Key takeaways
- The RTX 4090 leads native 4K raster by about 20% in Tom’s Hardware January 2025 averages. TechPowerUp puts the 5080 at about 89% of its pace.
- The 4090 has 24GB of VRAM against 16GB on the 5080, so it suits large textures and local AI work better.
- The 5080 adds DLSS 4 with Multi Frame Generation, which the 4090 lacks, and it draws 90 W less.
- At 4K the GPU sets the limit, so almost any modern six-core CPU feeds either card.
- The 4090 was discontinued in October 2024, while the 5080 is still in production at a 999 dollar launch price.
RTX 4090 vs RTX 5080: the short answer
For the RTX 4090 vs RTX 5080 bottleneck, the 4090 is faster in native games, and the 5080 is newer with better features. Tom’s Hardware measured the 4090 at 84.4 frames per second against 70.1 for the 5080 at 4K native raster. That is about 20% ahead, and TechPowerUp found a similar gap. For the tier below, see this: RTX 4080 vs RTX 5080 bottleneck compared. Check your own pair with the free tool: pc bottleneck calculator.
| Question | RTX 4090 | RTX 5080 |
|---|---|---|
| Faster in native games | Yes, about 20% at 4K | No |
| More VRAM | Yes, 24GB | No, 16GB |
| Newer features | No, DLSS 3 | Yes, DLSS 4 |
| Lower power | No, 450 W | Yes, 360 W |
| Still in production | No, ended October 2024 | Yes |
| Best for | Native 4K and AI work | New builds, DLSS 4 |
RTX 4090 vs RTX 5080 specs compared
A bottleneck is the part that finishes its work last, and it sets the frame rate. The 4090 is an Ada Lovelace card from October 2022, built on the large AD102 die. The 5080 is a Blackwell card from January 2025, built on the smaller GB203 die. The newer card is more efficient, yet the older card has far more silicon.
The 4090 has 16,384 CUDA cores against 10,752 on the 5080, which is 52% more. Its 384-bit bus gives 1,008 GB/s of bandwidth, against 960 GB/s on the 256-bit bus of the 5080. The 5080 answers with faster GDDR7 memory, fourth-generation ray tracing cores, and DLSS 4.
| Spec | RTX 4090 | RTX 5080 |
|---|---|---|
| Architecture | Ada Lovelace, AD102 | Blackwell, GB203 |
| CUDA cores | 16,384 | 10,752 |
| VRAM | 24GB GDDR6X | 16GB GDDR7 |
| Memory bus | 384-bit | 256-bit |
| Bandwidth | 1,008 GB/s | 960 GB/s |
| Boost clock | 2.52 GHz | 2.62 GHz |
| Power draw | 450 W | 360 W |
| Interface | PCIe 4.0 x16 | PCIe 5.0 x16 |
| Launch | October 2022, 1,599 dollars | January 2025, 999 dollars |

RTX 4090 vs RTX 5080 Bottleneck: Native Performance
Native raster means no upscaling and no frame generation, so it shows the raw hardware. Tom’s Hardware January 2025 averages put the 4090 at 84.4 frames per second and the 5080 at 70.1 at 4K ultra. That is a 20% lead for the 4090 in that suite. TechPowerUp found the 5080 at about 89% of the 4090 pace at 4K, which is an 11% gap.
The honest range is an estimated 10% to 20% at 4K, and it varies by game. At 1440p the gap is smaller, since both cards run fast and the CPU can limit them. In heavy ray-traced games the gap grows, since the 4090 has more cores and a wider bus. Use these figures as named estimates, not as our own tests.
VRAM: 24GB against 16GB
VRAM is the memory on the graphics card, and it holds textures, models, and frame data. When it fills up, the driver spills data to system RAM, and that spill causes stutter. The 4090 has 24GB, so it holds the largest texture packs and local AI models without spilling. The 5080 has 16GB, which is enough for most games today.
This gap matters most outside gaming. Local AI work, large 3D scenes, and heavy texture packs use far more than 16GB. For gaming alone at 4K, 16GB is enough in most titles today, yet 24GB has more room for future games. The 4090 is the clear pick when memory is the real limit.

DLSS and frame generation: where the 5080 fights back
DLSS is NVIDIA upscaling, and it renders at a lower resolution, then rebuilds the image. Frame generation adds extra frames between real ones, which lifts the shown frame rate. The 4090 supports DLSS 3 with single frame generation. The 5080 supports DLSS 4 with Multi Frame Generation. It adds up to three extra frames.
With Multi Frame Generation on, the 5080 can show a higher frame rate than the 4090 in supported games. That shown rate is not the same as rendered speed, since generated frames add latency. For competitive play, native frames matter more than generated ones. For story games, the smoother shown rate can feel better.
For the tier above the 5070, see this: RTX 4070 Super vs RTX 5070 Ti bottleneck compared. The lesson is the same there: features close a gap that raw speed opens.
RTX 4090 vs RTX 5080 Bottleneck: CPU Needs
A CPU bottleneck is when the processor finishes last, and it caps the frame rate. At 4K the GPU sets the limit for both cards, so almost any modern six-core CPU feeds either one. TechSpot paired the Ryzen 7 9800X3D with a six-core Ryzen 5 7600X on an RTX 5080, and the two posted identical numbers at 4K ultra.
At 1440p ultra the 9800X3D kept about a 9% lead in that TechSpot run, and the gap grew in CPU-heavy games. At 1080p the CPU matters most, since the cards finish frames fast. A Ryzen 7 9800X3D is the top pick for high-refresh play, while a six-core chip is enough for 4K. For the best CPU for the older card, see this: best CPU for RTX 4090.
What about 1080p and esports?
At 1080p both cards are far stronger than the resolution needs, so the CPU decides most of the result. Frame rates run very high, and a faster CPU shows a bigger gain than a faster card. Neither card is a sensible 1080p buy on its own, since cheaper cards already clear high refresh rates there.
Esports titles are the clearest case. They run at low settings, so the GPU finishes fast and the CPU carries the load. A Ryzen 7 9800X3D matters more than the choice between these two cards. Buy the CPU first for esports, and spend the card money where it shows.
The cards earn their price at 1440p high refresh and at 4K. That is where the GPU sets the limit, and the gap between them shows in the frame rate. If your monitor is 1080p, keep your current card and upgrade the monitor first.
Price, power, and availability
The 4090 launched at 1,599 dollars, and NVIDIA stopped building it in October 2024. It now sells only on the used market, where prices sit well above the launch figure. The 5080 launched at 999 dollars, and it is still in production. That is 600 dollars less at launch, for a card that is close in most games.
Power favours the 5080 too. It draws 360 W against 450 W for the 4090, so it needs a smaller power supply and runs cooler. A good 850 W unit suits the 5080, while the 4090 wants a 1,000 W unit as a rule of thumb. For the CPU picks for the newer card, see this: best CPU for RTX 5070 Ti and 5080.
Which flagship should you buy?
Buy the RTX 5080 if you are building new, since it is in production, costs less at launch, and adds DLSS 4. Buy a used RTX 4090 only if native speed and 24GB of VRAM matter more than warranty and features. Check the used price first, since a high used price erases the 4090 advantage.
For the full 5080 guide, see this: RTX 5080 bottleneck guide. Then check your exact pair before you buy, since your CPU and resolution decide how much of the gap you will see.
The bottom line on RTX 4090 vs RTX 5080
The RTX 4090 vs RTX 5080 bottleneck ends with native speed on the 4090 side, by an estimated 10% to 20% at 4K. The RTX 5080 is the smarter new buy. Speed and memory favour the 4090, while features, power, and price favour the 5080. Check your pair first on the cpu bottleneck calculator before you spend on new parts.
Frequently asked questions
Is the RTX 4090 still faster than the RTX 5080?
Yes, in native raster games. Tom’s Hardware January 2025 averages put the 4090 about 20% ahead at 4K ultra, and TechPowerUp found the 5080 at about 89% of its pace. The honest range is an estimated 10% to 20% at 4K. With DLSS 4 Multi Frame Generation on, the 5080 can show a higher frame rate.
Is a used RTX 4090 a better buy than a new RTX 5080?
Buy a new RTX 5080 for most builds. It is in production, it launched at 999 dollars, and it adds DLSS 4 with Multi Frame Generation. Buy a used 4090 only if native speed and 24GB of VRAM matter more than warranty. Check the used price first, since a high price erases the advantage.
Does 16GB of VRAM limit the RTX 5080 at 4K?
Yes, for most games at 1440p and 4K today. The 5080 handles current titles well with 16GB of GDDR7 memory. The 4090 has 24GB, which suits larger texture packs, local AI work, and heavy 3D scenes. For gaming alone, 16GB is enough now, yet 24GB has more room for future games.
Is a Ryzen 7 9800X3D worth it with an RTX 5080?
No, not at 4K. A modern six-core CPU feeds the 5080 there, since the GPU sets the limit. TechSpot found a six-core Ryzen 5 7600X equal to the Ryzen 7 9800X3D at 4K ultra on a 5080. At 1440p ultra the 9800X3D kept about a 9% lead, and at 1080p the CPU matters most.
Does the RTX 5080 beat the RTX 4090 with DLSS 4?
It can show a higher frame rate, but not a higher rendered speed. DLSS 4 Multi Frame Generation adds up to three extra frames between real ones, so the shown rate rises. Those generated frames add latency, so they help story games more than competitive play. Native speed still favours the 4090.
Why was the RTX 4090 discontinued?
NVIDIA stopped building the RTX 4090 in October 2024, as the Blackwell generation took over. The 5080 replaced it as the new high-end card, at a 999 dollar launch price. The 4090 now sells only on the used market, where prices sit well above its 1,599 dollar launch figure.
How this guide was prepared
This comparison uses the published specs of NVIDIA. It also uses the Tom’s Hardware January 2025 review averages, the TechPowerUp RTX 5080 review, and the TechSpot CPU and GPU scaling run. The performance gaps are the named sources’ figures, labelled as estimates, not our own tests. The prices shown are the makers’ launch figures.
Last updated: November 14, 2026
