Foreword: With the New Year just around the corner, it's time to treat yourself
For many of my friends, this period may be regarded as the most pleasant time of the year. Just after enjoying the romance of Christmas, followed by the New Year's Day holiday, and soon with the arrival of the Spring Festival holiday, after a year of hard work, it is finally time to take a breath and treat yourself.
Of course, it is not too extravagant to say that it is a "treat". For example, for players who love PC games and want to take advantage of the year-end bonus to upgrade their configurations, it is still necessary to "be budget-conscious".
That's why today we've found four of the most mainstream graphics cards on the market, which are biased towards 1080p high-brush gaming and 2K high-quality games. We hope that through a horizontal comparison evaluation, we can find out the choice with the highest "cost performance" in the current mainstream level and the most "worthwhile" for everyone's year-end bonus.
Product analysis: four graphics cards to be tested, two different design ideas].
First of all, let's introduce the four graphics cards that participated in the horizontal review, as well as their basic information.
First, we chose AMD's Radeon RX 6750 GRE. As a model designed for the "dessert" market, the RX 6750 GRE has been well received since its release, especially with AMD's consistent "chicken blood" drive, and it is quite interesting to see how much "power" it can have today.
And the RX 6750 GRE is available in two different models, 10GB and 12GB. Not only are they different in terms of memory capacity, but also in terms of core size, memory width, etc., which can be regarded as two different models of graphics cards to test the differences between them.
At the same time, we also got two NVIDIA graphics cards that are similar to the RX 6750 GRE, the GeForce RTX 4060 8GB and the GeForce RTX 3060 12GB. It's worth noting that the RTX 3060 12GB actually has two different core (GA106, GA104) versions, but because the number of stream processors and the main frequency of the two are exactly the same, it does not affect our performance rating.
I have to say that from the GPU specifications of AMD and NVIDIA, we can see their different emphases in the design idea of "dessert card".
Obviously, AMD's mid-range graphics cards have higher texture and pixel performance at this stage, and the large "infinite cache" combined with large capacity and large bit wide video memory will also give it a greater advantage in high-frame-rate or high-resolution games.
In contrast, NVIDIA is now clearly taking a completely different approach to their mid-range product line. Among them, the RTX 4060 in particular, it can be seen that its video memory is relatively small in terms of capacity and bit width, and the texture and pixel indicators of NVIDIA's recent generations of mid-range cards are not outstanding. But instead, they have significantly higher floating-point performance and much lower power consumption.
So what is this concept?To put it simply, AMD's mid-range graphics cards are designed with a "conventional" approach, and they honestly stack raster rendering performance and memory bandwidth, so they can theoretically perform better in more games, especially those that don't eat much "hashpower", as well as the latest games that are poorly optimized.
In contrast, NVIDIA's mid-range graphics cards are a bit of a "trick". They obviously rely more on "computational graphics" technology, which is called DLSS, frame generation, etc., which is not native "rendering", but relies on computing to increase the frame rate.
Is such technology good?Honestly, you can't say that they are bad, after all, "computational graphics" is indeed an important direction for PC games these days. However, if you take into account the daily needs of mainstream gamers and the compatibility with more existing games, then the relatively traditional design ideas adopted by the RX 6750 GRE series may be a more practical and at the same time more sincere choice for mainstream gamers.
Theoretical performance test: N card single precision computing power is high, A card graphics rendering speed is fast].
In the first part, we tested the theoretical performance and computing power of the four graphics cards that were tested. In order to avoid any additional bottlenecks for this test, we used a set of Ryzen9-7950X CPUs, DDR5-6400 32GB RAM, and PCIe 40 Flagship SSD test platform.
During the test, the system version is Windows 11 23H2 official version, and all drivers have been upgraded to the latest version. At the same time, we also turned on the PBO of the CPU and the ResizeBar option for the graphics card, and set all power modes to the "High Performance" setting.
First, we used AIDA64's GPGPU test to compare the general-purpose computing performance of the four graphics cards. In order to avoid errors caused by differences in system operating conditions, we use the 2Cu core display in the CPU as a "control group" to eliminate interference in each test.
It can be seen that NVIDIA's two dessert cards will indeed be higher in terms of single-precision floating-point measured performance. At the same time, however, their performance in double-precision scenarios is significantly worse. Of course, it can be said that "what kind of double precision does the game card want", but this deliberate "castration" will still be a little unpleasant.
Next, we used the familiar 3DMark to compare the scores of all the participating graphics cards in an all-round way.
As in the previous analysis, in the test scenario that is more demanding on "computing power" and relatively newer, N cards with higher floating-point computing power will have certain advantages.
But after all, the test of "eating computing power" is the same as the game that eats computing power, and at least for now, it is still not considered to be popular. Therefore, in the vast majority of 3DMark tests that are biased towards traditional raster performance, the two RX6750 GRE with higher texture and raster performance will still have a clear advantage, especially in the more "traditional" DX11 project.
In addition, in order to verify the basic stability of each graphics card under heavy load, we also used 3DMark's "speedway" scenario to stress test all graphics cards for 20 cycles. As a result, all of these "dessert cards" withstood the stress test, and none of the dessert cards had a core temperature of more than 70 degrees Celsius during the stress test.
Game test: A card traditional game has great advantages, and the ultra-high image quality is optimized
Next, let's enter the game test session, and use a variety of games to test the actual performance of these four graphics cards at 2K resolution and high image quality settings.
In the online game "Final Fantasy 14", which is not heavy, although it is a game with N card optimization, it is not architecturally based on the latest computational graphics, but still relies more on traditional vertex and texture drawing performance. As a result, as you can see, the two RX 6750 GRE, which have stronger encoder performance, will have a significant performance advantage.
A similar situation was seen in the test of Shadow of the Tomb Raider. While this is a game that doesn't support the DX12 API and isn't too old, it's clear that it's still more reliant on raster performance, giving AMD graphics cards a clear frame rate advantage. Not to mention, it also supports AMD's RSR super-resolution mode, so it's no surprise to see that the RX 6750 GRE can run more than 50% higher frame rates than the RTX 4060.
Next, it is "Machine Warfare Mercenary 6: Realm Heavenly Fire", which was very popular some time ago. It's a very "new" game that supports ray tracing, but it doesn't support either FSR or DLSS, so both graphics cards are still competing for raster performance after ray tracing is turned on, and there is not much "computational graphics" involved. As a result, the four graphics cards are not too far behind each other, and the frame rate rankings are basically in line with their respective market positioning.
Of course, speaking of fire, we also tested the popular martial arts mobile game Naraka: Infinite. This game supports both FSR21. It also supports DLSS, and also adapts to NVIDIA's frame generation and Reflex latency reduction technology. However, what is surprising is that although it seems to be adapted to the "green factory" with more graphics technology options, but from the perspective of the measured frame rate, under the premise of 2K high score + full image quality, it is the basic frame rate of the A card and the FSR "calculated graphics" The improvement effect is significantly better.
In our opinion, this may be due to the fact that high image quality + high resolution has touched the memory bandwidth bottleneck of the N card, or it may be that although the game itself seems to have relevant options, it is actually poorly optimized.
Finally, we also tested two ultra-high-quality games, Metro Departure and Cyberpunk 2077. In "Metro Departure", all the reference graphics cards this time did not run at 60 frames even if the computational graphics enhancement was turned on, but like the previous "Mecha Mercenary 6", the native frame rate performance of the four graphics cards basically coincided with their positioning.
In the end, it is naturally "Cyberpunk 2077" with 2K resolution + "Super Ray Tracing" + other "Cyberpunk 2077" with the highest full quality setting. I have to say that under such harsh conditions, the native rendering performance of the four graphics cards is "fully collapsed", and it is basically impossible to play if you don't turn on the enhancement of computing graphics. Judging from the final test results, it seems that additional frame generation processing is necessary.
However, it should be mentioned here that AMD's RX 6000 series has also received the blessing of frame generation (AFMF) technology since its driver in October 2023, but the games it is currently suitable for are still relatively limited. In other words, it's not entirely impossible to expect the RX 6750 GRE series to double the frame rate again in Cyberpunk 2077.
Summary: There are not many options for mid-range dessert cards, so you need to be more cautious
Overall, if you are in a relatively "new" game with heavy graphics, then the current mid-range dessert card, the new generation RX 6750 GRE series and RTX 4060 are more recommended choices. Because they all have high computing power, they can also support computing power super-resolution and frame generation processing, which will be of obvious help to improve the frame rate of high-quality games. In contrast, the RTX 3060, which does not support frame generation in hardware, is not worth much at the moment.
But if you still want to take into account online games, e-sports, and some more classic stand-alone old games, then in the current mid-range market, AMD's several "dessert cards" are likely to be significantly more cost-effective choices. After all, they have the objective advantages of raster performance and memory bandwidth, which is difficult to make up for by "computing power" alone.
After all, on the one hand, the current mid-range graphics card market itself is relatively limited, and mainstream players are often not "local tyrants", so they naturally hope that the graphics cards they have worked hard to buy can be more "all-round". From this point of view, it is not surprising that the Radeon RX 6750 GRE series, which is more "honest" in terms of basic design and can handle a wider variety of high-scoring and high-quality games, is more popular in today's mid-range market.