softmod 5900xt to quadro3000£¬ seems im the luck one

Hi Rui,

No, I can't read chinese. My best friend is Chinese-Malay, but he only speaks it, not read.

I'm trying my best to understand what I can by the pictures you guys put up, but it's a challenge.

Thanks for helping all you can Rui!

---

BluefireXP, on Rui's message board I managed to find some info. About halfway through the 41 pages I saw a rundown of where they were at that point.

I see two p177h-4 boards that succeeded, one GW-P/N@CVGG32210IP0B0:0 board succeeded,

Two p172fp297 failed, one 177 02 failed, one p177 -1n
failed, one GW-P/N@CVG732000IQ0J0:0 failed, and there is one p177h-4 board that failed as well.

There's even a guy on page 38 who did a 5700U -> FX1100 with UGS of 42.54. I've been jumping around pages, but I'll try to look through it tomorrow and see what else I can find.

I think this is as confusing as ever. Did they ever come to a conclusion on that board I wonder?

Augie

 

five case got succeeded,at least. three p177h-4n and two GW-P/N@CVGG32210IP0B0:0 board .
im trying to translate my "moded step " to english , will take a little time . you knowl , my bad english ...

and , that guy hardmod his 5700U , SO ,high score of UGS-03 not surprise me .

 

That is really interesting...If identical cards have different UGS...wtf.. :s3d:

 

Augie , welcome here :
http://bbs.hxsd.com.cn/post/view?bid=5&id=3150168&sty=1&tpg=1&age=30&ppg=42


that is my moded step , hope can help you .(english version)

 

Okay, here's my synopsis after reading pretty much everything on Rui's message board... (using Babelfish to translate, painstakingly, message by message).

177h-4n boards _MIGHT_ get you good UGS scores if you are lucky.
Even having an identical board does not guarantee good UGS.(my personal conclusion)

Page 25 of the message board, halfway down has a list of success/failures.
http://bbs.hxsd.com.cn/post/view?bid=5&id=3150168&sty=1&tpg=1&ppg=25&age=0#3150168

The only 5900XT boards that achieved a high UGS score so far (no hardmod):
High UGS: (succeed)
3 - p177h-4n (1x MSI FX5900XT-VTD128, 2x Chinese OEM under Albatron?)
2 - GW-P/N@CVGG32210IP0B0:0 board (Gainward 5900DT)

Low UGS: (fail)
2 - p177h-4n (1x MSI FX5900XT-VTD128 my bro's, 1x Unika 5918FXL 5900XT)
1 - 17702975 (Leadtek Winfast A380 TDH MYVIVO 5900XT)
1 - p177 -1n (?Õ‰F 5900XT -- unknown Chinese OEM)

This is the procedure given to me by Rui (I cleaned it up a bit):
5900XT -- > Quadro FX3000
(I'm posting this here for the heck of it. Made some links so if anybody out there wants to try it he can.)

(Download and install RivaTuner 14.3 NY if you don't have it)
http://downloads.guru3d.com/download.php?det=163

1. Download the 45.28 driver from NVIDIA :
http://download.nvidia.com/Windows/Quadro_Certified/45.28/45.28_win2kxp.exe
and Unpack this file to a directory,for example -- c:\45.28

2. Create a dir name "MOD",for example -- c:\MOD
3. Copy these 3 files from c:\45.28 to c:\MOD
nv4_mini.sy_
nv4_disp.dl_
nvoglnt.dl_

4. Copy the Unpack.bat (which Rui made) to c:\MOD, and run it.
unpack.bat is simply:
expand nv4_mini.sy_ nv4_mini.sys
expand nv4_disp.dl_ nv4_disp.dll
expand nvoglnt.dl_ nvoglnt.dll

( Download unpack.bat and nv4_disp.inf here http://jupiter.lunarpages.com/~augiem2/Modstep.zip )

5. In c:\MOD, delete
nv4_mini.sy_
nv4_disp.dl_
nvoglnt.dl_

6. Run RivaTuner
PowerUser --> Open Patch Script
Apply these four scripts:
*Softquadro4
*NV Strap Antiprotector
*AnisoBoosterOpenGL
*DetonatorFXAntiprotector
When the script ask for the location of the driver, just point to c:\MOD

7. In c:\MOD, del *.old ,and rename the rest:
nv4_mini.sys => nv4_mini.sy_
nv4_disp.dll => nv4_disp.dl_
nvoglnt.dll => nvoglnt.dl_

After that, copy them to c:\45.28, replacing the original files

8. Copy nv4_disp.inf to c:\45.28, replacing the original one.
(Allows 45.28 drivers to recognize the 5900XT - find it in modstep.zip listed in step 4.)

9. Download a 60.xx driver from guru3d, install it (run exe), reboot ,uninstall it, reboot., Your system is back to VGA mode.

10.In the moded driver dir c:\45.28, run setup.exe to install 45.28 driver, reboot.

11.Run RivaTuner, install NVStrap, but do not reboot. In NVStrap, Change the PCI Device ID to Custom --> 0338 Quadro 3000. Reboot.

12.When Windows asks to install a driver for Quadro 3000, press cancel.

13.In c:\45.28, run setup.exe, again, reboot.

14.Go to Display Properties -> Settings -> Advanced -> Nvidia panel -> Open GL Settings -> Vsync OFF by default

15. Download and run SpecViewPerf 7.1.1
http://www.spec.org/gpc/downloadindex.html

-----

I followed his instructions TO THE LETTER. I made sure there weren't any other settings he enabled.

My MSI results are the same as before. All good, but bad UGS. This MSI card can softquadro, but it cannot be a full FX3000.

I'm trying to confirm it again, but according to Rui, FyNi's MSI card that successfully got high UGS scores is physically identical to my brother's card. (I posted pictures of how his card looks and asked them to check some particular chip layouts that vary from board to board.)

5900XT -> Quadro FX3000 Conclusion:
Buy a p177h-4n or Gainward board
Have a small CHANCE at getting a full Quadro FX3000 - don't count on it though.
So far every other board revision tested fails UGS

Gainward board that succeeded looks like the one shown here:
http://www.hardsalon.com/2004/3-31/102411.html

MSI success and failure (both cards look the same also):
http://www.isso.com.cn/readarticle/27/27514.asp

2x success by Chinese OEM (under Albatron?)
http://www.pcpop.com/market/2004/5/18319.shtml

Phew... A lot of work.

Thanks for being so patient and helpful RUI! You don't find too many people like you out there these days!

Augie

 

For anyone who is undecided if s/he wants hardware antialiasing have a look at the pictures at this page, and decide for yourself if you need antialiasing :
http://fly.cc.fer.hr/~unreal/theredbook/appendixj.html

Figure J1 = normal wireframe (first picture).
Figure J3 = antialiased wireframe (third picture).

(Index of the book is here : http://fly.cc.fer.hr/~unreal/theredbook/)

From this page : Unigraphics Viewset (ugs-03)

Now something else.
While I was checking out this program PCItree (2 weeks ago), I bumped into this page http://www.pcidatabase.com/ (through the PCI links >> list of VIDs and DIDs).
When I key in "nvidia" into the "Vendor search" field, then click search, then click "NVIDIA Corporation", I get a list of all NVidia's PCI device IDs. There is a strange range 0x00C0-0x00CE :
0x00C0 Chip Number: NV41.0
Chip Description: ???

0x00C1 Chip Number: NV41.1
Chip Description: ???

0x00C2 Chip Number: NV41.2
Chip Description: ???

0x00C8 Chip Number: NV41.8
Chip Description: ???

0x00CE Chip Number: NV41GL
Chip Description: ???

Anyone ?

 

thx NxCPU, although this antialiased wireframe picture is really poor, the AA lines there are much thicker than i can accept. In 3dsmax the AA line are crisp and thin even without hardware. there is only 1 thing one should take care about: NEVER use the 3dsmax internal AA option coz then the lines will become thick also. Always force AA in the display properties- and even if not enabled in 3dsmax the lines WILL be antialiased- you can visually check this.

 

thank you ,Augie. it is great !!


BTW , Unika 5918FXL 5900XT was made by Palit , not the reference PCB design .

http://www.unika.com.cn/includes/product.php?id=453

 

Same here.
I have the book, and the pictures are better. It still doesn't look acceptable to me though.
I think most of the time antialiased wireframe mode (no matter h/w or s/w) is not needed.
Maybe only for a presentation or so.
I guess to me it's only eyecandy.

I don't use Max myself, but that's exactly what I like to see in wireframe mode. Thin and crisp lines displaying the line segments and vertices exactly as they are.

 

NxCPU i'm going to post a screenshot!

 

Ok, first a part of a wireframe scene in 3dsmax 6 without AA



Now with the matreme's 3dsmax internal AA, it's terrible




And now by disabling AA withn 3dsmax and enabling 16xAA with RivaTuner



I think the difference is obvious, thank you RivaTuner
Note that there is some image quality loss due to jpg-compressing.

 

@Bluefirexp :
Good example pictures.
Maybe now anyone feeling "bad" about having a low ugs-03 score, will not care anymore.

(I have to admit though, that the "16xAA with RivaTuner" mode looks quite good.)

BTW, looks like the Palit cards are the safest bet :

 

Is there any effect of the low-ugs when using the RT16AA in 3dsmax? BluefireXP, could you run a 3dsmax specview test for comparing the RT16AA and the no-AA speeds of 3dsmax?

 

Snowball, ok, i will post a comparison, i already did bench with 16AA + texture sharpening + 8x Anisothropic filtering! :

Run All Summary

---------- SUM_RESULTS\3DSMAX\SUMMARY.TXT
3dsmax-02 Weighted Geometric Mean = 17.96

---------- SUM_RESULTS\DRV\SUMMARY.TXT
drv-09 Weighted Geometric Mean = 37.53

---------- SUM_RESULTS\DX\SUMMARY.TXT
dx-08 Weighted Geometric Mean = 76.92

---------- SUM_RESULTS\LIGHT\SUMMARY.TXT
light-06 Weighted Geometric Mean = 26.52

---------- SUM_RESULTS\PROE\SUMMARY.TXT
proe-02 Weighted Geometric Mean = 18.68

---------- SUM_RESULTS\UGS\SUMMARY.TXT
ugs-03 Weighted Geometric Mean = 7.909



and here again without any AA:

---------- SUM_RESULTS\3DSMAX\SUMMARY.TXT
3dsmax-02 Weighted Geometric Mean = 23.79

---------- SUM_RESULTS\DRV\SUMMARY.TXT
drv-09 Weighted Geometric Mean = 128.9

---------- SUM_RESULTS\DX\SUMMARY.TXT
dx-08 Weighted Geometric Mean = 119.9

---------- SUM_RESULTS\LIGHT\SUMMARY.TXT
light-06 Weighted Geometric Mean = 28.77

---------- SUM_RESULTS\PROE\SUMMARY.TXT
proe-02 Weighted Geometric Mean = 25.88

---------- SUM_RESULTS\UGS\SUMMARY.TXT
ugs-03 Weighted Geometric Mean = 9.820



I can also say, that working with 16AA in 3dsmax is possible with excellent performance. I really never had any slowdowns that harmed my work.

 

So that means that 3dsmax users have absolutely no need for the real Quadro 3000, an ultra card would be much better (with lowUGS)?

 

well, i think so. But maybe some of the guys with TRUE Quadro clones could also run specviewperf 7.1 with 16x AA - this would show up the differences....It would be interesting to see, HOW high their UGS-03 will be with these settings!
Rui, could you plz. set 16x AA , texture sharpening and 8x Anisothropic filtering on your RivaTuner's OPEN GL settings and run Specviewperf?

 

I'm taking this back.
Because I checked the above link to Unika and it is a completely different card.
It just does not look like the one rui0317 posted pictures of here :
http://forums.guru3d.com/showthread.php?s=&threadid=91686&perpage=10&pagenumber=5.

So I started digging around the site a bit.
And found this menu link http://unika.com.cn/products/product_albatron.php (fourth menu option in the top menu, then third menu option in the dropdown).
Second option in the dropdown is http://unika.com.cn/products/product_unika.php
Both FX5900 family links from these pages lead to http://www.unika.com.cn/includes/product.php?id=453

But the sixth option in the dropdown is http://unika.com.cn/products/product_xyg.php
And the FX5900 family link from there leads to http://unika.com.cn/includes/product.php?id=449 A look-a-like of the pictures posted by rui0317 here : http://forums.guru3d.com/showthread.php?s=&threadid=91686&perpage=10&pagenumber=5 (first picture) Notice the name of the pictures "unika_xiaotyaog_5918fx-01.jpg".
xyg = xiaotyaog ? 5918fx not 5918fxl.

So maybe Unika is the one manufacturing cards under their own name, and also for Albatron and XiaotYaog ?
It just makes me wonder where Palit fits in to the story.

BTW, there is a nice FX 5900XT roundup review with many pictures here : http://www.hardwarezone.com/articles/view.php?id=1110&cid=3&pg=1

 

of course i will do that , after i go home . after ten days, for job reason . :O

"xiaoyao" G Turbo 5918FX made by Albatron :
http://unika.com.cn/includes/product.php?id=449

"supei" 5918FXL made by Palit :
http://www.unika.com.cn/includes/product.php?id=453

"UNIKA" put his name on those card and sell them , we call this "OEM"

 

The main problem with FSAA is that it disables the hardware overlays, a big no-no if you use Maya or Houdini.

 

The Unika 5918FXL (L) card failed the UGS test. It's not a reference design, so it's not surprising that it failed. Apparantly this is manufacturered by Palit and resold by Unika.

The Unika 5918FX (not L) card link Rui posted is identical to his card which can be seen at http://www.pcpop.com/market/2004/5/18319.shtml
Apparantly they're both the same card made by Albatron, but are resold by Unika and Colorful.

That's my impression anyway.

Augie