Following the launch of the first of three GPUs coming to the market soon, the RTX 4070 Ti SUPER is the second most anticipated by gamers. In fact, as we have seen, its performance-to-price ratio is expected to increase significantly, making it very attractive to gamers. Fortunately, just days before its arrival, this GPU has been spotted in a suite like 3DMark, showing results that can be categorized as “somewhat disappointing” for a high-end GPU. So, how does this RTX 4070 Ti SUPER perform in 3DMark benchmarks?
The RTX 4070 SUPER, the younger sibling of the 4070 Ti SUPER, showed a significant performance jump compared to its predecessor for just a few more euros. Many considered it what should have been launched initially, but as they also say, better late than never. That’s why there is a lot of interest in whether the next GPU in the lineup can maintain a similar performance ratio.
RTX 4070 Ti SUPER in 3DMark: half-empty or half-full?
Before we speculate on the performance, let’s first look at the leaked data for the RTX 4070 Ti SUPER in 3DMark. There are seven key benchmarks to consider, with the first being Speedway at 2K. This GPU achieves 6,236 points compared to 5,474 points for its predecessor, an increase of 13.92% in this benchmark.
As for Port Royal at the same resolution, the percentage differences are curiously pretty close. The new NVIDIA GPU manages to be 10.20% faster than the original model, but at the same time, the RTX 4080 outpaces both slightly.
In Time Spy at 2K, the data is a bit more disappointing, with just a 5.18% improvement, which is not enough to justify purchasing the new model. That said, depending on the benchmark, the differences can vary greatly and are not constant. But is this trend repeated in the following tests?
Time Spy Extreme and Firestrike confirm the performance
In Time Spy Extreme, the GPU’s differences should be more evident for obvious reasons. However, the improvement is only 7.72%, still too small to justify the expense for this GPU. What about the three Firestrike benchmarks at various resolutions?
In the standard test at 1080p, the situation doesn’t improve: only a 3.26% difference between the two graphics cards, with the RTX 4080 performing significantly better. In Fire Strike Extreme at 2K, there is some improvement, with a 6.20% increase, nearly double the difference with a single resolution jump, but still not enough. Can Fire Strike Ultra at 4K make a difference?
Not really, as the percentage difference remains at 6.18%. Averaging the performance across the seven benchmarks results in a 7.22% improvement, which is roughly half of what was expected, around 15%. Is there something amiss here? First, we should rule out driver issues. The tests and data were taken with GeForce 551.15, which is the driver NVIDIA supplied for reviews and is not yet publicly available.
The analysis then raises other questions. It’s said that these benchmark scores don’t truly reflect the potential of the graphics card. The primary reason behind this is its VRAM. NVIDIA appears to have opted for 16GB of VRAM with 4K gaming in mind to compete with the RX 7900 XT and close the gap with the RTX 4080.
Goal: real-world gaming beyond benchmarks
The objective is real-world gaming, where poorly optimized games can benefit from more VRAM, reducing the reliance on the interface and texture loading from RAM, resulting in a performance boost. The 256-bit bus also provides an edge in more demanding scenarios, be it due to good or bad game optimization, because at 4K and especially with Ray Tracing, any help is appreciated.
The additional ROPs should be noticeable, and along with the 256-bit bus and 16GB of VRAM, they should reduce the performance gap with its more powerful sibling. The issue with 3DMark benchmarks is that they don’t max out the VRAM, and therefore, they don’t show the improvements that should be apparent with a range of games.
It’s an unusual situation because the real-world performance of the card in well-optimized games should be around 10%, but due to the poorly optimized games, the difference should increase to around 15% with the RTX 4070 Ti SUPER’s features, something that 3DMark cannot demonstrate. We’ll know more in the coming days and will see if the arguments we’ve seen today are closer to reality.
Has NVIDIA been clever this time by adopting AMD’s strategy of more bus, ROPs, and VRAM? Will 3DMark have to develop a benchmark that maxes out a GPU’s VRAM to show a new real-world scenario as a result of the RTX 4070 Ti SUPER’s performance? We’ll find out soon.