Teclab pushes Nvidia’s RTX 4090 “SUPER” to its limit with HUGE memory overclock

Teclab goes to insane lengths to push the limits of Nvidia’s RTX 4090

Teclab have created what they are calling an “RTX 4090 SUPER” graphics card. Yes, Nvidia never released a GPU with this name, but this custom GPU is greater than any RTX 4090 on the market. With this in mind, the “RTX 4090 SUPER” name is appropriate.

The main upgrade for this GPU is its memory. The RTX 4090 uses 21 Gbps GDDR6X memory out of the box. Techlab have replaced this memory with 24 Gbps GDDR6X memory from an RTX 4080 SUPER. Nvidia limited this GPU’s memory speed to 22.4 Gbps on RTX 4080 SUPER GPUs, but the memory is rated for 24 Gbps speeds. Teclab than overclocked this memory to 25.8 Gbps.

Alongside this memory transplant and overclock, Teclab have also moved their RTX 4090 GPU to a PCB from a Galax RTX 3090 Ti HOF. This GPU has a 28-phase VRM and dual 16-pin power connectors, making it ideal for overclocking. Yes, you can put an RTX 4090 on an RTX 3090 Ti PCB!

Once Teclab added a GPU overclock and LOD adjustments to their custom GPU, they were able to achieve scores that are almost 40% higher than a stock RTX 4090.

Memory is a major limiting factor for Nvidia’s RTX 4090

With its memory overclock alone, Nvidia’s RTX 4090 can run over 13% faster in the Unigine Superposition 8K benchmark. In this test, memory performance is a limiting factor for Nvidia’s GPU performance. With this in mind, it is easy to see why Nvidia are moving to GDDR7 memory with their next-generation products.

Rumour has it that Nvidia’s RTX 5090 will have 28GB of 28 Gbps GDDR7 memory on a 448-bit memory bus. This will give Nvidia’s next-generation flagship a 55% boost in memory performance over Nvidia’s RTX 4090. This should enable higher levels of performance across a range of workloads.

You can join the discussion on TecLab’s insane RTX 4090 overclock on the OC3D Forums.