Basic FX59xxx bios changes

[jimmor]

DJ Sebas,

Attached below in PCX5900m.zip is a modded Gigabyte PCI-E 5900 bios which should be suitable for your card.

It includes timing changes and stock speeds of 300/420/750. And since the original bios had no GPU volts selection table, it was necessary to create one. This has been set with 2D at 1.2v, Throttling at 1.5v and 3D at 1.5v.

Give it a try and see how you get on ?

Or alternatively, why don't you try flashing with the BF5u5197.rom bios used by XL_ence. He said he has exact same card as yourself, and currently runs his at 505/1030. So I have also linked it here just in case you want to give it a try,

http://home.arcor.de/redinferno/5900XT-BIOS/MOD-BIOS/sev5950u/bf5u5197.zip

:cool: :)

[balthazor]

Hi Jimmor,

I have an aopen Aeolus 5700 LE 256 mb AGP card.
I tried some tips from the tutorial using X-bios editor and NiBiTor.
At the moment my FX5700 LE card runs at 2D = 450 mhz/500mhz and 3D = 477 mhz/500mhz Artefact free

The stock speed of this card are 2D=225/200 3D= 250/200

I get 3208 3D marks in 3Dmark2003

Width 1024
Height 768
Anti-Aliasing None
Post-Processing No
Texture Filtering Optimal
Max Anisotropy 4
Vertex Shaders Optimal
Force PS 1.1 in GT2 & GT3 No
Repeat Tests Off
Fixed Frame rate Off
3DMark Score 3208 3DMark Game Tests

And 30295 Aqua Marks [same configuration as above but with V High detail]
Would you help me, getting the last FPS out of my card?

I tried different voltages for 3D 1,29-1,35.
But without gaining any FPS
I donot know how to edit the ram timings table in Hex My ram has an timing of 2ns

I send you my bios file [ORIGINAL AEOLUS FX 5700 LE 256 MB AGP.zip]

My email address is govertgroenier@gmail.com

10X

[jimmor]

Hi Jimmor,

I have an aopen Aeolus 5700 LE 256 mb AGP card.
I tried some tips from the tutorial using X-bios editor and NiBiTor.
At the moment my FX5700 LE card runs at 2D = 450 mhz/500mhz and 3D = 477 mhz/500mhz Artefact free

The stock speed of this card are 2D=225/200 3D= 250/200

I get 3208 3D marks in 3Dmark2003

Width 1024
Height 768
Anti-Aliasing None
Post-Processing No
Texture Filtering Optimal
Max Anisotropy 4
Vertex Shaders Optimal
Force PS 1.1 in GT2 & GT3 No
Repeat Tests Off
Fixed Frame rate Off
3DMark Score 3208 3DMark Game Tests

And 30295 Aqua Marks [same configuration as above but with V High detail]
Would you help me, getting the last FPS out of my card?

I tried different voltages for 3D 1,29-1,35.
But without gaining any FPS
I donot know how to edit the ram timings table in Hex My ram has an timing of 2ns

I send you my bios file [5700LE.256MB.Rev.01.zip]

My email address is govertgroenier@gmail.com

10X

Sorry, but these days I don't get involved with fine tuning of everybody's bioses. And especially when it is for cards I don't have !!

Two things I will say however is:-

1) There is no reason to ever set 2D core speed higher than 300Mhz !

2) It appears that you have slightly tightened the 2D and 3D timings in the bios you linked, however you have entered them backwards. Your changed timings are currently at 0301040C000F0009000505020303, but should actually have been entered as 03030205050009000F000C040103. Although if just playing around, then you could also try 0303020404000A0015000A040203 !!

However, also remember that its not important if reducing the timings causes a reduction in max memory overclock, since it is only max performance that actually counts !

[balthazor]

I enclosed the wrong bios file sorry

This is the right one “[ORIGINAL AEOLUS FX 5700 LE 256 MB AGP.zip”
I hope there are way’s to improve the performance of this card.

Some additional info:

NIDIA VGA BIOS information:

Title : GeForce FX 5700LE BIOS
Version : 4.36.20.29
BMP version : 5.28
Core clock : 200MHz
Memory clock : 200MHz
Perf. Level 0 : 225MHz/160MHz/1.28V/100%
Perf. Level 1 : 250MHz/160MHz/1.29V/100%
VID bit mask : 00000001b
Voltage level 0: 1.28V, VID 00000000b
Voltage level 1: 1.29V, VID 00000001b

Reference clock: 27.000MHz
Memory clock : 250.714MHz (501.428MHz effective)
Core clock : 450.000MHz
Memory amount : 262144KB
Memory type : DDR (RAM configuration 06)
Memory bus : 128-bit
Hardwired ID : 0347 (ROM strapped to 0343)
Graphics core : NV36 revision A1


Thanks for the reply I understand (xmasmile)

I will try the timings
Wat do you mean by :

quote-
However, also remember that its not important if reducing the timings causes a reduction in max memory overclock, since it is only max performance that actually counts - end quote

[jimmor]

Wat do you mean by :

quote-
However, also remember that its not important if reducing the timings causes a reduction in max memory overclock, since it is only max performance that actually counts - end quote

Experience has shown that lots of people out there think that the purpose of tweaking the timings is to increase max memory overclock, in the mistaken belief that higher always means better.

A side effect of making certain timings more relaxed (larger) is that you will generally be able to overclock higher. Many people therefore see this as their answer to better performance, little realising that any increase in overclock is usually never enough to overcome the extra losses incurred from using the larger timing values.

On the other hand, although using more aggressive (smaller) timing values can definately improve performance, reducing the timings too far can also have an appreciable impact on just how high you can overclock. So the trick is usually to find the best trade-off (sweet spot) between aggressive timings and overclock for your particular card.

And therfore as previously posted, you should probably first try with the more aggressive values you entered backwards in your original linked bios. But this time enter the string into NiBiTor exactly as 03030205050009000F000C040103 !!

Then as a comparison you could also try the other suggested aggressive timing string of 0303020404000A0015000A040203 ?

[balthazor]

Hay Jimmor,

When I tried tightening the timings to 0101010101000A0011000A010101 (xmastong), it raises the frame rate a few fps .

Do this timings work as the timings of normal ram [ like 2 is faster then 5]

10x


PS Thanks for the reply(xmassign)

[jimmor]

Hay Jimmor,

When I tried tightening the timings to 0101010101000A0011000A010101 (xmastong), it raises the frame rate a few fps .

Do this timings work as the timings of normal ram [ like 2 is faster then 5]

10x


PS Thanks for the reply(xmassign)

Yes, your cards memory timings work just the same as for normal ram where 2 is faster than 5 !

And just for interest, does your chosen extremely aggressive timings have much of a hit on max memory overclock ??

[balthazor]

Hey.

I tried different configuration for ram timings. I settled for 03030205050009000F000C040103 2D /3d and throttle 312F3231B0E210B000000001656C.
The 0101010101000A0011000A010101 setting gave a higher 3d score but ran choppy I left the settings for 2D and 3D the same [2D 300/500; 3D 477/500] during testing. I mainly looked for a fps gain




3DMark Score 3215 3DMarks Timing 03030205050009000F000C040103
GT1 - Wings of Fury 121,8 FPS Game Tests
GT2 - Battle of Proxycon 20,6 FPS Game Tests
GT3 - Troll's Lair 17,1 FPS Game Tests
GT4 - Mother Nature 19,6 FPS Game Tests
CPU Score 621 CPUMarks CPU Tests
CPU Test 1 69,0 FPS CPU Tests
CPU Test 2 11,1 FPS CPU Tests
Fill Rate (Single-Texturing) 749,3 MTexels/s Feature Tests
Fill Rate (Multi-Texturing) 1356,0 MTexels/s Feature Tests
Vertex Shader 15,2 FPS Feature Tests
Pixel Shader 2.0 26,4 FPS Feature Tests
Ragtroll 11,9 FPS Feature Tests
No sounds 40,1 FPS Sound Tests
24 sounds 31,7 FPS Sound Tests
60 sounds 0,0 FPS Not supported Sound Tests

I also tried a different driver [81.95 to 66, 70] because it gave a little extra in performance.
The last thing I have tried is changing the agp bus speeds. I have an nforce 2 mobo with locked agp bus speeds. But with clockgen I altered them to 75 [normal 66 MHz] and 82. Overall 75 gave the best result
I forgot I also tried lowering the voltage from 1, 29 to 1, 28/1, 27 but I stick to 1, 29 volt

It was a busy weekend LoL

[balthazor]

Hi

I did a clean install of windows xp and tested different settings. The best I think was sequence 03030205050009000F000C040103 - 31.616 Aqua marks AGP 66.67

Settings:
DisplayWidth 1024
DisplayHeight 768
DisplayDepth 32
AntialiasingMode 1
AntialiasingQuality 4
AnisotropicFiltering 8
DetailLevel 5
EnableSound 0

Duration 172.6
DurationHighPrecision 172.6
NumOfFrames 5207
AvgFPS 301.616
MinFPS 71.492
MaxFPS 946.215
AvgFPS GFX 378.749
AvgFPS CPU 1.481.230
AvgTrianglesPerSecond 9079597
MinTrianglesPerSecond 1177846
MaxTrianglesPerSecond 19160210
AvgShadedPixelPerSecond n/a

I used 66.70 Force ware drivers
How does a sequence [03030205050009000F000C040103]
Translate in CAS tras etc.

It is very frustrating when doing multiple benchmarks after one and other sometimes with the same setting and the outcome varies. So I think I use this sequence 03030205050009000F000C040103.

Thanks for all the coaching

an nearly 100 % OC is not that bad

[alf_igel]

Hey folks, sorry I dug up this old thread but I need your help.

So I wanted to mod my 5900XT into a Quadro FX 3000 by flashing the FX 3000 bios (I know there's a soft-mod using rivatuner, but I still want that bios flash).
I used the nvidia.qfx3000.256mb.rev.02.rom (+ a couple of others) and the command "nvflash -4 -5 -6 new.rom" to flash my card. Everything went fine, I rebootet my system and a new graphics adapter was found. But instead of a qfx 3000, windows found a 5900 Ultra.

What did I wrong?

Another strange thing is: When I load my currently flashed bios into NiBiTor, everything is displayed correctly.

http://img9.imageshack.us/img9/9087/quadro.th.png (http://img9.imageshack.us/my.php?image=quadro.png)