ATI FireGL/FireProSoftmod - Discussion Thread for new members

Thanks so much for your clarification. I knew something was not right.
I re-read your instruction many times and could not find where this patch is required for XP32 so my assumption it was not required (my error). I will wait until you have time to look into this. Thank you much.

 

thanks...i downloaded both, but can you tell me which one performed better the fs 9250 or fp 8700

 

I have the sam card as you, HD4890. but my card is from gigabyte...The thing is that my ID is 9460 and that corresponds to the RV790 chip. Problem is, no other card uses that chip :bang: What to do? guess my luck just ran out.

 

Hi Veselinbg. I think that I know why the maximum frame
rate that I obtain in the Maya bike scenes is also 100 frames, like you.

What happens is that Firepro 8700/8750 is made based in the V770 gpu clock of the hd4870 , and not on the V790 GPU and memory clock of the hd4890!

hd4850 = 625MHZ +/-

hd4870 = Firepro 8700/8750 = 750MHZ = V770

hd4890 = +100MHZ = 850MHZ = V790


They are different as you can see. Probably, current's drivers and or XMl settings are not optimized to make that hd4890 takes all advantage using the firepro's driver with its clock.

This explain why a get the same results as you. I using a hd4890 and you a hd4850.

Cause de driver or XML `see` the two cards like same cards, a V770/4870 card type!

Maybe schoko can help us whit new script for firepro 8750 x 4890.

To be honest, I believe that this limitation ocour on XML; cause on SpecViewPro 10 I am getting now 300 frames on Maya test(1240-1024 resol); so, its not limited at all! On the same test, the real FirePro gets only 278 frames!

So, Why I can`t get that improved performance inside the real program?


I think that the XML adjust can be used to liberate the real performance on Maya for 4890 users.


I wish to have been helpfull and that it can be usefull to me also.

 

Fyi, the atiogl.xml is neither dependant on the type of card you use, nor the the clocks of the gpu. it just uses a setting for the opengl caps. those opengl capabilities only work when a certain hardware is available. the patch (atikmdag.sys or ati2mtag.sys) actually tells the driver to recognize a certain device id (card) as capable of using the advanced opengl settings.

the atiogl.xml uses a setting(mostly they are different) for each programm that is listed in the atiogl.xml.
when a programm is not listed it uses standard settings only. specviewperf for example runs on standard settings, while the atiogl.xml contains a profile for maya.

you would have to search for something else that makes the difference.

 

GPU Fan Speed Control

I am trying out ver. 8.603 at the moment (latest for XP32 I believe). Everything went fine as far I can tell. Now noticing that I lost the option of manual control of GPU fan speed - not available in the CCC nor Rivatuner 2.24c.:bang: So the fan is currently running up and down. I prefer constant speed.

Did anyone else experience this issue? Normal? Or did I not do this softmod correctly.

BTW, I am using Wesley8 patchscript. Thanks in advance for your reply.

Update:
Looks like I didn't have the correct 'rivatuner.cfg' for HD 4890 support. It's fine now. For the CCC issue, seems is normal for the Workstation cards.
It's all good now.

 

Ok, thanks for the info, friend but, just can't understand my hd4890 performance , cause I have the same system as Veselinbg at same clock(when not overclocked).

if the atiogl.xml or driver it's not dependant of the hardware clock and only 'open' certain capabilities when available on the video card, the real difference of performance , between both the video cards, 25-35%, would have to continue to exist, perhaps the default speed of my processor is limiting the total performance of the video card.


My spec:

core i7 920 clock default
HD 4890 clock default
2 giga Corsair 1333MHZ default
Windows 7-x64
Maya 2009-x64

 

anybody tried already this patch routine for mobile ATI Radeon HD3650 (dev id 9591 ) ?

 

Why not? Please don't ask for every model, every Radeon/Mobility Radeon tested from HD2/3/4xxx series worked so..

 

Does anyone has a patch script for XP32 for version 8.633 - equiv to the XP64 shown below? I searched but couldn't find. Perhaps there is not one created. Thanks in advance.

Code:

# ATI FireGL/FirePRO [B]XP64[/B] mod by gellmaR (gellmaR@yahoo.com)
# for FireGL drivers version [B]8.633[/B]
# test only, no ATIFuse enhancements!

[Common]
SrcFile = ati2mtag.sys
BakFile = ati2mtag.old
HlpFile = newbietechFireGLX2000.rth
MakeCRC = 1
Packed = 1

;common script for 669x+


DstVar0 = force ATI FireGL V8650 capabilities
DstVar1 = force ATI FireGL V8600 capabilities
DstVar2 = force ATI FireGL V7700 capabilities
DstVar3 = force ATI FireGL V7600 capabilities
DstVar4 = force ATI FireGL V5600 capabilities
DstVar5 = force ATI FireGL V3600 capabilities
DstVar6 = force AMD FirePro 2260 capabilities
DstVar7 = force AMD FirePro 2450 capabilities
DstVar8 = force AMD FirePro V3700 capabilities
DstVar9 = force AMD FirePro V3750 capabilities
DstVar10 = force AMD FirePro V5700 capabilities
DstVar11 = force AMD FirePro V7770 capabilities
DstVar12 = force AMD FirePro V8700 capabilities
DstVar13 = force AMD FirePro V8700 Duo capabilities
DstVar14 = force AMD FirePro V8750 capabilities
DstVar15 = force AMD FireStream 9170 capabilities
DstVar16 = force AMD FireStream 9250 capabilities
DstVar17 = force AMD FireStream 9270 capabilities
DstVar18 = force ATI FireGL V5700 AGP capabilities

Src0 = 0F BA A3 FC 00 00 00 0E 72 3C BA
SCM0 = FF FF FF FF FF FF FF FF FF FF FF
Dst0 = 0F BA A3 FC 00 00 00 0E EB 2F BA
DCM0 = 00 00 00 00 00 00 00 00 00 00 00

Src1 = 83 F8 55 75 22 48 8D 8F 78 01 00 00 FF 15 2E 6D 13 00 A8 10 74 11 33 C0 48 8B
SCM1 = FF FF FF FF 00 FF FF FF FF FF FF FF FF FF 00 00 00 00 FF FF FF 00 FF FF FF FF

Dst1 = 83 F8 55 75 00 48 8D 8F 78 01 00 00 FF 15 2E 6D 13 00 A8 10 74 00 33 C0 48 8B
DCM1 = 00 00 00 00 00 00 00 00 00 00 00 00 00 00 FF FF FF FF 00 00 00 00 00 00 00 00 

Src2 =   83 F8 40 75 09 66 81 7C 24 30 02 10 74 11 33 C9 E8 26 DC B0 FF
SCM2 =   FF FF FF FF 00 FF FF FF 00 00 FF FF FF 00 FF 00 FF 00 00 00 00

Dst2_0 = 0F B7 54 24 32 66 25 00 00 66 0D 0A 94 66 89 44 24 32 EB 0B 90
Dst2_1 = 0F B7 54 24 32 66 25 00 00 66 0D 0B 94 66 89 44 24 32 EB 0B 90
Dst2_2 = 0F B7 54 24 32 66 25 00 00 66 0D 11 95 66 89 44 24 32 EB 0B 90
Dst2_3 = 0F B7 54 24 32 66 25 00 00 66 0D 0F 94 66 89 44 24 32 EB 0B 90
Dst2_4 = 0F B7 54 24 32 66 25 00 00 66 0D 8C 95 66 89 44 24 32 EB 0B 90
Dst2_5 = 0F B7 54 24 32 66 25 00 00 66 0D 8D 95 66 89 44 24 32 EB 0B 90
Dst2_6 = 0F B7 54 24 32 66 25 00 00 66 0D CF 95 66 89 44 24 32 EB 0B 90
Dst2_7 = 0F B7 54 24 32 66 25 00 00 66 0D CD 95 66 89 44 24 32 EB 0B 90
Dst2_8 = 0F B7 54 24 32 66 25 00 00 66 0D CC 95 66 89 44 24 32 EB 0B 90
Dst2_9 = 0F B7 54 24 32 66 25 00 00 66 0D 9F 94 66 89 44 24 32 EB 0B 90
Dst2_10 = 0F B7 54 24 32 66 25 00 00 66 0D 9E 94 66 89 44 24 32 EB 0B 90
Dst2_11 = 0F B7 54 24 32 66 25 00 00 66 0D 46 94 66 89 44 24 32 EB 0B 90
Dst2_12 = 0F B7 54 24 32 66 25 00 00 66 0D 56 94 66 89 44 24 32 EB 0B 90
Dst2_13 = 0F B7 54 24 32 66 25 00 00 66 0D 47 94 66 89 44 24 32 EB 0B 90
Dst2_14 = 0F B7 54 24 32 66 25 00 00 66 0D 44 94 66 89 44 24 32 EB 0B 90
Dst2_15 = 0F B7 54 24 32 66 25 00 00 66 0D 19 95 66 89 44 24 32 EB 0B 90
Dst2_16 = 0F B7 54 24 32 66 25 00 00 66 0D 52 94 66 89 44 24 32 EB 0B 90
Dst2_17 = 0F B7 54 24 32 66 25 00 00 66 0D 05 94 66 89 44 24 32 EB 0B 90
Dst2_18 = 0F B7 54 24 32 66 25 00 00 66 0D 96 95 66 89 44 24 32 EB 0B 90 

 

which model will suit my gpu best v8700 or v8750 .(as both r R770 model). i hav visiontek 4870 512mb ddr5.........

 

Well rogerio, i have now made some tests with windows7 and the opimized atiogl.xml in win7. and the results clearly show one thing :

the performance with the modified atiogl.xml is mainly dependent to the performance of the cpu.

since win7 has a way better multicore cpu management than xp64 or xp32 the achieved performance boost is even higher with win7.

proof is in the numbers

testbed 1
maya 2009 sp1 - XP 64 HP 8510p with T7700(2.4GHZ),Hd2600 mobility(stock clocks 500/600MHZ), 4GB ram ( only 3GB Ram used to have hibernation working in xp64)

testbed 2
maya 2009 sp1 - Win7 x64 HP 8510p with T7700(2.4GHZ),Hd2600 mobility(stock clocks 500/600MHZ), 4GB ram
btw. additional info: turned off all the eyecandy in win 7 ( best performance setting), basicly it looks like XP now, and it is even that fast or faster in some demanding graphics intensive tasks




Results XP64 (3GB)

bike.ma :
shaded - 52.9 fps
wire - 65.7 fps

carbench :
textured - 7.9 fps
shaded - 9.0 fps
wire - 66.2 fps

mayascene2 :
textured - 5.2 - 10.7 fps
shaded - 6.1 - 15.7 fps
wire - 10.3 - 29.6 fps
-----------------------------------------------------------------------
Results Win7 x64 (4GB)

bike.ma :
shaded - 76 - 81 fps
wire - 79 - 83 fps

carbench :
textured - 8.1 fps
shaded - 9.3 fps
wire - 63.9 fps

mayascene2 :
textured - 5.6 - 11.5 fps
shaded - 7.5 - 20.3 fps
wire - 14.7 - 33.5 fps


Windows XP x64 / 3GB Ram
i have note the highest and the lowest fps per scene:

Win7 x64 / 4GB Ram
i have note the highest and the lowest fps per scene:

a hint for some people here at the board :
don´t forget i still have a crippled hd 2600 mobility chip, but when i look at the performance that this sweet little card is able to achieve it does not need to hide behind a desktop quadro fx 4600 or a modded desktop 8800GT(@ quadro fx 3700).


Quite to the contrary what a lot of people said about opengl performance in vista and win7 these results draw the opposite picture.
it seems that the opengl integration in win7 is as perfect a it is in xp. Infact maya uses quite much of the CPU, even when modeling in the standard viewports, und thus it might not surprise that an OS with a very good multicore-cpu integration likeWin7 is speeding up the viewport performance and simply blows good old XP out of the water.

guess these insights leave quite some room for discussions.

 

@ schoko

i just installed ur new patch for xp 32bit their was huge improvement while handling viewport....but modifiers n every other tool bar n tabs turned into grey color ... all came visible wen i run cursor over them n then again turned grey...and thanx man u made my day now i can run 3dsmax way better then i was from 2years (sorry for my bad english) ....

 

hmm, well, maybe i haven´t done that patch 100% right. will have another look into that.

what you describe should not happen. can you tell in which program you are experiencing these graphic glitches ?

 

in 3dsmax 2008 wen i started the program all tabs n menus goes gray n wen ran cursor over a menu it will become visible ....but after removing cursor from their it becomes gray again...

 

have updated the script.
hope it works now. since i don´t have the possibility to test xp 32 myself, please report back if it works this time.

btw- please delete the already patched ati2mtag.sys and rename the ati2mtag.old to ati2mtag.sys. after that you can patch it again.

 

Thanks much schoko. I'll report how it goes.

Are you saying to delete the ati2mtag.sys under the Windows Drivers folder? I found it in there but there is no ati2mtag.old file to rename per your instruction.

Do I still need to delete this ati2mtag.sys ?

 

i tired this script but same result ..btw wen i uninstalled drivers before going for new patch xp gives me notification that found new hardware ati firepro v8750 n i m installing v8700 bt before installing ur patchscript i was running my gpu as v8750 on gellmar patchscript (but had uninstalled it from device manager) is this becoz of that i m having this problem in 3dsmax 2008 ??

 

well, if you had done it right, you shouldn´t get such a notification.

the patching with rivatuner gave no errors ? is the performance in opengl as it should be ? have you tested with specviewperf 10 if the mod has worked out ?

if you can answer those question with yes, it is still possible that ati´s driver has some issues with max. it´s at least a possibility. or maybe a setting for the graphics card that is not working well with 3dmax?

please use the specviewperf so we can check if the mod is working.

 

i don,t hav specviewperf 10 (my download speed is 25 - 30kbps) i ll download it till tommorow n will run a benchmark to sort this confusion btw rivatuner had given me error...n i noticed that their is no difference in rendering bt viewport is running better then first it was ....