What is the throughput of an agp 8x video card: AGP 8x announced | Eurogamer.net

AGP 8x announced | Eurogamer.net

Yesterday CPU giant and motherboard chipset manufacturer Intel
proudly unveiled its

draft spec for AGP 8x (you will need Acrobat installed to read
that file), which has caused quite a noticeable stir amongst those
in the know, using all sorts of new techniques to increase the
bandwidth of the port. However, we would be naive to suspect that
amidst all the oo’ing and aar’ing there weren’t a number of people
scratching their chins and wondering, just what the heck is AGP
anyway? And from a gamer’s point of view, why should I be
interested?

You no doubt realise that AGP stands for Advanced Graphics Port,
and comes in several flavours, AGP 1x (or just AGP), AGP 2x, AGP 4x
and AGP Pro. The concept is thus; back in the day, all graphics
cards were either ISA or PCI, PCI being the more up to date and
common. A PCI graphics card though, was restrained in its
performance by several factors. Not only the data transfer rate
between the PCI Bus and the CPU (which was an incredibly limiting
factor by today’s standards), but also the problem that a lot of
cards were trying to share that narrow pathway of data.

So when the AGP port was created, it was heralded as a chariot of
gold so to speak for all the lovely data your graphics card wanted
to push back and forth between it, the processor and other areas of
your computer. By using this new, isolated transfer mechanism
running at 66MHz independant of the PCI bus, the amount of
bandwidth allocated to the graphics card became ludicrously big,
and AGP was clearly here to stay. With the introduction of AGP 2x,
the bandwidth doubled, and 2x is where a lot of people using the BX
chipset (Intel’s most widespread desktop motherboard chip) still
are today. Newer motherboards have now supported AGP 4x for quite
some time, but there were cries about its stability at first, and
as it doesn’t actually improve performance over AGP 2x by that much
in a handful of the top-selling graphics cards, not too many people
are all that bothered by it. AGP Pro is, well, an experimental
technology you could say, which hasn’t really been put to good use
just yet.

So AGP 8x (which may not be with us until 2003 at this rate) once
again improves the situation for this ham-fisted graphics card
producers. Of some interest is that the new bus will also support
Intel’s 64-bit processor Itanium and its successors, but more
importantly, thanks to a technique similar to that of the Athlon’s
DDR front side bus, the 66MHz bus frequency is to be effectively
doubled by using of the rising and falling edge of the clock
signal. Confused? Well, currently, a clock signal is sent at 66MHz,
and the card occupying the AGP slot can use one side of the clock
signal (which rises and falls) to transfer data. The Athlon
introduced a new technique which allowed motherboard manufacturers
to utilize the rising and falling edge of the clock signal as well,
doubling the transfer throughput as a result. Intel will now take
advantage of a similar technique to allow its AGP 8x spec to double
throughput. All this along with a proposed bandwidth of 2Gb/second
means AGP 8x will be a force to be reckoned with when the time
comes.

For gamers, this means more high resolution graphics at faster
speeds and with less jerking. With a high enough speed processor
and an AGP 8x graphics card, you could expect performance far in
excess of anything we can imagine today.

AGP 8X Graphics Cards | Cadalyst

31 Jan, 2003

By: Ron LaFon

---

Bigger Pipeline Speeds CAD Performance

Though
our roundup of current graphics cards features some very strong contenders, no
across-the-board leader emerged. What our tests did show quite clearly are the
individual strengths and weaknesses of the various cards. Overall, we found strong
3D performance, along with some weak OpenGL and 2D test scores. It’s not easy
to engineer and produce a graphics card that excels in both 2D and 3D, but the
lower 2D scores were a bit surprising-particularly because 2D still represents
a significant percentage of the workload in most CAD firms. Some results will
improve with later iterations of drivers, but drivers alone can’t make up for
a graphic chipset that offers weak 2D performance.

The most significant new feature is that three of these graphics cards support the AGP 8X interface standard, which can transfer 2GB of graphics information per second between the graphics card and system. This should greatly reduce at least one of the performance
bottlenecks in current high-end systems and result in higher performance scores, though the real benefits of AGP 8X may not be realized until faster systems become available.

HOW WE TESTED
Cadalyst Labs tested the graphics cards on an Xi Computer MTower 4800/3600X DPR system based on dual Intel Xeon 2.8GHz processors with 1GB of RAMBUS RAM. This system has a 120GB hard disk and runs Microsoft Windows 2000 Workstation with service pack 2. We used this system because it easily accommodates the 3Dlabs Wildcat4 7110 graphics card, a full-length card that requires two slot spaces (although only one is actually used-the card is thick), and because it has an AGP 8X slot.

We tested all graphics cards with MAXBench 4 running under discreet 3ds max 5 and the Pro/ENGINEER proe-01 test from SPEC ViewPerf v7.0. These tests are accompanied by the Cadalyst Labs C2001 benchmarks, which thoroughly test AutoCAD 2002-based systems. The C2001 test is not designed to necessarily showcase the fastest CPUs, hard disks, and graphics cards. Rather, it shows how these components all work together in relation to work typically done in an engineering firm. We are always under pressure from vendors to change the C2001 test procedures to showcase specific products, but never fear, our test procedure remains unchanged.

Once again, drivers often make or break the performance of a graphics card and can make a feature (or application) virtually unusable. Driver releases continue unabated, even over the short period of time encompassed by our testing.

Just after we finished this round of testing, Microsoft released DirectX 9, and ATI was first out of the gate with DirectX 9-compatible drivers. Though DirectX performance doesn’t typically affect our benchmark tests, new drivers can and often do. We’ve increasingly come to think of our benchmark test series as a snapshot of the state of graphics cards and drivers at the time we run our tests. As such, the tests give a good indication of the strength and weaknesses of the individual cards and their drivers, though your final mileage may vary.

With the range of drivers available on the Internet for various graphics cards, especially tweaked drivers and third-party performance drivers (often for specific applications), getting the best performance from your equipment can be a matter of doing some online research. Graphics cards change and evolve amazingly quickly, making it difficult to position these roundups to give a representation of the market. It seems that some significant new event or introduction is always just around the corner. For example, as we go to press with this article, Nvidia is about to introduce its first graphics card based on the NV30GL chip. None were available for testing and evaluation in time for this review.

Based on our experience, whatever is next for graphics cards will surely be faster, have more RAM, and offer more features. As always, Cadalyst Labs will evaluate them here.

Reviewer’s Report Card   Performance Price Warranty Extra Credit Star Rating Importance Factor (35) (33) (31)     3Dlabs Wildcat4 7110 B+ C A — 4 stars 3Dlabs Wildcat VP970 B- A- A — 4. 5 stars ATI Fire GL Z1 B A+ A — 4.5 stars PNY Quadro4 980XGL A A A Yes* *Additional application drivers.

— Spline

Home » Useful information » Articles, reviews and descriptions » Interface descriptions

Introduction Structural version Principles operation and the main advantages of AGP compared to PCI Physical characteristics AGP versus PCI Comparison of AGP and PCI graphics cards Development AGP AGP 2. 0 AGP Pro AGP 8X nine0015 What farther? Introduction The AGP standard was developed by Intel to without changing the established standard for the PCI bus, speed up data input / output to the video card and, in addition, increase computer performance when processing three-dimensional images without installing expensive dual-processor video cards with large volumes of both video memory and memory for textures, z-buffer, etc.. This the standard has been supported by a large number of firms that are members of the AGP Implementors Forum, an organization established on a voluntary basis to implement this standard. The starting version of the standard is AGP 1.0. nine0009 Structural version : Separate 3. 3V powered slot that looks like a PCI slot, but doesn’t really work with it incompatible. A normal graphics card cannot be installed in this slot and vice versa. How it works and the main advantages of AGP compared to PCI : 1. Physical characteristics AGP versus PCI nine0015 Data transfer rate up to 532 Mb/s, which is due to bus frequency AGP 66 MHz, the ability to cancel the bus multiplexing mechanism address and data (on PCI for the same physical lines are given an address first, and then data). The PCI bus has 33 MHz clock and 32 bits of data, so 33,000,000 x 4 bytes = 132 Mb/s. AGP has a bus frequency of 66 MHz and the same bit depth in standard mode (more precisely, «1x» mode) can skip 66,000,000 x 4 bytes = 266MB/s. In x1 mode, the clock signal itself is used as a strobe. frequencies. To increase the throughput of the AGP bus, the standard includes the ability to transmit data using additional special signals, used as strobes, instead of the CLK signal in normal mode (these are the «2x» and «4x»). In 2x mode, the bandwidth becomes 66,000,000 x 2 x 4 bytes = 532 Mbytes/s. In the «4x» mode (introduced in specification 2.0), the throughput the capacity increases accordingly, up to 1064 Mbytes/s. nine0009 In addition to the «classic» addressing method as on PCI — address is set first, then data appears on the same buses, in AGP sideband addressing mode, also called «sideband addressing», in which the address and data buses are separated and so they can be sent at the same time. Exchange rate in SBA mode significantly increases, since the time costs for transmission are eliminated bus addresses. In this case, special, absent in PCI, signals are used. SBA addresses ( S ide B and A dressing). The table below shows the test results 3DMark99 for ASUS graphics card V3400 TNT 16 MB SGRAM with and without SBA enabled. nine0129 16 MB textures 2 MB textures 4 MB textures 8 MB textures 32 MB textures SBA enabled fps 214. 1 159.2 103.5 58.9 28.3 SBA off, fps 202.4 144.6 95.3 49.3 22.9 The table shows that with an increase in the size of textures and, respectively, an increase in the amount of data transferred to the video card data difference in performance (in percentage terms) sharply increases. nine0009 AGP versus PCI. On PCI by exposed address after a delay, data appears. On AGP first a packet of addresses is set, to which a response with a data packet follows. (c) Intel Corporation Basic 3D image processing is performed in the computer’s main memory by both the central processing unit and the processor video cards. The mechanism for accessing the processor of a video card to memory is called nine0016 DI rect M emory E xecute (DIME — direct execution in memory). It should be mentioned that not all AGP video cards support this mechanism. Some cards still have only mechanism similar to bus master on PCI bus, i.e. uses DMA channels to quickly transfer data to video card. This principle should not be confused with UMA (Unified Memory Architecture), which is used in low-cost video cards, located, as a rule, on motherboard (for example, SP97-V firms ASUS TeK Computers). Main differences: An area of ​​the computer’s main memory that can be used by an AGP card (also called «AGP memory»), does not replace memory screen. In UMA, main memory is used as screen memory, and AGP memory is used only complements it. The memory bandwidth in the UMA video card is less than for PCI bus. Figure 2 . Block diagram of interaction AGP card and computer. For texture calculations, only the central processor and video card processor. The CPU writes data to the video card directly to the area of ​​\u200b\u200bconventional memory, which is also accessed by video processor. Only read/write operations in memory 9 are in progress0009 No arbitration on the bus (the AGP port is always the same) and time costs 2. Comparison of AGP and PCI video cards : In reality, an AGP card will be ahead of a regular card (if compare cards with similarly powerful video processors) only on tasks processing 3D images that require a large amount of memory for textures (more than 8 Mb). At the same time, it must be understood that the actual memory of the computer must be at least 32 MB, otherwise the AGP card will have nowhere to place the textures. nine0009 Various card tests by Tom Pabst in autumn 1997, showed that on conventional tests there is practically no difference between the appeared AGP cards and the Matrox Millenium II reference card are not currently available. On There is a difference in three-dimensional tests, but not very significant. Since then a lot of water flowed away and the situation has changed significantly, as discussed below. Windows 95 OSR2 version 2.5 already provides sufficient support for AGP and 3D test results for AGP are ahead of those for PCI, especially in scenes with large sets textures. From the Tom Pabst website you can see their results. Windows NT 4.0 does not support AGP and only in NT 5.0 (Windows 2000) will you be able to take advantage of AGP. nine0009 The AGP standard alone does not guarantee growth performance. Only in the case when the developer of the video card (more precisely, video card processor) uses all the possibilities of the bus, this gives an increase performance. For example, a Matrox video card Millenium II AGP does not support both DIME and «2x» mode, so it is almost impossible to find an application under which the AGP variant of this video card will somehow outperform the PCI option. nine0009 AGP video card can significantly outperform the same PCI video card only if either DMA and x2 or DIME and x2 are used. In modes without x2 win is almost non-existent. Check what mode the video card is in computer, you can use a small program pcilist , which can be copied from the EnTech Taiwan website. Development of AGP 1. AGP 2.0 In December 1997, Intel issued a preliminary version of the AGP 2.0 standard, and in May 1998 the final version. Main differences from the previous version: Transfer rate can be doubled by compared to 1.0 — this mode is called «4x» — and reach a value of 1064 MB/c. Address baud rate in «sideband addressing» mode can also be doubled further Added «quick write» mechanism F ast W rite ( FW ). Main idea — recording data/control commands directly to the AGP device, bypassing the intermediate storing data in main memory. To eliminate possible errors in the standard on bus introduced a new signal WBF# ( W rite B uffer F ull — the write buffer is full). If the signal is active, FW mode is not possible. nine0009 First graphics cards, supporting version 2.0 appeared at the end of April 1999. By appearance AGP video card connector can easily determine the presence of such support. View of the video card connector with AGP 1.0 View of the video card connector with AGP 2.0 additional slot for AGP 2.0. Because the corresponding connector on motherboard will have a plastic strip in under the second slot, the board with AGP 1.0 cannot be installed in such a slot, and vice versa — without any problems. 2. AGP Pro In July 1998, Intel released version 0. 9 of the AGP specification. Pro, which is structurally different from AGP 2.0. A brief summary of the differences in the following: nine0009 Changed the AGP connector — added pins on the edges of the existing connector for connecting additional power supply circuits 12V and 3.3V AGP 2.0 compatible from bottom to top only — boards with AGP 2.0 can be installed in an AGP Pro slot, but not vice versa. The AGP Pro is designed for ATX form factor systems only. Installing AGP Cards Pro in NLX system is not provided (the size of the board in AGP Pro is too large). nine0009 Since the AGP Pro card is allowed to consume up to 110 Wt (!!), the height of the elements on the board (taking into account possible cooling elements) can reach 55 mm, so two adjacent PCI slots must remain free. In addition, two adjacent PCI slots can be used by the AGP Pro board for their goals. From the point of view of circuitry, the new specification is nothing adds, in addition to special outputs informing the system about the consumption of the AGP board Pro. nine0009 It can be seen from the figure that the dimensions of the AGP Pro slot cause nostalgic memories of the late 80s when the controller board the display looked almost the same (although it was not more than one compartment in thickness). Of course, the specification for the AGP Pro stipulates the maximum dimensions and currents consumption, but after getting acquainted with it, a seditious thought comes to mind — and what does Intel now consider to be the main processor in a computer? AGP 8X November 2000 Intel has released a draft version of the next AGP bus variant — 8X. The main idea is to increase the bandwidth to 8×4 = 32 bytes per cycle system bus. This means that the data transfer rate on the bus will increase to 2 gigabytes per second. How this is done can be seen from the figure below: In addition, in the project of a new tire variant contains several fundamental changes that expand capabilities of the AGP interface. You can list some of them: Reducing the voltage level of the signals on the bus Calibration cycles Dynamic bus inversion Support for isochronous data transfer Support for multiple AGP 8X ports (previously only one was possible) port) New configuration registers for 8X bus What’s next? Actually reduced the cost of SDRAM synchronous memory (and its varieties — synchronous SGRAM graphics memory), as well as even faster DDR DRAM prospects for AGP are not as rosy as it once seemed. Primary goal, pursued by Intel — the creation of a cheap equivalent to professional video cards with large amounts of local memory loses its meaning at low memory cost. The bandwidth value in AGP 2.0 standard boards (1 GB/s), which is still only should appear, 2 times less than today’s actual bandwidth for local SGRAM memory up to 2 GB/s. Therefore, increasing the clock frequency and the width of the PCI bus can negate all the advantages AGP. nine0468 Another thing is that AGP has become in fact the only interface for video cards and only this fact makes it is practically impossible to switch back to PCI, so AGP will continue to develop henceforth, but as if in parallel with the development of PCI. Accelerated Graphics Port (AGP) — Accelerated graphics port

AGP (Accelerated Graphic Port) | AMD news

AGP (Accelerated Graphic Port) is a specialized 32-bit system bus (slot) for video cards. Tire introduced at 1996 year. Developed by Intel to increase the bandwidth between the GPU of a video card and the CPU. Replaced PCI.

History

The need to create AGP appeared in the early 1990s, when there was a need for a significant increase in the video memory capacity of graphic cards. This led to an increase in the cost of this type of component, and the only alternative to this development was to increase the bandwidth of the motherboard bus. At that time, the current PCI 1.0 slot could only offer a frequency of 33 MHz, a bandwidth of up to 133 megabytes and one data block per clock, which was clearly not enough. Even its PCI 2.1 receiver capable of operating at a frequency of 66 MHz and providing data exchange up to 266 Mb / s could not transmit more than one block of information, which could not meet the needs of the industry. nine0468 The first revision of the Accelerated graphic port was similar in capabilities to PCI 2.1 (both ports had a frequency of 66 MHz and a data exchange of 266 Mb / s), but almost immediately with it, Intel introduced AGP 2x (in fact, it was an improved version of AGP 1. 0), which, while similar to 1x predecessor characteristics were able to exchange two blocks of information at once per clock. This allowed the graphics processor of the video card not to alternately receive and send data to/from the bus, but to perform both processes simultaneously. As a result, the AGP 2x bus was able to transfer up to 533Mb/s, which made it possible to achieve the desired goal — to maintain performance without a significant increase in video memory. nine0468 In 1998, the second generation of AGP (AGP 2.0) appeared, which, while maintaining a frequency of 66 MHz, was capable of transmitting four blocks of information at once per clock. This technology began to be referred to by manufacturers of motherboards and video cards as AGP 4x, and the volume of data exchange increased to 1066 Mb / s. At the same time, the bus supply voltage dropped from the traditional 3.3V used in PCI and AGP 1.0 to 1.5V.
A year later, Intel introduced the most advanced revision of the technology — AGP 3. 0, which sends/receives up to eight blocks per cycle, is capable of transmitting up to 2133Mb/s (2Gb/s) and gets the «popular» name AGP 8x. The connector supply voltage drops to 0.8V, and the bus itself receives a unique (according to the developers) at that time opportunity for cross-streaming information exchange between two identical devices. In fact, it was nothing more than the SLI technology introduced by 3Dfx in 1998, at the time of the release of 3D accelerators Voodoo 2 PCI. Similar technologies, but under the AGP connector, the main manufacturers of ATI and Nvidia video cards called ATI CrossFireX and Nvidia SLI, respectively. CrossFireX/SLI involves the use of a motherboard with two AGP 8x (later PCI-Express) connectors and two identical video cards installed in them connected by a special cable. In this scheme, both GPUs work in parallel and draw each line of the image in turn, which significantly improves the overall performance of 3D rendering. nine0468 Intel has prepared for the release of another generation of the AGP connector — 3. 5, which turned out to be known thanks to Microsoft technical documentation. Such a connector is referred to under the working names AGP 3.5 and AGP Pro Universal, but was not implemented in end user products.

PCI and AGP bus characteristics

nine0005 66MHz

Bus type	Voltage	Frequency	Blocks per clock	Designation	Speed ​​Mbps
PCI	3.3/5V	33MHz	1x	PCI	133Mb/s
PCI 2.1	3.3/5V	33/66MHz	1x	PCI	133Mb/s 266Mb/s
AGP 1.0	3.3V	66MHz	1x	AGP_1x	266Mb/s
AGP 1.0	3.3V	2x	AGP_2x	533Mb/s
AGP 2.0	1.5V	66MHz	4x	AGP_4x	1066Mb/s
AGP 2. 0	1.5V	66MHz	4x	AGP_4x	1066Mb/s
AGP 3.0	0.8V	66MHz	8x	AGP_8x	2133Mb/s
AGP 3.5	0.8V	66MHz	8x	AGP_8x	2133Mb/s

Compatibility and support by manufacturers

Video cards for different generations of AGP have partial forward and partial backward compatibility, which mainly depended on the type of bus implementation by certain motherboard manufacturers. In order to clearly separate AGP revisions, Intel developed incompatibility keys that would not allow inserting cards with 1.5V power into 3.3V connectors and vice versa. nine0009

In fact, in the absolute majority of cases, the graphics device was able to work with any generation of the AGP bus and took the voltage it needed. To increase compatibility, video card manufacturers even created a special type of pad that had voids for both the 1. 5V slot key and the 3.3V slot key.

Video card with universal, compatible with AGP 2x and 4x contact pad.

With the advent of AGP 2.0, almost all motherboard manufacturers switched to the AGP Universal standard. Its feature was the complete absence of incompatibility keys, and the ability to install almost any graphics card. nine0009 AGP Universal — late revision of Accelerated Graphic Port with missing incompatibility keys.

Support for AGP versions by manufacturers and the decline of technology

Two seemingly different issues are closely related. The thing is that AGP is nothing more than a significantly modified PCI bus, there is even an accepted term «add-on» created by Intel. To implement Accelerated graphic port support, motherboard manufacturers had a choice:
a) Intel’s recommended method: integrate AGP protocols into the chipset; nine0468 b) significantly modify the classic PCI protocol to achieve the necessary behavioral characteristics of a separate graphics port (that is, the process of emulating PCI on AGP actually took place).