Nvidia only has two Ada graphics cards out right now, and they’re among the best graphics cards you can buy, but that lineup might soon expand to include not just one, but two new GPUs.
Specifications of the rumored RTX 4070 emerged today from a credible source. The card might be significantly slower than the RTX 4070 Ti, but there’s one way in which Nvidia could make it succeed.
Some updates of RTX 4070:
PG141-SKU336/337
AD104-250-A1
5888FP32
12G 21Gbps GDDR6X
36M L2
250W— kopite7kimi (@kopite7kimi) December 9, 2022
The rumor mill has recently been kept occupied with the upcoming AMD Radeon RX 7900 XTX and 7900 XT. On the Nvidia side, it’s all about the RTX 4070 Ti right now, which will most likely be the rebranded version of the “unlaunched” RTX 4080 12GB. The AMD cards are set to release on December 13, and the Nvidia offering is still just speculation, with a predicted January release date.
Today, another card joins the list of anticipated GPUs — the RTX 4070. Known Nvidia leaker kopite7kimi on Twitter shared some specifications of the card, and while nothing is confirmed, it’s possible that a lot of these specs will be what we see upon the release of the RTX 4070.
The RTX 4070 will feature a cutdown version of the AD104 die, and more specifically, the AD104-250-A1 GPU. The RTX 4070 Ti also features the AD104 die, but the 4070 will have the binned version, which means the dies didn’t make it through the lithographic process unscathed. These are still functional GPUs, but with some parts disabled to serve as lower-tier graphics cards.
According to kopite7kimi, the RTX 4070 will be equipped with 12GB memory across a 192-bit bus, clocked at 21Gbps, which means a bandwidth of 504GB/s. It will also come with 46 streaming multiprocessors (SM), which translates to 5,888 CUDA cores. It’s worth noting that the last-gen RTX 3070 comes with the same number of CUDA cores. On the other hand, the memory bandwidth on the RTX 4070 will be significantly higher.
By default, the RTX 4070 is said to have a total board power (TBP) of 250 watts, meaning 30 more watts than the RTX 3070. The rumored RTX 4070 Ti is said to have a TBP of 285W. If it really is going to be a rebranded RTX 4080 12GB, it will come with 12GB of GDDR6 memory across a 192-bit bus, so it will have the same memory bandwidth as this rumored RTX 4070. The 4070 Ti has significantly more CUDA cores (7,680) and 60 SMs.
Based on these specs and a comparison of the two cards, the RTX 4070 will likely be up to 25% slower than the RTX 4070 Ti. While not super impressive, the card could easily become a favorite if Nvidia was to alter its pricing strategy and make it somewhat affordable.
Based on the prices of the RTX 4090 and RTX 4080, Nvidia might price the 4070 in the $700 range, but that might turn out to be too much — especially when there are still so many great GPUs from previous generations up for grabs.
