NVIDIA GeForce RTX 5070 12GB Specs, Benchmarks & Pricing
The NVIDIA GeForce RTX 5070 is a mid-range desktop graphics card using the GB205 GPU based on the Blackwell architecture. It was announced at CES 2025 on January 6, 2025, and launched on March 5, 2025 with an MSRP of $549 USD.
Key specifications include 6,144 CUDA cores, 192 Tensor cores (5th generation), 48 RT cores (4th generation), and 12GB of GDDR7 memory on a 192-bit bus providing 672 GB/s bandwidth. The card has a TGP of 250W and requires a 16-pin power connector.
The RTX 5070 supports NVIDIA's latest features including DLSS 4 with Multi Frame Generation, which uses AI to generate up to three additional frames per traditionally rendered frame. It also features RTX Neural Shaders that integrate small AI networks into the shading process for higher quality rendering.
Unlike the RTX 5070 Ti, the RTX 5070 has a Founders Edition available directly from NVIDIA. Third-party models are available from ASUS, MSI, Gigabyte, PNY, Zotac, Palit, and Gainward.
NVIDIA positions the RTX 5070 as delivering RTX 4090-class performance in games with DLSS 4 enabled, at a significantly lower price point and power consumption.
- Release Date: March 5, 2025
- MSRP: $549 USD
- GPU Architecture: blackwell
- Hardware-Accelerated GEMM Operations:FP64 FP32 TF32 BF16 FP16 FP8 FP4 INT8 INT4
- CUDA Compute Capability : 12.8
Strengths
- Strong FurMark (GL) (4K) performance (top 21% of GPUs)
- Strong FurMark (GL) (1440p) performance (top 23% of GPUs)
Considerations
- Lower FP4 matrix performance compared to other GPUs
Specifications for NVIDIA GeForce RTX 5070
| Specification | Performance Ranking |
|---|---|
| FP32 TFLOPs | |
| FP16 TFLOPs | |
| Tensor Core Count | |
| Memory Capacity (GB) | |
| Memory Bandwidth (GB/s) | |
| Int8 TOPs | |
| FP8 TFLOPs | Data not available |
| FP4 TFLOPs |
Gaming Benchmarks for NVIDIA GeForce RTX 5070
Rankings based on 805,051 real-world benchmark results from OpenBenchmarking.org
| Benchmark | Performance Ranking |
|---|---|
| 3DMark Wild Life Extreme (4K) | |
| Counter-Strike 2 (4K) | |
| FurMark (GL) (4K) | |
| FurMark (GL) (1440p) | |
| FurMark (GL) (1080p) | |
| FurMark (VK) (4K) | |
| FurMark (VK) (1440p) | |
| FurMark (VK) (1080p) | |
| Quake II RTX (4K) |
Real-time NVIDIA GeForce RTX 5070 GPU Prices
Compare Price/Performance to other GPUs
Compare NVIDIA GeForce RTX 5070 to Another GPU
Price History
NVIDIA GeForce RTX 5070 Price History
This chart tracks NVIDIA GeForce RTX 5070 prices over the past 6 months based on eBay listings. The yellow line shows the average of the three lowest-priced listings each day—a useful indicator for finding deals.
Trend: Over this period, the lowest average price has been up 14.6%. The current lowest average is $704, compared to a typical $616 over the period. The lowest average price is calculated as the average of the three lowest-priced listings each day.
Product Identifiers
Manufacturer Part Numbers (4)
- NVIDIA Part Number
- 900-1G146-2570-000
- Board ID
- PG146
- Board ID
- PG147
- GPU DIE
- GB205-300-A1
Available from 7 Partners (14 products)
- PRIME GeForce RTX 5070 12GB GDDR7
- PRIME-RTX5070-12G(model number)
- PRIME GeForce RTX 5070 OC Edition 12GB GDDR7
- PRIME-RTX5070-O12G(model number)
- TUF Gaming GeForce RTX 5070 OC Edition 12GB GDDR7
- TUF-RTX5070-O12G-GAMING(model number)
- GeForce RTX 5070 12G Gaming Trio OC
- G5070-12GTC(model number)
- GeForce RTX 5070 WINDFORCE OC SFF 12G
- GV-N5070WF3OC-12GD(model number)
- GeForce RTX 5070 EAGLE OC SFF 12G
- GV-N5070EAGLE-OC-12GD(model number)
- GeForce RTX 5070 EAGLE OC ICE SFF 12G
- GV-N5070EAGLEOC-ICE-12GD(model number)
- GeForce RTX 5070 OC Triple Fan
- VCG507012TFXPB1-O(part number)
- GAMING GeForce RTX 5070 SOLID
- ZT-B50700D-10P(part number)
- GAMING GeForce RTX 5070 SOLID OC
- ZT-B50700J-10P(part number)
- GeForce RTX 5070 GamingPro
- NE75070019K9-GB2050A(model number)
- GeForce RTX 5070 GamingPro-S
- NE75070019K9-GB2050K(model number)
- GeForce RTX 5070 Phoenix
- NE75070019K9-GB2050X(model number)
- GeForce RTX 5070 Phoenix-S
- NE75070019K9-GB2050K(model number)
References
- https://www.nvidia.com/en-us/geforce/graphics-cards/50-series/rtx-5070-family/
- https://images.nvidia.com/aem-dam/Solutions/geforce/blackwell/nvidia-rtx-blackwell-gpu-architecture.pdf
- https://nvidianews.nvidia.com/news/nvidia-blackwell-geforce-rtx-50-series-opens-new-world-of-ai-computer-graphics
- https://www.nvidia.com/en-us/geforce/news/rtx-50-series-graphics-cards-gpu-laptop-announcements/
- https://www.nvidia.com/en-us/geforce/graphics-cards/compare/
- https://www.techpowerup.com/gpu-specs/geforce-rtx-5070.c4218
- https://www.asus.com/us/motherboards-components/graphics-cards/prime/prime-rtx5070-12g/techspec/
- https://www.msi.com/Graphics-Card/GeForce-RTX-5070-12G-GAMING-TRIO-OC
- https://www.gigabyte.com/Graphics-Card/GV-N5070WF3OC-12GD
- https://www.pny.com/geforce-rtx-5070-models
- https://www.zotac.com/us/product/graphics_card/zotac-gaming-geforce-rtx-5070-solid
- https://www.palit.com/palit/vgapro.php?id=5435&lang=en
- https://www.gainward.com/main/vgapro.php?id=1235&lang=en
Notes
- We use the peak TOPS/FLOPS specs using the sparsity feature.
- Blackwell GB205 is the chip (GB205-300-A1 GPU die variant).
- fp16TFLOPS is "Peak FP16 Tensor TFLOPS with FP16 Accumulate" (246.9 with sparsity) as published in the Blackwell GPU architecture whitepaper Table 6.
- int8TOPS of 246.95 represents non-sparse (dense) INT8 performance, derived from sparse INT8 of 493.9 TOPS ÷ 2 per NVIDIA Blackwell architecture whitepaper Table 6. With 2:4 structured sparsity, INT8 performance is 493.9 TOPS. Dense value used for consistent cross-vendor comparison.
- fp32TFLOPS of 30.9 is TF32 Tensor Core throughput without sparsity. Table 6 of NVIDIA's RTX Blackwell whitepaper lists the RTX 5070 TF32 as 30.9 dense and 61.7 with sparsity (https://images.nvidia.com/aem-dam/Solutions/geforce/blackwell/nvidia-rtx-blackwell-gpu-architecture.pdf). It equals FP32 vector (CUDA core) throughput: on GeForce Ampere, Ada and Blackwell, dense TF32 runs at the FP32 vector rate.
- MSRP of $549 USD sourced from nvidia.com official product page (https://www.nvidia.com/en-us/geforce/graphics-cards/50-series/rtx-5070-family/).
- NVIDIA Founders Edition is available for this model with NVPN 900-1G146-2570-000.
- Board variants include PG146 and PG147 based on available documentation.
- Third-party product identifiers sourced from manufacturer websites including ASUS, MSI, Gigabyte, PNY, Zotac, Palit, and Gainward official product pages.
- EVGA does not manufacture RTX 5070 cards as they exited the GPU market in 2022.
- fp4TFLOPS of 494 represents non-sparse (dense) FP4 performance, derived from the official NVIDIA AI TOPS figure of 988 (FP4 sparse, per nvidia.com/en-us/geforce/graphics-cards/compare/) ÷ 2. Dense value used for consistent cross-vendor comparison.
- NOT FOUND: fp8TFLOPS — NVIDIA does not publish a standalone dense or sparse FP8 TFLOPS figure for consumer GeForce Blackwell cards; searched the NVIDIA RTX Blackwell architecture whitepaper and the official RTX 5070/compare pages (URLs in references), which list only AI TOPS (FP4 sparse), Peak TF32 Tensor TFLOPS, Peak FP16 Tensor TFLOPS, and INT8 TOPS. Left null rather than guessed from the INT8 dense value.
- supportedHardwareOperations: INT4 removed per the locked GEMM-only definition (2026-09) — neither the consumer nor datacenter Blackwell documentation lists INT4 Tensor Core support (dropped starting with Blackwell per gpu-researcher.md's canonical Blackwell table); the array now matches the canonical consumer Blackwell baseline of FP4, FP8, FP16, BF16, TF32, INT8.