💪 GPU Battle #39 RTX 5080 vs RTX 3090 💪
GPU Battle #39 — RTX 5080 vs RTX 3090
The RTX 5080 is NVIDIA’s 2025 upper-high-end Blackwell card — 16 GB GDDR7, DLSS 4, and a $999 MSRP. The RTX 3090 is NVIDIA’s 2020 Ampere flagship — 24 GB GDDR6X, DLSS 2, and the card that once defined 8K gaming. One is new with the latest features; the other is a used former flagship with a massive VRAM advantage. The comparison is closer than the tier gap suggests — the 5080 is much faster in raw performance, but the 3090’s 24 GB of VRAM makes it uniquely valuable for AI and content creation.
Short version: The RTX 5080 is significantly faster in raster and RT, with DLSS 4 and lower power draw. The RTX 3090 is slower but has 24 GB of VRAM for AI and 4K workloads. For gaming, the 5080 wins. For AI on a budget, the 3090 is still compelling.
Specs
The RTX 5080 is Blackwell on TSMC 4N — 10,752 CUDA cores, 16 GB GDDR7 on a 256-bit bus, 360W TGP, PCIe 5.0. It supports DLSS 4 with multi-frame generation, and its RT hardware is fourth-gen.
The RTX 3090 is Ampere on Samsung 8nm — 10,496 CUDA cores, 24 GB GDDR6X on a 384-bit bus, 350W TGP, PCIe 4.0. It supports DLSS 2 with no frame generation, and its RT hardware is second-gen.
| Spec | RTX 5080 | RTX 3090 |
|---|---|---|
| Architecture | Blackwell | Ampere |
| CUDA cores | 10,752 | 10,496 |
| VRAM | 16 GB GDDR7 | 24 GB GDDR6X |
| Bus | 256-bit | 384-bit |
| TGP | 360W | 350W |
| PCIe | 5.0 | 4.0 |
| Upscaling | DLSS 4 (multi-frame) | DLSS 2 |
| RT hardware | 4th gen | 2nd gen |
| Launch price | $999 | $1,499 |
The two cards have similar CUDA core counts, but the 5080 is four generations newer. The 3090 has 50% more VRAM and a wider bus.
Rasterization
In raster, the RTX 5080 is roughly 60–80% faster than the RTX 3090 at 4K. Its newer architecture, GDDR7 memory, and much higher IPC give it a huge lead. At 1440p, the gap is similar.
The 3090 is still a capable 4K card by 2020 standards, but it’s been surpassed significantly. For pure raster, the 5080 wins by a wide margin.
Ray tracing
Ray tracing favors the 5080 even more. Its fourth-gen RT cores are dramatically faster than Ampere’s second-gen, and the gap is often 2–3x in RT-heavy titles. In path-traced games, the 5080 can run natively at playable framerates while the 3090 needs aggressive upscaling — and still struggles.
For ray tracing, the 5080 is in a different class.
VRAM and memory
This is the 3090’s one advantage. It has 24 GB; the 5080 has 16 GB. For AI workloads — running large language models, training, inference — the 3090’s extra VRAM is significant. It can hold larger models and datasets in memory.
For gaming, 16 GB is plenty at 1440p and adequate at 4K. The 3090’s extra VRAM rarely matters in games, but it matters a lot for AI and professional work.
For AI on a budget, the 3090 remains one of the best options — 24 GB for $700–900 used is unbeatable.
Power and thermals
Both cards draw around 350–360W. Efficiency is comparable in raw power, but the 5080 delivers much more performance for the same draw — so it’s effectively far more efficient per frame.
Both need good cooling and a 750W+ PSU. Neither is a low-power card.
Features
The 5080 supports DLSS 4 with multi-frame generation, DisplayPort 2.1b, and newer NVENC encoders. DLSS 4 can generate up to three additional frames per rendered frame, dramatically improving perceived smoothness in supported games.
The 3090 supports DLSS 2 with no frame generation — a major feature gap. It lacks DisplayPort 2.1 and newer media features. Its driver support continues but won’t gain new features.
For features, the 5080 is far ahead.
Price and value
New at $999 (if you can find it at MSRP), the 5080 is the more affordable card. But availability has been tight, and street prices often exceed MSRP.
Used at $700–900, the 3090 is cheaper and offers 24 GB of VRAM. For AI workloads, that makes it uniquely valuable at this price point.
For gaming, the 5080 is the better value. For AI, the 3090’s VRAM makes it the smarter buy.
The verdict
The RTX 5080 is the faster card in every gaming workload — 60–80% ahead in raster, 2–3x in RT. For gaming at 1440p or 4K, it’s the better card by a wide margin.
The RTX 3090 is the AI value pick. Its 24 GB of VRAM makes it uniquely capable for running large models at a price no other card can match. For content creation with large datasets, it’s still relevant.
The decision comes down to workload. For gaming, the 5080. For AI on a budget, the 3090.
Short version:
Choose the RTX 5080 if you’re gaming, want DLSS 4, and want the fastest card in its price class.
Choose the RTX 3090 if you need 24 GB of VRAM for AI or content creation and want the best value per gigabyte of VRAM.
Stop using slow, ad-bloated tool sites! 🤮
🔎 Search “KandZ Tools” on Google to use many professional utilities for free.
KandZ.me is the ultimate minimalist hub for:
✅ Finance (Mortgage, Interest, Inflation)
✅ Tech (Base64, JSON, Dev Suite, IP)
✅ Health (BMI, BMR, TDEE)
✅ Productivity (Timer, Workspace, QR)
⚡️ Fast & Private
🔒 No data leaves your device
💎 100% Free
🔗 Use it now: https://tools.kandz.me
🔖 Bookmark it—you’ll need it later!