Talk:Intel GMA
From Wikipedia, the free encyclopedia
I know its boring, but Intel's efforts at graphics from the ill fated i740, to the 810/15, should be here also. Timharwoodx 10:39, 13 May 2006 (UTC)
The section saying that Intel open sourced it's X11 graphics driver is a bit misleading. More accurately, they added support for the new 965 chips to the existing open-source 'i810' driver in the X.org source tree. Intel graphics products have had open source support for years now, but what is notable about the 965 support is that the release of this chip was a point at which Intel could have easily gone binary-only (as ATI did some years ago). They did have a binary-only driver called the IEGD (Intel Embedded Graphics Driver), but as the name suggests, this driver was not targeted at desktop and laptop end-users. Among other things, it lacked full 3D acceleration support. It was, however, capable of mode setting without the video BIOS (as the 'i810' driver was not) and driving third-party TV encoders. Now, this functionality is now available in open source and being worked on by Intel developers in a branch of the driver's freedesktop.org git tree. I am not aware of the fate of the proprietary driver, but it as far as I know it has no secrets or extra features anymore, or those that it does will soon be in the OSS driver. Simba B 00:03, 13 August 2006 (UTC)
Contents |
[edit] Many questions
Why does the chart list the X3000 as being able to do OpenGL 2.0? According to the documentation on Intel's site, it does only hardware-accelerated 1.5. Is the 2.0 stuff done in software? If so, we should note this on the chart. --Peter 19:16, 21 January 2007 (UTC)
Why did every section state the chip was not based on the PowerVR. This implies that some other chip was? Otherwise, why even mention it at all?
Why does the 950 section claim that it supports Shader Model 3, yet the X3000 claims that will be the first to do so?
- The 950 only supports SM3.0 in software; you'd never get usable performance figures from it. It'd only really be useful as a tool for developers to test their SM3.0 code on and ensuring that it was bug free. --DaveJB 13:57, 1 September 2006 (UTC)
Why does the 950 section claim that SM3 doubles the 3DMark performance?
Why does the 950 have a higher peak pixel fill rate if the clocks and pipes are the same as the 900?
- I take this one back, I mis-read the numbers. I have updated the page.
Maury 21:07, 25 August 2006 (UTC)
- Dave, do you know if the 950's shader support also worked on the 900's then? It's certainly possible Intel wouldn't back port, but given that it was all running on the host CPU, it seems equally possible they could have. Maury 13:25, 3 September 2006 (UTC)
-
- It's certainly possible, the 900 and 950 have similar enough cores. I think Intel's concern was that you'd need a dual-core CPU to get useful performance figures from software SM3.0 code (though it would still be nowhere near fast enough for any game), and since there's no way to modify 915G to support a dual-core processor, they just didn't see the point in allowing it to support SM3.0. --DaveJB 13:22, 8 September 2006 (UTC)
-
-
- Ahhh. Got it, thanks! Maury 15:51, 8 September 2006 (UTC)
-
[edit] Hardware details section
Could whoever wrote that section originally please indicate what chips are being talked about? Thank you. Simba B 21:18, 1 October 2006 (UTC)
-> "The X3000 architecture contrasts strongly with common external GPUs, like those from ATI or nVidia. In these systems the different functions are handled by different types of pipelines; one type to handle T&L or vertex shaders, another for pixel shaders, and ones for texturing"
This statement is no longer accurate, as the new GPUs use dynamic pipelines, capable of either shader or vertex usage as needed. Wozdog 07:47, 7 January 2007 (UTC)
[edit] Inquirer article
That thing (both the link and maybe the reference) needs to go--it is biased (and wrong as the article points out)--and IIRC it was based on a early version of the chip and drivers which probably has no bearing on what is availble on the market now. Objections? If not I'll remove it. Simba B 21:29, 1 October 2006 (UTC)
- No, go for it. What is annoying is that Intel makes what appear to be two very different GMA's and refers to them both as the 9000. This is going to lead to continued confusion. Maury 12:14, 2 October 2006 (UTC)
- This particular Inquirer article fails the smell test, and has problems to boot. There's just no way the silicon of the X3000 would implement the same funcationality as the 950 in a less performant fashion, it's based on the same design. The Inquirer's MO is to shoot from the hip, and sometimes they just plain miss. Be circumspect in using them as an external source, they're very much in the press world of calling things early, right or wrong, which is fine. But they _rarely_ come back and correct a link that turns out to be dead wrong. This particular one, remove. JoshuaRodman 12:27, 4 January 2007 (UTC)
[edit] Tense
This article is a messy mixture of present and past tense. I don't have the patience to fix it, but everything should be one or the other tense. For instance: "was basic even by contemporary standards, and lacks support" should be either "was basic even by contemporary standards, and lacked support" or "is basic even by contemporary standards, and lacks support". Foobazยทo< 01:16, 24 December 2006 (UTC)
-> We've been trying to make the wiki accurate, but our (Intel) posts keep getting erased. We are only trying to make the info accurate. We'll update it immediately when it changes if you'll allow us to. But don't compare graphics integrated into the memory controller hub to that imbedded on the motherboard. It will only make you look like you don't know much about the technical aspects of integrating computer graphics into a northbridge, while maintaining a good asp on a midrange graphics product.
[edit] ATI Xpress 200
Under the GMA950 section, why all the comparisons to the ATI XPress 200?
[edit] Tile based rendering vs. deferred rendering/texturing
The article claims that "the GMA series uses tile based rendering, which aims to ensure that only pixels that will end up on-screen will flow through the rendering pipeline." However, I think the term hidden surface removal is more appropriate to describe the technique of removing non-visible parts of the scenery. Tile-based rendering just means that the screen is divided into tiles.
PowerVR also uses the term deferred rendering to describe their HSR-before-texturing. Chithanh 00:18, 31 January 2007 (UTC)
