He overclocking It is a technique through which, by modifying the frequencies of the GPU or processor, additional performance can be obtained. The people at TecLab and Paulo Gomes went further and modified the memories of a NVIDIA RTX 4070 The SUPER. Thanks to this adjustment, they managed to make this graph perform better than the RTX 4080 SUPER.
The RTX 4070 Ti featured 12GB of GDDR6X memory with a 192-bit memory bus. But the 4070 Ti SUPER has 16 GB of GDDR6X memory with a 256-bit memory bus. In both cases, the memories actually have a working frequency of 21 Gbps.
Most users know that these memories are very suitable for overclocking. It is not at all difficult to increase the working frequency to reach up to 25 Gbps. This represents a notable additional performance, even if it is not achievable for all users.
Getting the most out of the RTX 4070 Ti SUPER
Responsible for this step are colleagues from TecLab and Paulo Gomes, a Brazilian hardware specialist. They teamed up to see how far they could take the RTX 4070 Ti SUPER and it was impressive. They have amended working frequencies at memories of this Graphic card so that they reach 26 Gbps.
TecLab used a Galax RTX 4070 Ti SUPER model, because it could not be otherwise, since the head of TecLab is a GALAX employee. Paulo Gomes, for his part, used a Manli model.
You should know that the memories of the two graphics cards have changed for this important step. More precisely, they changed the default 21 Gbps Micron (Code: 161-0479-900) for that of 24 Gbps (Code: 161-0480-900) also from this manufacturer, which is the only one to market GDDR6X.
If you plan to replicate it, forget it, because changing GDDR6X memories is extremely complicated. To begin with, it requires great welding skills and having specific equipment and tools. Additionally, Micron does not sell these memories to anyone who asks for them.
As if all this wasn’t enough, you need to have software knowledge because the BIOS needs to be changed. This is something very complicated in graphics cards and usually ends in disaster. We’re talking about leading hardware experts with the hardware, resources, and knowledge needed for this complex endeavor. Additionally, this completely voids the graphics card warranty.
The interesting part is that no liquid cooling or extreme liquid nitrogen cooling was used. Simply, the original graphics heatsinks including three fans were used. In both cases, the graphics cards perform within thermal limits, despite the modifications.
In addition to changing the memories, the frequencies have also been increased. Gomes modified the GPU frequency by increasing it by more than 200 MHz and that of the memories by more than 2,000 MHz to obtain 8,870 points. While the TecLab team increased the frequency of its GPU by more than 271 GHz and that of its memories by more than 2,850 MHz to obtain 9,133 points. In both cases, the Superposition benchmark in 8K resolution was used.
You should know that the stock RTX 4080 SUPER offers a result of 8736 points in the same test. Additionally, Manli’s RTX 4070 Ti SUPER gives 7,212 points and GALAX’s gives 7,028 points. They thus obtained an improvement in performance of between 20 and 30%, which is impressive, although impossible to reproduce at home.
