GeForce RTX 4070 confirmed with 5888 CUDA cores, 12GB GDDR6X memory
The first RTX 4070 has now been validated with GPU-Z software.
While the specs of RTX 4070 non-Ti are no secret to anyone at this point, we now have a proof of an actual custom RTX 4070 out in the wild being tested. As reported earlier, the RTX 4070 launches next week on April 13th, but reviewers will be allowed to share their initial test results a day sooner. This is probably how this GPU-Z validation has ended up online early.
The data confirms that the RTX 4070 made by Zotac will have 5888 CUDA cores, 184 Texture Units and 64 Raster Units. That said, not only would RTX 4070 have fewer ROPs than RTX 4070 Ti (80) but also a smaller L2 cache (48 MB vs. 36MB).
NVIDIA’s RTX 4070 memory specs are now confirmed to be 12GB GDDR6X across a 192-bit memory bus. The clock speed is 1313 MHz, which is translates into 21 Gbps effective speed (multiplied by 16). This particular card was running in PCIe x16 3.0 mode, but of course new 4070 supports Gen4 as well.
ZOTAC RTX 4070 validation, Source: GPU-Z
When it comes to clock speeds, the base clock is confirmed to be 1920 MHz, as we have seen in early leaks since 2 months ago. The only difference for this custom model is the boost clock running at 2535 MHz. This means that the card is factory-overclocked from reference 2475 MHz.
NVIDIA has recently confirmed it will launch the RTX 4070 GPU, by ‘accidentally’ showing it on a graph for one of the Reflex enabled games. The company is comparing their new XX70 class GPU to the last-gen XX60 series. The exact reason for this choice is unknown, however it is no secret that both RTX 4070 and RTX 3060 have a 192-bit memory bus, that is a narrower memory bus than last-gen RTX 3070 series.
| NVIDIA GeForce RTX 4070 vs 3070 Series | ||||
|---|---|---|---|---|
| VideoCardz.com | RTX 4070 Ti | RTX 4070 | RTX 3070 Ti | RTX 3070 |
| Picture |  |  |  |  |
| Board-SKU | PG141-SKU331 | PG141-SKU34X | PG143-SKU15 | PG142-SKU10 |
| Architecture | Ada (TSMC 4N) | Ada (TSMC 4N) | Ampere (SAMSUNG 8N) | Ampere (SAMSUNG 8N) |
| GPU | AD104-400 | AD104-250 | GA104-400 | GA104-30X |
| L2 Cache | ||||
| CUDA Cores | ||||
| TMUs/Tensors | ||||
| ROPs | ||||
| RT Cores | ||||
| Base Clock | ||||
| Boost Clock | ||||
| Max FP32 Compute | ||||
| Memory | ||||
| Memory Bus | ||||
| Memory Speed | ||||
| Memory Bandwidth | ||||
| TDP | ||||
| Interface | PCIe Gen4 x16 | PCIe Gen4 x16 | PCIe Gen4 x16 | PCIe Gen4 x16 |
| MSRP | ||||
| Release Date | January 5, 2023 | April 13, 2023 | June 10, 2021 | May 8, 2021 |
Source: TechPowerUP via @T4CFantasy
