Sometimes you just want a little more from your graphics card, though we’re not advising you to open it up for some hardware modification. Regardless, that’s what two teams from Brazil have done as a part of a friendly competition. The pair of them have been able to upgrade the RTX 4070 Ti Super in the memory department, showcasing a performance uplift in performance and stability testing benchmarks.
It definitely takes plenty of expertise, but they’ve been able to outperform a stock RTX 4080 with their tweaks, not that anyone should really be buying that anymore following the introduction of the 4080 Super. Either way, a commendable feat.
Modders upgrade RTX 4070 Ti Super memory up to 26 Gbps
When looking at the specs of the RTX 4070 Ti Super, you’ll see that it usually offers a memory clock of up to 21 Gbps. Now with some modding and overclocking from two Brazilian teams, Paulo Gomes and TecLab, they’ve boosted it up to 24 Gbps and 26 Gbps respectively. And with that, benchmark results show that it can even surpass the non-super RTX 4080. Both of these popular YouTubers actually work with GPU companies, Manli and GALAX to be exact.
This took place due to a little competition between the two teams, aiming to achieve new heights with some non-standard memory modification. It’s more than just physically replacing the memory though, as the GPU board itself also has to be configured in a way to support the better hardware. With faster GDDR7 memory expected for the next generation, this little experiment proves how much of a difference it can make.
Better benchmark results than the RTX 4080
Results of an 8K benchmark using Superposition demonstrate the performance uplift. Paulo Gomes was responsible for the Manli card, whereas TecLab tinkered with the GALAX model.
- GeForce RTX 4080 (stock): 8525 points
- Manli RTX 4070 Ti Super (stock): 7212 points
- Manli RTX 4070 Ti Super (GPU overclocked & 24 Gbps memory): 8870 points
- GALAX RTX 4070 Ti Super (stock): 7028 points
- GALAX RTX 4070 Ti Super (GPU overclocked & 26 Gbps memory): 9133 points
- GPU: AD103
- CUDA Cores: 8,448
- VRAM: 16GB GDDR6X
- Bandwidth: 672.3 GB/s
- Memory bus width: 256-bit
- Base clock speed: 2340 MHz
- Boost clock speed: 2670 MHz