On June 2nd, 2021 NVIDIA launched the GeForce RTX 3080 Ti Founders Edition with an MSRP of $1,199. It’s time to take that GeForce RTX 3080 Ti, overclock it, and see if it can outperform a factory overclocked GeForce RTX 3090!
The GeForce RTX 3080 Ti is basically a cut-down, but very slightly, cut-down GeForce RTX 3090. It is based on the same GA102 Samsung 8nm die as the GeForce RTX 3090 based on the Ampere architecture. It has 10,240 CUDA Cores, 112 ROPs, 320 TMUs, 80 RT Cores (2nd Gen) and 320 Tensor Cores (3rd Gen). The Boost Clock is set at 1665MHz. It has 12GB of GDDR6X instead of 24GB on the GeForce RTX 3090. It runs at 19GHz providing 912GB/s over the 384-bit bus. TDP is 350W.
Another factor that makes the GeForce RTX 3080 Ti Founders Edition unique is that it does not use the larger design format of the GeForce RTX 3090 Founders Edition. Instead, it uses a footprint size that is similar to the GeForce RTX 3080 Founders Edition. It still utilizes the flipped-fan design, with a push-pull configuration, just in the same smaller footprint as the GeForce RTX 3080 FE. This allows it to fit easier in smaller cases.
In our review, this video card was very fast. It actually kept up with or even surpassed (in some situations) our factory overclocked GIGABYTE GeForce RTX 3090 GAMING OC video card. This was mighty impressive, and our theory behind that was the increased GPU Boost frequencies. It could maintain a higher consistent boost frequency, compared to a GeForce RTX 3090, and that allowed it to compute faster, basically, despite the lower CUDA Cores, for games.
That is why we are very curious how it is going to behave when we overclock it. Will boosting the GPU frequency and memory even higher allow it to smash a factory overclocked GeForce RTX 3090? We will see.
Overclocking GeForce RTX 3080 Ti Founders Edition
Before we overclock the video card, we did need to confirm what it was running at by default. In our GeForce RTX 3080 Ti Founders Edition review we found out that the video card averaged around 1804MHz while gaming. This is well above the quoted boost clock of 1665MHz. GPU Boost is working well to boost the video card beyond that frequency. With such a large bump already from GPU Boost we were worried there might not be a lot of headroom left for overclocking. We were pleasantly surprised there was a lot of gas still left in the tank.
Here is how to overclock the NVIDIA GeForce RTX 3080 Ti Founders Edition video card. To overclock, we used the latest version of EVGA Precision X1 as it has official support for the GeForce RTX 3080 Ti. With Precision X1 we were able to increase the Power Target up 14% from the default 100% making it 114%. This gives us some headroom to overclock. While we could control the Voltage and increase it +100, we found there wasn’t a need to do so. It was best to leave GPU Boost to control Voltage dynamically. Increasing the Voltage only makes the Power Limit hit the TDP wall much quicker. We did set the fans to 100% to ensure the highest possible overclock.
With this video card, we managed to overclock up to +130 on the GPU Clock. We managed to increase the memory by +500. This puts the memory at 20GHz versus the default 19GHz. This increases the memory bandwidth up to 960GB/s from the default 912GB/s.
The memory overclock was really an interesting one. Technically we could overclock the memory higher. However, this caused issues. With GDDR6X it has better error correction, instead of just throwing out errors it starts to throttle performance if overclocked too high, so you have to find the right balance.
However, the main issue with this video card was the temperature of the memory and the power demand when overclocking. It already runs very hot and power-demanding without overclocking. Overclocking it at default fan speeds causes the memory temperature to increase to unconformable levels for us. It also causes the power demand to skyrocket, which reduces the total overall GPU clock frequency overclock because it hits the TDP wall faster.
It also just doesn’t really improve performance much, frankly, because memory bandwidth is already very high on this video card. It’s more advantageous to overclock the GPU frequency, over the memory frequency. Therefore, we intentionally kept it a little bit lower than its highest overclock potential, in order to allow the GPU clock to overclock to its highest.
In the graph above you can see what a +130 does to the GPU Clock. This boosts it quite a bit over the default frequency. Where we averaged 1804MHz at default, with our overclock we are now averaging 1953MHz. You can see it hovers around the 1945MHz area a lot. This is an 8% overclock of the GPU frequency.
According to GPU-Z sensor data, it achieved these frequencies with a 100% fan speed at 66c GPU Temperature and 80.5c Hot Spot Temperature. GPU Voltage topped out at 1.0750V. This put it at its maximum power consumption of 114% TDP, and Board Power Draw was 399.1W and GPU Chip Power Draw was 213.4W.