1 PCI bus in 3D application
AGP mainly aims at the data transmission bottleneck of PCI display card when dealing with animation and 3D drawing. As the processor becomes faster and faster, the bottleneck will become more serious, especially in the case of 3D images.
In 3D graphics rendering, not only image data, but also Z-axis distance data, TextureData and Alpha transform data are stored in the display memory of PCI display card. The more display memory for storing texture data, the better. From the point of view of the whole system, it is not as cost-effective to increase the main memory as to increase the video memory, and it is more effective to store the texture data in the main memory than in the video memory. In other words, when the application ends, the main memory space occupied by it can be recovered, and the texture data will not occupy the main memory space forever.
Unfortunately, when the texture data is moved from the main memory, the bottleneck of data transmission is also transferred from the memory bus on the graphics card to the PCI bus, and the transmission volume of texture data will exceed 100 MB/ sec. The existing PCI bus is far from meeting the requirements, so a new interface like AGP is needed to connect the main memory and the graphics card.
2.2 structure. Accelerated Graphics Port (which enables 3D graphics to be displayed at a faster speed on ordinary personal computers).
The purpose of AGP is to achieve high-performance 3D graphics rendering function at a relatively low price. To this end, Intel extended the three main specifications of PCI and defined AGP:
(1) pipeline processing of data reading and writing operations;
(2) The data transmission period is133mhz;
(3) The address signal is separated from the data signal.
The principle of AGP is to set the display chip independently on the system bus and connect the display chip directly with the memory controller circuit of the chipset. In this "point-to-point" connection, two edges (rising edge and falling edge) of the clock signal are also used for data transmission, so the speed is doubled. Because of point-to-point connection, a system can only have one AGP, so AGP will not replace PCI bus. The data transmission speed of the first generation AGP is 66MHz, which is twice that of PCI bus. The second generation AGP will reach 133MHz, which is enough to meet the requirements of playing DVD discs with software. The highest data transmission speed can reach 533 MB/ s, which is about four times that of current PCI. The comparison between PCI and AGP is shown in the following table:
Comparison between PCI and AGP
PCI bus AGP
Transmission mode synchronization
Memory priority access is not supported.
Data line bit width 32 bits 32 bits
Bus clock 33 MHz 66 MHz
The highest data transmission speed is 133 MB/ s 533 MB/ s.
The maximum number of expansion cards that can be connected is 5, 1.
Number of signal lines 49 65
The mapping processing of 3D graphics requires high data transmission speed between the display chip and the display memory. At present, most graphics cards use faster memory, but this will increase the cost of graphics cards. One of the compromise methods is to move the texture data from the main memory, which can reduce the capacity of the main memory and thus reduce the cost of the graphics card.
AGP is not only used for 3D graphics, but also effective for 2D graphics. Because the graphics card is connected to the main memory through AGP and chipset, the data transmission speed between the display chip and the main memory is improved, so that the texture data originally stored in the display memory can be directly stored in the main memory, which can improve the memory bus utilization efficiency of the main memory, improve the picture update speed and the transmission speed of data such as Zbuffer, and also reduce the load of PCI bus, which is beneficial to other PCI devices to give full play to their performance. You know, in the PC98 specification, ISA bus has been canceled, and ISA devices will eventually be eliminated. Therefore, it is a very necessary step to migrate the graphics card that occupies a lot of bandwidth of PCI bus to AGP.
AGP also has good performance in the transmission effect of image data. When MPEG-2 image data is decompressed by CPU, it needs to be written into the display memory through the bus. The decoded MPG2 image data of the whole picture needs to be transmitted at the speed of 15 ~ 20mB/ s. Although the actual data transmission speed of PCI bus is 27 ~ 33mb/ s, if the data transmission is not coordinated properly, the picture will be very unsmooth.
At present, there are still two factors that limit the development of AGP. One is the data transmission speed of main memory. The display chip supporting AGP needs to access the main memory when rendering 3D graphics, so it will increase the memory bus traffic of the main memory, which generally requires a speed of more than 800 MB/ s.. But at present, the data transmission speed of main memory is mostly 200 ~ 300 MB/ s. At this speed, even if AGP is used, it is impossible to make a detailed three-dimensional graphic description. In order to achieve the data transmission speed of 800 MB/ s, it is necessary to have high-speed DRAM, such as SDRAM, RDRAM or other devices above 100MHz, such as S-gram and VRAM. Another problem of AGP is the compatibility of graphics cards.
future
AGP is an open specification, and manufacturers do not have to pay patent fees. At present, 3D graphics card manufacturers such as 3Dfx, 3Dlabs, ATI, CirrusLogic, Rendition, S3, Trident, etc. Indicates support for AGP, and some prototype products have been launched. Intel not only signed a contract with Microsoft, but also encouraged many graphics card manufacturers to adopt AGP. At present, some high-performance PCs have been adopted first. Therefore, AGP can be popularized in a short time. Intel believes that by the year 2000, 90% of personal computers will be equipped with AGP display cards.
In order to give full play to the advantages of AGP, Microsoft has supported AGP function in its new versions of Windows 98 and Windows NT 5.0, and provided programming interfaces for software vendors through DirectDraw API.
Motherboards with AGP interface have been put on the market. For example, the latest motherboards of elite, Asus, Zhong Ling and other companies all adopt Intel 440LX and 440BX chipsets supporting Pentium II, while other chipset manufacturers such as VIA have also introduced chipsets supporting Socket 7 motherboard AGP of Pentium MMX CPU.
The graphics cards of AGP interface are all 3D graphics cards, using high-speed memory such as SDRAM or RDRAM. Trident's 3D images 985 and 875 both support AGP and have TVOut function.
From the perspective of prototype products, AGP will not significantly increase the cost of graphics cards, but it is much more powerful. For example, Trident's 3D Image 985, besides the chip itself, has an MPEG2-2 decompression chip which can play DVD discs, fully meeting the needs of future multimedia computers.
Talking about AGP technology
1.AGP is a "specific medicine" to improve the speed of graphics/video processing.
As mentioned above, in 3D graphics display, the bottleneck of high speed is "graphics texture processing", which needs to transmit a large amount of bitmap data at the transmission rate of 100Mbps (resolution 640×480 dots) ~ 150 Mbps (resolution 800×600 dots), while all of the existing.
In PC, 3D graphics processing can be roughly divided into "geometric transformation" and "drawing coloring". Both of these processes are undertaken by the overloaded CPU. So a three-dimensional graphics chip is used instead of CPU to deal with the "drawing and coloring" with a large amount of processing. In order to reduce the cost of graphics card, we must try to reduce the capacity of graphics memory, so we store texture data in main memory. However, in the current system, main memory and video memory are connected through PCI bus, and its maximum transmission rate is 133Mbps, while data is sent to main memory through HDD, LAN, sound card, etc. They are all through PCI bus, but the actual transmission rate is much lower than 133Mbps. Therefore, the special interface AGP for graphics data is introduced.
As we can see, AGP directly connects main memory and graphics memory. The width of AGP bus is 32 bits, the clock frequency is 66MHz, it can work at 133MHz, and the highest transmission rate can reach 533MBps. The primary purpose of AGP is to put texture data in main memory to reduce the capacity of graphics memory, so as to produce cheap and high-performance graphics cards. AGP is not only used for 3D image processing, but also for animation reproduction. The decompression of MPEG2-2 animation data requires a transmission rate of about 30Mbps, but PCI bus is not competent, and APG is more comfortable.
It is of great significance to adopt AGP in data transmission. The current PCI bus is the bottleneck of transmitting video and 3D graphics data. The transmission rate of AGP is 533Mbps, which is four times that of PCI. It is expected to become a new generation of buses to eliminate this bottleneck.
Intel Corporation, the overlord of PC CPU chip, declared in "Graphics Controller' 97" that starting from 1997, the following three devices will be standard in PC: 3D graphics drawing device equivalent to street game machine; Reproduce MPEG2-2 video equipment recorded in DVD-ROM with software; The video conference equipment that conforms to the technical standard of H.320/H.324 (ITU-T: Electrical Communication Standardization Department of International Electric Union) advocates the use of AGP and MMX to realize the above three kinds of equipment. Accordingly, X86 compatible chip manufacturers expressed support for MMX, and graphics control chip manufacturers also expressed the need to adapt to AGP.
MMX is an internal problem of the processor, while AGP will change the architecture of personal computers. In order to adapt to AGP, graphics control chip and memory /PCI control chipset must be redesigned.
Indeed, AGP is a "panacea" to improve the performance of 3D graphics. However, it must try to reduce the cost while improving the performance, so that it can be configured in a general-price PC.
Unfortunately, AGP sacrifices universality and scalability. The reason is that only 3D graphics control chips can be connected to AGP. Although PC is equipped with multimedia boards, such as graphics, MPEG2-2 decompression and video capture attached to 3D devices, the "beneficiary" of AGP is only the graphics card. Therefore, we dare not assert that AGP is "the best choice for a new generation of buses".
2.SGI's "unique method"
SGI Company proposed another alternative to AGP. In June 1996, 1 1, O2 graphic workstation with advanced UMA (unified memory architecture) was launched. O2 graphics workstation is the first system in the industry to adopt a unified memory structure. It relies on its 64-bit MIPS RISC microprocessor to integrate the capabilities of three-dimensional graphics and image processing, video, audio and compression, thus achieving super performance at a low price. It breaks through the traditional data transmission barrier based on bus, so that CPU graphics and image processing and I/O can directly access the memory at the speed of 2. 1Gbps, and transmit information quickly.
The focus of O2 graphic workstation is to reduce the cost and improve the performance as much as possible. Using UMA technology, four kinds of peripheral chips such as graphics controller and video processor and main processor can use main memory (SDRAM). Generally speaking, if UMA devices are used, when the access applications of multiple peripherals are concentrated in main memory, the performance will decrease. Therefore, in O2, an ultra-high-speed bus with a width of 256 bits and a clock frequency of 66MHz (the maximum transmission speed is 2. 1Gbps) is used to connect the main memory to suppress the performance degradation.
UMA plays an active role in all multimedia data operations, such as 3D graphics rendering, video reproduction and video capture. For example, the performance of 3D graphics depends largely on memory capacity and memory access performance, because Z buffer and texture data area are frequently accessed when processing graphics. According to Microsoft's calculation, in the popular color representation mode of 640×480 pixels, when using binary filtering and 24-bit Z-buffered texture mapping to draw 3D objects, memory bandwidth of about 30Mbps is needed. In addition, only Z buffer and texture data are stored at this time, which requires 4MB of memory. If UMA equipment is used and the graphics control chip uses the main memory as the frame buffer, then the special frame buffer can not be used, and the texture data area can be guaranteed to the maximum extent in the empty main memory area, which is expected to further improve the performance of 3D graphics.
UMA is particularly effective in video capture. Using a camera to capture video, and then pasting it as texture data of a 3D object, the video image can be reproduced in real time. Because of the use of UMA mechanism, as long as its memory pointer is passed to the graphics control chip as a pointer to capture data, the captured data can be sent to the main memory.
3.AGP is not a bus
It is the same as UMA, except that AGP is only a technical standard that enables peripheral devices to access memory at high speed. Specifically, the 3D graphics chip is connected with the memory /PCI chip, and the 3D graphics chip can use the main memory as a frame buffer to realize high-speed access. Strictly speaking, AGP is not a bus, it is just a "port" considering one-to-one (point-to-point) connection.
Therefore, the main goal of AGP is to draw three-dimensional graphics. The data bus width of AGP is 32 bits, and it has two working frequencies of 66MHz and 133MHz, with the highest data transmission rates of 266Mbps and 533Mbps respectively. The memory /PCI control chipset corresponding to AGP has a table named "GART", and the 3D graphics chip can freely map the main memory to its own address space in units of 4KB. The mapping area in main memory can be discontinuous, but it must be in 4KB units.
In addition, AGP has a positive effect on the reproduction of MPEG2-2 video. However, this is limited to the case where the processor is used to decompress MPEG2 video data without special decompression hardware. When decompressed by the processor, the decompressed video data can be transmitted to the video memory through AGP when the picture is displayed. However, if a dedicated MPEG2-2 decompression card is used, the decompressed data must be transmitted through PCI bus instead of AGP. In the MPEG2-2 specification, video with 7200×576 pixels and 30 frames per second is mainly used. Theoretically, the data transmission capacity of 36Mbps is needed to transmit decompressed data. The actual transmission rate of PCI is 30 ~ 40 Mbps. If you use PCI bus to transmit, the picture will shake. Intel recommends using the main processor to decompress MPEG2 video. In AGP, MPEG decompression card is no longer considered.
Video capture card can't be connected with AGP card, and it can't be used for texturing only by passing the memory pointer of data acquisition to a graphics control chip like O2.
4.AGP has a strong "patch" color.
Many PC graphics experts predict: "It will take two or three years for O2 architecture to be applied to PC." For example, relevant institutions have formulated the technical standard of PCI bus, with a width of 64 bits and a clock frequency of 66MHz. Its theoretical data transmission speed is 533Mbps, which is the same as AGP. In addition, VESA (video electronics standards association) has also planned to establish UMA technical standards for all external devices connected to PCI bus. If the UMA mechanism is installed on the PCI bus with a width of 64 bits and a clock frequency of 66MHz, its structure will become an O2 graphic workstation, so that all multimedia mechanisms can work smoothly.
Peripheral devices such as SCSI control chip, modem and serial/parallel controller do not need to transmit data faster than the current PCI bus, but they must work at the clock frequency of 66MHz. In this way, the manufacture of various such control chips not only increases the cost, but also is complicated to debug. However, if new equipment is introduced to replace AGP in the next 1 ~ 2 years, new machines must also be purchased, which will inevitably hinder the popularization of PC.
5.AGP is a feasible solution at present.
In fact, AGP is the most realistic solution to realize PC graphics and video processing functions. O2 is the technical standard for high-performance and high-price workstations exclusively formulated by SGI. It is very different from a PC made up of products from many manufacturers. For example, it connects the main memory to the bus with the highest data transmission speed of 2. 1Gbps, and integrates the reproduction mechanism of drawing 3D graphics and the main memory controller into one chip. These are technologies that can only be realized by closed independent manufacturers. In a PC with many manufacturers' products, it is really "reluctant" to realize a device completely corresponding to O2. And this is contrary to the spirit that PC regards the open environment as the "soul". On the contrary, AGP can be developed under the design idea of configuring AGP in low-cost PC, and the corresponding device (graphics control chip) is simple to manufacture and low in cost. For example, because AGP is only limited to connecting one device (except the main memory /PCI control chipset), the connected devices are easy to develop, and there is no need to install a special AGP arbitration circuit in the main memory /PCI control chipset, which can reduce the cost. In fact, the so-called PCI bus is the bottleneck of transmitting a large amount of data, which only refers to 3D graphics chips.
AGP is essentially an extension of PCI technical standards. This is also to simplify the development and design and make it similar to PCI bus. AGP is different from PCI bus in that its address lines and data lines are separated (PCI has 49 signals, while AGP has 65). Pipeline processing can be realized, and the actual data transmission rate can be improved; The address line and data line are separated, so there is no "overhead" of switching, which improves the performance of random access to main memory.
The memory /PCI controller chipset has a "transaction processing" queue to realize pipeline "processing". Once the graphics control chip sends a request to the main memory /PCI control chipset, it immediately releases the bus. The main memory /PCI control chipset can store multiple application commands in the queue and process and respond in turn according to the priority. The graphics control chip can accept the processing results within the data waiting time, so the overall use efficiency of the bus can be improved.
6. Thinking about the overall structure of PC.
Although AGP is a feasible scheme to realize the graphics and video processing function of PC, it is still a technical standard with strong "patch" color. Can AGP "take root and settle down" with a "permanent" device commensurate with the investment, or is it a flash in the pan like the VL bus in the past? It's hard to be sure now. On the contrary, AGP was put forward to popularize the demand of 3D graphics. If the demand for 3D graphics "shrinks", the failure of VMC(VESE Media channel) and SFBI (Shared Frame Buppzer Interconnect) may be repeated. How to use multimedia PC in the future is still inconclusive. Intel's prediction is only based on the judgment that the number of users playing "games" and MPEG2-2 video images on PC will increase sharply. More importantly, PC should be able to play "games", play MPEG2-2 videos and even play video capture. From this perspective, there will be new applications and services, and a brand-new multimedia world will be presented to us.
In order to further popularize PC and open up a huge home PC market, we should not only pay attention to immediate interests, but also have a long-term multimedia bus. Today, it is extremely urgent to design an ideal overall structure of multimedia PC.
Brief introduction to the characteristics of PCI Express
Low point-to-point connection mode
Compared with the traditional PCI bus, PCI Express has a fundamental innovation in its working mode-point-to-point bus connection. We know that the traditional PCI bus works with exclusive bandwidth, and only one device can communicate on PCI bus at any time. Once the number of devices on PCI bus increases, the problem of bus control contention will seriously restrict the performance of PCI devices.
PCI Express bus adopts point-to-point connection, and each device establishes its own transmission channel when it needs to transmit data. For other devices, this channel is closed, each channel does not interfere with each other, can work at the same time, enjoy bandwidth, and greatly improve data transmission efficiency.
● Serial transmission mode
The data transmission of PCI Express is serial, adopting "voltage differential signal transmission", that is, two lines, and the voltage difference between them is expressed as logic "0" and "1". Every two lines form a channel, and the theoretical transmission rate of each channel is 2.5 gbit/s. In fact, there can be two transmission channels, which are divided into uplink and downlink, so that PCI Express can work in duplex state.
Previous computer architecture diagram
● High speed transmission
PCI Express is divided into × 1, ×2, ×4, ×8, × 16 and ×32. The "× 1" structure has four lines, that is, two channels form a set of uplink and downlink. Therefore, the bandwidth of PCI Express × 16, which will replace the AGP interface, is 5GB/s (2.5GB× 16/8 = 5GB/s), but the actual effective bandwidth is 4GB/s (excluding 20% of embedded clock signals) due to the adoption of 8b/ 10b coding. Similarly, the "× 1" single channel that replaces the current PCI slot should have a bandwidth of 250 MB/s.
However, according to some foreign media, the next generation of Intel chipsets (i9 15, i925, etc. ) has not yet supported full-duplex mode, and there are still some technical problems to be solved in full-duplex mode, which may not be realized until the end of 2005, but even the single-channel transmission rate is much more advanced than the current interface technology.
Architecture diagram of PCI-E
● Hot plug support
The data transmission distance of PCI Express bus is as long as 3m, which makes the hardware subsystems completely separated from each other in space and connected only by cables. It supports hot plug function, and can monitor all access devices in real time, so that hardware manufacturers can design components whose shapes and sizes meet modular requirements. When users need to expand and upgrade hardware, they just need to unplug the old one and plug in the new one, without turning it off.
● Good compatibility
In addition, the PCI Express bus is compatible with the PCI specification at the software level, and can be used without updating the operating system and BIOS. In the future, motherboards with PCI Express bus can still support PCI slots, and expansion cards with various PCI interfaces can also operate normally in low bandwidth mode. This provides a foundation for the rapid popularization of PCI Express, without the embarrassment of waiting for software.
Influence of PCI Express on graphics card
As the graphics chip is updated every six months, the data processing ability of PC graphics card is getting stronger and stronger, and the required bandwidth is getting bigger and bigger. From ISA interface in 1980s to PCI interface in 1990s, and now to AGP interface, the bandwidth has developed from only 8.33MB/s to 133MB/s, and gradually climbed to 2. 1GB/s of AGP 8×.
model
full duplex
single
PCI-E × 1
500 MB/ s
250 Mbytes/sec
PCI-E ×2
1GB/ sec
500 MB/ s
PCI-E ×4
2GB/ s
1GB/ sec
PCI-E ×8
4GB/ s
2GB/ s
PCI-E × 16
8GB/ s
4GB/ s
PCI-E ×32
16GB/ s
8GB/ s
AGP 8×
-
2. 1GB/ sec
Interconnection standard between computer and its peripheral equipment
-
133.3 Mbytes/sec
personal savings account
-
8.33 Mbytes/sec
PCI Express × 16 can provide 4GB/s bandwidth in the early stage, which is much higher than the limit bandwidth of AGP 8×, and has obvious advantages. When the technology is mature, the duplex PCI Express or the next generation SPEC 2.0 specification can provide 8GB/s or even higher bandwidth, paving the way for the rapid development of graphics cards.
In fact, the contribution of PCI Express × 16 to graphics cards is not only to provide high bandwidth. On some high-end graphics cards, many products need external power supply, mainly because the AGP slot can't provide more power, only 30W. This is far from enough for the current high-end graphics card, so we have to supply power to the outside. Although PCI Express × 16 can't completely solve this problem, it can definitely alleviate the problem, because PCI Express × 16 can provide about 60W power supply.
At present, both NVIDIA and ATI have introduced graphics cards supporting PCI Express, but the implementation methods are different. NVIDIA is now promoting HSI bridging technology, and ATI is promoting products that chipset directly supports PCI Express × 16. Actually, if you think about it carefully, in the initial stage of PCI Express interface, the graphics card products are definitely not as rich as the mature AGP 8× interface products, so the transition mode in NVIDIA is still very good. All chips can be upgraded to support PCI Express only by a bridge chip, and then they can reach the level of AGP 16× through internal adjustment, and the bandwidth is the same as PCI Express × 16. ATI's complete PCI Express chip is more suitable for long-term development, but in the initial stage, the product line is inevitably not as rich as NVIDIA's, and it is very likely to lose business opportunities.
At present, NVIDIA's PCI Express products are about to enter the market, and ATI seems to be only the core of RV380 and RV370. Anyway, the product line is too monotonous. NVIDIA will launch NV4x core after PCI Express matures, which may be the core that officially supports PCI Express, and ATI will launch R423 core accordingly.
AGP Express Global Exclusive Disclosure
Nowadays, many people's computers are based on 1 to 2 GHz processor platform. For such a platform, replacing a better graphics card may not only improve performance more obviously, but also make the price more acceptable. Therefore, it seems that users who are going to upgrade their computers are not particularly interested in the platform of the 900 series chipset. & lt/FONT & gt； & lt/P & gt；
& ltP & gt& ltFONT face=Arial size=2>。 The elite noticed this a long time ago. They exhibited a motherboard based on 9 15 chipset at Computex 2004, but added an AGP slot next to the standard x 16 PCI Express slot. However, it was from then on that Intel made up its mind to completely give up supporting AGP in the 900 series chipset. & lt/FONT & gt； & lt/P & gt；
& ltP & gt& ltFONT face=Arial size=2>。 A feasible compatibility scheme is to set a bridge chip on the graphics card, so that PCI Express graphics card can be used in AGP slot. However, it is conceivable that this bridge chip will increase the manufacturing cost of the graphics card. Moreover, on the whole, this compromise scheme has lower performance than the scheme that the graphics card adapts to the pure PCI Express motherboard, so most people prefer to add more money to choose the latter. & lt/FONT & gt； & lt/P & gt；
& ltP & gt& ltFONT face=Arial size=2>。 The elite chose a solution closer to AGP, and they used an old PCI interface on the motherboard. At present, it seems that PCI interface is not the mainstream standard because of limited bandwidth, so this method is also a transitional solution. Then let's approach this motherboard. & lt/FONT & gt； & lt/P & gt；
& ltP align = left & gt& ltFONT face=Arial size=2>。 ECS 9 15P-A & lt； /FONT & gt； & lt/P & gt；
& LTPAlign = center & gt< font size = 2> & ltimg title = actual combat 4-head output-elite PCIE/AGP dual-slot motherboard style = border-right: purple1pXSolid; BORDER-TOP: purple 1px solid color; Border-left: purple 1px solid color; BORDER-BOTTOM:purple 1px solid src = upload files/2004 1029 1783580 . jpg align = no & gt； & lt/FONT & gt； & lt/P & gt；
& ltP & gt& ltFONT face=Arial size=2>。 ? In addition to PCI Express and two AGP-like graphics card slots, this motherboard is a very typical 900 series motherboard. The elite emphasizes the strong compatibility of the motherboard with different interface hardware. There are two PCI slots and two x 1 PCI Express slots on the motherboard. Users can choose to keep the original DDR400 memory or add DDR2-533 memory. Moreover, the PCB design of this motherboard is also suitable for Intel 9 15G chipset with integrated graphics card. & lt/FONT & gt； & lt/P & gt；
& ltP & gt& ltFONT face=Arial size=2>。 ? The 6-channel integrated sound card system based on C-Media AC97 coding chip can meet the general quality of multimedia applications. Elite provides coaxial digital output on this motherboard, but does not provide connector module. The network processing module selects Realtek 8 1 10S Gigabit Ethernet controller and works through PCI Express bus. & lt/FONT & gt； & lt/P & gt；
& ltP & gt& ltFONT face=Arial size=2>。 ? In terms of storage, Elite continues the solution integrated with intel chipset, supporting four serial ATA interfaces, and the other UltraATA/ 100 interface supports standard IDE devices. We found four fan connectors, which can make the airflow in the chassis flow more effectively, which is beneficial to heat dissipation and system stability. The north bridge chip did not use active heat dissipation, but only attached a heat sink. & lt/FONT & gt； & lt/P & gt；
& ltP align = center & gt& ltFONT face=Arial size=2>。 & ltIMG title= actual combat 4-head output-elite PCIE/AGP dual-slot motherboard style = border-right: black1px solid; BORDER-TOP: black 1px solid color; Border-left: black 1px solid color; BORDER-BOTTOM:black 1px solid src = upload files/2004 1029 178462 1 . jpg align = no & gt； & lt/FONT & gt； & lt/P & gt；
& LTPAlign = center & gt< font size = 2> & ltimg title = actual combat 4-head output-elite PCIE/AGP dual-slot motherboard style = border-right: black1pXSolid; BORDER-TOP: black 1px solid color; Border-left: black 1px solid color; BORDER-BOTTOM:black 1px solid src = upload files/2004 1029 1784360 . jpg align = no & gt； & lt/FONT & gt； & lt/P & gt；
& ltP & gt& ltFONT face=Arial>。 & ltFONT size = 2>& ltSTRONG & gt Dual PCI = AGP Expressway.
& ltP align = center & gt& ltBR clear=all>。 & ltFONT size = 2>& ltIMG title=4 simultaneous outputs-actual combat style of elite PCIE AGP dual-slot motherboard = border-right: black1pXSolid; BORDER-TOP: black 1px solid color; Border-left: black 1px solid color; BORDER-BOTTOM:black 1px solid src = upload files/2004 1029 178437 1 . jpg align = no & gt； & lt/FONT & gt； & lt/P & gt；
& ltP align = left & gt& ltFONT size = 2 & gt？ The technology of AGP Express slot is similar to that of traditional PCI slot, but the power consumption is twice that of the latter.
& ltP & gt& ltFONT face=Arial size=2>。 Undoubtedly, the biggest highlight of 9 15P-A motherboard is the AGP adapter slot. As we just mentioned, Intel has decided to completely abandon its support for AGP in the 900 series chipset. Therefore, users who want to choose the 900 series platform must obtain PCI Express hardware. Therefore, users who have just spent hundreds of dollars on high-end AGP graphics cards must abandon the expensive AGP graphics cards if they want to use the 900 series platform, which is a very nerve-racking problem. & lt/FONT & gt； & lt/P & gt；
& ltP & gt& ltFONT face=Arial size=2>。 It seems that elites have been thinking about this problem for a long time, and their solution is the motherboard in the picture. Users can use the weakened AGP slot on the motherboard as a transitional solution. Fans of pens and robes? Mechanical GP graphics card, wait until the price of PCI Express graphics card comes down before buying. Therefore, AGP Express was born. Based on the PCI interface standard, it concentrates the voltage collapse of the original two PCI slots in one AGP slot, which can not improve the performance of the AGP interface, but is just a PCI slot covered with AGP coat. & lt/FONT & gt； & lt/P & gt；
& ltP & gt& ltFONT face=Arial size=2>。 Although it is not a real AGP interface, AGP Express also provides compatibility for most AGP devices. Obviously, AGP Express has no DIME (Direct Memory Execution) function, and another function supported by the chipset, GART (Graphic Address Remapping Table), also depends on the system memory. & lt/FONT & gt； & lt/P & gt；
& ltP & gt& ltFONT face=Arial si