We use cookies for ads and analytics to support our work. You can accept or reject these non-essential cookies. Cookie policy.
NVIDIA's refreshed 1440p flagship built on the Ada Lovelace AD104 die with 7168 CUDA cores, 12GB of GDDR6X on a 192-bit bus, third-gen RT cores and DLSS 3 Frame Generation — all in a compact 2-slot Founders Edition that draws just 220W.
Photos via GSMArena · click to expand · ←/→ keys or swipe to navigate
Quick Specs
A near-RTX-4070-Ti-fast 1440p card that lands DLSS 3 and Ada efficiency at the RTX 4070's old price.
Buy it for
Think twice if
Overall expert editorial score
| Architecture | Ada Lovelace |
|---|---|
| GPU Die | AD104 |
| Process Node | TSMC 4N (5 nm class) |
| Transistors | 35.8 billion |
| CUDA Cores | 7168 |
| RT Cores | 56 (3rd generation) |
| Tensor Cores | 224 (4th generation) |
| ROPs | 80 |
| Texture Units | 224 |
| Capacity | 12 GB |
|---|---|
| Type | GDDR6X |
| Memory Bus | 192-bit |
| Memory Speed | 21 Gbps |
| Bandwidth | 504.2 GB/s |
| L2 Cache | 48 MB |
| Base Clock | 1980 MHz |
|---|---|
| Boost Clock | 2475 MHz |
| FP32 (Shader) | 35.5 TFLOPS |
| RT Core Performance | 82.6 TFLOPS |
| Tensor Performance | 568 AI TOPS |
| Pixel Rate | 198 GPixel/s |
| Texture Rate | 554 GTexel/s |
| DisplayPort | 3× DisplayPort 1.4a |
|---|---|
| HDMI | 1× HDMI 2.1 |
| Max Resolution | 4K @ 240 Hz / 8K @ 60 Hz (DSC) |
| Max Displays | 4 |
| Encode/Decode | Dual AV1 NVENC, AV1 NVDEC |
| Total Graphics Power | 220 W |
|---|---|
| Recommended PSU | 650 W |
| Power Connector | 1× 16-pin (12VHPWR) |
| Adapter | 2× 8-pin PCIe to 16-pin |
| Cooler | Dual-axial flow-through (FE) |
| Interface | PCIe 4.0 ×16 |
| Length | 244 mm |
|---|---|
| Width | 112 mm |
| Slot Width | 2-slot (dual) |
| Weight | ~1020 g |
| Design | Founders Edition aluminium shroud |
Configuration
Across 4 retailers · prices update frequently
Where to buy
Prices and stock are accurate as of the last update. TechSpecs may earn a commission on purchases made through these links, at no extra cost to you.
The RTX 4070 SUPER is the card the original 4070 should have been. By widening the AD104 die to 7168 CUDA cores, NVIDIA closes most of the gap to the 4070 Ti while keeping the 220W power envelope and a tidy 2-slot Founders Edition design. It is a genuinely excellent 1440p high-refresh card, and DLSS 3 Frame Generation makes even path-traced titles playable. The 12GB frame buffer on a 192-bit bus is the only real reservation — it is fine for 1440p today but leaves less headroom for 4K and future texture-heavy releases. For most high-refresh 1440p gamers, this is the sweet-spot Ada card.
Perfect 1440p card, runs cool and quiet
Upgraded from a 2060 and the jump is massive. Everything maxes out at 1440p 144Hz and DLSS 3 makes Cyberpunk with path tracing actually playable. The Founders Edition cooler barely gets above 65C under load and I can't hear it over my case fans. 220W means my old 650W PSU was fine too.
Great performance, 12GB is the only worry
Performance is excellent and the efficiency is impressive — pulls way less than my friend's 3080. My only hesitation is the 12GB VRAM. At 1440p it's a non-issue, but I do wonder how it holds up in two or three years at higher settings. For now it's flawless.
Fits my SFF build perfectly
The dual-slot FE design was the deciding factor for my small-form-factor build — most AIB cards are huge triple-slot bricks. The 12VHPWR adapter is a bit awkward to route but once seated it's solid. Dual AV1 encoders are a bonus for my OBS streaming.
Own this device? Share your experience.
Your overall rating
0 / 2000
Pick a star rating to continue.
Roughly 15 to 20 percent faster on average. NVIDIA enabled more of the AD104 die, taking the CUDA core count from 5888 on the 4070 to 7168 on the SUPER, while keeping the same 12GB of GDDR6X and a similar power draw. The result sits close to the RTX 4070 Ti in many games.
NVIDIA recommends a 650W power supply. The card has a 220W total graphics power rating and uses a single 16-pin 12VHPWR connector, fed by a bundled adapter that takes two 8-pin PCIe cables. ATX 3.0 PSUs with a native 16-pin cable can connect directly.
It has 12GB of GDDR6X on a 192-bit bus for about 504 GB/s of bandwidth. That is comfortable for 1440p in current titles. At 4K, or in a few VRAM-hungry games with maxed textures, 12GB can become a limit, so it is best thought of as a high-refresh 1440p card.
Yes. As an Ada Lovelace GPU it supports the full DLSS 3 suite, including Frame Generation, which uses the Optical Flow Accelerator and fourth-gen Tensor cores to insert AI-generated frames. It also supports DLSS Super Resolution and Reflex, and is compatible with the newer DLSS feature updates.
The Founders Edition ships with three DisplayPort 1.4a outputs and one HDMI 2.1 output. That supports up to 4K at 240 Hz or 8K at 60 Hz with DSC, and you can drive up to four displays at once.
It is a compact dual-slot card about 244 mm long, noticeably smaller than the triple-slot 4080-class Founders Editions. It fits comfortably in most mid-towers and many small-form-factor cases, though you should still check clearance and the 12VHPWR cable bend radius.
Share quick feedback - it's anonymous and separate from comments.
Comments
No comments yet - be the first to share your thoughts.
Leave a comment