The first benchmarks of NVIDIA's GeForce RTX 4070 Non-Ti graphics card have been leaked within the Geekbench OpenCL database.
NVIDIA GeForce RTX 4070 Graphics Card OpenCL Benchmarks Show On Par With RTX 3080 & 19% Slower Than 4070 Ti
The NVIDIA GeForce RTX 4070 is expected to feature the AD104-250/251 GPU core. The GPU will feature 5888 CUDA cores, and 12 GB GDDR6X VRAM clocked at 21 Gbps, driving the bandwidth up to 504 GB/s (a 12.5% increase over the RTX 3070), The GPU gets 36 MB L2 cache which is 9x higher than the RTX 3070's GA104 GPU.
The graphics card is expected to feature a 200W TGP so you're getting an 8% reduction in power compared to the RTX 3070. The slide also lists down average gaming power which is rated at 186 Watts or 13.5% lower than the RTX 3070 and 22.5% lower than the RTX 3070 Ti. This is in fact the same power consumption as the NVIDIA RTX 3060 while the card should deliver performance on par or faster than the RTX 3080 12 GB which consumed anywhere between 300-320W. That's a massive increase in efficiency.
In terms of performance, the NVIDIA GeForce RTX 4070 graphics card scored 177,594 points in the Geekbench 5 OpenCL benchmark which puts it just on par with the RTX 3080 but slower than its Ti sibling which is 16% faster in this specific benchmark. Although this is just one result and one that's reading "Invalid", the score should more or less be in the same range.
NVIDIA GeForce RTX 4090 OpenCL (Geekbench 5) Benchmark
The lower performance is to be expected considering that the NVIDIA GeForce RTX 4070 comes with 23% lower cores and also gets lower clock speeds for the AD104 GPU that it utilizes. VRAM specifications and the configuration remains the same as the Ti model. Once again, it looks like NVIDIA will be mainly targeting GeForce RTX 3080 12 GB performance with native rasterization & faster than 3080 Ti / 3090 using DLSS 3.
As for the compute performance figures, the NVIDIA GeForce RTX 4070 graphics card features 29 Shader FP32 TFLOPs, 67 RT TFLOPs, 300 OFA, and 400 Tensor TOPs. The compute performance is on par with NVIDIA's RTX 3080 12 GB (30 TFLOPs).
FP32 Compute Horsepower Comparisons (Higher is Better)
The NVIDIA GeForce RTX 4070 graphics card is expected to launch on the 13th of April with reviews for the FE variant landing a day earlier followed by custom AIC reviews on launch day. The graphics card will retail at an official MSRP of $599 US though certain custom models with factory overclocks and better cooling designs will come at a slight premium.
NVIDIA GeForce RTX 40 Series "Expected" Lineup Specs:
| Graphics Card Name | NVIDIA GeForce RTX 4090 | NVIDIA GeForce RTX 4080 | NVIDIA GeForce RTX 4070 Ti | NVIDIA GeForce RTX 4070 | NVIDIA GeForce RTX 4060 Ti | NVIDIA GeForce RTX 4060 | NVIDIA GeForce RTX 4050 |
|---|---|---|---|---|---|---|---|
| GPU Name | Ada Lovelace AD102-300(1) | Ada Lovelace AD103-300(1) | Ada Lovelace AD104-400(1) | Ada Lovelace AD104-250(1) | Ada Lovelace AD106-350 Ada Lovelace AD106-351 | Ada Lovelace AD107-400 | Ada Lovelace AD107 |
| Process Node | TSMC 4N | TSMC 4N | TSMC 4N | TSMC 4N | TSMC 4N | TSMC 4N | TSMC 4N |
| Die Size | 608mm2 | 378.6mm2 | 294.5mm2 | 294.5mm2 | 190.1mm2 | TSMC 4N | TSMC 4N |
| Transistors | 76 Billion | 45.9 Billion | 35.8 Billion | 35.8 Billion | TBD | TBD | TBD |
| CUDA Cores | 16384 | 9728 | 7680 | 5888 | 4352 | 3072 | 2560 |
| TMUs / ROPs | 512 / 176 | 320 / 112 | 240 / 80 | 184 / 64 | TBD | TBD | TBD |
| Tensor / RT Cores | 512 / 128 | 304 / 76 | 240 / 60 | 184 / 46 | TBD | TBD | TBD |
| Base Clock | 2230 MHz | 2210 MHz | 2310 MHz | 1920 MHz | 2310 MHz | TBD | TBD |
| Boost Clock | 2520 MHz | 2510 MHz | 2610 MHz | 2475 MHz | 2535 MHz | TBD | TBD |
| FP32 Compute | 83 TFLOPs | 49 TFLOPs | 40 TFLOPs | 29 TFLOPs | 22 TFLOPs | TBD | TBD |
| RT TFLOPs | 191 TFLOPs | 113 TFLOPs | 82 TFLOPs | TBD | TBD | TBD | TBD |
| Tensor-TOPs | 1321 TOPs | 780 TOPs | 641 TOPs | TBD | TBD | TBD | TBD |
| Memory Capacity | 24 GB GDDR6X | 16 GB GDDR6X | 12 GB GDDR6X | 12 GB GDDR6X | 8/16 GB GDDR6 | 8 GB GDDR6 | 6 GB GDDR6 |
| Memory Bus | 384-bit | 256-bit | 192-bit | 192-bit | 128-bit | 128-bit | 96-bit |
| Memory Speed | 21.0 Gbps | 22.4 Gbps | 21.0 Gbps | 21.0 Gbps | 18 Gbps | 18 Gbps | TBD |
| Bandwidth | 1008 GB/s | 717 GB/s | 504 GB/s | 504 GB/s | 288 GB/s | 288 GB/s | TBD |
| TBP | 450W | 320W | 285W | 200W | 160W/165W | 115W | ~75W |
| Price (MSRP / FE) | $1599 US / 1949 EU | $1199 US / 1469 EU | $799 US | $599 US | $399-$499 US | TBD | TBD |
| Price (Current) | $1599 US / 1859 EU | $1199 US / 1399 EU | $799 US | $599 US | TBD | TBD | TBD |
| Launch (Availability) | 12th October 2022 | 16th November 2022 | 5th January 2023 | 13th April, 2023 | May/July 2023 | July 2023 | 2H 2023 |
News Source: Benchleaks
About the author: A Software Engineer by training and a PC enthusiast by passion, Hassan Mujtaba serves as Wccftech's Senior Editor for hardware section. With years of experience in the industry, he specializes in deep-dive technical analysis of next-generation CPU and GPU architectures, motherboards, and cooling solutions. His work involves not only breaking news on upcoming technologies but also extensive hands-on reviews and benchmarking.
Follow Wccftech on Google to get more of our news coverage in your feeds.
