Get your X1950Pro modded BIOS here

[Ketxxx]

Nope, those screens and comments were from a few other people that requested the BIOS, and they verified the FPS increase. As to why your BIOS reports 256MB.. absolutely no idea when its come off a 512MB card. As long as your card reports all 512MB I wouldnt worry about it.

 

[dracosc]

Any 1950Pro BIOS should work on a GT as their both the same card, the pro is just clocked higher. As long as you get a BIOS made for the amount of memory on your card you should be ok. For details on how to flash see my post on P70.

Yes bu if i flash it with a pro bios, windows will recognise my card as pro. I need to be recognised as GT please. I am sending you my current "modded" bios and if you can add the voltage options please answer me!
Nice job of yours by the way with bioses!

 

[thegave]

Nope, those screens and comments were from a few other people that requested the BIOS, and they verified the FPS increase. As to why your BIOS reports 256MB.. absolutely no idea when its come off a 512MB card. As long as your card reports all 512MB I wouldnt worry about it.

I was referring to this post

http://forums.techpowerup.com/showpost.php?p=523249&postcount=1702

Heres a screen of the edited memory timings. As far as I'm aware these timings should work on any BGA memory, weather its Samsung, Hynix, Quimonda, etc. There should never be any doubt with a 1950Pro anymore Kinda shocked ppl didnt think another 10FPS with AA was possible.

Since I have Quimonda memory I was wondering if those timings would work on my card? Probably not though since I have my memory up at 837 now...

 

[Ketxxx]

They should do, though I would reduce your memory OC before flashing the modded timings.

 

[Bytor]

Ketxxx,

Those are Sapphire cards. They use blue PCB's on there cards.

 

[thegave]

I wouldn't call that blue.. More like emerald or turquoise... =/

Ketxxx what's the highest (mem) clock you've tried with those timings? Do they work at 800 for you?

I can push my memory up to at around 830 with the bios you gave me.

Is 650 considered a good OC on the core? I can't seem to get past 661...

 

[thegave]

Memory is artifact-scanning at 803 with your timings. having a go with X1950PRO0WNER's bios at the moment... if the vmod works should there be a noticeable temperature difference even at the same speeds? I'm supposedly at 1.5v but maxing at 60'C which was the same as stock

I noticed your Row precharge time is actually set slower than the default bios, is there a reason for that?

 

[MadCow]

Out of the box meaning no volt mod/Bios flash. never went above 35c...

So you sinked the caps and chokes, but not the MOSFETs? Watch out man, unsinked MOSFETs almost killed my card at 1.50v. It was all fine and dandy until one day it wouldn't POST at all, I took off the vmod and it POSTed, but I couldn't overclock without crashing.

Anyway, if I post my Sapphire GT-based PCB BIOS, will you be able to change the memory timings? I can't do it through RaBit or ATITool.

 

[thegave]

what temps were you getting at 1.5v?

I modded my bios to 1.5v but I'm maxing out at 59'C with the AC Accelero X2.. does that mean the vgpu probably hasn't changed?

 

[Ketxxx]

I wouldn't call that blue.. More like emerald or turquoise... =/

Ketxxx what's the highest (mem) clock you've tried with those timings? Do they work at 800 for you?

I can push my memory up to at around 830 with the bios you gave me.

Is 650 considered a good OC on the core? I can't seem to get past 661...

I tested to 803 when I done a pencil mod to my card. Any GPU speed over 600MHz is a good clock really as the GPU on a Pro is just an OCd GPU of a GT.

Memory is artifact-scanning at 803 with your timings. having a go with X1950PRO0WNER's bios at the moment... if the vmod works should there be a noticeable temperature difference even at the same speeds? I'm supposedly at 1.5v but maxing at 60'C which was the same as stock

I noticed your Row precharge time is actually set slower than the default bios, is there a reason for that?

The timings are tighter than stock, and offer far better performance at little memory OC cost. As a result thats why you cant OC as far, because the memory timings are tighter. The tradeoff is people can OC roughly around the same point, while at the same time gain 10FPS when AA and AF are used.

For the love of god for cying out fucking loud will a mod sticky this damn thread ppl want my help still and I cant do that effectively with it always dropping to god knows where. Its more than sticky worthy, over 110,000 views, over 70 pages long and almost 1800 posts. God theres actual stickies in this forum with way lower stats than that.

 

[Scrizz]

this should be stickied!

 

[LuxZg]

Agree! [bump ]

 

[Mussels]

+1 for sticky - after it is, delete ketxxx's giant red post tho

 

[LuxZg]

Accelero S1 is in the house I'll try some stuff this weekend, with timings and all, so we'll see where can this baby go with proper cooling

 

[Mussels]

Accelero S1 is in the house I'll try some stuff this weekend, with timings and all, so we'll see where can this baby go with proper cooling

with the side fan you have, you're about to get shocked at how well passive cooling can do!

 

[LuxZg]

Hehe, I hope so
I've got 120mm fan, and turbo module as well, so I'll test all three setups
Just waiting for my digital camera battery to fill, as I want to take photos of everything, which I'll post later on.. And that will take some time

I'll post all temps, and overclocking limits during this weekend hopefully

EDIT: Tubo module is from a friend, as I've bought 2 accelero's (1 for both of us), and a turbo for him.. So i'll test a bit to see what is the best way, and if turbo module is much better (silent & cool) than single 120mm fan I'll buy one myself..

 

[Mussels]

Hehe, I hope so
I've got 120mm fan, and turbo module as well, so I'll test all three setups
Just waiting for my digital camera battery to fill, as I want to take photos of everything, which I'll post later on.. And that will take some time

I'll post all temps, and overclocking limits during this weekend hopefully

i dont think you'll even need them... the coolers great with only light airflow.

Will be interested in the results however, so post away!

 

[LuxZg]

i dont think you'll even need them... the coolers great with only light airflow.

Will be interested in the results however, so post away!

That's true, by the reviews I've read it really needs just a bit of airflow..
But for testing purposes, I'll try to test all three setups, and with side fan turned on and off.. So people without side fans can compare a bit as well Deal?

Any other wishes? I'll maybe test with dual 120mm fans as well, just don't have second one with me at the moment, so that will come a bit later..
But I can test several different scenarios, like 120mm fan on top, or placed under 90° so it blows air from front of the card towards back of the case (and "exhaust" grills on the back)

If you've got any other suggestions I'm here to try it
And than we'll see what Kexx's BIOS changes can do, since this card clocks quite well even when cooled with stock cooler (which was 2C from overheating even on stock clocks )

 

[Mussels]

That's true, by the reviews I've read it really needs just a bit of airflow..
But for testing purposes, I'll try to test all three setups, and with side fan turned on and off.. So people without side fans can compare a bit as well Deal?

Any other wishes? I'll maybe test with dual 120mm fans as well, just don't have second one with me at the moment, so that will come a bit later..
But I can test several different scenarios, like 120mm fan on top, or placed under 90° so it blows air from front of the card towards back of the case (and "exhaust" grills on the back)

If you've got any other suggestions I'm here to try it
And than we'll see what Kexx's BIOS changes can do, since this card clocks quite well even when cooled with stock cooler (which was 2C from overheating even on stock clocks )

i've got one in my mini system, with just the 120mm intake from the case blowing in the back of it, i got 55C load in ~30C ambient on an 8800GT. I never even tested more... a friend with a less convenient intake uses a single 80mm fan on his GTS (G92) and gets the exact same 55C load. (both of us around 70C with no fan)

We know it cools a G92 core great, now its time to see how it fares on an x1950!

 

[LuxZg]

Just a quick answer; passive, just side fan pushing the air. on stock clocks (573.75/688.50) for now:

S1
36 gpu/40 ambient (idle)
<52 gpu/48 ambient (load; 40min of stalker non-stop, without loading levels or anything)

stock cooling was
47gpu/44 ambient (idle)
72gpu/52 ambient (load)

More tomorrow and next few days

Btw, should I put something on backside, I think that's PWM as well.. stock cooler was on that part but now I've got nothing there.. will post pics later if I find some of that part. (found them, below; I've marked them on pics, not all were covered so I hope it doesn't have to be cooled, works so far.. )
;

EDIT: Overslept a bit today, but I've tested my previous max OC 661,5/762,75 which was unstable in few "heavy" games, and now it seems 100% stable, and temps are 55,6C GPU, and 49C ambient after 20 minutes of running only firefly forest (3DMark06). Just for the record, on default clocks I got 0,5 higher GPU temp, and same ambient like I had yesterday playing stalker for 40 minutes.. so I guess some 20 loops of 3DMark fireflies will be an easier way to check the temps, specialy since it's best artifact-checker for memory stability that I've yet found
EDIT2: I've tested with all I've had, and 100% stable is 648/742.5; temps are always <54C for core, and 48C for ambient. Now that I've got stable overclock on passive I'll see what are temps with other setups, with the same clocks. Once I pick my setup I'll flash Kexx's memory timings, and see what memory clocks can I get with those.. It will take some time, but I'll get back to you ppl
Btw, is there some better ATITool than my 0.27 (beta something)? Cos in artifact checker it sometimes shows "dots" of corrupted pixels, but it doesn't react on them... so I had to stare at it myself for half an hour, lowering clocks bit by bit when needed :/

EDIT3:
Here are tested temps on my Radeon X1950PRO, 512MB, @ 648MHz core / 742,50MHz mem
Nothing special or scientific, I've just run firefly forest 3DMark06 in 20 loops every time, that pushes it to the max, maybe it could go like 0,5C higher if i'd run it for extra 15 minutes, but I didn't bother, so that's how it was tested.
all temps are with GPU first, and ambient temperature after, logged by Rivatuner, with max shown during those 20 loops (about 20-something minute test). First row is idle, second is max under load.. Idle isn't logged quite under same conditions, as sometimes I'd wait longer to stablize, sometimes less, but I wouldn't say there are much differences.

Stock cooler (only data took from earlier readings in games and other tests; on default clocks)
idle: 47/44
load: 72/52

S1 passive + side 250mm fan ON
35/39
53,5/48

S1 + 120mm fan + side 250mm fan ON
(120mm fan is GlacialTech Silent Blade; 19dB; 37CFM - bought here for <10$)
31/34
46/43

S1 + 120mm fan + side 250mm fan OFF
(I've put side of the case on a side of desk to "plug" the hole a bit, simulating closed case with no side fans/windows)
35/39
48,25/46

S1 + 120mm fan + 250mm ON
(120mm fan is turned for 90° so it pushes air from front of the card towards back of the case)
36,5/40
50/46

S1 + turbo module + 250mm side fan ON
35/38
46/44

Conclusion:
- turbo module and 120mm silent fan are about the same, but I'll give 120mm advantage cos it's cheaper, even quieter, and cools a bit better; not to mention that if you've got fan regulator of some sort you can tweak the speed/noise ratio to your preferences much better than with Turbo's..
- best cooling is with good case airflow (like my 250mm side fan ) and 120mm fan on S1, though even just good case airflow is enough to keep everything much cooler than with stock cooler
- over 35% cooler GPU (both idle & load) and about 20% cooler ambient temps - great!
- with 120mm fan with higer CFM, or 140mm one, or even dual 120mm's you'll have crazy cool graphics card - it's all about how much noise you can stand from all the extra fans (or how much you're willing to spend on a few high-CFM/low-noise fans)

I'm left to test completely passive cooling (no side fan, no fan on cooler itself), but with 120mm fans costing like 5-10$, it would be crazy to keep it with no extra airflow (IMHO)

I'll also try to get Zalmana ZM-F3, as it should have better CFM on 5V than my Glacialtech Silent Blade on 12V (with about same noise levels), and on 12V it's a hurricane So I'll repeat few tests with it on both 5&12V, and probably will try dual 120mm then (my Glacialtech + Zalman on 5V; so it's still pretty silent).

So anyone planning to OC this X1950 card, get yourself an Accelero S1 and one or two 120mm fans

Next step in OCing is flashing tighter memory timings, and fixing best cooling setup in this test (meaning 120mm+250mm, both turned on). Than I'll have to play with clocks a bit as well, to find max stable clocks for those new timings, and will post my results after that
Cya in a day or two!

 

[LuxZg]

[bump] - still no sticky?

 

[LuxZg]

Kexx, few question about those timings..

My original timings were 9/12/8/10/8/10/0x44/0x2C and yours were 9/11/8/11/8/12/0x24/0x29.. so I've tried both yours and lowest of both (meaning 9/11/8/10/8/10/0x24/0x29) but I've noticed that RaBiT won't actually save "memory refresh rate" and it leaves it on 0x44 anyway.. Tried with 2.2.0 and 2.2.1 - same thing. It saves it with 0x24, but when I reopen the file it's back to 0x44.. I guess that's part of reason why I'm not seeing any extra improvements..

So, question one is - is there any way to save that "memory refresh rate" to 0x24, and second question is - is it always "lower is better" or not? Meaning, would "lowest of both" be automatically better or it depends?
I've never been much into all those timings and stuff, never could force myself to learn all the things just to tweak memory a bit, but this time I'm interested a lot..

I've attachmed my BIOS (again) so maybe you could tweak those timings if you're doing it with RaBiT as well.. Maybe it will work for you, cos' it just doesn't work for me :/

Thanks!

EDIT: Just downloaded that rom file from page 70, and it has 0x24 Pls, tweak that thing for me so I can try it out! PLS!
Oh, and I forgot - how can I know if my card can be V-moded through BIOS? Voltage table isn't in current BIOS as much as I can see..
Btw, extra info about memory, it's Quimonda HYB18H512321AF-14, and about PCB - you can see it here and in the post above (back part)
;

 

[BlackStar]

A couple of questions

Hey, great thread! Come on, mods, this deserves to be stickied.

I'm planning to mod my X1950Pro's bios to add more granular fan control, because there are no applications that can do this on Linux. If I understand correctly, the safest procedure is to dump by card's bios, edit it with Rabit or another tool, then flash it back.

Questions:
- how do I dump the bios? Rabit can't do this under Vista x64 (which I'm currently using), I suppose I should boot to dos and dump it with Atiflash?
- the bios has only one powerstate (575/680) currently. How do I get to add a second 2d mode (e.g. 200/200)? How about three modes (2d/3d/3d OC)? Has anyone done this before?

I understand that this thread is partial to overclocking mods (I *do* overclock of course ), but there are significant power savings to be gained by underclocking (I think about 10-20W, but it's some time since I checked that). Temps, and consequently fan noise, can also be reduced *a lot*, too (I'm running at ~40-45C with 10-20% fan under 2d, which goes up to 60-65C and 60% on 3d/OCed).

 

[Mussels]

atitool can dump the bios. check on this websites homepage.

the cards should have two states, actually. Its just transparent, and changed by the driver. you cant really alter it that much, except to set the 3D setting (i think 2D is driver controlled)

 

[BlackStar]

Thanks, dumped the bios. It seems that ATITool does not support bios dumping on X1950 cards yet, but winflash does. I'll still use a dos-based utility to make sure the dump is 100% correct, but it looks good.

I'm still wondering about 2d/3d modes. You are correct that the driver is responsible for switching, but it seems there's more to it. Some X1950 cards *do* report more than one power state, while others report only one - this can be confirmed by running "aticonfig --power-states" on Linux - which leads me to believe that this information is stored inside the card's bios.

What's more, these extra states are usable not only by the driver, but by other programs too. My MSI card only reports one powerstate, and indeed, ATITool confirms that: it shows only one set of clocks, which stay the same when going from 2d to 3d modes and vice versa.

It would be *very* interesting if someone can upload a vanilla bios here (X1950 Pro 256MB, PCI-E) which does have multiple powerstates, so we can compare and find out if this hypothesis is correct. There are many linux users that would be interested in this finding, as the X1000 series cards tend to overheat and/or show higher power consumption on that OS (the windows drivers seem to have better power regulation).

Edit: Anyone know the offset where the clocks are stored?