Basic FX59xxx bios changes

[Daneel Olivaw]

[...]

[Unwinder]

Originally Posted by jimmor

Unwinder,
Now that I know, I could of course disagree or argue against the way Rivatuner does Low Power 3D testing, but it is your program, and therefore your right to do what you want with it.

If you can offer more logical (and the most important implementable) way of testing LP3D clocks - I'll gladly implement it. The only thing I see for improving LP3D clock testing is an ability to compare 2D/LP3D/3D voltages and choose either 2D or 3D mode testing if LP3D voltage matches with one of them. But it can only help 5600/5700 owners or ppl with modded BIOSes and it still will not help the owners of display adapters having LP3D voltage, wchich doesn't match either with 2D or with 3D voltage.

[jimmor]

Originally Posted by Unwinder

If you can offer more logical (and the most important implementable) way of testing LP3D clocks - I'll gladly implement it. The only thing I see for improving LP3D clock testing is an ability to compare 2D/LP3D/3D voltages and choose either 2D or 3D mode testing if LP3D voltage matches with one of them. But it can only help 5600/5700 owners or ppl with modded BIOSes and it still will not help the owners of display adapters having LP3D voltage, wchich doesn't match either with 2D or with 3D voltage.

Like it or not, there are very many people out there with modded bios's, who have changed throttling volts/speed to a higher level in order to reduce the effect of throttling back during stress of gameplay etc. And as it happens, it was actually one of those people who raised the problem of Rivatuner seemingly not doing a valid test of LP3D.

Also, as I stated in my guide to the layman, some 5700u manufacturers had supplied cards with throttling to same values as 3D; presumably for similar reasons now being used for 59xx's.

So as a first and very good stab at the problem; your comment "The only thing I see for improving LP3D clock testing is an ability to compare 2D/LP3D/3D voltages and choose either 2D or 3D mode testing if LP3D voltage matches with one of them", sounds like a much better way to go. However, since it is unlikely that people will set throttling at 2D volts/speeds, I would consider trimming your suggestion back to only checking for situations where throttling and 3D are same.

Nothing of course can properly be done for people who leave throttling at original defaults. But doesn't matter, since it is very unlikely any of them will ever want to test the limits of LP3D ?

[Daneel Olivaw]

________
this is of great interest to 59xx users still.
As always in this thread:
READ FIRST POST
I'll not let any1 forget this thread.
THE most important legacy thread by far.

[soulfly]

i currently have a pny5900se oc to 450/800 very stable, with powerstrip v3.54. do you think that i would be able to oc any higher with a modded bios from jimmor? if so what ones would you suggest? any advise would greatly be appreciated! thanks



thanks for sharing your knowledge!

[jimmor]

Originally Posted by soulfly

i currently have a pny5900se oc to 450/800 very stable, with powerstrip v3.54. do you think that i would be able to oc any higher with a modded bios from jimmor? if so what ones would you suggest? any advise would greatly be appreciated! thanks

thanks for sharing your knowledge!

Welcome to the nV News Forum


Since you can overclock stable at 450/800 with your original bios, there is no reason why you shouldn't be just as stable at min of 450/850 levels with a modded bios !

So why not try some from link below; and it doesn't matter which manufacturers version you use,

http://forums.guru3d.com/showthread.php?threadid=85394


Of course you could just mod your original bios iaw my 59xx bios modding guide at first posting -------> which is after all, the main purpose of this thread !!!

[soulfly]

thanks jimmor! i just wasnt sure if it would be better to mod my original bios or try flashing with one of your modded bios at guru3d




thanks for the help

[jimmor]

Originally Posted by soulfly

thanks jimmor! i just wasnt sure if it would be better to mod my original bios or try flashing with one of your modded bios at guru3d

I would recommend modding an original bios if there is a specific function like TV-out that must be retained, otherwise I believe it is better to checkout potential benefits from flashing with any of the modded bios's !!


And the modded leadtek a350xt at 300/450/850 is probably a good place to start,


http://home.arcor.de/redinferno/5900...D/A350XTDH.zip

[Daneel Olivaw]

________
this is of great interest to 59xx users still.
As always in this thread:
READ FIRST POST
I'll not let any1 forget this thread.
THE most important legacy thread by far.

[jimmor]

The following could turn out to be useful information for anybody with FX5800-5950u cards which are still using their original stock cooling solution.

My Aopen 5900nu, which is flashed with a modded MSI 5950u bios set at 300/530/970, has occasionally run the GPU at temps of 90C. So I decided maybe it was time to look for a better heatsinking solution than its current large black Nvidia reference design.

Although I had heard some indifferent reports on the real effectiveness of using a Arctic Cooling NV4 VGA Silencer, I decided to give it a try since currently at £16.39p from overclock.co.uk, I wasn't risking much.

It should be noted from the start however that this particular version of Arctic Cooler is for use with either FX58xxx or FX59xxx cards, and not FX5900XT types of card.

I wont go into the ramifications of fitting the Arctic NV4 cooler other than to say it was very straight forward. The only point probably worthy of a mention was the need to use a hairdrier to soften a thermal pad that had been used between the original heatsink and the GPU chip. And to ensure good thermal contact, I used the excellent and electrically safe Arctic Silver Ceramique between the new NV4 heatsink and GPU chip; instead of the supplied unknown thermal compound.

As a very simple but reasonably stressfull exercise, I ran 3 of the first game tests in 3dMark03. To help maintain a continuous temperature stress, and to ensure that temps had peaked out at their maximums, I repeated each test a number of times in succession. And I used Rivatuner's Hardware monitor to show all running temps throughout.

Below is listed the max Ambient and GPU temp's for each 3dMark03 test; covering both original heatsinking and the new Arctic NV4 silencer. The figures should however be taken more as intermittant peaks rather than continuous running temps. Although in general it could be said that the max average temps were no more than a few degrees below the noted peak figures.


1. Original Heatsinking fitted.

Battle of Proxycon, Ambient temp 36C, GPU temp 83C
Mother Nature, Ambient temp 36C, GPU temp 79C.
Troll's Lair, Ambient temp 37C, GPU temp 81C.

2. Arctic NV4 silencer fitted.

Battle of Proxycon, Ambient temp 35C, GPU temp 68C
Mother Nature, Ambient temp 35C, GPU temp 63C.
Troll's Lair, Ambient temp 36C, GPU temp 65C.


So for me it appears to be £16.39 well spent, since the NV4 works much better than the cards original heatsinking, with 15-16C reduction in max gpu temp. And not only does the NV4 provide better GPU cooling, but it has also demonstrated to me that it can achieve it at a much reduced noise level.

Something else to keep in mind if thinking of also going with a Arctic NV4 Cooler is that maybe a 20C reduction could have been obtained if the NV4 fan was connected directly to the 12v mollex supply line rather than the 5900 fan connector; which only gives max of 9v in 3D mode.

Take note however, the decision on whether to connect the fan to the card or directly to a 12v supply must be taken before fitting the NV4 assembly, since once fitted there is virtually no access to the card's normal fan connector.

[jimmor]

Since my Aopen 5900nu card was running very stable at 530/970 with its new Arctic Cooling NV4 (5950u) Silencer, I was determined to see if it could run stable at 550/970 ?

To help improve the speed/temp stability, I also decided to do a GPU voltmod to the card. And for the gpu voltmod, I used the method shown in the following linked thread,

http://www.xtremesystems.org/forums...?threadid=18766

Since I was only looking for improved stability, without adding more heat than necessary, I limited GPU volts in 3D to max of 1.6. And this only involved soldering a 39k resistor accross the points identified in the above linked thread.

Also in this instance, I decided to use a modded variant (with stocks at 300/550/970 plus slight timing changes) of the excellent performance BFG5950u bios from the first post of the "Moded 5900XT Bioses" thread at Guru3d.

http://forums.guru3d.com/showthread....threadid=85394


The following were the test results at 550/970,

Driver, 66.81 :-
Aquamark3, 55347
Codecreatures1, 67.3
3Dmark05 v120, 1482
3Dmark03 v360, 8077
3Dmark2001se v330, 20709

GPU temps during the tests, peaked around 74C, which is about 5-6C above previous configuration when running at 530/970 and no card voltmod.

Obviously since I only increased the GPU clock by 20MHz, it is not surprising that there are no great increases in scores.

However, one nice achievement is that my Aopen FX5950nu has now breached the 8000 barrier in 3dMark03 !!!

[Daneel Olivaw]