Although the AD104 is used by both the Geforce RTX 4070 and the RTX 4070 Ti, the level of performance is too great due to further changes for the RTX 4070 (here our XXL test) to reach the pure fps power of the RTX 4070 Ti could.
The Geforce RTX 4070…
- … is equipped with 14 fewer shader multiprocessors (CU/SMs) (46 vs. 60 units).
- … has 5,888 instead of the full 7,680 FP32 ALUs.
- … offers less level 2 cache with 36 MiB instead of 48 MiB.
- … is based on the same chip basis as the RTX 4070 Ti, but is limited more than just on the clock/power draw side.
- … like all stronger Ada cards, supports DLSS 3.0 (frame generation).
Nevertheless, it is exciting to what extent the differences between the models also with manual tuning such as classic overclocking (OC) or undervolting (for higher efficiency) fail – also with regard to older ampere models such as the Geforce RTX 3080 with 12 GiB.
To show the nuances, two example scenarios can be used:
- WQHD with ray tracing (Cyberpunk 2077, 1st page) as well
- UHD with pure rasterizing (God of War, 2nd page) – both with DLSS Quality.
David versus Goliath, literally: The weaker giant on the left (RTX 4070 Gigabyte Aorus Master, David versus Goliath, literally: The fainter giant on the left (RTX 4070 gigabyte Aorus Master, with the thickest cooler and power limit) must join the smaller map (RTX 4070 Ti Gigabyte Eagle OC) beaten on the right with a stronger AD104 expansion.No matter how hard you overclock and push the power limit, you won’t be able to beat the Geforce RTX 4070 up to the RTX 4070 Ti.But that efficiency level can very well be achieved (practically approximately) – see results.Source: PC Games Hardware
Example 1: Ray tracing
You have to give Nvidia one thing, in terms of power draw, or rather efficiency, the newer Ada models consume significantly less than the previous generation.Even with undervolting, our Geforce RTX 3080 (also with 12 GiB VRAM) sent into the race with approx. 288 watts of board power is still more power-hungry than all the tuning scenarios of the competitors.
Unique pattern: The efficiency level of the two AD104 models is on par, but with the full configuration (the RTX 4070 Ti) it is possible to achieve slightly more fps per watt, provided you create an effective, stable voltage curve yourself (targeted undervolting).In the standard configuration, the RTX 4070 scales beautifully with its larger sister card.The Ampere GPU has to admit defeat completely, and without the artificial intermediate frames.
Cyberpunk 2077: Ultra RTX, WQHD, DLSS Q – “Red Light Alley”
Cyberpunk 2077: Ultra RTX, WQHD, DLSS Q – Red Light Alley
|
Surname
1440p, DLSS Q (Average Fps)
Minimum fps
It’s lonely at the top: The Geforce RTX 4070 Ti mercilessly ignores any tuning of the other cards, with maybe one exception: The OC profile of over 3.0 GHz puts the RTX 4070 quite close to one undervolted RTX 4070 Ti back.That means: only if you let the smaller one work at its stable clock limit and the stronger one work more efficiently can one definitely speak of a similar frame rate.Only the efficiency suffers.In the standard configuration (stock) or even with overclocking, the RTX 4070 Ti is unattainable.Is it different with rasterizing?On the next page we look at God of War in pure (classic) rasterizing.