Is anybody else experiencing this?

I just went and bought a 2nd GTX460 for PhysX (overkill I know) and I modified its BIOS for clocks of 800/2000. Now, I've cleaned out the drivers thoroughly and then updated the BIOS using the following commands from DOS;

nvflash --eraseeeprom
nvflash -4 -5 -6 <filename>.rom

Rebooted, and installed drivers. Now whenever I open any program that tells me GPU and memory speeds they all say the clockspeeds are still stock (700/1800) :brick: I was wondering does anybody know if its possible the driver is overriding the BIOS clock settings?

Are you sure you DID flash BIOS?
Once I was tired and trying to flash 3-4 times not noticing what nvflash was showing.
Then I used nvflash -- protectoff (IIRC) before flashing and did flash that time. :D

--protectoff
-r
-4 -5 -6 newbios.rom

Yep positive. nvflash reports "update successful". I've tried the protectoff command and that didnt help either. I'll give the -r command a shot and see what happens :up:

Oh, if it matters I'm using nvflash 5.95. Tried 5.93 and 5.91 with the same results. 5.88 won't work because it doesn't recognise the 460.

Have you tried dumping your GPU BIOS and doing a file compare from the dump to your modified BIOS to see if they match? If they match then the flash was definitely successful.

Read bios with gpuz and check it back.

Why use it for physx and not sli?

@ josh, GPU-Z 4.5 doesn't support BIOS dumping on the card for some reason, I have to do it via nvflash. I'll flash then get a copy through nvflash and compare them.
@ pilsy, I can't do SLI on a 890FX, not without a "patch" at least ;) So for now its a PhysX card.

Ok, I've just run nvflash 5.95 from DOS and took the following procedure;

nvflash --protectoff
nvflash --eraseeeprom
nvflash --protectoff (unneccessary perhaps, but done to be sure)
nvflash -4 -5 -6 kfam.rom
---------------------------
nvflash --save backup.rom

I then restarted and have checked both KFAM and backup.rom files with FBE, settings within the BIOSes are the same and their checksums match.

Heres a couple screens I grabbed with my phone, everything looks to be in order. Will install drivers and see what happens.

all you did was modify your existing gpu bios and want to reflash it?

nvflash -j -y NEW.ROM

I'll try that. Just installed the drivers and it STILL says 700/1800 even though the BIOS mod settings are deffinately sticking.

Edit - -j didn't work either. Still 700/1800. Surely this has to be a driver issue? Completely cleaned (including registry) and reinstalled 259.32. Was using my customised 259.12s with the same results. This is bloody annoying. Theres nothing wrong with this card and I can't exactly return it citing modified BIOS clocks don't seem to work. I have successfully OCd this card with afterburner and nvidia inspector, so I know theres nothing physically wrong with the card either. It seems to me like some kind of driver clock lock is applying itself.

I know this is XS but MUST you overclock your PhysX card? :D

Try without erasing it first. Just protectoff and straight to flash.
Maybe, somehow, it restores original when it sees that BIOS was erased, as it seems there are some backup BIOS banks form what I've seen lately.

Quote:

---

Originally Posted by Oj101

I know this is XS but MUST you overclock your PhysX card? :D

---

Yes :D I work on the phillosopy of if you can, then why not :p:

Quote:

---

Originally Posted by donmarkoni

Try without erasing it first. Just protectoff and straight to flash.
Maybe, somehow, it restores original when it sees that BIOS was erased, as it seems there are some backup BIOS banks form what I've seen lately.

---

I'll give it a try. I think I have tried that and it still didn't budge.

What is Everest reading as your bios clocks? It looks like the system seems to have kept your clocks locked into the drivers/registry somewhere? DriverCleaner?

I have cleaned drivers out completely and reinstalled them 3-4 times now. I've just tried flashing the card again and atm clocks are holding. I don't trust it though so I'm going to uninstall drivers, totally clean them out, even completely reset my system BIOS, reflash the card with a few more tweaks I want to make anyway, and see if it still holds the clocks on a new driver install. It is nice seeing 800/2000 finally though in GPU-Z :D

So.... what changed in your method? :)

Cant be the drivers????? They have no MHz info of their own

Quote:

---

Originally Posted by K404

So.... what changed in your method? :)

Cant be the drivers????? They have no MHz info of their own

---

I thought the exact same thing you did. Drivers couldn't be the cause for the frequency change. But then I realized that the drivers do have some control, and they could be changing the speed to the defaults and giving this result. Basically the computer boots up with the new GPU speeds, then when nvidia drivers load the frequency changes back to defaults. To us the GPU speeds never change.

Do you mean the reference MHz for the GPU model or the default for the card (factory OC etc)

Are the reference MHz for each card supported by a driver included in the drivers data set? I've never seen any sign of it?

Indeed. The drivers override the BIOS settings basically. Oddly enough my system has just crashed for whatever reason, and clocks have been reset from BIOS modded 800/2000 back to 700/1800. Certainly looks like driver controlled clocks to me based on that.

Maybe its coz its showing 2d clocks and not 3d clocks ?

In the nvidia driver if you play with the power management settings it can do a lot of funky things like you are reporting ESPECIALLY as you arent running both cards in SLI. So whats happening probably is the phyx card is not in use so it goes to lower clocks. Set it to max performance and reboot and see if it holds the clocks.

use XtremeTuner HD to flash your card via windows
this method is very safe

Is it possible that your OC Utility was setting the clocks @ startup?

I uninstalled afterburner once I verified 800/2000 was fine. I'm using GPU-Z to verify clocks. Gonna test a couple things and will report back.

Have you tried the card in question by it's self, i.e. remove your older card from the equation.

Also what is the make/model of the card, I read somewhere about some makers gtx460 that sports dual bios/flash chips.

And also perhaps the bios structure is different, or its using a different 3d clock domain, and FBE is just not presenting it correctly. I have
nibitor 5.8 (support for 460) if you post your orig bios i'll check it out to see if there is anything strange.

Quote:

---

Originally Posted by Ket

Yes :D I work on the phillosopy of if you can, then why not :p:

---

Because you can't :rofl: I'm sorry :D

Quote:

---

Originally Posted by Ket

@ josh, GPU-Z 4.5 doesn't support BIOS dumping on the card for some reason, I have to do it via nvflash. I'll flash then get a copy through nvflash and compare them.
@ pilsy, I can't do SLI on a 890FX, not without a "patch" at least ;) So for now its a PhysX card.

---

gpuz 0.4.4 can do the bios dumping

Quote:

---

Originally Posted by CrazyNutz

Have you tried the card in question by it's self, i.e. remove your older card from the equation.

Also what is the make/model of the card, I read somewhere about some makers gtx460 that sports dual bios/flash chips.

And also perhaps the bios structure is different, or its using a different 3d clock domain, and FBE is just not presenting it correctly. I have
nibitor 5.8 (support for 460) if you post your orig bios i'll check it out to see if there is anything strange.

---

Cool thanks :up: I bought another KFA card to match my existing one. If there is a dual BIOS.. I guess the only option is to find out how to flash the backup chip as well. I've only got the 1 card installed atm as well.

The only thing left I can think of is maybe the 1 domain FBE doesnt touch listed [GPU default] voltage table screws with the card in some way. Anyway, heres a copy of the original BIOS and the modified BIOS.

Ok I edited your stock bios with NiBitor to 800/1600/2000, and removed the voltage cap.

Also when you dl anything from this forum i think you should use firefox, or chrome, anytime I use IE it corrupts my dl's

I'll give it a shot now :up:

Quote:

---

Originally Posted by CrazyNutz

Ok I edited your stock bios with NiBitor to 800/1600/2000, and removed the voltage cap.

Also when you dl anything from this forum i think you should use firefox, or chrome, anytime I use IE it corrupts my dl's

---

I don't suppose you could do the two 768mb bioses posted recently in the 460 bios thread/ :D

Just tried flashing again and the BIOS is deffinately taking but clocks still report stock of 700/1800. One thing I did notice when erasing the eeprom was something about the card has no preservation table? Must be something new I've never seen that before on other nvidia cards.. Heres a couple screens.

Hmmm??? Weird! Dude are you shure you dont have a bad floppy? I have nvflash 5.95 for windows it works fine, maybe you can try it?

preservation table, I've read about this, I thinks it can be lost when you erase eeprom. There was a german article about unlocking a 465 to a 470, and it mention somthing about clearing the preservation table.

Quote:

---

Originally Posted by lowfat

I don't suppose you could do the two 768mb bioses posted recently in the 460 bios thread/ :D

---

Here you go, removed voltage cap:up:

Quote:

---

Originally Posted by lowfat

I don't suppose you could do the two 768mb bioses posted recently in the 460 bios thread/ :D

---

And almost missed your other one here you go.

Quote:

---

Originally Posted by CrazyNutz

Hmmm??? Weird! Dude are you shure you dont have a bad floppy? I have nvflash 5.95 for windows it works fine, maybe you can try it?

preservation table, I've read about this, I thinks it can be lost when you erase eeprom. There was a german article about unlocking a 465 to a 470, and it mention somthing about clearing the preservation table.

---

I use a bootable USB stick ;) I've also flashed from DOS and the Windows flasher, neither "appear" to work even though the flash goes fine. Maybe I just have to wait for nibitor and a newer nvflash for my card :shrug:

Quote:

---

Originally Posted by CrazyNutz

Here you go, removed voltage cap:up:

---

Quote:

---

Originally Posted by CrazyNutz

And almost missed your other one here you go.

---

Dood. You are awesome. :up:

Quote:

---

Originally Posted by Ket

I use a bootable USB stick ;) I've also flashed from DOS and the Windows flasher, neither "appear" to work even though the flash goes fine. Maybe I just have to wait for nibitor and a newer nvflash for my card :shrug:

---

Man this is freaky. I can't imagine this being a driver issue either, the Nvidia default clocks are't even 700mhz, and I doubt the drivers have any vendor specific logic...

I wander if Nvidia has given vendors a way to keep us from putting modded bios images on these things? With the potential of the 460's hurting there profits on the high end parts...

Have you tried to put your modded bios from your first 460 on the new one?

Or maybe you just have a botched card???

Quote:

---

Originally Posted by lowfat

Dood. You are awesome. :up:

---

No Problem, hope you enjoy. BTW If you guy's donate just $5 to Mavke he will send you NiBiTor 5.8

Its same here with Galaxy GTX 460.
Modified clock and voltages do not stick even though flash goes fine.
And KFA is basically Galaxy. So its brand specific. They have done something to this card that is making this happen.

Quote:

---

Originally Posted by Funky

Its same here with Galaxy GTX 460.
Modified clock and voltages do not stick even though flash goes fine.
And KFA is basically Galaxy. So its brand specific. They have done something to this card that is making this happen.

---

Uh ohh! bastards! I guess we'll have to wait to see if other vendors follow suit

One way to check what is actually going on is to try and flash bios from some other manufacturer (stock bios).
If problem is at flashing step, it should still stick to 700/924 clocks which is default for KFA and Galaxy GC Ed. The only thing stopping me from doing that is the fact that this board uses custom PCB. Dont really know what will happen and at the moment I do not have backup card. Will give it a try after few days when I get my HD3450 back from a friend.

Ok, I tried it. Still nothing.
Whichever bios I flash, speeds remain 700/924 :P

I checked the original bios dump and before and after flash dumps using hex editor.
Bios is definitely getting flashed as new dumps after flashing the card are identical to the bios I flashed.
Still for some reason clocks and voltages are not sticking.
I tried flashing it with stock and modified bios of evga, ASUS, Palit and MSI.

Nothing happens to clocks :@

Quote:

---

Originally Posted by Funky

Ok, I tried it. Still nothing.
Whichever bios I flash, speeds remain 700/924 :P

I checked the original bios dump and before and after flash dumps using hex editor.
Bios is definitely getting flashed as new dumps after flashing the card are identical to the bios I flashed.
Still for some reason clocks and voltages are not sticking.
I tried flashing it with stock and modified bios of evga, ASUS, Palit and MSI.

Nothing happens to clocks :@

---

The clocks must be hard set i.e. some where on the board there may be a section an array of 0ohm resistors (bridges) that override the bios clocks.

@Ket since you have two of theses cards, can you examine the PCB's on the two and look for differences i.e. added/missing resistors (black very small with a 0 on them)

I dont think that is the problem.
I can clearly see two bios chips.
There are two PM25LV512 one each on front and back of the card. And they have done something to the chip on the front. There is huge blob of solder on the chip which is on the top near the pin 0. I will try to take better pic of the same.

Ok as promised, the pics.

Here is the root cause of the problem. The two bios chips. (Sorry for large pics I just wanted to make sure the pics are clear enough )

http://img820.imageshack.us/img820/8589/frontvs.jpg
http://img706.imageshack.us/img706/4392/backap.jpg

Very interesting. I assume this filled in hole is what prevents the bios from sticking? This must be on very late batches of manufacturing after they realized what everyone was doing to previously manufactured cards to get the massive performance increases we are all getting from these little gems.

Or... One is a backup which can't be flashed (and therefore should theoretically never go bad)
in case of emergency? Think Gigabyte's Dual BIOS but with one of the chips write protected.

The PCB is way too complex for you to just superglue a second BIOS chip on somewhere, I
think it was designed like that right from the beginning.

ok dual bios as I suspected erlier.

The top chip with the solder blob on it, looking at the datasheet pin 1 is CE (Chip select/enable). I wonder if the solder is by accedent?
Can you flake it off? For that blob of solder to do anything the would have had to drill down to the pin one terminal.

Ket has a two of these, One is ok the other no go on bios mod, if he looks at his two boards we will know exactly how to bypass this lock.

And here is the datasheet for the serial flash chip (bios)

Ok, so theoretically if we unsolder the BIOS chip on the front of the card it should automatically force the card to use the other BIOS chip, so if you flash it.. VIOLA! As theres no backup chip theres nothing for the card to revert to. Alternatively we need a way to flash the backup chip as well. Ideas anyone?

ED - My other card I haven't flashed either yet. But as its a KFA as well (Galaxy rebrand) then it will have that damn backup BIOS chip as well.

Quote:

---

Originally Posted by Ket

Ok, so theoretically if we unsolder the BIOS chip on the front of the card it should automatically force the card to use the other BIOS chip, so if you flash it.. VIOLA! As theres no backup chip theres nothing for the card to revert to. Alternatively we need a way to flash the backup chip as well. Ideas anyone?

ED - My other card I haven't flashed either yet. But as its a KFA as well (Galaxy rebrand) then it will have that damn backup BIOS chip as well.

---

Thats what i was thinking.
And that solder may be holding the CE pin low (forceing the front bios chip to be enabled)
First thing I would do is touch the tip of your soldering iron to that blob and suck it out with a desoldering tool. Then the internal pullup resitor (if it has them) in the flash chip would pull the CE signal high, and disable that bios chip. Or heat pin1 and lift it off the pad with a needle

I'm all over that. I'll get on it once I've taken a shower. Kinda didn't open my business today because I had new PC parts arriving and I look like a proper bum atm :p:

I took the markings off of the front BIOS chip (blue dot and a white dot), without any effect. What I'm unsure of now is which BIOS chip is the main BIOS chip? Or won't it matter which BIOS chip I snip a leg off of? Oh, and if anyone wants to have their GTX460 always run in PCIE 2.0 mode instead of switching between PCIE 1.1 (desktop) and PCIE 2.0, just take the resistor off shown in this pic;

Do bare in mind though if you take it off your card will always run 3D clocks.

ket dont snip the leg off :) Unless you can repair it.
you dont know witch one is the main bios.

Snipping would be the only option. You cant heat the pad from the other side or anything because the other BIOS chip is right there.

In physx games with two GTX 460s, is it better to have them in SLI mode, or with the second GPU dedicated to Physx?

Quote:

---

Originally Posted by Ket

Snipping would be the only option. You cant heat the pad from the other side or anything because the other BIOS chip is right there.

---

You dont heat from the otherside, you need a 15-25 watt small tip solder iron,
just heat the leg until the solder flows then with the other hand use a sewing needle to pry the leg up.

And if you really want to know whitch one is main bios put you volt meter on the pin 1 leg of the bios on back of card, fire up the machine. If it is pulled low you'll see 0v (or less than 1v) that indicates chip enable (i.e. the bios it is using), if high > 2v, then its not the main bios.

Well my 2nd card has just started acting even more strange so its being sent back.

I have similar issues with two GTX 460 1GB GC Galaxy cards; let me describe the problem

- Two identical cards in SLI. Card #1 on top Card #2 bottom
- Successfully flashed the same mod bios in both cards using the process described here
- nVidia inspector correctly reports the new upper limit voltages 1.212v. (This is the only tweak in the modded bios)

Next is when things became interesting, because after rebooting the computer or crashing the cards with a high overclock the upper limit voltage in the card #1 is back to 1.087v. The only way to bring back the upper limit voltage to 1.212 is by reflashing the bios. Card #2 behaves normally and keeps the modded bios.

I will include here both bios in case somebody is curious enough to compare them

I found very similar with the KFA cards. Only for me reflashing doesn't work. I have to physically remove the card and put it back in, then it'll use the modded clocks.. until I reboot.

Below are the original bios from my cards

Im wondering if somebody can mod each bios to brake the upper limit so I can try them to flash it back into its original card.

Also the min fan RPM to 0 --Tks

I think its safe to say there could be something fundamentally wrong with the engineering design of KFA/Galaxy cards, at least in early batches. If thats not the case then we need a way to disable one of the BIOS chips, or find a way to flash the backup chip as well.

Dudes, these cards have dual bios, Its just supposed to work that way, i.e.
safety net incase you screw up your bios, however it's not very desirable and does seem to be a pain in the arse:) You guys should contact your vendor and get instructions on how to flash the backup bios, there has to be a way. They should be ok with giving out this information.

I mean hell they put dual bios on these cards they did that so you could tweak one and have one to fall back on, they made this card for tweakers/modders they should give you guy's the tool or information.

The cards seem to have a mind of their own, choosing which BIOS to boot with at whim. Hence my comment about there could be something wrong with early batches.

Something else that I would like to add is that after erasing the eeprom in both cards I rebooted the computer, but the video card was able post. This may confirm the presence of a back up bios.

Ket can you please mod my original bios attached above with the min rpm to 0% and the voltage upper limit to 1.212v? Tks

Quote:

---

Originally Posted by Ket

The cards seem to have a mind of their own, choosing which BIOS to boot with at whim. Hence my comment about there could be something wrong with early batches.

---

I agree with what you are saying, It's flaky!

Things that are supposed to help you not screw up, tend to be flaky, or get in the way, just all out nuisance, especially for people that actually know what they are doing.

Yeah, real pain in the arse. If anyone has the proper equipment to try it, I would say try outright removing one of the BIOS chips and see what happens.

Quote:

---

Originally Posted by askimo

Something else that I would like to add is that after erasing the eeprom in both cards I rebooted the computer, but the video card was able post. This may confirm the presence of a back up bios.

Ket can you please mod my original bios attached above with the min rpm to 0% and the voltage upper limit to 1.212v? Tks

---

Clicky

Quote:

---

Originally Posted by bhavv

In physx games with two GTX 460s, is it better to have them in SLI mode, or with the second GPU dedicated to Physx?

---

This might be game specific just try both and see what works best :shrug:

Quote:

---

Originally Posted by Zloyd

This might be game specific just try both and see what works best :shrug:

---

I dont have any physx games though, except for Sacred 2, and they recommend not use physx in that because it is too buggy.

The weird part is, we know the bios is sticking to at least one chip. Once you flash the bios and read it again via GPUZ, it has same checksum and same voltages as the modified one which I flashed. So bios is definitely sticking.
Results are same with even when I use xtremetuner HD for flashing as well which is tool made by galaxy themselves.

I dont have any problems like others with this card though. No pci express 2.0 bus issues what so ever, it clocks nicely, remains cool. If only I could unlock the voltages successfully. Seems like very nice GPU as it is 100% stable at 920/2225 in vantage, Kombuster as well as games with max temp of 63°C (sure fan is at 100% at these clocks).

As I said earlier, somebody needs to try removing one of the BIOS chips from the PCB completely, that way the board will have no choice but to use the one BIOS remaining.

after trying to flash twice and it saying successful still no modded bios for me either.so i grew a pair and yanked the bios chip off the back leaving the one on front with the solder,or whatever that is on it.restarted and presto your modded bios was there waiting when rebooted.but sadly it wasn't stable and this galaxy looks to be a poor overclocker.so i reverted back to stock bios for now.any chance you could mod one that has gpu@800 and remove the volt cap but leave everything else at stock?pls pls pls .it was either the 1.1v or the memory @4005 that kept making the driver crash i think but i'm not sure which.

hehe nevermind about modding a bios i think i did it.thx for the thread and the one with all the modded bioses,much appreciated.

Success?

as far as being able to flash the bios and it stick yes.but this card is not a high o'clocker at all.about 860 is as high as i can get it.anything over 1.05v and it just wont run.