backtop


Print 51 comment(s) - last by Griswold.. on Jul 2 at 11:14 AM


Image quality comparison between GMA 950 (left) and GMA X3000 (right)

Intel G965 block diagram

A quick comparison of 945GM (left) and GM965 (right), specifications not final
DX9, Pixel Shader 3.0 and hardware T&L

DailyTech previously reported rough details of the integrated graphics core in the upcoming Intel G965 Express chipset—Intel Graphics Media Accelerator X3000. Early roadmaps showed the graphics core as being Microsoft Windows Vista Premium compatible with support for Aero Glass. It was also said the graphics core would have Clear Video Technology to improve video playback.

DailyTech recently had the opportunity to pick apart some of the specifications for Intel's GMA X3000 architecture, and compare it to some of the other architectures available today. Perhaps the largest improvement of GMA X3000 over GMA 950 is the move away from a fixed function pipeline in favor of a programmable pipeline.  NVIDIA and ATI abandoned fixed function pipelines in 2001.

Intel’s latest motherboard update has more detailed information on the Graphics Media Accelerator X3000. DirectX 9 features such as Pixel Shader 3 and Vertex Shader 3.0 are supported. This time around the Vertex Shader 3.0 units are hardware based instead of the software based shaders found in previous GMA900/950 and Extreme graphics cores. A hardware transform and lighting engine has also been integrated and a significant improvement over the previous software T&L engine. High dynamic range is also supported for great realism in gaming. Lastly the GMA X3000 graphics core will be clocked up to 667 MHz -- quite a bit higher than current budget ATI and NVIDIA offerings.

Video output capabilities of the GMA X3000 are limited to a native VGA output. HDMI, DVI, UDI, component, composite and S-Video can be added through the SVDO port or with an ADD2 expansion card like the previous GMA900/950 graphics cores. This more or less indicates HDCP compliance will be left up to the motherboard manufacturer or ADD2 card manufacturer. GMA X3000 will support resolutions up to 2048x1536 including 720p/1080i/1080p in 16:9, 4:3 and letterbox aspect ratios.

Intel Clear Video Technology will provide algorithms and features to improve video playback. Clear Video Technology will have plenty of graphics power to simultaneously playback one high definition and one standard definition video stream for picture-in-picture. Hardware acceleration for high definition MPEG2 and VC1 is supported. However, it doesn’t look like Intel will offer hardware acceleration for H.264 at this time. An advanced de-interlacing algorithm is also integrated for improved video quality of interlaced sources such as DVD’s and cable programming. The built-in advanced pixel adaptive de-interlacing algorithm supports standard and high definition video content up to 1080i lines of resolution.

Intel previously had a single integrated graphics core for all of its integrated core-logic but this time around there are at least two different graphics cores. While the consumer level G965 Express receives the GMA X3000 graphics core, Q965 Express chipsets receive the GMA 3000 graphics core. Differences between the GMA X3000 and GMA 3000 include the lack of Intel Clear Video Technology on the GMA 3000. This isn’t too surprising as the Q965 is part of Intel’s Stable Image Platform Program aimed towards business and corporate users. Aside from the lack of Clear Video Technology the GMA 3000 graphics core retains compatibility with Windows Vista Premium with Aero Glass interface like the GMA X3000. There’s no word if the GMA 3000 graphics core clocked as high as GMA X3000.

One thing that is important to remember about the GMA X3000 family is that it is a completely programmable pipeline architecture -- meaning Intel only needs to update the microcode to add support for features like SM 4.0.  This opens the door to a few possibilities with where Intel can go with the architecture.  For example, since the Santa Rosa notebook platform is based on G965, but will not launch until next year, Intel may take the opportunity to add better features to the core.

G965 Express is expected to launch the last week of July with Core 2 processors while Q965 Express is expected to launch the first week of September with Intel’s vPro business platform.



Comments     Threshold


This article is over a month old, voting and posting comments is disabled

RE: Unified Shaders
By NextGenGamer2005 on 7/1/2006 4:14:37 AM , Rating: 2
It would HAVE to have unified shaders if there is any truth to the DirectX 10 support through a BIOS update. DX10 adds a new shader, the geometry shader (it sits between the pixel and vertex shader units), so GMA X3000 would have to be using unified shaders that already contain support for the geometry shader if only a BIOS update is needed to make it DX10 compliant. My guess is that at launch, it will have 8 unified shaders but 6 will be "locked" as pixel shaders and 2 will be "locked" as vertex shaders. Then, closer to Vista launch, the BIOS update will "unlock" these to full unified shaders that can dynamically change between pixel, vertex, and geometry operations depending on the current graphics load. 4 TMUs and 4 ROPs sounds about right to me as well. And a 90-nm process would be required for this to operate passively (in fact, I wouldn't be too surprised if this was actually using a 65-nm process, just because the northbridge still has to contain the memory controller and PCI Express lanes as well; that's a lot to fit into one chip).


RE: Unified Shaders
By squeezee on 7/1/2006 7:09:09 AM , Rating: 2
Shaders don't have to be unified at the hardware level, that isn't a DX10 requirement. Having separate or semi-separate PS/GS/VS units in hardware doesn't affect the ability to bypass or disable the GS in software(Driver or Firmware).

Regardless of how they've done it hopefully it will make a big jump the abysmal performance of their current IGPs.


RE: Unified Shaders
By Zoolook on 7/1/2006 5:45:10 PM , Rating: 2
Have to agree, 667 MHz on an igp sounds like a 65-nm process, Intel already has 3 running and will add one more 65 and a trial 45 in the autumn. Sounds like they are finally turning up the heat on ATI and Nvidia on the integrated front, I've always wondered why they haven't pushed it harder before. Maybe they have been to complacent with their large market share.


RE: Unified Shaders
By ltcommanderdata on 7/1/2006 9:33:53 PM , Rating: 2
Chipset production is switching to 90nm now that those fabs are done producing processors. The 65nm fabs are needed to supply current processors, especially the Core 2's since Intel is deciding to do 3 launches (Woodcrest, Conroe, and Merom) in 3 months. Not to mention Celeron D production is switching to 65nm with the Cedar Mill based ones, Celeron M is on 65nm with the Yonah based parts, and the 65nm Tulsa for Xeon MP is launching in Q3. Needless to say, the 65nm fabs are overworked.

The 96x series of chipsets are the first to be on the 90nm process, which may be why they're a bit slow to market since now only does the logic need to be debugged and optimized but they have to accomodate a process transition too. Evem with the 90nm process, with 8 shaders on the IGP and all the other northbridge features, the die seems to be very large if the picture in the HKEPC article I originally linked is any indication. I just hope the G965 can be passively cooled. The Inquirer found that the P965, the graphics-less version, consumes 25% more power than the 945P.

http://www.theinquirer.net/default.aspx?article=32...

Hopefully, that's only an early engineering sample, possibly a 130nm trial, or else it doesn't bode well for the G965. It'll be even more worrisome for the mobile version of the G965 in Santa Rosa.

In regards to the implementation of unified shaders in a DirectX 9 environment, I don't think it's necessary to actually "lock" the pipelines as either PS or VS. ATI's upcoming R600 can determine whether the incoming data is vertex, pixel or geometry and assign it to an open pipeline. They plan to use this same mechanism in DirectX 9 games just that it only needs to decide between vertex and pixel data, bringing the flexibility of unified shaders to DirectX 9 games. Intel can just use the same type of mechanism to avoid having to "lock" their pipelines. Seeing the good relations between ATI and Intel, I wouldn't be surprised if they're actually working together on unified shader implementation.

http://www.theinquirer.net/default.aspx?article=32...


"Paying an extra $500 for a computer in this environment -- same piece of hardware -- paying $500 more to get a logo on it? I think that's a more challenging proposition for the average person than it used to be." -- Steve Ballmer














botimage
Copyright 2014 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki