- you find that the improvements in the RTX 5090 match what you need
Also Compared
Frequently Asked Questions
Should I buy the RTX 4080 or the RTX 5090?
The RTX 5090 wins more of the areas we compared, so it’s the safer pick unless you care more about 255 W lower board power than anything else.
Is the RTX 5090 worth it over the RTX 4080?
It’s one of the more worthwhile generational jumps: the RTX 5090 offers 115% more theoretical FP32 compute, 16 GB more VRAM and 124% more shader units compared with the RTX 4080.
Is the RTX 4080 bigger than the RTX 5090?
Neither by much. The RTX 4080 and the RTX 5090 have nearly identical dimensions.
Does the RTX 4080 perform better than the RTX 5090?
The RTX 5090 has the edge in performance thanks to 115% more theoretical FP32 compute and 124% more shader units.
Is the RTX 4080 cheaper than the RTX 5090?
Going by launch prices, the RTX 4080 costs $800 less than the RTX 5090. Prices as of January 2025; current offers may differ.
Executive Summary
If you’ve settled on this lineup, the RTX 4080 or the RTX 5090 is the last call to make. Here’s mine. Practically speaking, the RTX 5090’s biggest advantage is 115% more theoretical FP32 compute. Add 16 GB more VRAM to the list, and the RTX 5090 looks even better. I’ll admit the RTX 4080 has a point with 255 W lower board power.
Think about your week. Heavy on power & thermals? The RTX 4080. More about gpu & performance? The RTX 5090.
My pick: the RTX 5090 is the sensible default. Only one thing changes that: if you’d rather have 255 W lower board power on the RTX 4080.
Personal opinion based on the data above. Your priorities may differ.