Technical characteristics and specifications
|Lithographic process||8nm Samsung|
|Die size||392 mm2|
|SM / CU||48|
|Hidden L2||4 MB|
|Reloj base game mode||1575 MHz|
|Reloj Boost game mode||1770 MHz|
|Basic clock OC mode||No|
|OC boost mode Reloj||No|
|Reloj VRAM||1188 MHz (19 Gbit / s)|
|VRAM bandwidth||608.3 GB / s|
|FP32 performance||21.75 TFLOPS|
|Food||1 x 12 pins|
|NVLink / SLI / Crossfire||No|
|Video outputs||1 x HDMI 2.1 and 3 x DisplayPort 1.4|
|Introductory price||$ 599|
|Release date||June 10, 2021|
Unpacking and external analysis
- Intel Core i7-8700K
- ASUS Maximus X Formula
- ASUS Ryujin 240
- Corsair Vengeance Pro RGB DDR4-3200 MHz Cas16 (Samsung B-Die, XMP enabled)
- Corsair AX1200i Platinum
- Corsair MP510 960 GB
- ASUS STRIX Helios
- EK Vardar EVO RGB (x 4)
- Tacklif HM02
- Key DNM-51
Thermal and acoustic performance
With this NVIDIA RTX 3070 Ti FE 8GB, we have mixed feelings, and it’s no wonder. Aesthetically, we have the RTX 3070 heatsink with the RTX 3080 and RTX 3080 Ti pull system, a very curious mix that NVIDIA saw fit to use given the high TDP of this card for the range in which it is found.
It is true that she is closer to the top than to the average high, and perhaps opting for the same version as her big sisters would have been more correct, but they are the ones who are and are who are as they say. We have already seen the result: it gets the highest temperature recorded so far, the loudness is equal to the RTX 3080 Ti which gets better temperatures with a higher TDP and by far, ergo it is clear that this radiator is at the limit.
It is true that the difference between stock and overclock in these terms is minimal and there it regains some ground, but this is not what is really wanted because the average user does not even understand the term. Finally, among what we didn’t like, we have again the problem of the drivers which we have already seen in the RTX 3080 Ti, although in this case it has been mitigated in many titles. Others like BFV in DX11 lose a lot of FPS going to 60% on average, and in DX12 the stuttering is obvious compared to what we see with the previous API.
Like its older sister Ti, the higher the resolution or the graphics load, the GPU evolves in percentage of use and maintains the FPS despite the greater weight of the resolutions, at least until reaching the desired 99%. It’s clear that NVIDIA is still working on fixing the issue, and it’s more than likely that in a few releases it will be fixed at the rate they evolve.
Leaving the bad side, it is clear that the build quality is unbeatable, as always at Nvidia with its radiators of the moment. The overclocking is very good (+175 MHz on Core and +450 MHz on the memos), the performance is high, exactly what was expected, one step ahead of the RTX 3070 and one step below the RTX 3080.
This is a graphics card focused on 2K Ultra or UWHD, a bit tight for 4K in some titles unless you lower the graphics settings. In short, at the price at which it is released on the market and given that it is currently it is perhaps the best GPU in performance / price of the NVIDIA range. The problem is logically with the RX 6800, which although it is normally higher, costs 20 dollars less, consumes less, heats less and is quieter.
In favor of this 8GB RTX 3070 Ti FE, we have NVIDIA’s proprietary technologies such as DLSS, Ray Tracing and G-SYNC which are far superior in performance, smoothness and FPS than what is offered by AMD. In this type of case where the price and performance are close, you have to look at these details, because they make the difference and in this case the cat is taken by NVIDIA.