RTX 3090 vs RTX 4070
The complete head-to-head: full specs, real per-game FPS with 1% lows, ray tracing, a live CPU-bottleneck check, thermals, power and value, with a straight verdict.
GeForce RTX 3090
GeForce RTX 4070
avg, 1080p
The short answer
The RTX 3090 is still the faster card here, about 17% ahead on average, and it doubles the 4070's VRAM at 24 GB against 12 GB. That surprises a lot of people who assume the newer card always wins. The 4070 answers back hard on every other front: roughly a third less power draw, a much lower price both new and used, and DLSS 4 upscaling plus AV1 encoding that the 3090 simply does not have. Buy the 3090 if you need the extra VRAM or raw speed and don't mind the power bill. Buy the 4070 for a modern, efficient, better-value card that is still genuinely fast.
This is a genuinely useful comparison because it does not go the way most people expect. The RTX 3090 was a $1,499 flagship, and even against a current-generation card two tiers below its original class, it still wins on raw speed and holds double the VRAM. The RTX 4070 fights back with efficiency, price, and every modern NVIDIA feature the 3090 launched without.
The data below covers the full spec sheet, real frame rates with 1% lows, ray tracing and DLSS behavior, a live CPU-bottleneck check, thermals, power draw and price history, all built to answer whether the 3090's 17% speed lead and 24GB of VRAM are worth roughly $220 more and 150 extra watts than the 4070.
At a glance
Seven dimensions, two cards. Bigger area is not automatically better, each card owns different corners.
Strengths radar
Each axis scored 0-100 relative to the pair
Performance vs price
Up and to the left is better value
Grey dots are neighbouring cards for context.
Which should you buy?
Pick your use case, the winner changes with what you actually do.
24 GB gives real headroom for texture mods, high-res modding, and 4K max settings that 12 GB can run tight on.
Still a strong 1440p card, at roughly half the wattage and a third of the price of the 3090.
24 GB versus 12 GB is decisive for larger local models and batch sizes, capacity wins over raw speed here.
AV1 hardware encoding and DLSS 4 come standard on the 4070, the 3090 has neither.
At 200 W the 4070 fits a 650 W PSU and a much smaller cooler, the 3090 needs 750 W and a lot more airflow.
936 GB/s of bandwidth and 24 GB of headroom give the 3090 the edge in the most demanding emulation cores.
Already own a 3090? The 4070 is a sidegrade, not an upgrade.
Short answer: probably not
Moving from a 3090 to a 4070 would cost you about 17% of your frame rate and half of your VRAM, in exchange for lower power draw and newer features. If your 3090 is running fine and 24 GB matters to you at all, there is no performance case for switching.
If efficiency and modern features are the real draw, look at an RTX 4070 Super or 4070 Ti Super instead, both close most of the speed gap to the 3090 while keeping the same low power draw and DLSS 4 support. A straight 4070 alone is a reasonable buy only if you are shopping fresh, not trading down from a 3090.
Full specification comparison
Every spec side by side. The highlighted figure wins each row.
| Specification | RTX 3090 | RTX 4070 |
|---|---|---|
| Architecture | ||
| GPU / architecture | GA102 · Ampere | AD104 · Ada |
| Process node | Samsung 8 nm | TSMC 4N (5 nm) ▲ |
| Release date | Sep 24, 2020 | Apr 13, 2023 |
| Launch MSRP | $1,499 | $599 ▲ |
| Compute | ||
| CUDA cores | 10,496 ▲ | 5,888 |
| RT cores | 82 · 2nd gen ▲ | 46 · 3rd gen |
| Tensor cores | 328 · 3rd gen ▲ | 184 · 4th gen |
| Base / boost clock | 1395 / 1695 MHz | 1920 / 2475 MHz ▲ |
| L2 cache | 6 MB | 36 MB ▲ |
| Memory | ||
| VRAM | 24 GB GDDR6X ▲ | 12 GB GDDR6X |
| Memory bus | 384-bit ▲ | 192-bit |
| Bandwidth | 936 GB/s ▲ | 504 GB/s |
| Power & form factor | ||
| TDP | 350 W | 200 W ▲ |
| Power connector | 2× 8-pin | 1× 16-pin (12VHPWR) |
| Suggested PSU | 750 W | 650 W ▲ |
| PCIe interface | 4.0 ×16 | 4.0 ×16 |
| Features | ||
| Upscaling | DLSS 2 / Super Res | DLSS 4 + Frame Gen ▲ |
| AV1 encode (NVENC) | No | Yes ▲ |
| Display outputs | 3×DP1.4a · HDMI 2.1 | 3×DP1.4a · HDMI 2.1 |
Average FPS by game
Switch resolution, metric, ray tracing and upscaling, the bars update live. Hover a bar for the avg / 1% / 0.1% breakdown.
Average FPS · 1080p · Rasterization. Representative data, hover a bar for avg / 1% / 0.1%. Note the RTX 4070's 1% lows fall off at 1440p and up from its 8 GB VRAM.
Smoothness & frametimes
Average FPS hides stutter. This is one run plotted as frametime, a flat line is smooth, spikes are hitches. Watch what the RTX 4070's 8 GB VRAM does in a texture-heavy game.
Frametime: Alan Wake 2, 1440p Ultra + Ray Tracing
Lower and flatter is better · 16.7 ms = 60 fps
Counterintuitive but real: the slower RTX 3090 delivers a smoother line here because 12 GB holds the texture pool the 8 GB RTX 4070 keeps evicting, those spikes are VRAM stutter, not raw speed.
Lab scores
Clean apples-to-apples tests, useful, but weaker at predicting real games than the FPS above.
Ray tracing & upscaling
Ray tracing tanks both cards at native resolution. What makes it playable is upscaling, and here the RTX 4070 has a real edge.
Is frame generation "real" performance?
The RTX 3090 predates DLSS 3 entirely, Ampere has no Frame Generation hardware, so every frame it shows you is fully rendered. The RTX 4070 supports DLSS 3 single Frame Generation, inserting an AI-made frame between two real ones to smooth motion and nearly double the counter in supported titles. Those inserted frames do not react to input, so latency rises a little, NVIDIA Reflex offsets most of it.
Use Frame Generation on the 4070 for single-player and cinematic games, and skip it for competitive shooters or when your base frame rate drops under about 50 fps. It only works in DLSS 3 supported titles, and the 3090 has no equivalent at all, though its higher native frame rate often narrows the practical gap anyway.
Will your CPU bottleneck these cards?
A faster GPU only helps if your CPU keeps up. Pick your CPU and resolution to see expected FPS and bottleneck for each card.
CPU + GPU pairing check
The willitbottleneck engine, built right into the comparison.
Estimated from a 43-CPU model across 1440p. A weak CPU mostly bites at 1080p; at higher resolutions the GPU becomes the limiter and the bottleneck shrinks.
Thermals & noise
The efficiency gap shows up here: the newer card runs cooler and quieter, day in and day out.
Real draw, efficiency & PSU
TDP is the rated figure; what matters is real gaming draw and how many frames you get per watt.
Both use a single 8-pin connector. If your PSU already ran the RTX 3090, it will run the RTX 4070 easily. Leave about 30% headroom for spikes.
Will it fit, and what to pair it with?
Size to scale
Card length matters for small cases
What monitor to pair
RTX 3090: 1440p at 144 Hz and up with room to spare, and a strong 4K card at 60-100 Hz. RTX 4070: 1440p at 100-165 Hz, and a genuinely capable 4K card with DLSS in most current titles. Both are comfortable well past 1080p.
Running an older PCIe 3.0 board?
Both cards run a full PCIe 4.0 ×16 link, so neither loses meaningful performance even on an older PCIe 3.0 board, under 5% either way. Neither supports PCIe 5.0, but at this performance tier that would not matter yet.
Price history & dollars-per-frame
Raw price is only half the story. Cost per frame is the number that decides it, and it keeps moving as prices fall.
New price over time
Street price of new cards (USD)
The 24 GB vs 12 GB question
RTX 3090 · 24 GB
- 24 GB is double the 4070's 12 GB, decisive for large local LLMs, big Stable Diffusion batches, and heavy 3D scenes.
- 936 GB/s of bandwidth backs that capacity up, nothing here is starved for data.
- For any workload where 12 GB is genuinely tight, the 3090 has no real equal at this price.
- The compute processing that memory is a generation behind, so raw throughput per GB still favors the 4070 in lighter tasks.
RTX 4070 · 12 GB
- 12 GB comfortably covers 1440p gaming and most creator workloads through 2026.
- 36 MB of L2 cache and DLSS 4 reduce effective memory pressure well below the raw 12 GB figure.
- Half the 3090's capacity is a real ceiling for the largest local LLMs and multi-GB diffusion checkpoints.
- The 192-bit bus and 504 GB/s bandwidth, while efficient, cannot match the 3090's raw data throughput in memory-bound tasks.
Emulation & memory bandwidth
The narrower bus is a real downgrade for emulation
The 3090's 936 GB/s over a 384-bit bus and its 24 GB of headroom make it the stronger pick for the most demanding high-resolution emulation, PS3, Wii U, and heavier Switch cores lean on exactly this kind of bandwidth. The 4070 remains capable at 1080p-1440p emulation, but the 3090 is the one with room to spare when internal resolution gets pushed hard.
Creator, AI & streaming
The newer card is quicker per task, but the older card's extra VRAM unlocks larger jobs it can't fit at all.
Streaming: the AV1 encoder is the differentiator
The RTX 4070 carries the same AV1 hardware encoder and DLSS 4 feature set found on the RTX 4090. The RTX 3090 predates AV1 entirely and is limited to H.264 and HEVC. For anyone who streams or exports AV1-aware content, the 4070 is the clear pick despite being the slower card overall in raw gaming terms.
Overclocking & undervolting
The 3090 has real undervolting headroom at its 350 W stock draw, worth doing before you consider an overclock. The 4070 is already efficient with less to give back, though its factory clocks leave modest room for a mild core overclock.
Pros & cons
RTX 3090
- 24 GB VRAM, double the 4070's capacity, for large AI and creator workloads
- About 17% faster on average, a strong 1440p and 4K performer
- Wide 384-bit bus and 936 GB/s bandwidth avoid data bottlenecks entirely
- 350 W draw needs a serious PSU and runs notably hotter and louder
- No DLSS 3 or 4 Frame Generation and no AV1 encoder, both of which the 4070 has
RTX 4070
- AV1 hardware encoding, DLSS 3 Frame Generation, and DLSS 4 upscaling
- 200 W draw, far cooler and quieter, and a lower-tier PSU is enough
- Costs roughly a third less both new and used, and wins clearly on cost per frame
- Still about 17% slower on average than the older, bigger 3090
- 12 GB VRAM is half the 3090's capacity, a real ceiling for the largest AI workloads
Buying pitfalls
The 3090's price premium isn't about speed alone
A used 3090 usually costs more than a 4070, and most of that premium buys VRAM capacity, not frame rate. If your workloads never come close to 12 GB, that extra money is not doing much for you.
The 4070's power connector needs the same care as bigger Ada cards
The 4070's 16-pin 12VHPWR connector should be pushed in until it clicks and kept free of sharp bends near the plug, the same guidance that applies to the 4080 and 4090, just at a lower power level.
RTX 3090 vs RTX 4070: your questions answered
On raw speed and memory, the 3090 does, it's about 17% faster on average with double the VRAM. The 4070 answers with better efficiency, a lower price, and modern features like DLSS 4 and AV1 encoding, so which one is 'better' depends on what you're optimizing for.
Pick the 3090 if 24 GB of VRAM matters for AI or creator work, or if raw speed is worth the higher power draw to you. Pick the 4070 for a modern, efficient, better-value card that still handles 1440p and 4K comfortably.
Yes. It still outruns the current-generation 4070 in raw performance and carries more VRAM than almost anything short of a 4090 or 5090. Age hasn't caught up with it yet.
No. You would lose about 17% of your frame rate and half your VRAM in exchange for lower power draw and newer features. If you want the 4070's efficiency without giving up speed, look at a 4070 Super instead.
The RTX 3090, on the strength of its 24 GB of VRAM, double the 4070's 12 GB. Capacity tends to be the limiting factor in local AI and large rendering jobs, and that's exactly where the older card pulls ahead.
The RTX 4070, because of its hardware AV1 encoder and DLSS 4 support, neither of which the 3090 has. AV1 means cleaner streams at a given bitrate and quicker exports in AV1-aware software.
It supports DLSS 4 upscaling and DLSS 3 single Frame Generation, but not the multi-frame generation exclusive to the RTX 50 series. The RTX 3090 predates DLSS 3 entirely and has no frame generation at all.
Surprisingly close. The 4070's newer 3rd-gen RT cores are efficient enough to match or slightly edge past the 3090 in RT-heavy synthetic tests, even though the 3090 leads comfortably in raw raster performance.
The 3090 is the faster card, so it is more likely to hit a CPU ceiling at 1080p. Use the pairing checker on this page, a Ryzen 7 7700X or Core i7-13700K class CPU and up keeps either card fed at 1440p and 4K.
The RTX 4070 wins clearly here: 200 W against the 3090's 350 W, which translates into noticeably lower temperatures and less fan noise under load. Both go silent at idle with fan-stop.
A quality 750 W unit covers the RTX 3090, while 650 W is enough for the RTX 4070. Both are common PSU sizes, but the 3090 leaves less headroom if you're also running a power-hungry CPU.
The 3090 runs around 313 mm and 3 slots, while the 4070 is a noticeably more compact 244 mm and 2 slots, so check your case clearance either way. On a PCIe 3.0 board, either card loses under 5%, not something to worry about.
No, the laptop RTX 4070 is a separate, notably weaker chip with a lower power budget. Don't compare laptop and desktop numbers directly, use the mobile version of this comparison instead.
Yes, both are capable 4K cards with DLSS help in the heaviest titles. The 3090's extra VRAM and bandwidth give it a slight edge in 4K texture-heavy and modded scenarios.
The RTX 3090 is used-market only at this point. The RTX 4070 can still turn up new in some markets, though used prices sit close enough that a well-tested used card with a return window is often the smarter buy for either one.
Yes, for years yet. Both architectures, Ampere and Ada, still get Game Ready driver support from NVIDIA, so neither card is close to end-of-life for a 2026 purchase.