After several disappointing rumours revolving around the NVIDIA GeForce RTX 4070 and its allocation of graphics memory and CUDA cores, it seems that the GPU brand may be feeling some form of remorse and has reportedly been preparing to modify the card’s internal specification. At least, to a measurable degree. Now, instead of the initial 10GB GDDR6X and a 160-bit memory bus, the yet-to-be-released graphics card could launch with 12GB of graphics memory instead, as well as a 192-bit memory bus.
The rumours come, yet again, by way of prominent leakster kopite7kimi (@kopite7kimi), who shared their latest details about the RTX 4070 on Twitter. According to their tweet, the card is expected to be fitted with a PG141-SKU331 GPU, have 7680 CUDA Cores, 12GB GDDR6X graphics memory, running at a frequency of 21Gbps, and a 300W TDP. If the majority of these specifications seem familiar, it’s because they’re nearly identical to the specs that Kopite posted about the Ti version of the GPU, back at the start of the month.
There is an update of RTX 4070.
PG141-SKU331
7680FP32
12G 21Gbps GDDR6X
300W
TSE >11000— kopite7kimi (@kopite7kimi) August 4, 2022
At the time of writing, the only difference between the RTX 4070 Ti and the non-Ti version of the GPU would be TDP; the former is rumoured to have a TDP of 400W, while the latter is said to consume 100W less power.
Getting back on point, if the rumours are to be believed, Kopite says that the new specs of the alleged RTX 4070 could push the Lovelace-based GPU’s score in UL’s 3DMark Time Spy Extreme benchmark up, from 10000 points to the 11000 points average.
As always, and like so many of Kopite’s earlier tweets, all of the specs written here about the RTX 4070 are simply rumours and not to be taken too seriously until NVIDIA itself breaks its silence on the matter, which is expected to be soon and more precisely, either in this month or October.
(Source: Videocardz)
