Talk:Accelerated Graphics Port

Latest comment: 6 years ago by Immigrant laborer in topic "is a high-speed" in the lede

Backwards compatibility

edit

AGP is backward compatible from 8x to 2x, but the bus must exactly look like on picture http://en.wikipedia.org/wiki/Accelerated_Graphics_Port#Compatibility, so bus on graphic card must have two grips (3,3V and 1,5V). AGP 8x and AGP 4x is more compatible than AGP 4x and AGP 2x (you can put AGP 4x card to AGP 8x bus and vice versa, but not AGP 4x or 8x graphic to AGP 2x bus and vice versa; result: wrong keyed)

Which?

edit

a ge-force FX 5200 or radeon 9200SE? which is better

Picture

edit

A picture like that in the PCI article qould be good


edit

http://www.gcsextreme.com/agpfaq.htm gives me a 404 error.

To do

edit

The article does not mention AGP 2.0 nor AGP 3.0.

The article does not talk about compatibility (e.g. can I put an AGP 4x card into an AGP 2x slot?).

The article does not talk about visual identification of AGP cards from the features of the connector.

Advanced features

edit
The article is also missing the advanced features of 2.0 and above (FastWrites, etc).
However, I doubt anyone else here is going to expand id. Ah well, time to take some photos of my motherboard and graphics card... the latter can be dumped into Radeon as well...

Kiand 20:58, 9 Nov 2004 (UTC)

edit

I found a very comprehensive link on AGP compatibility, with clear diagrams: http://www.playtool.com/pages/agpcompat/agp.html


Incorrect information about clock frequency in AGP 4X, and 8X modes

edit

In the section titled `Newer Versions of AGP' it is stated that AGP 4X operating at 133MHz double pumped and 8X at 266MHz double pumped. This is incorrect. All AGP modes run at the same frequency, i.e., 66MHz. In 4X 4 32-bit data words are transferred per clock cycle. In 8X 8 32-bit data words are transferred per clock cycle. Note, it is essentially only the AD and CBE busses that are doing this. The PCI derived control signals are still running at the base 66MHz frequency.

The above can be verified by checking the AGP spec which should be available on Intel's web site.

Questions/requests.

edit

Can someone descripe AGp 1.0, 2.0 3.0?

Also explain AGP multiplier (in BIOS sense).

Any photos of the different AGP connectors would be good.


Rich Farmbrough 16:23, 5 August 2005 (UTC)Reply

Introduction and early Support

edit

The article does not mention that AGP followed the footsteps of Vesa Local Bus (VLB) common on 486 era systems, or that it was first introduced on Socket 7 (second gen. Pentium) and Slot 1 motherboards (early Pentium II systems, IIRC), or that OS support was first introduced in Windows 95 and NT 4.0 (I'm guessing here). MaxEnt 01:44, 9 April 2006 (UTC)Reply

I think I'm going to add this section myself, since I found good refs. MaxEnt 01:44, 9 April 2006 (UTC)Reply

Done. But I'm too pooped to add references. Here are the windows I had open at the end of my search, in addition to what I listed above. Embedded these into a comment within the main page text as well.

MaxEnt 03:23, 9 April 2006 (UTC)Reply

AGPGART excised

edit

It was kind of disgusting yet pertinent, but then to my shock I discovered an Agpgart page so I moved my Linux-centric footnotes there. MaxEnt 04:30, 9 April 2006 (UTC)Reply

Early AGP cards

edit

Many AGP 1x cards were merely an existing PCI design reworked to plug into an AGP slot and do not use AGP texturing or any of the other advanced AGP features.

Running NT on a Matrox Mystique I didn't pay attention to that at the time. One angle I would like to see further developed in this article is a sense of technologies introduced into the PC industry and their life story.
  • How long was the PC universe pushing the new technology with no immediate benefit to the end consumer?
  • Which advanced features ended up making a difference, at what point, under what circumstance?
Another issue which was still big at the time was BIOS level PCI bus enumeration and Plug and Play/Pray factor which was almost beginning to work around the time that AGP came along.
  • Did AGP contribute anything toward cleaning this mess up?
  • Is there any distinction in the IO port, DMA, or interrupt resources required of an AGP card as opposed to a PCI card?
The material I added was just enough notes and keywords that someone motivated could fill in more of the blanks themselves. AGP is well on its way to becoming a historical footnote. My feeling is that an encyclopaedic treatment should read more like a good history article and less like a dictionary of AGP ratifications (which is excellent material in its own right that I would never remove).
The most interesting story to me is how the industry shifted demand to a new design that didn't benefit the end user until the *second* PC they purchased featuring the new technology. When the AGP enhanced video cards finally showed up, did they actually work to their full potential in the AGP systems sold in the first six months after introduction, or did you have to replace your system board at the same time with one designed and tested to work to full potential with the generation of video cards that finally supported the new feature set? From a school of business perspective, the most interesting thing is how often the PC industry managed to get away with selling some new technology to the end user that promised to enable the future that ended up proving incompatible with the promised future when it finally arrived. See also: Bus Mastering DMA, USB, MMX, HDTV, HDMI, RDRAM, DDR, DDR2 all of which sucked for the early adopters. It's not like we (the gullible public) have learned much. MaxEnt 13:29, 20 April 2006 (UTC)Reply

Cripple Bits

edit

I'm getting a little off topic here, but take a look at this: http://www.pcguide.com/ref/mbsys/cache/char_Cacheability.htm

  • The amount of RAM that the system can cache is very important if you are going to be using a lot of system memory. Almost all modern fifth generation systems can cache 64 MB of system memory. However, many systems, even newer ones, cannot cache more than 64 MB of memory. Intel's popular 430FX ("Triton I"), 430VX (one of the "Triton II"s, also called "Triton III") and 430TX chipsets, do not cache more than 64 MB of main memory.
  • The keys to how much memory your system can cache are first, the design of the chipset, and second, the width of the tag RAM.
  • The only mainstream Pentium chipset to support caching over 64 MB is the 430HX "Triton II" chipset by Intel. In actual fact, caching over 64 MB on this chipset is considered "optional"; the motherboard manufacturer has to make sure to use an 11-bit tag RAM instead of the default 8-bit. The extra 3 bits increase cacheability from 64 MB to 512 MB
  • Pentium Pro PCs use an integrated level 2 cache that contains the tag RAM within it, so none of this is really a concern for these machines. The Pentium Pro will cache up to 4 GB of main memory, basically anything you can throw at it. The Pentium II uses an SEC daughtercard. It has the same general architecture as the Pentium Pro, but due to a design limitation will "only" cache up to 512 MB.

Around the time that AGP was introduced, the industry was still busy shaving three bits off a tag RAM at a point in time where cheaper memory was by far the most cost effective upgrade for any PC with a graphical desktop. That was the reason I got into the PPro early: fewer cripple bits. People who knew that the early AGP designs didn't help you much ended up buying AGP systems regardless in order to gain other advantages from less crippled chipsets. Some of those Pentium era chipsets (I'm trying very hard not to type cripsets) already had Steve Yzerman's late career knee cartilage the day they were first introduced. MaxEnt 14:14, 20 April 2006 (UTC)Reply


Current capacity

edit

I would like someone to find the max current (amps) that can be delivered through the AGP interface. I see this information varies on web pages. A total wattage would also be benificial.

I'm not sure, but it is at least 19Amps, as there are AGP editions of PCI-e cards that use that much, such as X1950s.Alexander 06:57, 20 August 2007 (UTC)Reply

Correction to Multiple AGP ports

edit

Noticed someone just made a correction to the idea that a board could possibly support more than one AGP slot.

This was once debated on the linux-kernel list. The two most informative posts were these:

It does appear to be true that there are no known boards with more than one AGP port, but it wasn't fully decided whether it was possible to make such a chipset or not (some voices speculated that it is). MaxEnt 15:07, 1 May 2006 (UTC)Reply

After some searching I found the "Micron Samurai".
Article - [1]
Photo - [2]
This is possibly another dual-agp board:
http://www.watch.impress.co.jp/akiba/hotline/20041127/image/np8sa1.html
The Samurai board was dual-processor with a seperate northbridge for each. This gave it two PCI buses, memory banks, and AGP slots. There must have been some sort of link between the two northbridges to coordinate things. Anyway, it seems that at least one motherboard had dual AGP slots.
Imroy 15:29, 1 May 2006 (UTC)Reply
As reviewed at Tom's Hardward, this board did not sport both AGP ports:
The Samurai DDR chipset was a concept chipset from memory maker Micron to push DDR technology into the marketplace.
Micron could afford to do this, as their margins were exceptional:
I don't think this chipset was ever commercialized with both AGP slots available, and I remain doubtful than viable OS support materialized.
Micron had also purchased Rendition to drive their "embedded DRAM" technology into the graphics space. More games to sell DRAM. MaxEnt 16:39, 1 May 2006 (UTC)Reply
Another source of confusion about "dual AGP"
this was a single AGP board with dual onboard GPUs MaxEnt 17:13, 1 May 2006 (UTC)Reply

If you guys agree (Stonehead, Imroy) I'll add a section "Dual AGP" and dispel various sources of confusion about this term. I first contributed to this article because I was interested in capturing more of the history of the AGP lifecycle and deployment. Micron's largely unsuccessful efforts to use AGP to sell more DRAM is interesting to me from that perspective. MaxEnt 17:41, 1 May 2006 (UTC)Reply

I'm not an expert on the subject. However, the spec and the definition claim AGP to be a point-to-point channel. As such, it is not a bus (point-to-multipoint). Therefore no motherboard can theoretically support multiple AGP slots. At least, not with one and the same AGP port :) Any mention of motherboards with support for multiple AGP ports on websites has proven to be an error (and I've seen quite some), which was one of the reasons to make my change in the first place. However, it looks like Imroy is right that it's possible to have several independent AGP ports on one and the same board (I wonder whether that offers any improvements over PCI anyway.) However, I'll leave it up to him. Thanks for your help! Stonehead 21:07, 4 June 2006 (CEST)

Not SO Slowly

edit

"AGP is slowly being phased out in favour of PCI Express."

It's actually happening very quickly...

In new PCs, yes. But a Valve software survey of people running their games in March showed only 26% using PCI-Express systems: http://www.steampowered.com/status/survey.html Mark Grant 21:36, 24 July 2006 (UTC)Reply

Lists and sections

edit

I believe that the current lists look a lot better than lengthy text will look, and are also much more readable. —The preceding unsigned comment was added by 67.165.235.73 (talkcontribs).

The 'Hardware and Software Adoption' section should certainly be rewritten into a few paragraphs of text. The list at the end of 'Versions of AGP' could probably also be rewritten into a subsection. That list isn't even introduced - it has no context. Some of the sections could be rearranged too, 'Hardware and Software Adoption' could become 'History' and moved to the top. --Imroy 07:03, 17 November 2006 (UTC)Reply

Bandwidth Asymmetry?

edit

I've come across some information about the asymmetry of bandwidth over AGP. http://www.edn.com/article/CA6351293.html?industryid=2283

This is one aspect of AGP not mentioned here. —The preceding unsigned comment was added by 66.69.192.103 (talkcontribs) 16:41, 3 December 2006 (UTC).Reply

nitpicky question about speed and bandwidth

edit

I warned you, this is very nitpicky. This article states that AGP1x is "A 32-bit channel operating at 66 MHz resulting in a maximum data rate of 266 megabytes per second (MB/s)". However, 4 bytes * 66 MHz does not equal 266 MB/s, but rather 264 MB/s. Is the maximum bandwidth actually 264 MB/s, or is the frequency rounded in the article, or is there something I'm missing? (The official spec is very hard to read.)

66 2/3 MHz. 66.666666667 MHz. So, technically, it's 267 MB/s. But, for whatever reason (maybe the typical clock fluctuation) it is stated as 266 MHz. --Swaaye 04:15, 8 February 2007 (UTC)Reply

AGP - not a bus

edit

Well, since AGP = Accelerated Graphics Port, it is not a Bus like PCI so the phrase "AGP bus" all over the article is basically wrong. Why does it matter? Because acceleration was one of the reasons for moving from bus (shared resource among many devices) to port (dedicated point-to-point connection between graphic card and the CPU). Cheers! 212.25.124.154 14:26, 20 February 2007 (UTC)Yoad Snapir.Reply

True, but it's regularly referred to as the 'AGP Bus', so not referring it to that way is probably more confusing :). Mark Grant 02:36, 4 March 2007 (UTC)Reply
I've added a clarification on this matter to the article (near the top). The fact that it is not a bus and only supports a single device or slot (and furthermore only one type of device) should be mentioned as a disadvantage (over PCI) in the article.--ChrisJMoor 19:07, 9 June 2007 (UTC)Reply
Yes and SLI never existed? Like many Buses, there are protocol standards and interface standards. Compact PCI, Cardbus, AGP, etc. use PCI protocols but differing physical attachment. Although bus buffer specs for AGP didn't indicate multiple devices should be supported, the PCI control signals are there. Like PCI cards each AGP card has Request/Grant signals that connect to the PCI subsystem, allowing multiple AGP cards. AGP cards with AGP to PCI hubs have been sold with one or more PCI devices and even PCI to PCI hubs internal to the AGP chipset. (Similarly Cardbus cards with internal Cardbus to PCI hubs to support multiple PCI chips e.g. modem and ethernet.) Look at Intel Northbridge chipset specs ( See reference http://download.intel.com/design/chipsets/datashts/29074602.pdf ), although physically different i845G allows PCI64-66MHz or AGP 4x using the same signal pins (due to bus loading only one device per i845G so PCI64-66 is not a bus?).
The Wikipedia definition of "Bus" allows it to be non-contiguous and to be point-to-point. You are stuck on ISA bus: multiple devices attached to the same copper wire. PCI bus isn't, several signals are private to the controller. And there are multiple instances going through hubs, not sharing the same copper trace. PCIe "bus" has none of the data lines shared. USB is point-to-point, sharing data through hubs. Is Ethernet a bus? It used to be. Is SCSI a bus? It used to be. Now Serial Attached SCSI and SCSI over Ethernet and iSCSI. Shjacks45 (talk) 05:21, 14 February 2011 (UTC)Reply
Although AGP may share lines with the PCI bus, there's no simple way to connect multiple devices. There's no defined way of addressing these devices are arbitrate busmastership among them.
Ethernet is only a bus for 10BASE2 and 10BASE5. From 10BASE-T onwards it's point-to-point eletrically, the bus is only logical through repeaters or switches. Zac67 (talk) 11:41, 23 December 2012 (UTC)Reply
edit

hi,

thought this might help develop this wiki page - I am no expert in video graphic cards - but I thought this might help - it explains the difference in AGP version 2.0 and 3.0

In layman's terms AGP 2.0 is AGP 2x and AGP 4x. While AGP 3.0 is an AGP 8x or 4x.

[3] —Preceding unsigned comment added by Krisketapetatum (talkcontribs) 03:53, 12 October 2007 (UTC)Reply

About the AGP & AGP Pro Keying Diagram

edit

The image http://en.wikipedia.org/wiki/Image:AGP_%26_AGP_Pro_Keying.svg should be updated with the slots aligned in such a way showing that AGP Pro is a slot longer than AGP in both direction, not just in the right direction.

It is currently showing all slots aligned left. -- Gafex (talk) 08:31, 16 July 2008 (UTC)Reply

I suggest the AGP pro slots to be moved in left and alligned right toward above AGP slots. In this way, we can more easily compare 3.3V key and 1.5V key and I'll like the locations of this keys to be alligned in same vertical lines like the slot. If the drawn video card is moved vertically down, it'll no fit in AGP pro slots, because they're misplaced in right. I'll also like the captions to be moved in right of the slots. Here is sample:
        ®®®®®® AGP
®®®®®®®®®®®®®® AGP pro
I think that after is shown videocard and above slots are alligned in correct way, the rest slots have to be alligned too or someone to post comment here and to explain that I'm wrongly thinking. I boldly require answer, because above coment from another wikipedia user is not answered so long time and this diagram is confusing and misleading for me in current way.

This has bothered me as well - changed! Zac67 (talk) 11:31, 23 December 2012 (UTC)Reply

AGP still going with HD 4670

edit

The newer ATI Radeon HD 4670 will be available as AGP, apparently, http://arstechnica.com/news.ars/post/20080910-ati-launches-new-hd-4000-series-at-sub-100-prices.html 150.237.47.14 (talk) 14:47, 9 October 2008 (UTC)Reply

it is with a pci-express bridge chip though Markthemac (talk) 05:25, 16 May 2009 (UTC)Reply

http://www.techfresh.net/his-agp-based-radeon-hd-4670-iceq/ && http://www.techpowerup.com/95987/PowerColor_launches_HD_4670_AGP_Passive_HD_4770_and_more_ATI_cards.html 88.104.2.141 (talk) 13:25, 1 August 2009 (UTC)Reply

64 AGP ports

edit

Is this a misprint: the HP AlphaServer GS1280 has up to 64 AGP slots? Why would any computer need 64 video outputs? Danceswithzerglings (talk)

No, it's not a misprint. It's right there on the cited datasheet. As to what they'd be used for, I can only imagine two scenarios:
  1. A multi-user setup where everyone needs accelerated 3D graphics. The datasheet does talk about partitioning the system, so it could be a kind of thin-client setup. An awfully expensive one though.
  2. A large visualisation setup with a huge tiled display. Each AGP graphics card drives a screen or projector. With up to 64 Alpha's and 512GB of RAM, this system could have some serious grunt and work through really large datasets e.g seismic survey data for oil exploration.
--Imroy (talk) 08:55, 13 July 2009 (UTC)Reply

Laptop AGP

edit

Some laptops come with an "AGP" port (eg. Dell Inspiron 8100) but this is obviously incompatible with desktop AGP. Does this special port have a specific name? How dies it differ electrically from the desktop version? 2fort5r (talk) 19:27, 12 July 2009 (UTC)Reply

Laptops can use AGP in the same way they use PCI and ISA - to connect the various chips together. It's the same with all the "on-board" functionality of modern desktop motherboards e.g IDE/SATA, audio, network, USB, etc. They connect directly to the buses/interconnects, without the edge connectors. --Imroy (talk) 08:21, 13 July 2009 (UTC)Reply

No, the Inspiron 8100 laptop has a removable graphics card, which uses a completely different port to desktop AGP. So I assume "AGP" refers to a set of electrical protocols rather than to the physical connector? 2fort5r (talk) 18:23, 13 July 2009 (UTC)Reply

Ah, ok. I wasn't familiar with this laptop. It could be specific to Dell - they could have chosen a suitable connector, devised a pin-out, and had the various manufacturers make parts. It doesn't have to be part of a standard. Large companies are known for creating their own non-standard connectors. Sorry I can't be of more help. --Imroy (talk) 19:26, 13 July 2009 (UTC)Reply

Throwing around terms when we don't know what they mean

edit

"point-to-point"? nonsense. Load a non-plug and play OS like DOS (or FreeDOS or DRDOS). Load a debugger. Motherboard BIOS is in control. "Peek" the PCI I/O memory above 0xC000, you will see the AGP card off a PCI Bridge off the main PCI bus.

Original AGP cards were 5 volt powered, hence the 3.3 volt "key" towards the front of the card. original TTL logic levels were .4-2.8 V out, .8 -2.0 V in with logic threshold 1.4 volt. Original 5V CMOS was .5-4.5 out with 2.5 v threshold. Updated 5V TTL (S, LS) and CMOS (HCT) logic levels were .4-3.5 out. 3.3V CMOS input thresholds were 1.5 V. Most newer chipsets are powered by 3.3 V or (even 1.8 V to reduce power usage) and would be tolerant of 5V (i/o) powered chips signal levels (original Intel i440LX AGP chipset).

Should be noted that some motherboard came with a tape blocking the AGP slot saying '.8 volt only' or "damage will result to the motherboard".

Multiple Video Card problem is a BIOS issue. DOS accessed video through processor interrupts using legacy i/o and memory addresses. Legacy I/O addresses: 3B0-3BB Mono display/printer adapter, 3BC-3BF LPT, C0-3CF EGA/VGA adapter, 3D0-3DF CGA/EGA/VGA adapter. Legacy Memory addresses: 0xA0000 - 0xAFFFF VGA, EGA; xB0000 - 0xBFFFF Mono, CGA, VGA, EGA; 0xC0000 - 0xC7FFF Video BIOS ROM. In protected mode, most newer software accesses the video card in PCI extended I/O and Memory Addresses (above physical RAM addresses). The Motherboard disables the duplicate legacy addresses on all secondary cards to avoid the consequences of short-circuited chips. Video cards that require the Video BIOS ROM to be active will not work as secondary cards. Several computers I have been using (especially EFI BIOS based systems) have some or all of these legacy addresses missing. I have a laptop that will boot to a USB to Video adapter (per usb video standard support in bios, although not Microsoft supported per http://www.microsoft.com/whdc ).

windows 95 and windows 98 vga.txt files list multiple card issues, could be referenced since multiple video adapters are in part OS specific. http://flammingfiles.tripod.com/files/biosdl.pdf http://www.playtool.com/pages/agpcompat/agp.html http://www.directron.com/15agpguide.html http://www.manualshark.org/manualshark/files/28/pdf_29582.pdf HP Vectra manual http://www.onlinecomputerbooks.com/blog/free-book-upgrading-and-repairing-pcs-16th-edition/ apparently free version

Shjacks45 (talk) 00:09, 12 April 2010 (UTC)Reply

"you will see the AGP card off a PCI Bridge off the main PCI bus." Just because you see something doesn't make it real. PCI was built on the assumption that there would be a root PCI bus with devices either directly connected to that bus or connected to it by bridges. More recent systems don't physically follow that topology but they still have to pretend to follow it for compatibility reasons.
"Updated 5V TTL (S, LS) and CMOS (HCT) logic levels were .4-3.5 out." you seem to be misleadingly representing maximum zero voltage and minimum logic one voltage as a range of voltages. When connecting stuff together you have to consider maximum logic 1 voltage which for outputs is usually specced to be the same as power supply voltage and for inputs may or may not be greater than the supply voltage. Plugwash (talk) 17:54, 23 December 2012 (UTC)Reply
"Original AGP cards were 5 volt powered, hence the 3.3 volt "key" towards the front of the card." AGP cards are powered by 3.3, 5, and 12 V. The key notches are for I/O voltage which has never been 5 V. Zac67 (talk) 19:12, 23 December 2012 (UTC)Reply

Power

edit

How much power does an AGP-slot provide?--78.48.205.180 (talk) 16:05, 10 March 2011 (UTC)Reply

edit

Hello fellow Wikipedians,

I have just added archive links to one external link on Accelerated Graphics Port. Please take a moment to review my edit. You may add {{cbignore}} after the link to keep me from modifying it, if I keep adding bad data, but formatting bugs should be reported instead. Alternatively, you can add {{nobots|deny=InternetArchiveBot}} to keep me off the page altogether, but should be used as a last resort. I made the following changes:

When you have finished reviewing my changes, please set the checked parameter below to true or failed to let others know (documentation at {{Sourcecheck}}).

 Y An editor has reviewed this edit and fixed any errors that were found.

  • If you have discovered URLs which were erroneously considered dead by the bot, you can report them with this tool.
  • If you found an error with any archives or the URLs themselves, you can fix them with this tool.

Cheers.—cyberbot IITalk to my owner:Online 13:50, 28 March 2016 (UTC)Reply

edit

Hello fellow Wikipedians,

I have just modified 2 external links on Accelerated Graphics Port. Please take a moment to review my edit. If you have any questions, or need the bot to ignore the links, or the page altogether, please visit this simple FaQ for additional information. I made the following changes:

When you have finished reviewing my changes, you may follow the instructions on the template below to fix any issues with the URLs.

This message was posted before February 2018. After February 2018, "External links modified" talk page sections are no longer generated or monitored by InternetArchiveBot. No special action is required regarding these talk page notices, other than regular verification using the archive tool instructions below. Editors have permission to delete these "External links modified" talk page sections if they want to de-clutter talk pages, but see the RfC before doing mass systematic removals. This message is updated dynamically through the template {{source check}} (last update: 5 June 2024).

  • If you have discovered URLs which were erroneously considered dead by the bot, you can report them with this tool.
  • If you found an error with any archives or the URLs themselves, you can fix them with this tool.

Cheers.—InternetArchiveBot (Report bug) 21:07, 25 June 2017 (UTC)Reply

"is a high-speed" in the lede

edit

It was high speed, now it is not. But it still exists, so the lede should remain in present tense. "is a formerly high speed" feels... judgy. Just take "high speed" out? - Immigrant laborer (talk) 17:26, 30 May 2018 (UTC)Reply