The Russian modder, VIK-on, updated the NVIDIA GeForce RTX 3070 graphics card with 16 GB GDDR6 memory. Modding required some technical experience, but in the end, VIK-on made the RTX 3070 run at twice its original memory capacity.
NVIDIA GeForce RTX 3070 modified and updated with 16 GB GDDR6 memory, runs a little faster than the original RTX 3070 version
We’ve seen modders replace memory modules on NVIDIA GeForce graphics cards before. Teclab did this in 2019 by modifying the GeForce RTX 2080 Ti to support 16 Gbps modules, since it had 14 Gbps modules in stock. Here, we don’t see an increase in the speed of memory, but, instead, an increase in capacity. The NVIDIA GeForce RTX 3070 originally comes with 8 GB of GDDR6 memory and although we have heard several rumors of a higher-capacity Ti variant in the past, this no longer seems to be the case.
So VIK-on decided to do the job with his own hands and replaced the 8 GB memory modules with 16 GB memory modules. The memory modules used are the K4ZAF3258M-HC14 manufactured by Samsung, which has a density of 16 Gb which makes up the capacity of 2 GB per module. Each of the 8 memory modules in the RTX 3070 are also Samsung K4Z803256C-HC14 chips, but have densities of 8 Gb for a total of 1 GB per module.
To make the GeForce RTX 33070 video card work, it must be configured via the BIOS, but since NVIDIA has a very tight BIOS lock that not even manufacturers can modify, a different approach was needed.
The modder decided to bypass the set of belts that needed to be manually configured. Each set of bracelets comes with a unique number of bits that determine which memory modules are being used. The 00000 bit bracelet is for Samsung 8 Gb (1 GB) modules, 00001 is for Micron 8 Gb (1 GB) modules, while the 00110 bracelet set supports 16 Gb (2 GB) modules. After the appropriate belt configuration is defined for the card, a different memory configuration can be enabled (16 GB in the case of this RTX 3070).
The initialization of the NVIDIA GeForce RTX 3070 graphics card was successful, but it had some errors in which it did not load a different application without Furmark running in the background. Since Furmark is a very heavy stress test for the GPU, the final performance of the video card was not so good. But a solution for this comes in the form of the EVGA Precision’s clock locking method, which fixed the problem and let the card run at its maximum potential with a capacity of 16 GB.
The final results, as reported, are slightly better than the NVIDIA GeForce RTX 3070 reference graphics cards. Certainly they are not a huge leap in performance, which could be the reason why NVIDIA decided to discard the dual-capacity variant , but we still hope to see a Ti variant with a higher-bay GA104 GPU with 16 GB GDDR6 VRAM in the future.
News sources: VIK-on, Videocardz