The RTX 4070 Ti is a mid-range graphics card in the upper price segment
Product test

The RTX 4070 Ti is a mid-range graphics card in the upper price segment

Kevin Hofer
23.1.2023
Translation: Patrik Stainbrook

Nvidia’s third graphics card featuring Lovelace architecture offers plenty of power. However, like all current cards, it’s simply overpriced.

If you game in 1440p at max settings, the Gigabyte GeForce RTX 4070 Ti Aero OC 12G should be the perfect choice. However, at over 900 francs/euros, the mid-range card inexplicably hangs around the upper price bracket. I’d therefore advise against it at the moment.

Design and ports: not super-sized, for once

The RTX 4070 Ti is compact compared to its big sisters the RTX 4080 and 4090. At 33.6 × 14 × 5.8 centimetres, the card should fit most ATX and even some Mini-ITX systems. Three 100-millimetre fans and a heat sink are supposed to keep temperatures low. The card is white and has only one illumination zone: the Aero lettering.

The Gigabyte GeForce RTX 4070 Ti Aero OC 12G only has one illumination zone.
The Gigabyte GeForce RTX 4070 Ti Aero OC 12G only has one illumination zone.

The RTX 4070 Ti uses the PCIe 4.0 interface. It offers one HDMI 2.1 port and three DisplayPort 1.4 ports. The ports provide enough throughput for the card to perform.

Specifications: RTX 4080 into RTX 4070 Ti

The third desktop graphics card with Ada-Lovelace architecture is manufactured at TSMC in a 4 nm process. There are 60 streaming multiprocessors active on the AD 104 chip. This results in 7689 Cuda cores and corresponds to the chip’s full configuration. The specifications are identical to the RTX 4080 12 GB introduced last September. That card wasn’t released due to negative feedback. Now Nvidia is giving it another go with the RTX 4070 Ti.

The Gigabyte model reached 2640 MHz in tests – 30 MHz more than the reference model. Its maximum power consumption is 285 watts, Gigabyte recommends at least a 750-watt power supply. The card gets supplied by a 16-pin 12VHPWR power connector. A corresponding adapter for two 8-pin power plugs is included.

Compared to its predecessor RTX 3070 Ti, the RTX 4070 Ti has 25 per cent more shaders, twelve times as much L2 cache and 50 per cent more memory. In return, it has a slower memory interface. Price-wise, the card costs just under 40 per cent more.

Here are the specifications in comparison with other already-tested latest-gen graphics cards.

Test set-up and methods

I used the following components for this review, which were provided to me by the manufacturers for testing:

Intel Core i9-13900K (LGA 1700, 3 GHz, 24 -Core)
Processors
EUR518,21

Intel Core i9-13900K

LGA 1700, 3 GHz, 24 -Core

ASUS ROG MAXIMUS Z790 HERO (LGA 1700, Intel Z790, ATX)
Motherboards
EUR491,63

ASUS ROG MAXIMUS Z790 HERO

LGA 1700, Intel Z790, ATX

G.Skill Ripjaws S5 (2 x 16GB, 6000 MHz, DDR5 RAM, DIMM)
RAM
EUR121,52

G.Skill Ripjaws S5

2 x 16GB, 6000 MHz, DDR5 RAM, DIMM

Crucial P5 (1000 GB, M.2 2280)
SSD

Crucial P5

1000 GB, M.2 2280

PNY XLR8 CS3030 (1000 GB, M.2 2280)
SSD

PNY XLR8 CS3030

1000 GB, M.2 2280

The system runs on Windows 11 version 21H2 (22621.1105). I’m using BIOS version 0703 and enable XMP. Otherwise, I leave everything on default and Resizable BAR is activated. For the graphics card, I use driver version 528.02.

Here’s an overview of the different benchmarks:

  • Games: «Anno 1800», «Forza Horizon 5», «Assassin’s Creed: Valhalla», «Civilization VI», «Cyberpunk 2077», «Far Cry 6», «Gears of War 5», «Red Dead Redemption 2», «Shadow of the Tomb Raider»
  • Games with ray tracing: «Cyberpunk 2077», «Far Cry 6», «Shadow of the Tomb Raider»
  • 3DMark: Fire Strike, Fire Strike Ultra, Time Spy and Time Spy Extreme
  • Blender Benchmark
  • Photoshop
  • Premiere

I ran each benchmark three times and took the best result. For the games, I used the highest possible presets. Otherwise, I left everything at default except for the resolution. I left DLSS or FSR deactivated. In this review, I’m looking at the rasterisation and ray tracing performance of various games without additional tools.

Performance in games: almost as strong as an RTX 3090

On the following slides, you can see the mathematical mean of frames per second (FPS) for the nine benchmark games compared to other already-tested graphics cards. Clicking through the gallery, you’ll see the results from individual games.

In 1080p resolution, the RTX 4070 Ti is only twelve per cent slower than the current fastest graphics card, the RTX 4090. Compared to AMD’s best card, the RX 7900 XTX, there’s only an eight per cent difference – same goes for its big sister, the RTX 4080. It’s even enough for a narrow victory, with a two per cent edge in FPS against the RTX 3090.

However, the gap between it and higher models grows with increasing resolutions. At 1440p, the RTX 4070 Ti is just over one per cent behind the RTX 3090. At 2160p, it’s already nine per cent. Compared to the top model, this becomes 29 and 61 per cent respectively. It places closer to its big sister RTX 4080, 16 and 24 per cent removed.

The RTX 4070 Ti feels most comfortable in 1440p. Here, one FPS on the tested model costs an average of 6.6 francs. Frames on the RTX 4080 are considerably more expensive at 8.9 francs. One frame costs 8.05 francs on the only competing AMD card we tested (all data from 20.01.2023). Thus, the RTX 4070 Ti is clearly the top dog in this race.

In terms of frame times in percentiles, the RTX 4070 Ti is roughly on par with the RTX 3090. Compared to the other cards, however, it has to admit defeat in some cases. The percentile values are usually frame times measured in milliseconds. These are the time intervals from image to image, or frame to frame. 99th percentile means that 99% of all data is faster than the value provided.

As expected, the RTX 4070 Ti performs worst in the three ray-traced games. However, it does score a success against AMD’s model in 1080p.

3DMark benchmarks

The graphic shows the mathematical mean of the four 3DMark benchmarks, followed by the individual results. I’ll only be giving the values for the graphics card. Differences in the overall ranking from tests are just too big.

In contrast to game benchmarks, the RTX 4070 Ti broadly beats the RTX 3090 in 3DMark. Even if only by a narrow two per cent margin and only in the Fire Strike benchmarks. The RTX 3090 regains its crown in Time Spy. However, the tested graphics card doesn’t stand a chance against the other models.


What the benchmarks test
3DMark’s synthetic game benchmarks render game-like scenarios. From this, they calculate a score that indicates theoretical in-game performance. I’m only giving you the values for the GPU.


Productivity benchmarks: brilliant performances in Blender

On the following slide, you can see the mathematical mean of results from three productivity benchmarks. If you click through the gallery, you’ll see the results of the individual benchmarks.

The productivity benchmarks once again show the superiority of Nvidia cards over AMD in this area. Even the mid-range RTX 4070 Ti puts the flagship RX 7900 XTX in its place. Compared to the in-house competition, it’s at least enough for a victory over the RTX 3090. However, this is mainly thanks to an excellent result in Blender.


What the benchmarks test
The Blender Benchmark renders three scenes in the 3D graphics suite in version 3.3 and higher and calculates three scores. I’ve added these up to give end scores for each.
The photo editing and video editing benchmark UL Procyon simulates various workloads in Adobe Creative Suite. Finally, the benchmark calculates a score.


Power consumption, volume and temperature: not enough for the efficiency crown

The maximum thermal design power (TDP) of 285 watts is rarely reached by the RTX 4070 Ti. On average, the power consumption remains at 230 watts in 2160p across the entire benchmark suite. Thus, the Nvidia card requires the absolute minimum in power out of all the cards in this generation I’ve tested. Relative to the speed, however, things look different. In 2160p, the RTX 4070 Ti needs 2.6 watts per FPS. Its big sister RTX 4080 is more efficient with 2.45 watts per FPS. Nevertheless, the leap in performance compared to the last-gen RTX 3090 is enormous. That one needs 3.45 watts per FPS.

When gaming, temperatures reached a maximum of 59 degrees Celsius. On average, I measured 53 degrees Celsius across all games on the open testbench.

All other data can be found in the following graphs.

When idle, watching Netflix and browsing, the card isn’t audible, the fans remain still. In games, I measured 40 dB from a distance of 30 centimetres. That’s pretty quiet. In rare cases, like Time Spy Extreme, they can max out at 42 dB.

Verdict: good, but unconvincing

Performance-wise, the RTX 4070 Ti takes up a similar spot in the RTX 40 series as its predecessors do in their families. This makes perfect sense for Nvidia’s graphics card portfolio. However, its price seems anything but reasonable for a mid-range card. Over 900 francs/euros. That’s around 300 to 400 francs/euros more than the previous mid-range model cost. A steep markup.

Depending on the resolution, the Gigabyte GeForce RTX 4070 Ti Aero OC 12G is between eight and 24 per cent behind its big sister RTX 4080 and AMD’s flagship RX 7900 XTX. Compared to the RTX 3090, it’s almost on par in 1080p and 1440p. In UHD, however, it’s again slower than the last-gen card. That’s not bad in itself, and the generation-to-generation performance leap is greater than before. However, its performance isn’t anywhere near 80 per cent better than the RTX 3090 Ti, as Nvidia advertised in the run-up to release. Still, we can assume that DLSS 3.0 upscaling technology was enabled in these comparisons.

The target resolution for the RTX 4070 Ti is 1440p. Frame rates beyond 120 FPS are the order of the day here. If you reduce the level of detail, you can even expect values around 120 FPS in UHD.

Compared to the previous generation, the change in manufacturer from Samsung to TSMC and the improved process are also evident. With 2.6 watts per FPS, the RTX 4070 Ti requires 0.8 watts per FPS less than the RTX 3090. Efficient, but it doesn’t match the RTX 4080, which only needs 2.45 watts per FPS.

12 GB of memory is sufficient for my benchmark games. But there are other games where this might hinder performance. Nvidia is being stingy with 12 GB.

All in all, the RTX 4070 Ti isn’t a bad graphics card. However, the model I tested is too expensive compared to its predecessor. If you don’t care about that and you’re gaming in 1440p, I can recommend this card. However, only if the RTX 4090 is too expensive for you. That model remains the best of all new cards in terms of value for money.

Cover picture: Kevin Hofer

36 people like this article


User Avatar
User Avatar

From big data to big brother, Cyborgs to Sci-Fi. All aspects of technology and society fascinate me.


These articles might also interest you

Comments

Avatar