The Only NVIDIA GeForce RTX 4090 SUPER In The World: Teclab Details The Building of The Fastest Custom “Ada” GPU With 26 Gbps G6X Memory
The Only NVIDIA GeForce RTX 4090 SUPER In The World: Teclab Details The Building of The Fastest Custom “Ada” GPU With 26 Gbps G6X Memory

NVIDIA GeForce RTX 4090 SUPER may not be real, but Teclab made one themselves with the AD102 "Ada" silicon & faster G6X memory.
Last month, Teclab showcased how they converted the GALAX GeForce RTX 3090 Ti HOF into a new graphics card using NVIDIA's AD102 "Ada" GPU. This new graphics card was titled the GeForce RTX 4090 SUPER and why it is named so because of its customized design, which uses binned Ada silicon and the fastest GDDR6X memory available. Also, this card's performance is quite a spectacle, so let's go through the whole building process of this impressive project from team Teclab.
The NVIDIA GeForce RTX 4090 hasn't got a "SUPER" release yet, nor do we expect to see any such card given how close we are to the next-gen RTX 50 series. However, remnants of the card exist in the form of prototype coolers and PCB which highlight a stronger variant featuring a faster AD102 GPU SKU.
So starting with the details, Teclab explains that the project was first realized back when the RTX 3090 Ti was introduced. The RTX 3090 Ti was launched very late in the RTX 30 series, actually a few months before the release of the RTX 40 series. The card featured the GA102 GPU which was also used by the Non-Ti model (in a different config) however one different thing was the PCB layout.
The 3090 Ti cards had a slightly different PCB which ended up being more in common with the RTX 4090 (AD102) than the RTX 3090 (GA102). When the RTX 4090 launched, it was found that the 3090 Ti and 4090 PCBs had small differences such as memory straps and a 1.8V power supply. The RTX 4090 introduced two different voltage controllers. The 1.8V controller powered the memory & the rest of the PCB components while an additional 1.2V controller was used to power only the GPUs.
So why specifically select the RTX 3090 Ti HOF OC LAB variant? The reason is that the older model comes with a water block that allows for better cooling without the use of exotic cooling setups like LN2. The RTX 4090 did receive a HOF variant but not a HOF OC LAB so there was no custom water block for this card available besides reusing the one from the HOF 3090 Ti OC LAB. Other changes to the 3090 Ti HOF to convert it into the NVIDIA GeForce RTX 4090 SUPER card include the AD102 GPU, 24 Gbps GDDR6X memory, and changes to the memory straps & also the voltage controller to supply 1.2V to the GPU.
For the memory modules, each GDDR6X IC was hand-selected by Teclab to ensure the best quality dies were used in the project. The specified memory dies are Micron's 24 Gbps GDDR6X (ZF) modules (161-0480-900). While you may think that the process is a similar GPU/Memory swap, it is way more complicated than that. This custom NVIDIA GeForce RTX 4090 SUPER project required a lot of expertise and hard work. Since Teclab is associated with GALAX, they have the right tools and schematics for the PCBs, the proper tools/equipment to modify/swap the components, and use the proper power routing, memory straps, & voltage limits to ensure that everything goes as planned.
For cooling, the same water block as the GALAX GeForce RTX 3090 Ti HOF OC Lab was used with a few modifications such as the thermal putty used for each of the memory dies and high-quality TIM for the GPU. This ensures great contact between the water block which uses a nickel-plated base plate and is designed specifically for this PCB.
The cooling is provided by two 360mm radiators that are set up in a push/pull configuration along with a water pump that runs at 234 L/h. The ambient temps were 18.7C while the coolant temps were reported at 20C. The result is that even with a heavy overclock above 3 GHz, the card will be able to run at around 54-55C.
The custom NVIDIA GeForce RTX 4090 SUPER by Tecblab was tested in Unigine Superposition 8K benchmark using a 3105 MHz GPU overclock and 12,900 frequency for the G6X memory which is clocking the modules up to 25.8 Gbps. Teclab has already achieved 26 Gbps speeds with its RTX 4070 Ti custom memory swap mod in the past and the 25.8 Gbps speeds were used since the RTX 4090 has a wider memory interface.
End result of countless hours of modding, tuning, and testing is that the custom NVIDIA GeForce RTX 4090 SUPER by Teclab was able to outperform several LN2 RTX 4090 scores with just water cooling. You can see the score comparison below:
This is a 39% improvement over a stock NVIDIA GeForce RTX 4090 GPU with overclocking applied and a 16.5% improvement without the LOD tweaks which is super impressive. Teclab says that they have more testing of the card planned for the future such as LN2 benchmarks and more so stay tuned for some impressive custom projects from the team.
What's Your Reaction?






