Top video cards always attract attention, especially when the next generation of graphics adapters offers fundamentally new gaming or computing capabilities. Obviously, it is interesting to find out “what progress has come to” and compare the achievements of such models with your needs. However, the four-digit price tags of such adapters do not bother except the most demanding players, while the general public is much more interested in more affordable models. The new line of video cards from NVIDIA has been replenished with the GeForce RTX 5070 Ti . Will the new series be the best choice for demanding buyers who, however, are used to counting money? We will deal with the question by experimenting with the ASUS TUF Gaming GeForce RTX 5070 Ti OC .
NVIDIA GeForce RTX 5070 Ti
The GeForce RTX 5070 Ti uses the same Blackwell architecture GPU as the older GeForce RTX 5080, the GB203. But this time it’s the GB203–200 variant, which offers 8,960 active compute units, instead of 10,752 in the RTX 5080 variant. The chip is manufactured by TSMC using the 5nm process in an improved version (4N) with some optimizations for NVIDIA’s needs. The GPU’s silicon die has an area of 378 mm² and contains 45.6 billion transistors. The chip has a base operating frequency of 2,300 MHz with an average boost of 2,452 MHz.
| GeForce RTX 5070 Ti | GeForce RTX 4070 Ti SUPER | GeForce RTX 4070 Ti | |
| Graphics processor | GB203-300 | AD103-275 | AD104-400 |
| Architecture | Blackwell | Ada Lovelace | Ada Lovelace |
| Technology | 5 nm, TSMC | 5 nm, TSMC | 5 nm, TSMC |
| CUDA-computers | 8960 | 8448 | 7680 |
| TMU/rasterizers | 280/128 | 264/96 | 240/80 |
| Tensor kernels | 280 (5th generation) | 266 (4th generation) | 240 (4th generation) |
| RT blocks | 70 (4th generation) | 66 (3rd generation) | 60 (3rd generation) |
| GPU clock speeds | 2300/2452 MHz | 2340/2610 MHz | 2310/2610 MHz |
| Memory capacity | 16 GB GDDR7 | 16 GB GDDR6X | 12 GB GDDR6X |
| Memory bus | 256 bits | 256 bits | 192 bit |
| Effective memory frequency | 28,000 MHz | 21,000 MHz | 21,000 MHz |
| Memory bandwidth | 896 GB/s | 672 GB/s | 504 GB/s |
| Graphics card power consumption (TGP) | 300 W | 285 W | 285 W |
| Starting recommended price | $749 | $799 | $799 |
| Release date | February 2025 | January 2024 | January 2023 |
How about comparison with predecessors? Here we need to consider which model is considered the predecessor - GeForce RTX 4070 Ti SUPER or RTX 4070 Ti. The older one also uses the AD103 GPU from the model one step above - RTX 4080, which has very similar technological parameters (379 mm², 45.9 billion transistors). While for the GeForce RTX 4070 Ti, a lightweight version of the AD104-400 processor with a die area of 294 mm2 and 35.8 billion transistors on board was used. Interestingly, the declared operating frequencies of the graphics processor in the case of the RTX 5070 Ti are even slightly lower than those of the previous generation models. This especially applies to the average acceleration (-160 MHz).
Returning to the functional features of the GeForce RTX 5070 Ti, we note that the new model has retained the 256-bit bus and received 16 GB of memory. The video card also uses GDDR7 chips with an effective operating frequency of 28,000 MHz. In this configuration, the memory bandwidth is 896 GB/s. And this is a third more than the RTX 4070 Ti SUPER had (672 GB/s). If we compare the RAM with that of the basic RTX 4070 Ti (504 GB/s), then the difference is 78%. Increased memory bandwidth will never be superfluous, but will certainly affect the final performance results. Especially in high-resolution modes.
The GeForce RTX 5070 Ti has a claimed total power consumption of 300 W, and for a system with this video card, it is recommended to use power supplies with a capacity of 750 W or more. The previous RTX 4070 Ti/4070 Ti SUPER models had a TGP of 285 W.
As for the recommended price, the MSRP for the base modifications of the GeForce RTX 5070 Ti line is $749. A potentially attractive value, but the actual price tags of most adapters in the new series will obviously be higher.
In the case of the GeForce RTX 5070 Ti, the developers do not plan to release a reference Founders Edition model, relying entirely on partners. Of course, the latter have already prepared many original modifications, and we will get to know one of these versions better.
ASUS TUF Gaming GeForce RTX 5070 Ti OC
ASUS, as expected, offered several GeForce RTX 5070 Ti variants at the start of sales. Among the presented versions, one of the most interesting is the ASUS TUF Gaming GeForce RTX 5070 Ti OC model. A hefty representative of the TUF Gaming family with an initially overclocked GPU.
The video card comes in a relatively compact box with a bright exterior. The graphics adapter comes with a short installation manual, adapters from three 8-pin PCI-E connectors to one 16-pin (12HVPWR), a magnetic sticker with the TUF logo, a velcro strap, and a stand to support the adapter.
ASUS TUF Gaming GeForce RTX 5070 Ti OC received a factory overclock. Instead of the recommended frequency formula of 2300/2452 MHz, the GPU has a combination of 2300/2588 MHz. The GDDR7 memory chips operate at the recommended 28,000 MHz (effective frequency)
The recognizable design makes it quite easy to guess that the video card belongs to the TUF Gaming series. The combination of brutal chopped shapes and restrained exterior design with predominant metal accents. The model turned out to be quite large – 348×146×72 mm. Yes, it is somewhat more compact than the ASUS ROG Astral GeForce RTX 5080 OC, which we recently reviewed, but even the declared compliance with the 3.6-slot design requires the presence of appropriate space inside the case.
For cooling ASUS TUF Gaming GeForce RTX 5070 Ti OC uses a three-fan cooler with a massive combined radiator, consisting of two stacked cassettes of aluminum plates. The blocks are interconnected by nickel-plated heat pipes. The surface of the heat sink has an additional super-finish (MaxContact) for maximum contact with the graphics processor crystal. In addition, a thermal interface with a phase transition is used to accelerate heat transfer.
To blow the structure, three 100 mm fans with Axial-Tech technology are provided, which involves the presence of a ring connecting the blades on the outer radius. Models with double ball bearings with increased reliability are used (80,000 hours of operation are claimed).
One of the features worth noting is that the central fan rotates in the opposite direction to the side ones. In this way, the developers are trying to reduce the effect of air flow turbulence.
To connect additional power, a 16-pin 12VHPWR connector is provided on the top panel. In general, the TGP for the GeForce RTX 5070 Ti is 300 W, so NVIDIA suggests using an adapter with two 8-pin PCI-E directly from the PSU if there is no cable with a 12VHPWR connector. But if we are talking about a forced version of the video card, a splitter with three PCI-E connectors will not be superfluous. That is why the ASUS TUF Gaming GeForce RTX 5070 Ti OC is equipped with such an adapter. As for the recommended PSU power, the recommendations here coincide with those provided by NVIDIA - from 750 W.
ASUS TUF Gaming GeForce RTX 5070 Ti OC is equipped with two BIOS chips with the ability to switch between them using a toggle switch located near the top of the PCB. The basic “P Mode” provides a standard mode with improved cooling, while “Q Mode” allows you to slightly reduce the noise level of the cooling system by using a CO algorithm with lower fan speeds.
In general, the strict TUF Gaming style allows for some “relaxation” when it comes to additional lighting. The series logo on the side panel is an unnecessary reminder that the video card belongs to the manufacturer’s corresponding product line. Modest and tasteful.
The back of the ASUS TUF Gaming GeForce RTX 5070 Ti OC is covered by an aluminum plate with holes for through-flow of the radiator by one of the fans. Such a design fully justifies itself. For the top models of the new ROG Astral line, the developers even provided an additional fan to enhance the airflow, but here they did without large holes.
We should also note the presence of a metal bracket in the GPU location area to balance the pressure force of the radiator block.
The interface panel, which is covered by a stainless steel cover, has five video outputs - three DisplayPort 2.1b and a pair of HDMI 2.1b. The video card allows simultaneous display on four screens, but the larger number of ports adds flexibility to switching.
Test system configuration
- Motherboard: ASUS ROG STRIX Z890-F GAMING WIFI
- Processor: Intel Core Ultra 7 265KF (8P+12E; 3.9/5.5 GHz + 3.3/4.6 GHz)
- Cooling: ASUS ROG STRIX LC III 360
- Memory: G-SKILL Trident Z5 RGB DDR5-6400 32GB (F5-6400J3239G16GX2-TZ5RK)
- Graphics card: ASUS ROG Strix GeForce RTX 4080 SUPER 16GB GDDR6X OC Edition (ROG-STRIX-RTX4080S-O16G-GAMING)
- Drive: KINGSTON NV3 2 TB (SNV3S/2000G)
- BZ: ASUS TUF Gaming 1000W Gold
- Case: ASUS TUF Gaming GT502 PLUS
To test the ASUS TUF Gaming GeForce RTX 5070 Ti OC, we used an Artline PC based on the Intel LGA1851 platform, which was equipped with a 20-core Core Ultra 7 265KF processor and an ASUS ROG STRIX Z890-F GAMING WIFI motherboard based on the top-end Intel Z890 chipset. The system has already worked during a recent test of the more powerful ASUS ROG Astral GeForce RTX 5080 OC, so it remains only to recall its features.
To cool the CPU, an ASUS ROG STRIX LC III 360 liquid cooling system with a three-fan layout, a radiator block of appropriate dimensions and a fairly powerful pump was used. The platform was complemented by a dual-channel G-SKILL Trident Z5 RGB DDR5-6400 RAM kit with a total capacity of 32 GB, and a KINGSTON NV3 2 TB drive was used for data storage.
The power supply was provided by the ASUS TUF Gaming 1000W Gold “kilowatt”, which allowed us to play around with the graphics card overclocking without any problems.
The PC was assembled in an ASUS TUF Gaming GT502 PLUS case. The modular two-chamber model with good ventilation of the internal space provides the necessary temperature regime for a fairly powerful platform.
In operation
Let's start our study of the capabilities of a video card by evaluating its temperature performance in different operating modes.
In idle mode, the GPU warmed up to only 32C, the temperature of the memory chips was kept at 42C. In this mode, the cooling system fans stop, and accordingly the video card operates silently. The GPU operating frequency was reduced to 217 MHz, the DRAM chips to 50 MHz (800 MHz effective). We note that the power consumption level was about 16 W, which is slightly more than we observed in the ASUS ROG Astral GeForce RTX 5080 OC.
Under prolonged gaming load, the GPU and GDRR7 chips warmed up to the same limit – 64C. To maintain this temperature, the fan speed was increased to 1580 rpm. The operating frequency of the graphics processor at the peak increased to 2872 MHz.
When using Q Mode, the temperature of the GPU and memory chips rose to 70C, but the fan speed only increased to 1100 rpm. The video card works noticeably quieter. Of course, if the difference is not covered by other sources of noise in the system. In general, the temperature indicators are quite acceptable, so the user has the opportunity to independently choose the appropriate mode, depending on their own priorities.
GeForce RTX 5070 Ti test results
To evaluate the results of the GeForce RTX 5070 Ti 16 GB, we compared its performance with that of the GeForce RTX 4070 Ti SUPER 16 GB. Despite the fact that the new model is formally the successor to the GeForce RTX 4070 Ti 12 GB, it is the SUPER modification with increased memory capacity that is relevant on the market and has a similar recommended price. In addition, the GeForce RTX 4070 Ti SUPER model is the base video card for progressive configurations of our traditional “PC of the month”, so here we can immediately assess the potential changes for powerful gaming systems.
Since we recently tested the GeForce RTX 5080, comparing the results with those of the GeForce RTX 4080 SUPER, for convenience, the charts will also present the performance of these models. The same test platform was used for their review, so the performance will be relevant.
We suggest you familiarize yourself with the screenshots of the test results first. For ease of perception, the performance of the ASUS TUF Gaming GeForce RTX 5070 Ti OC, as well as the GeForce RTX 4070 Ti SUPER, GeForce RTX 4080 SUPER and GeForce RTX 5080 video cards is presented in the diagrams. All of the listed models have a certain factory overclocking, which usually increases the results by ~2%. For this reason, the video cards of the corresponding series are displayed with the “OC” mark.
In the popular 3DMark tests, the new model is close to the GeForce RTX 4080 SUPER in terms of performance. Moreover, in the Port Ryal stages with ray tracing and Speed Way, the graphics adapter even slightly outperformed this video card.
The Geekbench test suite at the OpenCL calculation stage may need additional optimizations, since in the current version the GeForce RTX 5070 Ti even slightly lost to the GeForce RTX 4070 Ti SUPER. Although in the “RTX 5080 vs. 4080 SUPER” confrontation, no such anomalies were observed.
Blender was more lenient towards the new product, allowing it to outperform the RTX 4070 Ti SUPER by 6%. Although it still needs to accelerate by as much as 14% to reach the RTX 4080 SUPER.
But V-Ray with the RTX engine allowed the GeForce RTX 5070 Ti to rehabilitate itself in computational tasks. Here, the new product not only outperformed its predecessor by 21%, but also surpassed the performance of the RTX 4080 SUPER by almost 5%.
Gaming performance
Graphics cards of this level are usually considered excessive for Full HD gaming. However, if you use the maximum image quality settings in certain projects without additional scaling, even the capabilities of the GeForce RTX 5080 do not seem excessive. Examples include Black Myth: Wukong and even STALKER 2: Heart of Chornobyl.
In 1080p mode, the new product turned out to be 6.5% more productive than the GeForce RTX 4070 Ti SUPER on average. However, in most games at this resolution, the top video cards have a small difference in performance. In cases where the platform is not limited by processor performance, the GeForce RTX 5070 Ti gravitates towards older models, confidently outperforming its predecessor.
With the transition to 1440p mode, the average difference in performance increases to 9%. In some cases, the GeForce RTX 5070 Ti even manages to catch up with the GeForce RTX 4080 SUPER. Although there are also reverse cases, when the difference with the RTX 4070 SUPER is minimal.
In general, models of the GeForce RTX 5070 Ti 16 GB level are an ideal option for gaming in 1440p mode, where they offer very good overall performance and have a good memory reserve for all occasions. Allowing you to actively experiment with ray tracing and other resource-intensive options that improve image quality.
Architectural improvements and significantly increased memory bandwidth allow the GeForce RTX 5070 Ti to further improve the results in 4K mode. In such conditions, the average advantage of the new model over the GeForce RTX 4070 Ti SUPER is already about 12%. In many cases, this allowed us to approach the performance of the GeForce RTX 4080 SUPER. Especially considering that we use the results of accelerated versions of video cards for comparison. At least we can definitely say about close gaming capabilities with the base GeForce RTX 4080.
Given this balance of power, we can extrapolate the results and note that in 4K mode, the GeForce RTX 5070 Ti 16 GB video card will be 25–30% more productive than the GeForce RTX 4070 Ti 12 GB. Recall that the latter model was introduced two years ago with a recommended price tag of $799. However, in current realities, it is still worth considering not the formal names of video cards, but the existing offers with a specific price/performance ratio. The GeForce RTX 4070 Ti SUPER 16 GB also received a price tag of $799 and a year ago actually replaced the basic GeForce RTX 4070 Ti 12 GB.
The appearance of the GeForce RTX 5070 Ti 16 GB will probably not be a reason for owners of the GeForce RTX 4070 Ti SUPER 16 GB to upgrade, but users of earlier generation models should take a closer look at the new product, especially if they want to play with frame generation and get all the benefits of DLSS 4. For example, the ~100% advantage over the GeForce RTX 3070 Ti 8 GB, which could be purchased for $600 3.5 years ago, already looks like a pretty convincing argument for a possible upgrade.
If we talk about competition with AMD solutions, then the results of the GeForce RTX 5070 Ti allow us to position it between the Radeon RX 7900 XT 20 GB and the Radeon RX 7900 XTX 24 GB. Probably even closer to the flagship model. AMD has its own plans in this niche, so it will be very interesting to see the Radeon RX 9070/9070 XT, which, even judging by the new name, should potentially complicate the situation of the GeForce RTX 5070 Ti. Or rather, they will help balance the offer in this segment.
DLSS 4 + MFG
Returning to the practical capabilities of the GeForce RTX 5070 Ti 16 GB, let us recall that the “highlight” of the RTX 50 line is full support for DLSS 4 and Multi Frame Generation (MFG) technology. This is approximately how the balance of power between the GeForce RTX 5070 Ti 16 GB and the GeForce RTX 4070 Ti SUPER 16 GB looks in projects where MFG has already been implemented.
Multiple generation allows you to “pull out” seemingly hopeless cases. Maximum quality tracing in 4K mode on GeForce RTX 5070 Ti? Yes, such an option is available with MFG.
Overclocking
The possibility of additional overclocking of the GeForce RTX 5070 Ti aroused particular interest, given the relatively low standard frequencies of the graphics processor. Since the video card uses GB203, and we already know that this chip has good potential, at least as part of the RTX 5080, practical experiments made more sense. First, let's pay attention to the fact that the ASUS TUF Gaming GeForce RTX 5070 Ti OC allows you to increase the initial power limit by 10%. Taking advantage of this opportunity, we began the experiment.
Our initial expectations were met. The base frequency of the GPU was increased by 450 MHz at once - from 2295 MHz to 2745 MHz. At the same time, the average acceleration value increased to 3035 MHz. The memory chips were also significantly accelerated - from 1750 MHz (28,000 MHz) to 2125 MHz (34,000 MHz). A relative increase of 21%, and an amazing bandwidth of 1088 GB/s for a video card of this level. Moreover, the figure of 2125 MHz is clearly not the maximum, since the possibility of further increasing the memory frequency is still limited at the driver level. We hope that this temporary misunderstanding will continue to allow developers to continue to allow enthusiasts to search for the real technological limit of the new GDDR7 chips.
What changes have been made to the monitoring parameters? The peak GPU frequency has increased to 3322 MHz. We experimented with overclocking in the standard “P Mode” mode. In this case, even after a significant increase in frequency, the temperature of the GPU increased to 64C, the memory chips to 68C, while the fan speed increased to ~1680 rpm.
Since we previously increased the Power Limit by 10%, the power consumption of the video card accordingly exceeded the initial limits. After overclocking during high load, the maximum value was fixed at 337 W. Therefore, the presence of an adapter from three PCI-E to 12VHPWR in the kit is strategically correct. We note that according to GPU-Z, the consumption through the PCI-E x16 slot was 10 W, while the rest was provided by an additional 12 V line.
What about performance after additional overclocking? Since the increase in operating frequencies is quite significant, the changes in the results are also significant. In the 3DMark Time Spy Extreme test, the final Graphics Score increased by 11% - from 13,631 to 15,144 points.
In real games, the situation is similar. For example, in Cyberpunk 2077 in 1440p mode, the performance increased from 104/127 to 118/143 frames/s. The average fps increased by 12.6% . It seems that we already miss such acceleration from self-overclocking when it comes to video cards of this level.
Power consumption
As already noted, the GeForce RTX 5070 Ti 16 GB has a claimed power consumption of 300 W. According to the results of testing the forced version of the ASUS TUF Gaming GeForce RTX 5070 Ti OC, we received 305 W under load in normal mode and 309 W during experiments in Q Mode. After additional GPU/memory overclocking with an increase in the power limit by 10%, the power consumption of the video card reached 337 W.
As for its predecessor, the GeForce RTX 4070 Ti SUPER in the basic version has a TGP of 285 W. However, the model used in the comparison test consumed up to 300 W under load.
Recall that the top-end GeForce RTX 5080 with standard overclocking required up to 390 W of energy, and the GeForce RTX 4080 SUPER – 320 W. In general, there are no significant improvements in energy efficiency here, which is not surprising, given the immutability of the GPU manufacturing process for this generation.
