Graphics card usage algorithm between memory capacity and bandwidth, bit width, frequency

The capacity has nothing to do with the other three, it only determines how many material library files can be held in the video memory.

Memory frequency x bit width) 8 = bandwidth.

Calculation example: GTX280 bit width is 512 bits, frequency is 2200 MHz, and the bandwidth is calculated = 512 bits*

9800gtx+ bit width 256 bits, frequency 2200 MHz, this is calculated as bandwidth = 256 bit*

Note: 1MB=8MBit, 1GB=8GB, 1 big B is equal to 8 small B, so divide the result by 8, for example, broadband 2M, where 2M is 2MB=2M 8=256KB, so our **speed is about 200KB S.

There are three important metrics for a graphics card: capacity, frequency, and memory bit width.

Capacity. The video memory is responsible for the data exchange between the system and the graphics card and the data cache when the display chip is computing 3D graphics, so the video memory capacity naturally determines the amount of data that the display chip can process. Theoretically, the larger the video memory, the better the performance of the graphics card.

However, this is only a theoretical calculation, and the actual performance of the graphics card is constrained by many factors, such as: display chip speed, video memory bit width, video memory speed, etc.

2.Clock cycle and operating frequency.

Clock cycles and memory operating frequency are very important performance indicators of video memory, which refers to the time that elapses for each time video memory processes data. The faster the memory speed, the larger the amount of data exchanged per unit of time, and the performance of the graphics card will be significantly improved under the same circumstances. The clock cycle of video memory is generally measured in ns (nanoseconds) and the operating frequency is measured in MHz.

The memory clock cycle corresponds to the operating frequency, and the relationship between them is: operating frequency 1 clock cycle 1000.

Common memory clock cycles are 5ns, 4ns, and more. For DDRSDRAM memory to describe the operating frequency of the macro family, the equivalent operating frequency is used. Because data can be transmitted on both the rising and falling edges of the clock cycle, the memory bandwidth is twice that of SDRAM with the same operating frequency and data bit width.

In other words, in the case of the same clock cycle of the memory macrobook, DDR

SDRAM memory actually works twice as often as SDRAM memory.

The capacity has nothing to do with the other three, but you can decide how many material library files the video memory can hold.

Memory frequency x bit width) 8 = bandwidth.

Calculation example: GTX280

The bit width is 512 bits, and the frequency is 2200 MHz, so the bandwidth is calculated = 512 bits*9800gtx+

The width of the bits is 256 bits, and the frequency is 2200 MHz, so the bandwidth = 256 bits*

Note: 1MB=8MBit, 1GB=8GB, 1 big B is equal to 8 small B, so the result is divided by 8, for example, broadband 2M, where 2M is 2MB=2M 8=256KB, so it is good to quarrel with us**The speed is about 200KB s.

Yours doesn't have all the information, because this diagram contains two graphics cards with different cores, and the stream processor is missing, which is the most critical.

The first thing to talk about is bit width, which is equivalent to the memory frequency, that is, in the case of the same core frequency, number of stream processors, and video memory capacity, if the bit width is halved and the memory frequency is doubled, this performance is theoretically unchanged.

In addition, the core frequency and memory frequency are generally more important than the memory frequency, which means that the increase in the core frequency of 100MHz is more important than the increase in the memory frequency of 100MHz.

You didn't list the number of stream processors, I won't talk about it here, the more the merrier, this should be equivalent to bit width.

In addition, in terms of video memory capacity, the more the merrier, but in the case of sufficient video memory, consumers should pay more attention to the frequency and bit width of the video memory, and the video memory is very secondary to the graphics card.

Although HD5670 is said to have a relatively high memory frequency and core frequency, the number of stream processors is lower than the previous three, and the performance can only be roughly regarded as the level of HD4750, that is, the level of HD4830.

4870 has the strongest performance, but the brand of Shuangmin is a bit ...... thatIt's best if you have Blaupunkt Dilan.

The one with a bit width of 128 can be ignored.

The performance of the remaining three models from high to low is Shuangmin Dilan Hengjin Sapphire, which has a large memory capacity but a low memory frequency.

Take 3DMark scores, or load your favorite 3D games, hang FRAPS, and test FPS

High scores and high FPS are the last word.

The 4870 in your column, that is, the 2nd one, is significantly stronger than the rest.

Judging from the data provided by the landlord, Shuangmin's 4870 is the strongest performance. It is still necessary to consider comprehensively.

Upstairs is complete nonsense, Xiaobai and you both have the same idea. Note that the "bandwidth" here refers to the product of the memory "bit width" and the memory frequency, and then divide it by 8. For example, a graphics card is 256bit, the memory frequency is 2000 (for example, 9800gt, with GDDR3 video memory, the frequency is average), then the memory bandwidth is 256x2000 8=64000, if the other graphics card is 128bit, but the memory frequency is 4600 (for example, HD6750, with GDDR5 video memory, the frequency is very high), then the memory bandwidth is 736000, theoretically the performance of the second graphics card memory is stronger, Therefore, buying a graphics card is not only based on bit width, but also on frequency, and the combination of the two - bandwidth is the key.

And the video memory capacity is the last factor to consider, a 2G capacity GT210 is not as good as a 512M capacity GT240, in addition to the poor performance of GPU cores, the memory bandwidth of 210 is much lower than 240 is also an important reason. Nowadays, many profiteers like to use large video memory to fool lay buyers, and even there is a 4G capacity GT440 garbage card, but the video memory he uses is GDDR3, the video memory frequency is very low, so the memory bandwidth is low, if it is compared with the standard GT440 (1G video memory, with GDDR5 video memory, high frequency, high bandwidth) performance is actually worse! So when you buy a graphics card, don't stare at the video memory capacity, profiteers just take a fancy to the problem that laymen like large video memory capacity, so they make a big fuss.

You remember, there are two most critical factors that determine the performance of a graphics card: GPU core performance and memory bandwidth! Video memory capacity is the last thing to consider, just like a truck, you can't have a good engine, you can't do the road, the speed can't work, can you run with a 100-ton capacity?

The biggest impact is capacity, followed by bit width, then frequency! And finally, the manufacturing process!

Since the memory capacity of a graphics card directly affects the performance of the graphics card, people now measure the performance of a graphics card by the memory capacity.

Bit width refers to the amount of data transmitted per unit time when the graphics card is working, and in layman's terms, it reflects how fast it is. What he affects is the processing speed of the graphics card. Most of today's graphics cards are game-oriented, and the graphics of the game itself have already been rendered.

So you don't need a graphics card to be too fast. Therefore, the bit width of the gaming graphics card is not large. Only professionally drawn graphics cards require high-speed processing.

At present, the memory of game graphics cards on the market is 512MB-2GB, and the bit width is 64bit-256bit. (A few enthusiast level up to 512bit).

Frequency is also a measure of processing speed, as is bit width.

The final manufacturing process is also a technical indicator (also known as video memory technology). The higher the technology, the better the quality and overall performance of the card.

The current production process has DDR3 GDDR3 4 5, and the best is GDDR5. GDDR is a memory-specific technology. DDR is a general-purpose memory technology that can also be used for video memory.

Well, that's all that said. I'm sure you should know a little bit about it!

Remember to adopt! Thank you!

The core is the foundation of a graphics card. The strong core is the premise of ensuring performance, followed by the memory particles on the board PCB, not the capacity, but the bit width and frequency. If you queue from the above three factors alone, bandwidth and frequency are equally important, and capacity is secondary.

Overall, the core process, i.e., the manufacturing process and the core architecture, are the most important.

Importance order: bandwidth - frequency - capacity.

Decisively looked around the 1st floor and appeared theism again How did he get to level 8.

Generally, memory bit width is more important. Nowadays, the frequency of mainstream video memory of the same generation is not too different, and the width of video memory is particularly important. The width of the memory bit is like the width of the river, and the larger the width, the greater the instantaneous traffic and the larger the amount of GPU data exchange.

The memory frequency is like the speed of a river, and the benefits of a wider river are clear.

High-end graphics cards are all high memory bit width. Memory bit width literally seems to be a parameter of video memory, but it is the GPU core that determines the memory bit width, because the memory bit width controller is integrated into the GPU. High-end graphics GPUs can support higher bandwidth, so a high memory bit width means a higher-end GPU with stronger performance.

Memory bandwidth refers to the rate of data transfer between the display chip and the video memory. The memory bandwidth is determined by the memory bit width and the memory frequency, the memory bandwidth = the memory bit width x the memory frequency 8, and the memory frequency is determined by the memory speed, the rated memory frequency = 1000 memory speed x 2.

The bit width of the video memory is the number of bits of data that can be transmitted by the memory in a clock cycle, and the larger the number of bits, the greater the amount of data that can be transmitted in an instant, which is one of the important parameters of the video memory. Insufficient memory bit width can cause a choppy image at high resolutions, or an unstable frame rate, because the graphics card chip does not have enough data to calculate. Basically, 128-bit can cope with the mainstream high resolution.

Memory bandwidth = memory frequency x memory bit width 8, then in the case of the same memory frequency, the memory bit width will determine the size of the memory bandwidth. For the same 128-bit and 256-bit video memory with a video memory frequency of 2200MHz, then their memory bandwidth will be: 128 bits = 2200MHz * 128 8=, and 256 bits = 2200MHz * 256 8=, which is twice that of 128 bits, which shows the importance of memory bit width in video memory data.