Video cards of the GeForce RTX 4060 Ti line have been caught at a bad time by users who are generally not satisfied with NVIDIA’s latest actions in the discrete graphics market and directly with the company’s pricing policy regarding GeForce RTX 40xx line devices. An additional reason for criticism was the, at first glance, unclear technical specifications of the video card – 128-bit bus, 8 GB of memory, PCI-E 4.0 x8 interface. And this is the middle class now?! Angry fans of high-speed games got a great object for mockery and an opportunity to let off steam, and numerous opinion leaders – a great opportunity to hype and enjoy the number of views on the counters. After the first wave of indignation seems to have subsided, it is time to evaluate the capabilities of the GeForce RTX 4060 Ti without unnecessary emotions. MSI GeForce RTX 4060 Ti GAMING X TRIO 8G will help us with this – a unique model with a giant cooler and substantial factory GPU acceleration.
NVIDIA GeForce RTX 4060 Ti
The new model uses the AD106 graphics processor. This is a fairly compact chip with the ADA Lovelace architecture, which is manufactured by TSMC according to 5 nm standards (TSMC 4N). The crystal has an area of only 188 mm2 and includes 22.9 billion transistors.
The GeForce RTX 4060 Ti uses the AD106-350-A1 version with 4352 active CUDA cores, 136 texture modules, and 48 rasterizers. The GPU works with a frequency formula of 2310/2535 MHz and offers a theoretical computing power at the level of 22 TFLOPS (FP32). Also, note that the GPU has 128KB of L2 cache per thread multiprocessor (SM) and a total L2 buffer of 32MB.
|GeForce RTX 4070||GeForce RTX 4060 Ti||GeForce RTX 3070||GeForce RTX 3060 Ti|
|Architecture||Ada Lovelace||Ada Lovelace||Ampere||Ampere|
|Technology||5 nm, TSMC||5 nm, TSMC||8 nm, Samsung||8 nm, Samsung|
|Tensor cores||184 (4-generation)||136 (4-generation)||184 (3-generation)||152 (3-generation)|
|RT-cores||46 (3-generation)||34 (3-generation)||46 (2-generation)||38 (2-generation)|
|GPU clock rates||1.92/2.475 GHz||2.31/2.535 GHz||1,5/1,725 ГГц||1.41/1.665 GHz|
|Memory||12 GB GDDR6X||8 GB GDDR6||8 GB GDDR6||8 GB GDDR6/GDDR6X|
|Memory bus||192 bit||128 bit||256 bit||256 bit|
|Effective memory frequency||21 000 MHz||18 000 MHz||14 000 MHz||14 000/19 000 MHz|
|Memory bandwidth||504 GB/s||288 GB/s||448 GB/s||448/608 GB/s|
|TGP||200 W||160 W||220 W||200/225 W|
The GeForce RTX 4060 Ti is equipped with a 128-bit memory bus and 8 GB of GDDR6 with an effective frequency of 18,000 MHz. This configuration allows you to get bandwidth at the level of 288 GB/s. A similar configuration of the memory subsystem for a mid-range video card has received devastating criticism. Indeed, the specified GPU does not look very attractive, even against the background of the specifications of the predecessor GeForce RTX 3060 Ti. We remind you that the latter, although it also has a memory capacity of 8 GB, is equipped with a 256-bit bus with GDDR6 chips (14,000 MHz) and a resulting bandwidth of 448 GB/s.
To answer the hot question: “What the hell is going on here?!”, which was heard from everywhere even before the official start of sales of GeForce RTX 4060 Ti, NVIDIA had to call the explanation team.
To summarize, the developers draw attention to the radical increase in L2 cache capacity in GeForce RTX 40xx models, which allows to significantly increase the overall efficiency of the memory subsystem.
When the graphics processor needs certain data during operation, such a request is initially sent to the first-level L1 cache, which is located directly in each streaming multiprocessor. If the necessary data is not found there, the GPU turns to the L2 cache memory. The presence of data in this buffer (Cache hit) significantly reduces the time and speed of processing. If the data is not found in L2 (Cache miss), the graphics processor addresses the memory of the video card and tries to get it from there. If data is found, it is transferred over the same infamous memory bus. If the necessary data is not found at this stage, requests are first sent to the system RAM, and then to the storage device (SSD).
Of course, various algorithms and mechanisms for saving the most requested data, which are used with the greatest frequency, in the cache memory work here.
One of the architectural features of the models with the Ada Lovelace architecture is a radical increase in the capacity of the L2 cache memory. For comparison, the GA104 chip used for the GeForce RTX 3060 Ti has 4 MB of L2, while the AD106 in the configuration for the GeForce RTX 4060 Ti is already equipped with 32 MB of L2. That is, the capacity of the cache buffer of the second level has increased by 8 times. Increasing L2, respectively, allows you to significantly increase the frequency of successful calls to the cache, and, accordingly, to access the memory of the video card less often. In addition, according to the developers, in solutions based on Ada Lovelace, the bandwidth of the L2 cache itself has been significantly increased. This also affects performance.
NVIDIA engineers experimented using a pair of GeForce RTX 4060 Ti graphics cards, one of which was equipped with a test GPU with only 2 MB of L2 activated. As it turned out, the increase to 32 MB of L2 allows various test tasks and real games to reduce data transfer on the memory bus by up to 50%. In addition, thanks to the larger capacity of L2, the overall performance increased by ~17% in modes with classic rendering, up to ~25% – with active ray tracing, and up to 35% when using DLSS 3.
In such conditions, NVIDIA, evaluating the capabilities of the memory subsystem, uses the term “effective bandwidth”, assuring that this figure for the GeForce RTX 4060 Ti is 554 GB/s. However, this selectivity of terminology is somewhat surprising, given that such definitions are not used, for example, for older RTX 40xx models, which are also based on GPUs with even larger L2 cache capacity. One way or another, the results of practical tests will put an end to this issue.
In their post, the developers also had to justify the memory capacity of 8 GB, which is not optimal for this class of adapters. This value is a consequence of using a 128-bit bus. At this bit rate, the GPU supports 4 or 8 memory chips. The most common DRAM chips of GDDR6/GDDR6X standards have a capacity of 1 GB (16 GB) and 2 GB (32 GB). For the basic modification, four microcircuits of 2 GB are used, which gives a total of 8 GB. The upcoming 16GB RTX 4060 Ti modifications, due soon, will already use eight 2GB chips. Microcircuits of different densities cannot be mixed. That is, if there is a 128-bit bus, the video card can have either 8 GB or 16 GB. There is also an option with 4 GB, but this is unlikely to be of any interest to anyone. But the version with 12 GB is technically impossible. It requires a 192-bit bus and a set of six 2GB chips. Older GeForce RTX 4070/4070 Ti models already have this configuration.
Returning to the GeForce RTX 4060 Ti equipment, it should also be noted that eight PCI-E 4.0 lines are used for the video cards of the line. For platforms with this version of the interface, the bandwidth will be sufficient, but on systems with PCI-E 3.0, a certain decrease in performance is possible, albeit within a few percent.
But the strength of the GeForce RTX 4060 Ti is improved energy efficiency. The total power consumption of the video card (TGP) is declared at the level of 160 W. For systems with such a graphics adapter, the developers recommend using power supplies with a power of 550 W or more.
NVIDIA offers its version of the GeForce RTX 4060 Ti Founders Edition, and this version will be available only for modifications with 8 GB of memory, while for models with 16 GB, original models from partners are expected. Of course, the latter offered a wide variety of improvements for the GeForce RTX 4060 Ti 8 GB, one of which we will examine in more detail this time.
MSI GeForce RTX 4060 Ti GAMING X TRIO 8G
MSI has developed more than a dozen different models on the AD106, among which the MSI GeForce RTX 4060 Ti GAMING X TRIO 8G version stands out.
An oversized three-fan cooler seems excessive for devices of this series, but we already know that it is usually these coolers that allow you to get almost silent graphics cards with excellent temperature performance. However, we will leave it for the practical part of the review.
Initially, it can be noted that in addition to the “monstrous” cooler, the MSI GeForce RTX 4060 Ti GAMING X TRIO 8G also received a noticeable factory overclock. The GPU received a formula of 2310/2670 MHz instead of the recommended 2310/2535 MHz. An additional +135 MHz for average GPU acceleration is one of the biggest figures for the tuned RTX 4060 Ti models.
MSI GeForce RTX 4060 Ti GAMING X TRIO 8G comes in a fairly large cardboard box, which is not surprising considering the overall dimensions of the video card – 338x141x52 mm. A metal bracket is provided with the adapter for additional support of the adapter. The model weighs 1167 g, so such an element will no longer be superfluous, besides, it can additionally form a visual series of system design.
The video card is equipped with a large branded TRI FROZR 3.0 cooler with TORX FAN 5.0 fans. First of all, we can estimate the dimensions of the cooler and the total area of dispersion of an infinite number of aluminum plates equipped with two radiator cassettes. Both units are interconnected by five heat pipes formed in the heat sink area with a rectangular profile for maximum contact area.
The design also includes a solid copper plate with additional nickel plating. Such a “substrate” allows you to eliminate the formation of local GPU overheating zones at the seams between the heat pipes. In addition, the plate contacts the memory chips through thermal pads. Usually, the profile provides for the formation of plates to remove heat from the VRM power elements.
As we mentioned, the cooler is equipped with three 100-mm axial fans, made according to Torx Fan 5.0 technology. They are equipped with nine blades, with each triad joined by additional arcs on the outer radius.
In combination with a certain angle of inclination of the blades, it is possible to maintain high static air pressure even at low fan speeds.
Despite the large overall dimensions of the video card, the printed circuit board is only about 140 mm long. The GPU power stabilization subsystem is made according to an 8-phase circuit, which includes OnSemi NCP302150 DrMOS power assemblies, designed for an operating current of up to 50 A. The UPI uP9512R controller is used to control the VRM parameters.
The memory chips are powered by a single-phase circuit with a Nikos PKE96BB MOSFET (50 A) and a control controller labeled 7212.
As you can see in the photo, the video card is equipped with four memory chips, which are located in an L-shaped manner near the GPU. In this case, Samsung chips marked K4ZAF325BC-SC20 are used. According to the specification, the chips are designed to operate at an effective 20,000 MHz. Given that the RTX 4060 Ti uses the 18,000 MHz mode, such chips allow you to count on a certain potential for additional acceleration.
To connect additional power, the MSI GeForce RTX 4060 Ti GAMING X TRIO 8G is equipped with the usual 8-pin connector. Therefore, an additional adapter for 12VHPWR will not be needed, even if the power supply does not have a corresponding connector.
Even though the graphics processor of the tuned version of the adapter from MSI is significantly overclocked by the manufacturer, the developers did not change the power specifications of the video card. The declared level of energy consumption is 160 W, and the recommended power of the power supply unit starts from 550 W.
On the reverse side, the video card is covered with an aluminum plate – additional protection against mechanical damage. It is not used to cool the elements on the PCB. However, the shape with a huge hole for almost half of the plane allows a large radiator cassette to be blown through. Of course, this should improve the temperature regime of the elements.
Please note that an additional bracket is used to fix the radiator in the GPU area. Such a bracket allows you to balance the uniformity and force of pressing the cooler to the GPU crystal.
The composition of interface connectors is standard for models of the series – three DisplayPort 1.4a and one HDMI 2.1. The panel has openings for partial removal of heated air.
MSI GeForce RTX 4060 Ti GAMING X TRIO 8G has additional RGB lighting. Additional illumination is provided for the game series logo on the top panel. Also, the segments of the world conductors are placed diagonally next to the central fan. The developers decided not to make the backlight intrusive, but at the same time outlined the gaming orientation of the device.
The proprietary MSI Center utility is used to adjust the backlight. A fairly wide set of available modes is provided here, and it is also possible to adjust the speed of dynamic effects and the intensity of the LEDs.
Test bench configuration
- CPU: Intel Core i5-13600KF (6/12+8; 3.5/5.1 GHz + 2.6/3.9 GHz)
- Cooling: ASUS ROG RYUO III 360 ARGB
- Memory: Kingston FURY Renegade DDR5-7200 32 GB (KF572C38RWK2-32 2×16 GB)
- Drive: Kingston KC3000 1024 GB and Samsung 980 Pro 1TB
- PSU: ASUS ROG STRIX 1000W GOLD
The video card is expected to support hybrid cooling technology with the fans stopping when the GPU heats up slightly. In the interpretation of the developers, this mode is called Zero Frozr.
At minimum load, the GPU temperature was kept at 38C at 25C indoors. Stopping the fans implies operation in passive cooling mode and, accordingly, the absence of noise from the video card.
After the game battles, we also evaluated the maximum indicators. It turned out that the graphics processor warmed up to 58C at 70C in the hottest point (Hot Spot). This is a very good indicator, especially considering that the fans at the same time rotated with a frequency of only up to 1100 rpm. At the same time, the MSI GeForce RTX 4060 Ti GAMING X TRIO 8G worked very quietly. It is not surprising that this model does not have an additional mode such as Quiet Mode, because it is almost silent as standard. The cooling system with a huge radiator unit and three fans fully justifies itself.
Note that during game loads, the frequency of the graphics processor increased to a peak of 2835 MHz.
The first approach to assessing the performance of the MSI GeForce RTX 4060 Ti GAMING X TRIO 8G is made using current game “synthetics” – 3DMark tests. The Time Spy Extreme stage is quite difficult even for top video cards, however, it usually allows you to get a fairly clear graphic hierarchy according to the capabilities of the test participants.
According to the obtained indicators, the tuned version of the GeForce RTX 4060 Ti is 6% inferior to the GeForce RTX 3070. The latter is about 15% faster than the GeForce RTX 3060 Ti, so we can predict that the difference between the models of the xx60 Ti line of different generations will be ~10%. A significant difference (~27%) is observed when comparing the performance of the older GeForce RTX 4070 12 GB model. However, it should be noted that the Time Spy Extreme stage is performed in 4K mode, which is, potentially the most “problematic” for models with a relatively small memory bandwidth.
At the Speed Way stage, the positions of the video cards moved further apart. The difference between the RTX 4060 Ti and the RTX 3070 grew to 8%, and the GeForce RTX 4070 12 GB went into the gap altogether, increasing the advantage to 35%.
The 3DMark Port Royal subtest with ray tracing brought the RTX 4060 Ti and RTX 3070 close together. The latter had an advantage of only 2%. But the older model is also out of reach for novelty at this stage.
We started the tests in real games with the 1080p mode, for which the developers directly recommend using the GeForce RTX 4060 Ti. According to the results of the experiments, the MSI GeForce RTX 4060 Ti GAMING X TRIO 8G provided an average of 145 fps against 143 fps in the GeForce RTX 3070 8 GB. That is, the performance of both video cards in modes with classic rendering turned out to be quite close. However, it is important to note here that we are talking about the GeForce RTX 4060 Ti 8 GB model with a significant initial GPU overclock (+135 MHz), which of course affected the overall results.
If you look at the actual fps, then, despite the almost equal arithmetic averages, the ratio of forces in the projects may differ. In some, the accelerated RTX 4060 Ti had an advantage, in others it was inferior to the GeForce RTX 3070 both in terms of average and minimum fps. A good example is Cyberpunk 2077.
However, in general, performance at the level of GeForce RTX 3070 8 GB cannot be called something that is not worth paying attention to, as it is often highlighted by ardent critics of the new model. More recently, the GeForce RTX 3060 Ti and GeForce RTX 3070 with 8GB of memory were considered excellent options for optimal gaming configurations, especially for 1080p mode.
Has everything changed so dramatically in the last few months? Of course, with the growing need for “heavy” projects on engines like Unreal Engine 5, the extra memory certainly wouldn’t be superfluous, even for Full HD at maximum settings. But this is where DLSS intelligent scaling support, which is almost guaranteed to be implemented in resource-intensive projects, will come in handy.
We suggest you look beyond the limits and see how the GeForce RTX 4060 Ti 8 GB will feel for owners of screens with a resolution of 2560×1440. To be honest, we expected the worst. The forced version here tries to compete with the GeForce RTX 3070 8 GB, and in certain cases, it even succeeds. Although in most cases the older model of the previous generation turns out to be faster. Sometimes the advantage is more than 10%.
However, it would also be an exaggeration to consider the GeForce RTX 4060 Ti 8GB completely useless at 1440p. This is indicated by the average of 106 fps based on the results of tests in 13 projects.
And what should you expect if you get brave and try to enable ray tracing in 1440p? Consider such a situation using the example of Dying Light 2. Well, at first it can be “painful”, even with the older model GeForce RTX 4070 12 GB, which does not provide an average of 60 fps in such conditions. Therefore, 42 fps with a forced GeForce RTX 4060 Ti 8 GB is the expected result.
However, after connecting DLSS in Performance mode, we already have an average of 90 frames/s, and the additional activation of DLSS 3, which, by the way, is not available for models of previous generations, allows you to count on 130 frames/s. Here’s a video card that “barely pulls Full HD”.
Breaking all canons and regulations, we try GeForce RTX 4060 Ti 8 GB in 4K mode. Well… we didn’t want to. Of course, no miracles happen here. This is a mode for more powerful solutions, so you shouldn’t count on a good performance. Especially when using settings with maximum picture quality. The MSI GeForce RTX 4060 Ti GAMING X TRIO 8G managed to score an average of 59 fps in the aggregate of all game stages, but 35 fps in Cyberpunk 2077, 44 fps in Dying Light 2 or 48 fps in Red Dead Redemption II are not the indicators you will put up with having just bought a new video card. Especially considering that these are not the heaviest projects already available.
Note that the 4K model GeForce RTX 3070 8 GB has already confidently taken the lead, in some projects ahead of MSI GeForce RTX 4060 Ti GAMING X TRIO 8G by more than 10%. We can also see that with a similar nature of the load, even a much larger L2 cache (32 vs 4 MB) is not always better than “brute force” with an actual bandwidth of 448 GB/s compared to 288 GB/s.
The desire to get a 4K gaming experience with ray tracing and maximum quality settings can be shattered by the harsh reality of the system requirements in this mode. For example, in the same Dying Light 2 under similar conditions, even the RTX 4070 provides only 26 fps. Yes, this is as much as 30% more than the RTX 4060 Ti, and will it be easier. In such cases, DLSS support becomes crucial.
First, we get as much as 52 fps on the RTX 4060 Ti by activating DLSS 2 Performance, and only then add Frame Generation (DLSS 3) and get an average of 70 frames/s. After that, it is possible to restore order after the zombie apocalypse in comfortable conditions. This will not always be the case, but the presence of DLSS 3 will give RTX 40xx owners confidence.
The GeForce RTX 40xx line is generally quite energy-efficient, even if we talk about the top models, which, although they do not have such low energy consumption, are justified by high performance.
In the case of the GeForce RTX 4060 Ti, a quite modest TGP of 160 W is declared for the video card. For the predecessor GeForce RTX 3060 Ti, this figure is 200 W. Given that the new model is faster, it has better performance per watt.
If we evaluate the MSI GeForce RTX 4060 Ti GAMING X TRIO 8G directly, then, despite the significant additional acceleration of the GPU, the video card corresponds to the declared energy consumption values. The monitoring records the maximum values at the level of 160.5 W. In general, the values are very good.
However, if we evaluate the performance per watt, then the older model – GeForce RTX 4070 12 GB will still have a slight advantage. Well, a comparison with models of previous generations does not make much sense. Their energy efficiency is noticeably worse.
Despite the significant overclocking of the GPU, it is far from all the capabilities of the MSI GeForce RTX 4060 Ti GAMING X TRIO 8G in terms of frequency potential. The ability to increase the Power Limit to 121% also hints at this. So, although in express mode, we still experiment a little with additional acceleration. And quite successfully.
It was possible to add +120 MHz to the already available standard acceleration of the graphics processor, so the final formula took the form of 2430/2790 MHz. The memory chips, which worked stably at an effective 21,600 MHz instead of the standard 18,000 MHz, were perfectly overclocked. Here we have a real frequency increase of up to 20% – 2700 MHz instead of 2250 MHz.
After additional overclocking, we also monitored the performance of the video card by loading it with Cyberpunk 2077. In this case, the GPU temperature increased by one degree to 59C. Although the GPU was already accelerated to 2940 MHz. But the maximum speed of the fans remained at the previous level – 1100 rpm. Peak power consumption also increased minimally to 163 W. Obviously, an effective cooler allows you to not notice changes in the operation modes of the video card, doing an excellent job of heat dissipation.
But what did not go unnoticed was the increase in productivity. In the 3DMark Time Spy Extreme test, the additional GPU/memory acceleration made it possible to increase the Graphics Score by as much as 8% – from 6290 to 6802 points. This is already a noticeable increase, which will not be superfluous, especially if the cooler can cool the overclocked video card, almost without changing the temperature regime.
In the case of the GeForce RTX 4060 Ti 8 GB, the price caused a lot of complaints from potential customers. The recommended price of the model for the American market is $399. The predecessor GeForce RTX 3060 Ti 8 GB had the same value at the end of 2020. However, the new model is technologically simpler in many ways – it has a GPU with half the crystal area (188 vs. 398 mm2), a 128-bit bus vs. a 256-bit bus, 4 memory chips vs. 8 chips, a PCI-E 4.0 x8 interface vs. PCI -E 4.0 x16. That is, the pricing questions of the RTX 40xx models remain unanswered. Unfortunately, they are not difficult to find, considering the distribution of the discrete graphics market with the total dominance of NVIDIA.
In Ukraine, the current starting prices of the GeForce RTX 4060 Ti start at 17,500 UAH (~$470) for the basic models, while the most equipped modifications ask for up to 23,000 UAH (~$620). The price difference is significant, but the platforms in these cases also differ significantly.
Of course, the MSI GeForce RTX 4060 Ti GAMING X TRIO 8G with an effective overall cooler and high basic overclocking refers to more expensive modifications. The estimated cost of the considered model is UAH 21,500 (~$570).