Link Copied
anhchuong:
I moved from Ati x200M to Intel X3100 as I saw on intel website's documents that introduce who brilliant x3100 would be.
I guess this one:
http://software.intel.com/en-us/articles/common-misconceptions-of-intel-integrated-graphics
First of all, BSOD happens almost everydays will Memory Corruption issue (Vista). [...] Fix all of the bugs expecially memory corruption and BSOD on Vista.
Please read this posting:
http://software.intel.com/en-us/forums//topic/57863
And post your Laptop (vendor + model) and VBIOS version. You find VBIOS version in igfxcfg utility clicking information.
I think it is caused by HW T&L.[...] Disable HW T&L if it is totally useless and concentrate on developing SW shader mode.
You can disable HW T&L and go software T&L it if you want. And indeed - depending on the application - it's even faster.
Read this one including the pdf linked:
http://software.intel.com/en-us/forums/showthread.php?t=52983#40332
To improve x3100 as well as other GM965 IGPs, I hope you allow the fixed memory to at least 32MB (as GMA945) and extend max Vram to 25% total memory (as possible)
Well it allows to use 384MB dynamically. In my cases the average usage is at 170MB during Aero Vista. If your chipset is not very powerful at fillrates (like the X3100) then there is no point in spending more memory.
- Re-map the core-clock + memclock to exact as FSB rate ( I put 2x1Gb 667Mhz but it can only run at 533Mhz)
Does the core clock impose restrictions on the maximum DDR2 speed? Didn't know that. Using DDR2-667 at it's full speed here with GM965 on Dell XPS-M1330. Is you SPD programmed correct?
Allow End-user to update Vbios as not many retail companies care about it (TOSHIBA, IBM and Dells)
I'd like to have that feature too. However I don't see that happening at other companies either (in case of SATA BIOS included in the BIOS). In the case of VBIOS it's just more difficult since the BIOS can be customized quiete a lot. Have a look at crl_1588.bsf by opening it in notepad. It's included in vbios_gm965. You have to download the .zip version of the Windows driver. It includes some VBIOS developer information. At least for DOS it provides a facility to TSR the new VBIOS. Well - I know that's not very helpful.
This is the way I'd propose VBIOS updates:
. intel compiles all configuration data in VBIOS and puts it in BIOS. That means the actual "BMP Script file" data has to be included in VBIOS.
. at (Windows-)driver initilization "BMP Script file" is read from VBIOS and VBIOS is replaced by uptodate version in RAM. Whether you use the MMU or other facilities for replacing VBIOS doesn't matter here.
. done. Your Windows driver now uses the VBIOS all (regression) testing and verification has previously been done for.
Intel definitely has a problem here and I'm wondering whether they are aware of that. If that's not sufficient Intel's benefit would be that bug finding is easier since they always know which VBIOS has been used together with which driver.
Do not discontinue developing it or you are really have no match with ATI
More simply put: Intel spend more $$ on driver development, verification and testing.
Maybe a first step could be to introduce some platform for intel Windows driver team to get awareness of issues/whishlists etc. Things like these 24p, HDMI topics would be reported there. Why's that necessary? If you use intel IGP you always feel like being the first one hitting a problem. In case of nVidia/AMD/ATI you're never alone. Someone else found that already and for some reason communication channels to driver developers work there.I kind of agree.NEXT TIME, TEST YOUR PRODUCTS CAREFULLY BEFORE YOU INTRODUCE IT TO EVERYONE AND TELLUS EXACTLY WHAT YOUR PRODUCTS CAN DO RATHER THAN EXEGRATING THEM
Then the core clock and mem clock are only at 200MHz not as advertised on intel site (500Mhz).
Where did you get the 200MHz number from?
I just used GPU-Z 0.2.3. It has some intel IGP support and tells:
GPU Clock 667 MHz
Default Clock 667 MHz
I do not know whether I can trust that numbers. If I look in the datasheet for the GM965 it says 500MHz
http://download.intel.com/design/mobile/datashts/31627303.pdf
Page 14 section 1.2.2 Integrated Graphics
500 MHz core render clock at 1.05V core voltage
Wikipedia tells 667 MHz:
http://en.wikipedia.org/wiki/X3100#Table_of_GMA_graphics_cores_and_chipsets
You will also find (inspecting the .bsf file I mentioned) that graphics core clock is customizable by the OEM:
Combo $Render_Freq_Switch, " Enable Intel Dual-Frequency Graphics Technology:", &No_Yes_List,
Help "This feature determines whether the Intel Dual-Frequency "
"Graphics Technology (DFGT) is enabled or disabled. Intel "
"DFGT is a graphics power savings technology that adaptively "
"lowers the render clock frequency when the system is running "
"in battery mode. Note that this technology is only active "
"when the system is running in battery mode and the LFP is the "
"only active display device. When Intel DFGT is enabled, the "
"Default Render Clock Frequency should also be selected."
Combo $Render_Freq_Default, " Default Render Clock Frequency:", &Render_Freq_List,
Help "This feature allows the OEM to set the default render clock "
"frequency when the Intel Dual-Frequency Graphics Technology "
"feature is enabled. "
" "
Furthermore I just recognized that there's indeed some relationship between FSB/Memory/Clock and Render Clock. See table 31 in section 13.3 "Host/Memory/Graphics Core Clock Frequency" p.132.
Last but not least the chipset offers Thermal Management throttling according to seciton 11.
About GPUz, I am not interested in it much as it doesn't show very detailed and precise information about core+memclock. I tested it with an overlocked GPU but it could only show the stock information.
I'm not interested in GPU-Z either as I said. I'd still like to know where you got those 200 MHz from.
Here you are refering to an overclocked GPU. I assume Not an Intel GPU. I wouldn't be surprised if GPU-Z wouldn't detect overclocking since some overclocks only apply to some demanding 3D mode. In Vista Aero GPUs run in some lower-power mode and not at full clock.
About the table 31 you mentioned, isn't it better if Intel supports another combination of Host: 533Mhz Mem: 667Mhz =>333Mhz as ATI does. (a thought for other people).
What makes you think this would be better?
There is already a Host 533 MHz / Memory DDR2-533 / Render Clock 400 MHz configuration in place, which would outperform a 533/533/333 configuration. For GM965 as well as GL965. There is no reason for synchronous operation or integer clock multipliers what so ever.
About Vbios update, have you ever tried it on XP. I don't have XP on my laptop to test but when I checked with notepad, I realized that the Vbios update can only be done in XP. Why doesn't Intel support it in VISTA?
There is no VBIOS update!
There is a Terminate and Stay Resident application to load the VBIOS to RAM (but is NO VBIOS update itself). But it's useless in Vista/XP:
crl_1588.exe /?
IntelGM965/PM965/GL960 PCI Accelerated SVGA BIOS
Build Number: 1588 PC 14.34 01/24/2008 14:38:24
DECOMPILATION OR DISASSEMBLY PROHIBITED
Copyright 1999. Intel Corp. All Rights Reserved.
64K RAM Video BIOS.
There is also a generic VBIOS file crl_1588.dat. If you'd try to patch your notebook/desktop BIOS with this one, it may very likely break your notebook/desktop, since some important configuration data is missing.
Bugs on BSOD can be solved by editing the reg (using regedit) which I am testing now. If I dont have BSOD for 1 week, I will post the bugfix here.
My questions are still open:
. What type of notebook?
. What driver revision?
. What VBIOS revision?
. GM965 or GL965 chipset?
. What sort of BSOD? When?
I used Powerstrip to see the core+mem clock and it is exactly the same as in the table you showed me.
If I start Powerstrip, I'm getting the same information:
Memory clock - 200,00 MHz
Engine clock - 200,00 MHz
It doesn't change when starting some more demanding 3D application, but I simply don't trust that numbers. I don't trust that numbers the same way I do CPU-Z.
By the way: I know Powerstrip has support for intel chipsets and especially custom resolutions, but check back whether it's not broken. I'm pretty sure this is still valid (although an older posting):
http://www.avsforum.com/avs-vb/showthread.php?t=947830
Probably best source is powerstrip forum.
The reason is because the setting you mentioned never appear on my friend's system (CPU t2370 bus 533 + ram 2gb bus 667 but core + mem only at 200Mhz. How can you get 400 Render clock there)
I am not getting 400 MHz clock here. I was referencing table 31.
And the settings I posted before were read by CPU-Z, which I don't trust either. I'm still pretty confident that the X3100 runs at a decent speed even if no tool is able to read accordingly.
That is exactly as what I got when I run crl_1588.exe. However. for some reason, it will not work due to lack of access to the Vram + Bios.
The .exe is not meant to work on XP/Vista at all. It's useful for MS-DOS only and is no permanent BIOS update.
Because vendors are kinda lazy updating the Vbios to fix bugs, isn't it better that Intel supplies user update manually (at risk maybe)
I agree to the first part, but not necessarily to the second. The solution I'm thinking off I've been posting before.
In the meantime I can think of compatiblity lists of drivers and minimu VBIOS. Or at drivers not installing on too old VBIOS.
