GeForce RTX 4070 vs GeForce RTX 5070: specs and benchmark comparison
ANVIDIA · Ada Lovelace
NVIDIA GeForce RTX 4070
- 12 GB GDDR6 / GDDR6X
- 200 W
- PSU 650 W+
April 13, 2023 · $599 at launch
Short verdict
NVIDIA GeForce RTX 5070 leads in every scenario with comparable benchmarks: rendering and compute. This does not imply an advantage in tasks without benchmarks.
Compare with AIRendering and compute
B +19.7%NVIDIA GeForce RTX 5070 leads in this scenario (up to 19.7% in comparable benchmarks).
No comparable benchmarks yet: gaming: rasterization, gaming: ray tracing, power efficiency.
Advantages
Only measured and rated values: a noticeable benchmark gap and specs where more is better.
Where GeForce RTX 4070 is better1
Board power 50 W lower200 W (TGP) vs 250 W (TGP)
Where GeForce RTX 5070 is better2
Blender Benchmark 4.5.0 18.3% faster6,180 samples/min vs 5,226 samples/min
Launch price $50 lower$549 vs $599
Benchmarks
Results are compared within one benchmark, one version and the same conditions.
- B ahead by 18.3%
Blender Benchmark 4.5.0
GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) · samples/min, higher is better
GeForce RTX 40705,225.8 samples/minGeForce RTX 50706,179.8 samples/minPosition among tested models (67, OptiX)
A · 5,226B · 6,18005,00010,00015,000Each dot is one model’s median in this benchmark. The scale starts at zero, samples/min. NVIDIA GeForce RTX 5070 is ahead by 18.3% in this benchmark.
About the measurement and source
- A · NVIDIA GeForce RTX 4070
5,225.8 samples/min · community measurement, sample 248 · November 14, 2026
Blender Open Data (Blender Foundation)
Median of 248 submissions (14 outliers excluded), interquartile range 5058.13–5339.47; submissions 2025-07-17 to 2026-07-07; OPTIX backend. Snapshot 2026-11-14.
- B · NVIDIA GeForce RTX 5070
6,179.8 samples/min · community measurement, sample 843 · November 14, 2026
Blender Open Data (Blender Foundation)
Median of 843 submissions (71 outliers excluded), interquartile range 6033.97–6296.89; submissions 2025-07-17 to 2026-10-04; OPTIX backend. Snapshot 2026-11-14.
difference = |A − B| ÷ the smaller value × 100%. The 5% noticeable threshold is a UX heuristic, not statistical significance.
Cycles render in Blender Benchmark. Median of the sum of samples per minute over three scenes after removing outliers.
Other comparable benchmarks (6)
| Benchmark | A | B | Difference |
|---|---|---|---|
| Blender Benchmark 5.1.1GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 5,265.7 | 5,995.9 | B +13.9% |
| Blender Benchmark 5.2.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 5,114.2 | 5,823.6 | B +13.9% |
| Blender Benchmark 4.3.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 5,113.6 | 6,086.9 | B +19% |
| Blender Benchmark 4.4.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 5,172.3 | 6,191.2 | B +19.7% |
| Blender Benchmark 5.0.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 5,108.7 | 5,654.4 | B +10.7% |
| Blender Benchmark 5.1.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 5,526.5 | 5,903.6 | B +6.8% |
Each benchmark version has its own scale: numbers are compared only within a row.
Reference results (14)
| Benchmark | A | B | Difference |
|---|---|---|---|
| Blender Benchmark 3.6.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | 3,719.9 | — | no pair |
| Blender Benchmark 4.0.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | 3,435.7 | — | no pair |
| Blender Benchmark 4.2.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | 3,387.9 | — | no pair |
| Blender Benchmark 4.5.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | 3,291.8 | — | no pair |
| Blender Benchmark 5.1.1GPU, sum of scenes monster + junkshop + classroom (CUDA) | 3,114.6 | — | no pair |
| Blender Benchmark 4.2.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 5,153.6 | 6,075.8 | reference |
| Blender Benchmark 3.2.1GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 5,138.5 | — | no pair |
| Blender Benchmark 3.3.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 5,812 | — | no pair |
| Blender Benchmark 3.4.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 5,890.6 | — | no pair |
| Blender Benchmark 3.5.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 5,999.6 | — | no pair |
| Blender Benchmark 3.6.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 6,034.8 | 6,492.2 | reference |
| Blender Benchmark 4.0.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 5,403.1 | 6,701.7 | reference |
| Blender Benchmark 4.1.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 5,432 | — | no pair |
| Blender Benchmark 5.3.0GPU, sum of scenes monster + junkshop + classroom (OptiX / HIP / oneAPI / Metal) | 5,906.2 | 6,452.3 | reference |
Each benchmark version has its own scale: numbers are compared only within a row.
Specifications
| Specification | A · GeForce RTX 4070 | B · GeForce RTX 5070 | Difference |
|---|---|---|---|
| Chip | |||
| Architecture | Ada Lovelace | Blackwell | differs |
| Shader unitsShader unit counts are comparable only within one architecture. | 5,888 (CUDA cores) | 6,144 (CUDA cores) | not directly comparable |
| Clocks | |||
| Base clock | 1,920 MHz | 2,330 MHz | B: 410 MHz more |
| Boost clockClocks of different architectures do not reflect performance directly. For laptop chips, the clock depends on the laptop’s power limit. | 2,480 MHz | 2,510 MHz | B: 30 MHz more |
| Memory | |||
| Video memory | 12 GB | 12 GB | equal |
| Memory type | GDDR6 / GDDR6X | GDDR7 | differs |
| Memory bus | 192 bit | 192 bit | equal |
| Power and dimensions | |||
| Board powerTGP (NVIDIA) and TBP (AMD/Intel) are calculated differently; higher is not better. For laptops, the chip maker sets the range and the laptop maker picks the value, so the same chip behaves differently in different laptops. | 200 W (TGP) | 250 W (TGP) | B: 50 W higher |
| Recommended PSU | 650 W | 650 W | equal |
| Interface | no data | PCIe 5.0 | no data for A |
| Length (reference / Founders Edition)Partner cards have different dimensions; check the specific SKU. | 244 mm | 242 mm | A: 2 mm higher |
| Thickness (slots) | 2 | 2 | equal |
| Power connectors (reference) | no data | 1x 16-pin (adapter to 2x 8-pin included) — Founders Edition | no data for A |
Chip
Architecture
AAda Lovelace
BBlackwell
differsShader units
A5,888 (CUDA cores)
B6,144 (CUDA cores)
not directly comparableShader unit counts are comparable only within one architecture.
Clocks
Base clock
A1,920 MHz
B2,330 MHz
B: 410 MHz moreBoost clock
A2,480 MHz
B2,510 MHz
B: 30 MHz moreClocks of different architectures do not reflect performance directly. For laptop chips, the clock depends on the laptop’s power limit.
Memory
Video memory
A12 GB
B12 GB
equalMemory type
AGDDR6 / GDDR6X
BGDDR7
differsMemory bus
A192 bit
B192 bit
equal
Power and dimensions
Board power
A200 W (TGP)
B250 W (TGP)
B: 50 W higherTGP (NVIDIA) and TBP (AMD/Intel) are calculated differently; higher is not better. For laptops, the chip maker sets the range and the laptop maker picks the value, so the same chip behaves differently in different laptops.
Recommended PSU
A650 W
B650 W
equalInterface
Ano data
BPCIe 5.0
no data for ALength (reference / Founders Edition)
A244 mm
B242 mm
A: 2 mm higherPartner cards have different dimensions; check the specific SKU.
Thickness (slots)
A2
B2
equalPower connectors (reference)
Ano data
B1x 16-pin (adapter to 2x 8-pin included) — Founders Edition
no data for A
“More” does not always mean “better”: higher power draw or size is a system requirement, not an advantage (such bars are dimmed).
Compare with AI
An AI model reviews the specs and benchmarks of this pair and gives a plain-language verdict on who each model suits.
Questions and answers
How much video memory do NVIDIA GeForce RTX 4070 and NVIDIA GeForce RTX 5070 have?
NVIDIA GeForce RTX 4070: 12 GB GDDR6 / GDDR6X, 192-bit bus. NVIDIA GeForce RTX 5070: 12 GB GDDR7, 192-bit bus.
What power supply does the manufacturer recommend?
For NVIDIA GeForce RTX 4070, 650 W or more; for NVIDIA GeForce RTX 5070, 650 W or more. This is a recommendation for a typical system; also check the power connectors of the specific card model.
Which is faster: NVIDIA GeForce RTX 4070 or NVIDIA GeForce RTX 5070?
NVIDIA GeForce RTX 5070 leads in every scenario with comparable benchmarks: rendering and compute. This does not imply an advantage in tasks without benchmarks.
See also
Data sources (3)
- NVIDIA — specifications of NVIDIA GeForce RTX 4070, specifications of NVIDIA GeForce RTX 5070checked November 7, 2026
- Blender Open Data (Blender Foundation) — benchmark Blender Benchmark 3.6.0, benchmark Blender Benchmark 4.0.0, benchmark Blender Benchmark 4.2.0, benchmark Blender Benchmark 4.5.0, benchmark Blender Benchmark 5.1.1, benchmark Blender Benchmark 5.2.0, benchmark Blender Benchmark 4.3.0, benchmark Blender Benchmark 4.4.0, benchmark Blender Benchmark 5.0.0, benchmark Blender Benchmark 5.1.0, benchmark Blender Benchmark 5.3.0, benchmark Blender Benchmark 3.2.1, benchmark Blender Benchmark 3.3.0, benchmark Blender Benchmark 3.4.0, benchmark Blender Benchmark 3.5.0, benchmark Blender Benchmark 4.1.0checked November 14, 2026
- Wikipedia (English) — release date and launch pricechecked November 14, 2026
How we compare: methodology. Where the data comes from: sources.