Repost from:
Http://www.xfx.com.cn/forums/announce/announce.asp?boardid=222&id=17893
Overclocking is a DIY eternal topic, in a DiYer player's chassis, there is almost no accessories work in it should be. As a DIY fan, I am standing again to talk about some shallow understanding of overclocking.
If you have the following steps, it is still difficult to find the overclocking master, the overclocking master is on site guidance. Because the lower level is limited, just a rookie who loves YY overclocking.
Note: Any changes made to the graphics card, including the replacement of the radiator and refresh the graphics card BIOS, it is possible to lose the graphics card warranty. It is recommended to save the original graphics card cooling and the original card BIOS backup to prepare from time to time.
The key to overclocking is to identify bottlenecks affecting the rate of increased, and resolve this bottleneck in a targeted manner.
1, heat dissipation:
The low-end graphics card generally has a low frequency, so many models of products are only equipped with a heat sink or a smaller heat dissipation fan, generally not equipped with a sink. After overclocking, the graphics card is heated greatly, and the original heat sink is often unable to compete.
When the graphics core frequency is improved, the program is running with a 3DMark, and the program is running after a period of time; or the display frequency exceeds its default frequency or more, such as 3.6 ns, more than 550MHz, run the test program, starting a flower screen after a test program It is generally considered to be due to fever. To overclock to higher frequencies, it is necessary to replace a better heat sink.
Of course, when replacing the heat sink, pay attention to the installation of the heat sink, especially the non-standard radiator that uses its DIY, and improper installation is likely to cause damage or heat dissipation of the graphics card.
TT monster 3 radiator, Wenyodi friends use this stuff to exceed 5700LE
In fact, in most of the time, replacing a better heat sink, there is often a significant effect. Anyway, the heat dissipation problem is always placed first.
2, design of the graphics card itself:
Since the low-end graphics generally does not design a core, the independent power supply circuit is not displayed, and sometimes the core, the memory can be exceeded, but the two frequencies are combined, but it is very unstable. At this time, it can be appropriate Reduce the frequency, the other can be more than one.
So, is our overclocking should be prioritized to super-core or how to survive?
The display core determines the triangulation rate and pixel filling rate of the graphics card.
We know that 3D images are composed of multilateral rows, and the so-called polygons refer to triangles. Because the triangular graphics is the simplest polygon, and after the first triangle is generated, the spatial position of the next triangle can be determined as long as one end is specified. In summary, the triangle is the skeleton of the 3D image, and the triangle has determined the resolution of 3D graphics. For graphics cards supporting T & L or VS, the calculation formula of the triangular production rate is: core frequency * vertex unit pipeline / 4. Of course, this is just theoretical value, the actual triangulation rate of the graphics card does not reach this value.
Pixel filling refers to a skeleton of a 3D image consisting of a triangle, posted a layer of "skin", for the triangular production rate, the amount of pixel filling is larger, the cycle is longer, so the pixel filling is generated compared to triangles The rate is more important for the number of game frames. The formula of the pixel filling rate is: the number of core frequencies * pixel rendering pipeline.
Therefore, in the case of the graphics architecture, it is necessary to improve the triangular generation rate and pixel filling rate, it can only improve the graphics core frequency.
Save, also known as frame caches, in general, the display core stores the processed data in "frame" in the form of memory, then outputs the VGA interface to the graphics card via the RAMDAC (digital mode conversion module) (or by memory Direct output to the DVI interface). I believe that everyone will understand why the memory is overclocking, it is easy to cause the flower screen. To know, if the memory work is unstable, it will also reduce the image quality, so for the reduction in overclocking, you still have more Let's find the reason on the memory, don't die, Ramdac or low pass. Now the 3D game is increasing in the texture, and it is also a large number of light and shadow information, plus the Z buffer, frame buffer, and texture buffer, and the memory bandwidth is once a bottleneck that restricts graphics performance. The calculation method of the memory bandwidth is: the memory frequency * memory bit width / 8. In the case where the display bit is wide, you want to want the graphics card to have a larger data throughput, which can only improve the memory frequency.
Therefore, the core and the frequency of memory have a certain dependency, and the core frequency is more important. When the increase in the increase of the monitoring frequency exceeds the limit of the core data processing power limit, the graphics can not be improved.
In the form of a current form, for the 4-line, 128-bit display bit wide N card or A card, its core / memory ratio is 1: 1.6 to 1: 2, it is the most capable of playing graphics . 8 pipeline 256 bit wide graphics card is similar. However, different graphics cards are different in different environments, and the core execution efficiency will also be different.
In general, for low-end graphics, especially for 64-bit memory cards, its performance bottleneck is mainly in the memory bandwidth, so it can be prioritized to overclock. For medium and high-end graphics cards, the super-core is generally improved in the super-core graphics card.
In addition, due to overclocking, the power consumption of the graphics card increases, and the heat generation of the power supply circuit will increase, and heat dissipation is made of power supply circuit, which not only has the advantage of extending the life of the graphics, but also reduces thermal loss, improve the output power of the graphics power supply circuit. . Increasing the heat sink is effective and easier on the power field effector (MOSFET) or regulator.
3, power supply:
Don't think that the flower screen must be caused by memory, in fact, many situations will cause a flower screen. Power supply problem is more common. As for the power supply, in addition to the flower screen, what other problems will be caused, I can say that as long as the problems you want, the problems you can't think of.
There are now many power supplies, including some domestic major factory products, and the virtual standard is serious. So don't believe how much W. w on merchants or advertisements when you buy a power supply. The actual power of the power supply can be estimated by the 5V output current value * 10 on the nameplate on the power source; for the 1.3 standard power supply, it can be estimated with ( 5V output current 4) * 10. Alternatively, you can use OCCT to test, see if your power is moving your machine. In short, a strong power supply is essential. Even if it is not overclocking, in order to system security and stability, it is also strongly recommended to match a good power.
For so-called "power supplies of the magnetization", it is an inductive coil, which is a regulator technology, and does not increase the output power of the power supply. In fact, this technology is used in a lot of brands, it is not a patent for a brand power supply. (Only on behalf of personal views)
4, AGP power supply:
The lack of AGP power supply will also affect the overclocking ability of the graphics card.
The AGP power is not referred to by the power supply of the graphics card core or memory by the AGP slot. The power supply of the graphics card core or the memory is generally provided by the power supply. There is a specific definition in the AGP foot position, typically 3.3V, 5V, and 12V, and the transformer is completed by the internal power supply circuit of the graphics card. Generally, the AGP power supply refers to the power supply of the AGP signal, that is, the GPU and the AGP bus are exchanged. Since the AGP interface has a large bandwidth, it is very important to increase current to keep signal strength. In the case where the power is constant, the reduction voltage can effectively increase the current intensity. Therefore, its voltage is reduced with the update of the AGP version, generally: AGP1X / 2X: 3.3V, AGP4x: 1.5V, AGP8X: 0.8V (compatible with 1.5V). It can be seen that the AGP power supply causes an error in the data signal when the GPU is exchanged with the AGP bus data.
After overclocking, the increase in GPU data processing capabilities, the data throughput of the AGP bus has also increased, and it is natural to supply AGP power supply. If the system is unstable after the graphics card overclocking or using non-standard outer frequency, try to increase the main board AGP voltage; or try to drop the AGP standard, such as using the AGP8X to 4x (although in the test The AGP8X and AGP4x performance gaps are small, generally only 1% -5%, but it can be sure that the instantaneous data traffic of AGP8X may be much higher than AGP4X, and this momentary power consumption increases, it is possible to cause crash) or off Falling, such as AGP, maybe it will improve.
The main board with independent AGP power supply is very helpful to the overclocking of the graphics card. When purchasing the motherboard now, everyone often focuses on the main board's design of the CPU, but many people ignore the memory and AGP power supply. The AGP 4x, 8x voltage is only 1.5 or 0.8V, and when the high performance graphics card, the stability of the maintenance current is very important. This is also why the same graphics card may have different overclocking performance on different motherboards.
In general, the AGP power supply design of the two groups of MOSFET choke coils is best; second is a set of MOSFET choke coils, a set of MOSFETs. However, these two design costs are relatively high, generally not common. Nowadays, the epidemic design is: dual MOS tube or single MOS tube (double MOS design), with control chips. This design typically supports AGP voltage adjustment, and can provide current intensity of about 10-15A. The motherboard below this design is best not to buy again, and the boss of the stealing and reduction is not a good thing.
This topic related pictures
Sheng KV8PRO, AGP power supply is full
This topic related pictures
EPOX 8HDA5 , single MOSFET design
5, graphics card BIOS:
Updating the graphics card BIOS does not change from essentially.
Higher memory delay settings can bring better overclocking capabilities, but the high delay itself will reduce performance, but the frequency of memory is more wide, so the range of adjustable is more extensive, so overclocking flexibility is also greatly improved. .
In addition, through the study of BIOS, it shows that sometimes the high-end graphics card BIOS will be higher, such as 5950 BIOS, the display card core 3D voltage is increased from 5900 to 1.4V to 1.5V. This also has a great help to showing the improvement of core frequencies.
Moreover, refreshing BIOS is to solve hardware, such as refreshing 9550 BIOS is to relieve the ATI driver to the graphics card overclocking; 9500 hard change 9500PRO, or early NVIDIA graphics hardware is quadro card, often need to graphics BIOS Reconscence.
When choosing BIOS, you should pay attention to: the same or similar board type, it is best to be public version; the package, the particles are from one. But even if so, some may still be generated, such as S-TV-OUT that cannot be used, a flower screen, AGP 8X cannot be opened. Ritai Company has always been to the production public version of the graphics card and independently developed BIOS. Its BIOS has a good compatibility, and the memory delay setting is relatively conservative and is deeply loved by the majority of DIY players. In addition, I feel that the BIOS of XFX is also very good. If you use the graphics card of Xun, there is no need to go back to the Bios.
As for a card, who will know more about ATI than ATI? Find the original card BIOS.
Of course, some players like to make changes to the original BIOS to achieve optimal compatibility. Players who have no related experience should not take this insurance. If there is no special need or use the software overclocking, after all, the overclocking function of NV drive is now relatively perfect.
6, CPU:
Many friends have found that after overclocking, the performance before overclocking is basically no change or test score increases. There are two possibilities here.
Your CPU may become your computer's performance bottleneck. Your CPU's data processing power has reached the level. At this point, you should want to further improve performance, you need a more powerful CPU.
There is also a possibility that in some middle and high-end and high-end graphics cards, a "forced frequency" is added. When your graphics temperature exceeds a certain critical value, the graphics card will automatically work in this "forced frequency", actually The upper is to reduce the frequency. Only restart can be restored. You can view the actual operating frequency of the display with RivaTruner. In general, in the 4x.xx series drive, this mandatory reduction is performed in strict, and there is less appearing in 5x.xx and above.
7, graphics card version:
I often see that there are many players in the forum, such as: "Why is a friend's card that exceeds the XX frequency, but I don't go?" Or "I see the" I can exceed the XX frequency on a certain website. " My card is unable to exceed. " In addition to the graphics card self-chip body and website, there is also a possibility: the version of the graphics card is different.
Why will the manufacturer launched a different version of the graphics card? In general, there are several situations:
A, fix the BUG of the original graphics card.
Many of the first versions of the graphics card, because the technology is not mature, there are some designs defects, such as the early Ti4200 low temperature screen, or some early 5600 graphics card upgrade driver, request to insert an external power interface, and the graphics card is not designed such The classes of the socket. The new version of the product generally does not have such problems.
B, reduce costs.
Since the first batch of graphics cards are mainly to seize the market, manufacturers generally adopt products with lower technical requirements, simultaneous stability and compatibility with strong public versions. The second batch of products is mainly in order to expand the market. While price cuts, manufacturers' engineers generally have enough time to design a lower cost. For example, the early 5200U or 5600U, generally uses the 8-story PCB P140 public version, while most manufacturers have changed from the 6-laminate of P141, or 6-layer board design.
However, it is not possible to reduce costs. On the contrary, non-public low-cost design often represents strong research and development capabilities representing the manufacturer. With the smallest cost to achieve the desired purpose, this is the highest realm pursued by the designer. Personally appreciate the magic version of XFX, especially 5200ultra magic version, excellent workmanship, and materials are also good. You must know the 9550 design of the HY 2.8NS 6-layer version of the HY 2.8NS 6-layer version, which can reach more than 600MHz. (Note: Even is not a gunman)
This topic related pictures
ASUS 6800GT, non-english design, effective control cost, fully reflecting ASUS's super high design level of course, I don't say that garbage manufacturers who have spanning the work are vendors with high R & D capabilities, but they have those who have no product life. Under the premise of stability, performance, display quality, and safety, a newly developed and effectively controls cost-capable vendors for the PCB board. Non-English design does not necessarily have better or worse overclocking capabilities.
C, the upgrade of the product.
This is mainly to expand the market. In the case of the price constant, the manufacturer will use a higher speed memory, or more excellent materials to attract more consumers to buy.
D is affected by the upstream manufacturer.
Among them, the greatest influence is to store manufacturers and capacitors, especially in memory vendors. In the cost of the graphics card, the graphics core core is of course the highest, followed by memory. Due to the update of memory, the price fluctuations are also relatively large, and the storage of memory is not easy. So the general graphics manufacturer will not have a lot of inventory. The case of the capacitor is also similar. Moreover, the capacitance generally has a certain validity period, even if the capacitor is not used, its performance indicators will decline, plus the manufacturers like Nichicom, Sanyo, Rubycom it also limits the shipments of high-end capacitors. So once these upstream manufacturers have adjusted or control the shipments in the price, graphics card manufacturers often have to adjust their products.
Therefore, different versions of graphics card work materials will be different, and overclocking ability is naturally different. But even the same version of the product, because the chip batch factor, overclocking ability will have a certain difference.
8, motherboard BIOS setting:
In fact, the main board BIOS setting is generally not very great to show the improvement of the frequency of the core and the frequency of memory, but it is also related to the system's stability and CPU overclocking, and talks.
The first is a problem with non-standard outer frequency. Many people think that the use of non-standard outer fry will effectively improve the overall performance of the system, so that the AGP data is transmitted faster. This kind of saying is incorrect. The AGP mount is rising from 4X to 8X, the theoretical bandwidth has been doubled, but it shows that the promotion of subsystem performance is almost no in practical applications, and only 5% increase in some high-end graphics cards. It is difficult to have practical effects only on the point frequency that relies on boost AGP. I have done related tests on this person: on a motherboard of a GA-60XT-A (GIGATIT 815EPT, supporting AGP / PCI manual division), fixed CPU and memory frequencies on the outer frequency of 133MHz, through different division, Make the AGP on different frequencies, 3DMark test, test the graphics card for ELSA MX400, and the results test score have little variation. In addition, the use of non-standard outer fuses to produce considerable influence on graphics card life and system overall stability, generally recommended standard extravreads.
Second is the FSB Spread Spectrum option in the BIOS, and many people don't know the functions of these two items. These two options represent FSB spreadsheets and AGP spreads, which will reduce the generation of peak EMIs to reduce the interference of the computer to other electrical equipment (such as televisions, radios, etc.), generally said that the motherboard manufacturer is passed The CE certification is only added in the BIOS, which will affect the stability of overclocking after opening.
Finally, the CPU Interface option is generally Nomal and Overclock options or disable and enable options. This work is related to the work of the clock generator and the intensive device. According to different settings, different frequency modes are generated, and it is set to NOMAL or disable motherboards to have higher stability, and set to overclock or enable, CPU will Better overclocking.
9, operating system and drive:
Many old overclocking players have found that when the operating system is upgraded from Win98 to XP, the CPU or graphics card will seem to be reduced. Similarly, after replacing the graphics driver, the overclocking ability of the graphics card sometimes has some subtle differences. In fact, your CPU and graphics card have not changed, but the XP system is much higher than the hardware stability requirements, and some security hidden dangers that are not reflected in 98 are exposed under XP. The stability of 98 is not a real "no problem", but the problem is not discovered. The same truth of the graphics driver. So, don't look for the so-called more overclocking drive or system because of overclocking, because your hardware is actually unstable, so you can use only hardware life. Still old, realize the frequency, or enhance heat dissipation. It is also believed that the NVIDIA driver is limited to overclocking, and it is better to use Rivaturner or PowerStrip. Indeed, in some older version of the drive, I did hit the video card driver could not correctly identify GPUs and memory frequencies and could not perform overclocking or overclocking error. However, in some older versions of Rivaturner and PowerStrip, the graphics card still decreases the default voltage at 2D frequently due to the operation of the FX graphics card 2D / 3D, which causes the overclocking to run 3D games. The exhibition of overclocking ability. The new version of the driver and overclocking software did not find this image, at least in my 5900XT graphics card, three overclocking capabilities. Personally think that there is nothing in nature.
10, surrounding environmental factors:
The ground is poor, dust, static electricity, urban pressure is unstable, strong electromagnetic field is a terrible computer killer, temperature and humidity to overclocking, vibration, while brings noise, ruthlessly destroy your hardware . Give your computer to find a clean space, plus the quality chassis, a good ventilation environment, whether it is for your computer, or a great advantage for yourself.
11, other:
If there is no effectiveness through the above efforts, it indicates that the chip used by this graphics card is not good, or the circuit design and materials of the graphics card are not suitable for overclocking. It is generally no way to solve this situation.
Of course, some advanced DIY players will consider replacing filter capacitors, or by changing the resistance ratio of 2 external resistors at both ends of the load, the core chip, the IO cache, or the terminator voltage can be adjusted to make the graphics have better. Overclocking ability. But the players who have this ability are not used to see YY.
This topic related pictures
9700p used in step-down, pressurized usually used 2B pencils to be completed, but there is quite skill, not "painting"
Telling so much, saying, the core content is only two points: heat dissipation and power supply. Because the wiring design of the graphics card and the problem of chip quality are unable to solve it through DIY. So far, I have been in contact with the overclocking master, and it is basically from these two aspects. Even before I have heard that there is a fanatical overclocking enthusiast, overclocking by polishing chips (pair of chip polishing), saying that the bottom is also better heat dissipation effect.
In addition, each piece of graphics card, because the results of various internal or external causes will have different overclocking performance. The overclocking result of players show in the forum, and the overclocking value of the tests on each website can only be used as a reference data. It does not mean that your graphics card has the same overclocking ability, even the same version, the same batch of graphics card!
In summary, the current architecture of the graphics card is more able to determine the speed of the graphics card. In addition to a few products, the overclocking graphics card will not have quality flying. Tens of megabytes of frequencies cannot give game performance to change. The overclocking ability is not strong, and the manufacturer cannot be blamed. For the manufacturer, it is a qualified product that can be stabilized in the preset frequency. Finally, give overclocking final level of internal strength method -
Shun it naturally, can not force
