What is your verification environment?
The verification environment this time uses the same verification environment as the previous RTX 3080 Ti FE verification environment. Since only one week has passed since the previous review, we will try to compare the performance of the new GeForce again by inheriting the previous verification results.
In addition, we added the RTX 3070 FE and RTX 2070 SUPER FE as comparison targets this time. As for the driver environment, the RTX 3070 Ti FE and RTX 3080 Ti FE are measured with a beta driver for review, and the others are measured with GeForce 466.47. In addition, all GPUs except RTX 2070 SUPER were tested in an environment where Resizable BAR can be used.
[Verification environment] | |
---|---|
CPU | AMD "Ryzen 9 5950X" (16 cores/32 threads , 3.4-4.9GHz), |
CPU cooler | Corsair "iCUE H115i RGB PRO XT" (simple water cooling, 280mm radiator) | < /tr>
Motherboard | GIGABYTE "X570 AORUS MASTER" (AMD X570, BIOS F33i) |
Memory | < td>G.Skill "Trident Z RGB F4-3200C16D-32GTZRX" (DDR4-3200, 16GB x 2) x 2|
Video card | NVIDIA "GeForce RTX 3090 Founders Edition", NVIDIA "GeForce RTX 3080 Ti Founders Edition", NVIDIA "GeForce RTX 3080 Founders Edition", NVIDIA "GeForce RTX 3070 Ti Founders Edition", NVIDIA "GeForce RTX 3070 Founders Edition", NVIDIA "GeForce RTX 2070 SUPER Founders Edition" |
Storage | Corsair "CSSD-F1000GBMP600" (NVMe M.2 SSD, 1TB) |
Power supply unit | Super Flower "SF-1000F14HT" (80PLUS Titanium, 1000W) |
OS | Microsoft "Windows 10 Pro 64bit version" (May 2021 Update) |
NVIDIA's claim was confirmed "3DMark"
First It is a score comparison of "3DMark". The RTX 3070 Ti FE has more CUDA cores and memory bandwidth, but the memory bus width is left at 256 bits, and the performance gain tends to slow down for models with fewer CUDA cores, so it landed closer to the RTX 3070 FE than the RTX 3080 FE. can be expected to
3DMark score
First, when the score of the RTX 3070 Ti FE is 100%, the RTX 2070 SUPER FE is about 73% to 67%, so NVIDIA's claim that it has 1.5 times the performance of the RTX 2070 SUPER is almost achieved. (1÷0.67=1.49) was confirmed.
Next, looking at the confrontation between the top and bottom of the RTX 3080 Ti FE, the RTX 3080 FE is about 11 to 31% higher than the 3070 Ti FE, and the RTX 3070 FE is about 6 to 8% lower than the 3070 Ti. The results were obtained. The RTX 3080 Ti had a performance closer to the high-end GPU (RTX 3090), but the RTX 3070 Ti had a lower performance than the RTX 3080. It is believed that the 256-bit memory bus specification makes it difficult to enjoy the benefits of increasing memory bandwidth and increasing the number of CUDA cores by adopting GDDR6X.
"Rainbow Six Siege" where the difference from the high-end GPU is clear in the high-resolution area
From here onwards, it will be verified using the actual game. In "Rainbow Six Siege", the API is Vulkan, and the image quality is "best" based on the render scale of 100%. Measure the frame rate using the in-game benchmark function.
"Rainbow Six Siege" Vulkan, frame rate at 1920 x 1080 dots
"Rainbow Six Siege" Vulkan, frame rate at 2560 x 1440 dots
"Rainbow Six Siege" Vulkan, frame rate at 3840 × 2160 dots
While there was almost no difference between the RTX 3080 Ti and 3090, the RTX 3070 Ti FE has an average frame rate of about 7-9% higher than the 3070 FE. This difference tends to widen as the resolution increases, and it can be seen that the increase in memory bandwidth due to the adoption of GDDR6X has had a certain effect.
However, the difference from the top RTX 3080 FE is not small, and the RTX 3080 FE is about 13% higher in full HD and about 22% higher in 4K. The difference in the number of CUDA cores itself is also large, but it is thought that the difference in the fundamental specifications of the memory bus 320bit vs. 256bit produces this result. However, I would like to emphasize that since it is a light drawing game, even the RTX 3070 Ti FE has a frame rate that cannot be used on an ordinary gaming display.
"Apex Legends", which seems to be cool even in 4K
Next, in "Apex Legends", set the image quality to the highest image quality, and explicitly remove the frame rate upper limit in the startup options "+fps_max unlimited" added. Using "CapFrameX", we measured the frame rate when taking a series of actions at the shooting range.
"Apex Legends" frame rate at 1920 x 1080 dots
"Apex Legends" frame rate at 2560 x 1440 dots
"Apex Legends" frame rate at 3840 x 2160 dots
Currently, the specifications are such that 300fps cannot be achieved no matter how hard you try, so the GPU above the RTX 3080 FE is almost in a dumpling state, but below the 3070 Ti, the difference in GPU specifications is easy to see numerically. The average frame rate of the RTX 3070 Ti FE is about 3-10% higher than the 3070 FE and about 5-22% lower than the RTX 3080 FE, especially in 4K where the memory bandwidth tends to be a bottleneck, and the RTX 3070 Ti FE is out of breath. It has been shown to be easy
However, even in 4K, we managed to keep 60fps, so if the image quality is slightly reduced, it seems that we can obtain a frame rate that is sufficient for play. However, if you want to ensure high frame rate play, it would be better to operate the RTX 3070 Ti in full HD to WQHD.