Computer graphics card knowledge science (notebook and how to see graphics card parameters)
A brief description of notebook graphics
Mobile graphics, as the name suggests, are the graphics cards of applications on mobile platforms. Due to the high space requirements of mobile platforms, discrete graphics cards often appear in the form of integration into the motherboard, although it looks like 'centralized display', but this is a veritable 'unique display'. Because, even if it is integrated into the motherboard, it is divided into a separate area in the form of a whole. That is to say, the mobile platform simply fuses the motherboard PCB and the graphics card PCB together, eliminating the need for connected slots and gold fingers and display output interfaces. The display is different, the display chip does not appear independently, generally integrated in the north bridge of the motherboard (now integrated in the CPU), generally there is no independent video memory, which is the main difference from the unique display. However, now the notebook also has a DIY trend, many mobile platforms will also be distinguished from the motherboard, into a real unique display, the eye interface is MXM, in the barebones system book is more common.
'Single display' integrated into the motherboard (black area is roughly the graphics area)
MXM interface mobile graphics card GTX770M
There is no essential difference between notebook graphics and desktop graphics, but the core is really not manufactured on a production line. Notebook core focuses on low power consumption and low heat generation, desktop graphics card core focuses on high performance, so the frequency of many notebook graphics cards is very low. Some players like to overclock the notebook graphics card of the same parameters to the desktop level, which is actually very dangerous, the power supply of the notebook graphics card is much worse than the desktop graphics card, the core process is not focused on high-frequency high performance, the result of this is likely to be that the core bears load beyond its range, and the life is greatly shortened. However, it is not that the notebook graphics card cannot be overclocked, as long as the voltage is not added when overclocking is guaranteed, and the temperature does not exceed the limit of the machine, such overclocking is also acceptable.
How to easily distinguish the performance from the graphics card model and parameters
Initially use the software to see the parameters
Let's continue by introducing the GPU-Z recognition parameters
The image above is a screenshot of a GTX770 desktop graphics card
Don't think that the above are all in English, do not understand, in fact, each item has a Chinese explanation, move the mouse to each parameter to see the introduction of the Chinese, here will not be repeated.
Among these parameters, the parameters we need to pay attention to are as follows (shown in the red box of the figure):
Name
GPU (core codename)
Shaders (number of stream processors)
Memory Type
Bus Width
Memory Size
Bandwidth
Default Clock (core default frequency)
Memory
Boost (can be understood as turbo frequency)
The video memory section has just been covered, now let me explain the core codename and the stream processor.
The core code number refers to the development code of the graphics card's display core (GPU). The so-called development code is the corresponding basic code number given by the display chip manufacturer to a series of display chips in order to facilitate the unification of the management and drive architecture of the display chip in the design, production, and sales. Different display chips have corresponding development codes.
The first time the term stream processor appeared in people's sights, it dates back to December 4, 2006, when NVIDIA officially released a new generation of DX10 graphics card 8800GTX, in the technical data sheet, you can not see the two parameters that are usually used: Pixel Pipelines (Pixel Rendering Pipeline) and Verrtex Pipelines (Vertex Shading Unit), replaced by a new term: streaming Processor, Chinese translated as a stream processor (also called SP unit, a meaning) Its role is to process the data transmitted by the CPU, and after processing, it is converted into a digital signal that can be recognized by the display.
It can be said that the main difference between the high-end and low-end graphics cards in addition to the different cores is the number of stream processors, but there is a point to note, that is, the number of NV and AMD graphics card stream processors is not comparable, their two graphics card core architecture is different, can not be compared by comparing the number of stream processors to see the performance. For the same generation of graphics cards, the difference in performance can be roughly estimated by the difference in sp number.
Graphics cards after the Kepler architecture added boost functionality so that their cores can increase frequency within the range of power allowed at higher loads to improve graphics performance. The Boost frequency displayed on GPU-Z is not the actual maximum, if you want to know how high the boost frequency of your graphics card can be, you can turn on a rendering that comes with gpu-Z, as shown below
Once you've started, switch the GPU-Z tab to the "Sensors" item and look at the GPU Core clock to see what the maximum boost frequency is for that graphics card
For example, my GTX770M, which is displayed as boost 797MHZ, is actually 928MHZ
(PS: here is a hint, if it is a notebook graphics card, GPU-Z wants to view the unique display frequency need to right-click the program, select the graphics processor - high-performance NVIDIA processor, in order to let the unique display run this program, otherwise the running rendering is the core display, can not view the unique display frequency)
Simply judge the performance of the graphics card according to the model
Now let's look at the order of the parameters when selecting the graphics card.
First of all, we can look at the graphics card name. The name can represent the performance of the graphics card to some extent.
Nvidia graphics card name example: GTX780TI
GTX is a graphics card version, performance ranking GTX> GTS> GT>G
The previous old model graphics cards also have GS, GSO, GX2, etc., GX2 is a dual-core card, GS is lower than GT, GSO is higher than GS.
TI is the graphics card name suffix, which may represent strengthening or weakening, performance sort TI> no >SE> =LE. Some graphics card names are followed by M, which is the notebook graphics card code. Desktop and notebook graphics cards are not easy to compare, but generally the same model of desktop (desktop) graphics is better than mobile (notebook) graphics. Sometimes there is also the '+' sign, which represents the transformation of the original graphics card, which is generally better, such as GTX260+ > GTX260, but GTX460+ is inferior to the same frequency GTX460, which is stronger than the public version of the low-frequency GTX460.
The suffixes that already exist are TI, +, LE, SE, M, MX. MX is a plus version of the same model, and the performance has been improved.
Put a few typical graphics card GPU-Z screenshots
GTX780TI
GTX460+ (sometimes called GTX460 V2)
GTX560SE
GT640M THE
GTX670MX
In the number part, the first digit represents the graphics card series (if GT is placed last, the graphics card of that series is older than the graphics card placed in the front, for example, the 9600GT is older than the GT240), and the second digit represents the high and low end of the graphics card of the same series. Mainly look at the size of the second digit. In the same first digit, the larger the second digit, the better; the second digit is the same, the larger the first digit is better, but this comparison often does not work for the low end, such as GT440< GT240, GT520< GT220; if the two digits are different, it is not easy to judge at this time, and the main one with the large difference is selected, such as
GTX280>GT440
GTX280<GTX560TI
However, it still fails on the low-end, for example
GT610<GT220
GT630<GTS250
Example of AMD (ATI) graphics card: HD7970
HD is the graphics card version, now A card is no different, unified with HD beginning. The M at the end also stands for the mobile version
The first digit represents the graphics card series, the second digit represents the high and low end of the graphics card of the same series, and the third digit generally represents the high and low end caused by different specifications of the same core code (not necessarily the same core code, such as HD677O and HD6790). Mainly also look at the size of the second digit. The comparison method is similar to the N card, with the third place being the last factor to consider.
Now amd name has been replaced, using R5/R7/R9 naming, typically R9 290X, R9 M270.
The number after R determines the high and low end of the product, and the "M" stands for the mobile graphics card, which is similar to the one described above.
Further determine the performance of the graphics card according to the core parameters
However, looking at the core code to speculate on the performance of the graphics card is still too superficial, and many times it is not accurate.
Therefore, we also have to look at the core code number and the number of stream processors.
The core code indicates that the core is old and new, and sometimes the graphics card core code with different names is the same, and sometimes the graphics card core code with the same name may be different. The core generation determines the maximum number of SP.
Stream processors, the number of SPs is definitely the larger the better, but the premise of comparison is that the core codes are the same or the core codes are all in the same series. A card is not easy to distinguish whether it is the same generation, but the current A card different series directly compare the number of SP can also approximate the conclusion of good or bad, such as 800SP HD5770 is greater than 320SP HD3870. However, if the difference is only within 200SP, it is more difficult to distinguish, generally the performance of the new generation of graphics cards will be better, such as the 640SP HD7770 is greater than the 800SP HD6770
Compared with the A card, the N card core code is easy to identify, the core letters of the same series are the same, as long as the number size is compared, the core is good or bad. Note, however, that the last digit is as small as possible, for example the maximum number of stream processors for the GK104 is larger than the maximum number of stream processors for the GK106. The same series can directly compare the number of SP, the larger the graphics card is better, such as 192SP (GF106/GTS450, GF116/GTX550TI) is weaker than 336SP (GF104/GTX460, GF114/GTX560). Different series of graphics cards, the same number of SP, the older the series, the better (that is, the N card SP efficiency is getting lower and lower), such as 96SP's 9600GSO(G92)> GT240(GT216)> GT440(GF108), 384SP GTX560TI(GF114) > GTX650(GK107)
Below is a screenshot of the relevant GPU-Z
In addition, the new and old cores, the level and level can also be determined by the Technology (production process), Release Date (release date) and Transistors (number of transistors) displayed on GPU-Z.
After comparing the name and SP, you can look at the core frequency, and it is certain that the higher the frequency, the better...
After the core part is compared, the parameters of the video memory also need to be compared. For example, GT650M D5 Edition, GT745M, GT740M GK208 Edition, the number of stream processors in these three core aspects is exactly the same, the GT740M is the highest in terms of frequency, followed by GT745M, and the GT650M D5 is the lowest. But in reality, performance is arranged in reverse.
Inside is the difference in the memory parameters. The GT650M D5 uses GDDR5 memory, which has a high frequency and high bandwidth, more than double the bandwidth of the GT745M. Kepler 384SP is still relatively bandwidth-hungry, so the GT745M is obviously limited in bandwidth and causes performance to shrink, and ultimately not as good as the gt650M performance. The GT740M is less than half the bit width of the GT745M, and according to the bandwidth formula, the bandwidth is half less, and the performance drops sharply. According to previous reviews, such a reduction in bit width at the same frequency leads to a performance drop of up to 40%, which is almost half of the performance loss, which shows how important bandwidth is to the performance of the graphics card.
(But here it is still necessary to explain that it is meaningless for graphics cards of different architectures to directly compare bandwidth, the ability of the core first determines the level of performance, and bandwidth is only whether it can limit the core performance)
The above judgment basically allows you to compare the advantages and disadvantages of the two graphics cards (limited to the core of the same brand, N to N, A to A). The comparison of different brand cores can generally only rely on professional evaluation to compete, and the comparison of CPUs is the same.
Just introduced here, there is a supplementary welcome message.