Please note that this post is tagged as a rumor .
New specs for NVIDIA RTX 4070
With rumors focusing entirely on the ‘unlaunched’ RTX 4080 12GB and its ‘reincarnation’ as RTX 4070 Ti, we have not heard much about the RTX 4070 non-Ti SKU for a while. Today hardware leaker @kopite7kimi has an update on this new graphics card specifically.
It is alleged that RTX 4070 graphics card which was supposed to feature 10GB of memory and 160-bit interface, has now been upgraded to 12GB 192-bit configuration. These are new specs that have not been mentioned before.
According to Kopite, this SKU is now believed to feature 46 Streaming Multiprocessors, so 5888 CUDA cores. That’s significantly fewer than previously rumored 7168, but also the same configuration as the current RTX 3070 SKU. At the same time, this card is supposedly getting ‘an upgrade’ to 12GB 192-bit spec, so the memory bandwidth would be identical to RTX 4070 Ti.
Some updates of RTX 4070:
PG141-SKU336/337
AD104-250-A1
5888FP32
12G 21Gbps GDDR6X
36M L2
250W— kopite7kimi (@kopite7kimi) December 9, 2022
Furthermore, the RTX 4070 is to have TDP of 250W, so just 35W less than the Ti variant. That said, both models are to feature AD104 GPU, similar thermal power requirements, and identical memory specs, meaning board partners are likely use the same board design for both SKUs.
This also raises a question when we should expect this model to launch. Given that the RTX 4070 Ti has been pushed back to January, and it features very similar specs, the non-Ti variant might launch just as soon. The only reason not to launch this SKU right is overstock of RTX 3070/3080 cards and no sign of AMD Radeon RX 7800 series.
| RUMORED NVIDIA GeForce RTX 40 Series Specs | ||||
|---|---|---|---|---|
| VideoCardz.com | GeForce RTX 4090 | GeForce RTX 4080 | GeForce RTX 4070 Ti | GeForce RTX 4070 |
| Picture |  | |  |  |
| Board-SKU | PG136-SKU330 | PG136-SKU360 | PG141-SKU331 | PG141-SKU336/337 |
| Architecture | Ada (TSMC 4N) | Ada (TSMC 4N) | Ada (TSMC 4N) | Ada (TSMC 4N) |
| GPU | AD102-300 | AD103-300 | AD104-400 | AD104-250 |
| GPU Transistors | 76.3B | 45.9B | 35.8B | 35.8B |
| GPU Die Size | 608 mm ² | 379 mm ² | 295 mm² | 295 mm² |
| SMs | 128 | 76 | 60 | 46 |
| CUDA Cores | 16384 | 9728 | 7680 | 5888 |
| Boost Clock | 2.5 GHz | 2.5 GHz | 2.6 GHz | TBC |
| Memory | 24 GB G6X | 16 GB G6X | 12 GB G6X | 12 GB G6X |
| Memory Bus | 384-bit | 256-bit | 192-bit | 192-bit |
| Memory Speed | 21 Gbps | 22.4 Gbps | 21 Gbps | 21 Gbps |
| Bandwidth | 1008 GB/s | 736 GB/s | 504 GB/s | 504 GB/s |
| Default TGP | 450W | 320W | 285W | 250W |
| MSRP | 1599 USD | 1199 USD | TBC | TBC |
| Release Date | October 12th, 2022 | November 16h, 2022 | January 5th, 2023 | TBC |
Source: @kopite7kimi