Enthusiasts in the graphics card community have recently taken the NVIDIA GeForce RTX 4070 Ti SUPER to new heights by tuning its memory all the way up to 26 Gbps and achieving performance levels that exceed that of the RTX 4080 SUPER. The GeForce RTX 4070 Ti already presented a notable upgrade at its launch when the original 12 GB memory configuration was replaced with a substantial 16 GB, wielding 21 Gbps GDDR6X memory modules.
These modules are no strangers to overclocking potential, and it has been shown that achieving 24-25 Gbps isn’t out of reach with the right conditions. The experimenters, including a team from TecLab alongside Ronaldo from GALAX, have outdone these expectations by pushing memory speeds to 26 Gbps on the NVIDIA GeForce RTX 4070 Ti SUPER GPU.
For the purpose of this trial, a GALAX GeForce RTX 4070 Ti SUPER card and a Manli GeForce RTX 4070 Ti SUPER card were chosen. To achieve these extreme overclocking results, the teams replaced the default 21 Gbps Micron memory dies with higher-rated 24 Gbps ones. This process is certainly not recommended for the average user, as it involves advanced soldering skills and a deep understanding of PCB and BIOS intricacies. Such modifications also come with the risk of voiding warranties.
Remarkably, the teams did not rely on elaborate liquid nitrogen (LN2) cooling systems. Instead, the GPUs maintained operation within thermal limits under their original custom triple-fan cooling solutions, capable of occupying a 2.5 to 3.0 slot size. This highlights the effectiveness of the cooling solutions crafted for these GPUs.
When it came to testing, both the core GPU clock and memory were overclocked, resulting in some impressive statistics. The Paulo Gomes team enhanced the GPU clock by +200 MHz and the memory clock by +2000 MHz, notching up a score of 8870 points. On the other hand, the TecLab and GALAX collaboration went a step further, boosting their GPU by +271 MHz and their memory clock by +2850 MHz, culminating in a score of 9133 points in the Unigine Superposition benchmark, edging out the competition with a 2 percent lead.
Comparison of the results compiled by sources illustrates a 20-30 percent performance improvement through memory upgrades and clock speed rises:
– GALAX GeForce RTX 4070 Ti SUPER (OC 26 Gbps) – 9133 Points
– Manli GeForce RTX 4070 Ti SUPER (OC 24 Gbps) – 8870 Points
– NVIDIA GeForce RTX 4080 SUPER (Stock) – 8736 Points
– NVIDIA GeForce RTX 4080 (Stock) – 8525 Points
– Manli GeForce RTX 4070 Ti SUPER (Stock) – 7212 Points
– GALAX GeForce RTX 4070 Ti SUPER (Stock) – 7028 Points
This considerable performance enhancement from the overclocked GPUs hints at what may be possible with the upcoming GDDR7 memory standard, which is expected to start at 28 Gbps. Including potential GPU architecture enhancements and process node refinement, these advancements suggest that the GeForce RTX 50 “Blackwell” series, including the RTX 5090 and RTX 5080, might bring significant gains upon their anticipated release later in the year.
