Intel GMA X3000

[colo]

Hello there,

I've been looking for a free as in speech solution for state of the art 3D-graphics for quite some time now, and Intel's newest offering seems promising. I've however failed to dig up some info about whether Intel is going to provide drivers for the latest GMA graphics adapter series licensed under the GNU GPL, or if they are pursuing a proprietary approach for the X3000 series. Has anyone gained insights on that matter as of yet?

Please feel free to post anything else related to the new GMA in here, as well :)

[1c3d0g]

...you know, due to Intel licensing the GPU tech from Imagination Technologies, I doubt they'll be able to provide Open Source drivers. But you never know.

On another note, I really wish that the new GPU will perform well, 'cause frankly, I'm tired of dealing with Intel's horrible IGP performance. :mad:

[colo]

Yes, it's the 3rd party "Intellectual Property" (note the quotes) that#s worrying me, too. But maybe Intel can put enough weight behind its possible demands (and my certain wishes ;)).

If not, I just won't buy their products any more. Easy as that.

[Svartalf]

Yes, it's the 3rd party "Intellectual Property" (note the quotes) that#s worrying me, too. But maybe Intel can put enough weight behind its possible demands (and my certain wishes ;)).

If not, I just won't buy their products any more. Easy as that.

I guess that it's not a big issue any longer- they provided open source drivers and apparently the needed tech info for the X3000 GPU yesterday.

I'm hopeful that this GPU will do better in regards to things 3D and will prompt other vendors to follow suit.

[1c3d0g]

Colo: congrats, now your wish has come true. ;)

[colo]

Yeah, the board's already ordered ;)

A pity though there are some hooks left in for proprietary code to enter kernel-space... some b$ about CSS and/or Macrovision for MPEG2-on-DVD-accel (or so... did not care that much, as all GL-specific stuff is free as in speech :)) required this, from what I've read.

[Michael]

colo, how is the board running?

Just ordered an Intel GMA 3000 board for running a couple tests.

[niniendowarrior]

I hope to see some benchmarks soon.

Always wondered how non-nV,non-ATi Linux drivers fare.

[Michael]

I hope to see some benchmarks soon.

Always wondered how non-nV,non-ATi Linux drivers fare.

Hope to finish up some tests by next weekend or so. We will likely compare them to some ATI and NVIDIA components as well.

If there are any specific tests or areas that anyone has special interest in seeing numbers on, please report them asap.

[drag]

Hope to finish up some tests by next weekend or so. We will likely compare them to some ATI and NVIDIA components as well.

If there are any specific tests or areas that anyone has special interest in seeing numbers on, please report them asap.


Great! I've been looking forward to something like that for a long time.

I've given up on propriatory drivers in Linux and right now Intel is the only company providing open source drivers for their video cards. Right now I have a GMA 950 card and am ok happy with it. It's only realy able to handle games up to 'Return to Castle Wolfenstein'-style complexity. I'm hoping that the X3000 will provide better performance then the GMA 950.

One of the things you probably need to do is contact the folks at http://intellinuxgraphics.org/ and find out on the status of the current drivers and tweaks to get the best performance. I'd bet they'd love to work with you on this subject.

For instance with the GMA 950 with cvs drivers I found that running games with 'export INTEL_BATCH=1' nearly doubled 3d performance in some cases. Also specifing the amount of VideoRam and other tweaks shown noticable performance impact. Also check out driconf which is a gui python app for controlling DRI configuration options. You can do it for general configurations or you can specify options for specific games.

The major 'selling point' for the X3000 is the fact that it's ability to handle hardware accelerated features that were software rendered with older Intel graphics. Such as shaders, vector, antiscopic filtering etc etc.

I would be very interested on how much of these new hardware features are currently supported by CVS drivers and what they plan on supporting in the future. Also about what sort of things they have in that binary-only add-on for the drivers. (macrovision support?)

Also find out a bit about their programmable graphics core and ask about how well the on-the-fly pipelines changing from texture pipelines to vector pipelines, etc etc. I don't realy understand about how all that realy applies to this video card, maybe your curious about it too?

Maybe check out:
http://www.intel.com/products/chipsets/gma3000/gma3000.pdf#search=%22gma%20x3000%20whitepaper%22

Also exploring AIGLX would be interesting and current driver's support for it.

I have X800 video card with Free software DRI/DRM r300 drivers from Mesa's CVS stuff. With that I have AIGLX enabled and I am able to play games and use apps remotely using X over ssh, just for a lark. So a lot at the difference in accelerated indirect rendering vs accelerated direct rendering.

Used Blender over wireless. Played Xmoto over wireless. Played Quetoo (optimized Quake2) over wireless (10Mb was to slow, but 54Mb seemed 'ok' with only noticable mouse lag over the busier scenes). On gigabit ethernet I was able to play 1600x1200 resolution for 'RTCW' over ssh and it worked pretty well. Ranged from 10-60 fps at highest quality settings.

[Michael]

I had sent off a message to some Intel contacts, and am hoping for their reply in time for this article -- to comment both on the features implemented/not-yet-implemented and any performance tuning.

I have read the GMA 3000 white-paper when it was initially released as well as extensively investigating other areas of its architecture as well.

I'll likely comment on Compiz and AIGLX in the article as well.

I'm hoping to compare the GMA 3000 performance to the R300 open-source drivers with an ATI Radeon X300 and X800, and more if time permits.

[Michael]

Here is just what Keith Packard had to say:

As noted in the driver release announcement, this is just our first
release of the code and there remain significant possible performance
improvements that can be made. Some of them are low-level chip
optimziations, while others are ongoing efforts to improve the Mesa
project as well. Right now, there aren't any performance tuning
parameters available for the driver.

[1c3d0g]

Interesting. I'm really looking forward to a big performance boost compared to the GMA950, which couldn't even play UT2k4 with all the bells and whistles properly.

[niniendowarrior]

I'm sorry, but it's either the benchmarks on the Intel chipset hasn't been posted or I just missed it... I'm not sure where it is, Michael.

By the way, I also skimmed a bit on the whitepapers of the 965G/GMA x3000.

[Michael]

Not posted yet, there was some problems with the motherboard (Q965) and both Fedora Rawhide and Ubuntu Edgy Eft Beta 1. The benchmarks will be out soon.

[drag]

I posted this in the motherboard forum also, but with the G965-based motherboard and Windows XP it performed very badly in 3d performance.

http://www.reghardware.co.uk/2006/10/04/review_supermicro_pdsba_g965/

[Svartalf]

I posted this in the motherboard forum also, but with the G965-based motherboard and Windows XP it performed very badly in 3d performance.

http://www.reghardware.co.uk/2006/10/04/review_supermicro_pdsba_g965/

I'd take ANY Windows site review with a grain of salt (A lot of them,
actually- but not for the reasons you'd normally think...) because Intel
has said that they've not got the full support in the drivers that are
shipping for the motherboard and that a later driver would have full
support of everything. No use of the T&L. No use of the independant
rendering pipelines. It's being used as a GMA 950 with those drivers.

Of course it's going to perform very badly. The GMA 950 performed
VERY badly.

What I want to know is if with everything turned on, is it lackluster
or does it stink on ice. Nobody can tell anyone that just yet.

[Michael]

We hope to have our initial results out next week after the official release of Fedora Core 6. Communicating with Intel's Linux driver team, there aren't any performance tuning options available at this time and there is still much work ahead.

[drag]

Found this:
http://deadmoo.com/articles/2006/09/28/intels-new-onboard-video-benchmarked

The X3000 on a similar setup to the person's blog that is posted above posted compared favorably to a Nvidia 7300GS with UT2004 and beat ATI's propriatory drivers with the X550.

So at least it seems now that Intel has gotten to low-end discrete video cards.

[colo]

colo, how is the board running?

My Intel DG965WH (I got two of them, actually) don't run at all, because of incompatibilities between the Board's BIOS and a certain charge of MDT/MCI DDR2-SDRAM. Neither Intel, nor MDT are willing to shell out information about when or if the BIOS-fix for the widely known issue will be implemented, whilst other manufacturers provided updates to the BIOS of their boards weeks ago already.

I'm really, really pissed by now.

If anyone of you posting here is in the position to get to this information, I'd be glad to have it - or rather more if one could provide me with a beta-BIOS of some sort to make my damn boxes running at last (all the hardware is lying around here for 4 weeks or so now, that really is frustrating - but I can't afford new components as of now), but that'd be too cool to come true, I guess.

[DavidC]

Performance of GMA X3000 on windows drivers version 14.25

http://www.forum-3dcenter.org/vbulletin/showthread.php?t=321049&page=4

Google translation:
G965: “a diagram driver” approved yet by Intel. “supports the numerous hardware features”. In the comparison to the old driver the values are improved around approximately 10-15%. Only Doom 3 jumps from 4 fps to 10 fps. Benchmarks in 1024x768.

E6400, 2x 1 GiB 6400-555.

Asrock Conroe945G-DVI / Asus P5B-VM
3dMark 2003 Patch 350: 2101 / 1760
3dMark 2005 V 1.2.0: 717 / 878
UT2004 Botmatch dm-ranking: 41 / 37
Quake 3 high detail: 139 / 107
Doom 3 medium: 7 / 10
Windows Vista Grafik für Aero: 3,2 / 3,9

[Michael]

My Intel DG965WH (I got two of them, actually) don't run at all, because of incompatibilities between the Board's BIOS and a certain charge of MDT/MCI DDR2-SDRAM. Neither Intel, nor MDT are willing to shell out information about when or if the BIOS-fix for the widely known issue will be implemented, whilst other manufacturers provided updates to the BIOS of their boards weeks ago already.

I'm really, really pissed by now.

If anyone of you posting here is in the position to get to this information, I'd be glad to have it - or rather more if one could provide me with a beta-BIOS of some sort to make my damn boxes running at last (all the hardware is lying around here for 4 weeks or so now, that really is frustrating - but I can't afford new components as of now), but that'd be too cool to come true, I guess.

I can work some sources and try to get you a Beta BIOS or at least more details when it will be released.

[colo]

That would be most kind. Great thanks in advance for your efforts :)

German is my mother tongue, and I happen to own the PC mag they got their numbers from right now in my hands - the new drivers is said to sport features that enable "numerous hardware features". One of the users muses about how the next update may include major speed improvements as well, but on no profound basis whatsoever.

At the time being, GMA 3000 seems to be the overall faster solution than the GMA X3000. I hope this is going to change with an upcoming release, and the X3000 still got untapped power-resources of some sort.

On a side note, The Intel-board tested in the shootout performs incredibly well in everything concerning I/O-performance. Audio Quality is top-notch as well, at least concerning playback functionality.

If there's anything else you'd like to have translated, feel free to ask :)

[colo]

I can work some sources and try to get you a Beta BIOS or at least more details when it will be released.


Did you happen to dig up some info on this already? :)

[Michael]

Still working on it, I'll let you know if I come up with anything. If you happen to find anything, be sure to report back.

[colo]

Well, today there was another BIOS update released (I check for those on a near-daily basis), and the changelog doesn't speak of anything remotely regarding memory compatibility. F*ck it, I'm going to get new DIMM right now.