Understand the gap via parameters of 30 series 7 graphics cards
Understand the gap via parameters of 30 series 7 graphics cards. 30 series 7 graphics cards on sale in this article to understand the gap between different parameters.
As Bitcoin once again breaks through the 52,000 dollar mark, although the current graphics card prices are rising all the way, most users are holding a wait-and-see attitude.
But there are still some users who just need to buy a graphics card. After all, they spend their own money. How to choose the configuration they want has become a problem.
After all, according to the current RTX 30 series graphics cards, there are a total of 7 graphics cards on sale, which is the largest number of graphics cards sold at the same time in the past.
The graphics card is a component that undertakes the task of outputting and displaying graphics. There are many performance indicators to measure it, and the relationship between these parameters is the most disturbing thing for players. Which parameter is the most important? What is the relationship between them? Is the big video memory that the store often advertises must be good?
Let me talk about the conclusion first, the only standard to measure the performance of the graphics card is the running graph, or the ladder graph.
Because it is a comprehensive manifestation of the various parameters of the graphics card, even under different cores, the flagship card of the Turing architecture has better performance than the entry card of the latest Ampere architecture.
However, if the graphics card of the same model is used, the frequency is generally more important. After all, in most cases, the frequency will be higher only if the internal materials and workmanship can keep up.
But there are exceptions to this situation. Some products will forcibly increase the Boost calibration frequency when the internal quality is not enough. Although the value on GPUZ is up, there may be frequency instability during actual use. In this way, the score may not be as good as the lower frequency graphics card.
At present, each generation of graphics cards is basically divided into 60, 70, and 80 models of graphics cards (Ti series are counted as prototype cards), so the official has actually given a clear positioning:
- RTX 60 models: online games (1080P/2K 144Hz), 3A games (1080P 144Hz), late stage (1080P)
- RTX 70 model: online games (1080P/2K/4K 144Hz), 3A games (1080P/2K 144Hz; 4K 60Hz), late stage (1080P/2K)
- RTX 80 model: online games (1080P/2K/4K 144Hz), 3A games (1080P/2K/4K 144Hz), late stage (1080P/2K/4K)
- RTX 90 model: online games (1080P/2K/4K 144Hz; 8K 120Hz), 3A games (1080P/2K/4K 144Hz; 8K 120Hz), late stage (1080P/2K/4K/8K)
Architecture is actually the most important indicator that affects the performance of graphics cards. The emergence of a new generation of architecture often means an iterative upgrade of products, such as Maxwell→Pascal→Turing→Ampere.
In terms of parameter performance, such as: GP107; TU106; GA102; Vega 10; Navi 10, as well as some derivative codes of this model, such as GA102-225-A1.
How do you understand the architecture? It is like the hub of the entire graphics card. It is the same as the road planning of a city. In the same city, whoever has a good road plan will not be so prone to traffic jams. At the same time, you The places you want to go can also be reached faster.
However, it is good for us to simply understand the code name and release time of the architecture, because only the architecture engineer can understand the deeper internal transistor layout and the like.
As long as we roughly know that GA102 is the core code of RTX 3080 and GA106 is the core code of RTX 3060 , you can infer from it which is better.
The core of the graphics card can basically be distinguished by all players, because NVIDIA will name the graphics card model according to the chip model when it is sold, such as RTX 3060, RTX 3070, and everyone can know which performance is better at a glance.
The CPU has a manufacturing process, and so does the graphics card. At present, the minimum size for N card is 8nm (nanometer), and the minimum for A card is 7nm. A smaller process can bring less power consumption and heat generation, so the smaller the number, the better.
But the overall performance of the graphics card is still not determined by this value, so just take a look.
As far as this generation of A/N graphics cards are concerned, AMD uses all 7nm processes, while NVIDIA all uses 8nm, but the RTX 3090 must be more powerful than the RX 6600 XT.
In terms of stream processors, there is a slight difference between N card and A card, which is why in our earlier understanding, A card is suitable for graphics and design, and N card is suitable for playing games.
The specific difference is easy to find on the Internet, so I won’t explain it too much. Simply put, the higher the value, the better.
certainly! This is not absolute. If there are many computing units inside a GPU, but there are not enough other resources to support these computing units in parallel, then these computing units will be idle, wasting computing power.
However, this situation is not a player’s consideration. The manufacturer will calculate these at the beginning of the design. Waste of computing power is a waste of cost, and it does not exist.
30 series 7 graphics cards on sale in this article to understand the gap between different parameters
In addition, the number of stream processors should also be compared in the same camp. AMD RX 6900 XT has 5120 stream processors and NVIDIA RTX 3070 has 5888 stream processors, but obviously the performance of RX 6900 XT is better. powerful.
The base frequency of the graphics card
The basic frequency of the graphics card usually refers to the operating efficiency of the stream processor. Generally, the higher the value, the faster the core operating frequency, and the acceleration frequency can be understood as a dynamic frequency. When we play some 3A games, the frequency will be higher than the indicated basic frequency Much higher, just like the standard brightness and peak brightness of a mobile phone.
The value of the frequency is not endlessly high. High frequency will inevitably bring about high heat and high energy consumption, and it is also prone to problems such as blurred screens.
In addition, the frequency can only be compared under the same model, and it is meaningless to compare the frequency of different graphics cards with each other.
Take the NVIDIA public version frequency as an example. For example, the acceleration frequency of the RTX 3070 Ti is 1770MHz, while the RTX 3080 is 1710MHz. Although the frequency of the RTX 3070 Ti is higher, we cannot say that the former has better performance than the latter.
Video memory capacity is the main routine used by unscrupulous merchants to pit Xiaobai in the early years, and large video memory is often marked in the eye-catching position of the e-commerce page.
The video memory capacity can be simply understood as the temporary storage warehouse of the graphics card. If there is too much data, the warehouse will burst, the game will inevitably be stuck. At present, the video memory capacity of the RTX 30 series graphics card is mainly 8/10/12 GB, and it is also like the super flagship RTX 3090 has reached 24GB of video memory, 8GB is enough for most players.
In fact, the video memory capacity has the closest relationship with the game resolution. At 1080P resolution, the effective memory usage rate of 8GB is sufficient, while the 12GB large video memory like RTX 3080 Ti is basically prepared for 4K games.
The types of video memory are mainly GDDR6X and GDDR6 in the 30 series graphics cards, the 20 series are GDDR6, and the 10 series graphics cards are GDDR5. The type of video memory mainly affects the performance.
A typical example is RTX 3070 Ti. Although its video memory is still 8GB, it has been changed from GDDR6 of RTX 3070 to GDDR6X. The difference between the two is mainly in frequency and bandwidth. It is possible to achieve higher memory bandwidth under the same frequency. Reduce cost and power consumption. In addition, GDDR6X can transmit 2 times more data than GDDR6 in the same time.
This is particularly important for tasks that require a large amount of data load, such as ray tracing games, AI learning, and 8K video rendering.
In terms of performance in this respect, GDDR6X>GDDR6>GDDR5X>GDDR5.
Video memory bit width
The video memory bit width is the “memory” of the graphics card, which means the throughput of the graphics card for processing data instantly. The larger the better, it has the same complementary effect as the video memory. If the video memory capacity is large and the bit width is small, it indicates the data throughput capacity. Very small.
Simply put, the video memory bit width is equivalent to the lanes in the city, the large bit width is equivalent to three or even four rows of lanes, while the small bit width can only allow single-row lanes.
Video memory bandwidth
The video memory bandwidth is determined by the video memory frequency and the video memory bit width. The calculation formula is “Video memory bandwidth (GB/s) = actual video memory frequency (MHz) * video memory data magnification * video memory equivalent bit width (bit)/8”, but currently Almost none of the graphics cards are stuck on the bandwidth, basically enough.
As shown in the figure below, the RTX 3070 Ti is substituted according to the above formula: 1188MHz x 16 × 256bit ÷ 8 = 608.256 GB/s, which is approximately equal to 608.3 GB/s.
The RTX 3070 also uses 8GB video memory, but uses GDDR6 video memory type, so even if the video memory frequency is the same, the bandwidth should be smaller: 1188MHz × 8 × 256bit ÷ 8=304.128 GB/s.
But in reality, the RTX 3070 video memory frequency that uses GDDR6 video memory will be higher, and the public version is 1750MHz, so the final video memory bandwidth is 448 GB/s.
Generally, the parameters of video memory capacity, video memory bit width, and video memory bandwidth are put together. At present, the highest of these items is NVIDIA’s RTX 3090 graphics card.
As the basic parameters of the graphics card, the players who are just getting started understand the above are basically enough, but you will find that they are still confused when choosing a graphics card. It is basically impossible to see the performance of the graphics card from the cold parameters.
If players want to really understand the performance of the graphics card in actual use, they also need to look at the graphics card running score, ladder diagram or evaluation.