NVIDIA GeForce GTX Titan User-modded BIOS

sk3tch

2[H]4U
Joined
Sep 5, 2008
Messages
3,389
There's a big thread over at Overclock.net (link) - I've summarized everything (including file download links and links to original posts/threads) here in case you just wanna get up and go...of course, USE AT YOUR OWN RISK and YMMV...

EDIT: these are the “Naennon” and “maarten12100″ user mods for those that are already familiar with the situation...
 
Last edited:
I don't recommend doing this. Chances are this issue will be fixed in either a future driver or partner released BIOS update.
 
Well so far no one has bricked their $1000 card :) and several have felt brave enough to flash their Bios and gain stable 1200 clocks.

I'm a OCN regular for years and have only joined here in recent days though I have lurked here for years. Looking forward to grabbing a Titan or two if things stabilize with the clocks/voltage in both 2D and 3D profile wise in the Bios.
 
Do the 1200 clocks push ahead of the 690?

Not quite. Using this bios I was able to reach 1215/3700 without any throttling. I scored 78.5fps in Valley at 1080p. I didn't do much more because even at 85% max fan speed, my card was passing 80C. So this bios is great because it seems to works nearly perfectly, but the card gets hot and loud. My water block will be here later this week, and then I can have some more fun.

GTX 690 scores I've found range from 80-90fps depending on stock or overclock. So if I can get a little bit more out the Titan after I put it under water, I'll be within GTX 690 territory.

I'm quite happy now with the card!
 
but the card gets hot and loud

I actually reverted my vmodded evga 680 sc sig to "stock" (plain jane non-oc) rom values after I looked at the Kill-A-Watt readings last night. Not only did it drop a non-trivial amount of wattage, it's much, much quieter too.

I don't think I would vmod my own Titan's bios presently.
 
Not quite. Using this bios I was able to reach 1215/3700 without any throttling. I scored 78.5fps in Valley at 1080p. I didn't do much more because even at 85% max fan speed, my card was passing 80C. So this bios is great because it seems to works nearly perfectly, but the card gets hot and loud. My water block will be here later this week, and then I can have some more fun.

GTX 690 scores I've found range from 80-90fps depending on stock or overclock. So if I can get a little bit more out the Titan after I put it under water, I'll be within GTX 690 territory.

I'm quite happy now with the card!

Interesting, I figured around 1200 would get you there. Post your clocks under water.
 
Interesting, I figured around 1200 would get you there. Post your clocks under water.

Same, but Titan vs GTX 690.

2688 vs 3072 CUDA cores = 14% less
48 vs 64 ROPs = 33% less

Leads me to believe Titan may need 20% on the clock to match at GTX 690. And while Titan has more memory bandwidth and VRAM, neither matters at 1080p. Maybe the gap starts to close you go to higher resolutions (kind of like how the 7970 pulls away from the GTX 680 starting from 1440p onward) but I haven't really been able to find any results of GTX 670 SLI or GTX 690 to compare to.

I don't think I would vmod my own Titan's bios presently.

I was just sick of the throttle and wanted to find out what my card was capable of. The more I overclocked my memory, the more my core throttled. But at least now I have a very good idea of what my card can do once this issue is fixed (officially).
 
I don't recommend doing this. Chances are this issue will be fixed in either a future driver or partner released BIOS update.
BIOS update ?
Has there ever been official BIOS update for video cards either from Nvidia, AMD/ATi and/or their partners?
I know BIOS updates for motherboard. But I've never seen BIOS update for video cards...
 
Same, but Titan vs GTX 690.

2688 vs 3072 CUDA cores = 14% less
48 vs 64 ROPs = 33% less

Leads me to believe Titan may need 20% on the clock to match at GTX 690. And while Titan has more memory bandwidth and VRAM, neither matters at 1080p. Maybe the gap starts to close you go to higher resolutions (kind of like how the 7970 pulls away from the GTX 680 starting from 1440p onward) but I haven't really been able to find any results of GTX 670 SLI or GTX 690 to compare to.



I was just sick of the throttle and wanted to find out what my card was capable of. The more I overclocked my memory, the more my core throttled. But at least now I have a very good idea of what my card can do once this issue is fixed (officially).
if you are going to double some specs then double them all which would mean the 690 actually has more memory bandwidth. 384 GB/sec for 690 and 288 for Titan.
 
BIOS update ?
Has there ever been official BIOS update for video cards either from Nvidia, AMD/ATi and/or their partners?
I know BIOS updates for motherboard. But I've never seen BIOS update for video cards...

EVGA has released official firmware updates before to unlock fan speeds on 580's, but that wasn't straight from nvidia, iirc.
 
Thanks, OP! Your summary rocks and I used the first BIOS you posted.

With this BIOS I'm getting rock-solid 1202 MHz on my core. I can finally OC the memory, too. Turns out the memory overclocking is taking tons of power. Probably not worth it, but it's there now for bandwidth-heavy games.
 
Last edited:
I myself get no throttling. My Titan is not overclocked. Power is set to 106%. Temp set to 90. I have it set to prioritize temps rather than power.
 
Anyone who has flashed these. Did you get mismatches? I didn't follow through as my titans are already under water and I really don't want to kill a card and end up taking the whole loop apart again.
 
Flashing a BIOS is easy and takes 5 minutes. The only worry is a power outage, really.
 
Flashing a BIOS is easy and takes 5 minutes. The only worry is a power outage, really.

I've flashed bios before, but the 1.25v rom threw up about every mismatch possible. Which made me hesitate.
 
I used this on one of my Asus Titan's. Proceed with extreme caution. When using this I was stuck at maximum voltage the entire time, even at an idle desktop.

That is pretty much insane IMO and I had to go back to the normal BIOS. Don't like idling at 85C on the desktop.....
 
Last edited:
You really gotta hate throttling to flash a user-made BIOS onto a thousand dollar card.
 
I used this on one of my Asus Titan's. Proceed with extreme caution. When using this I was stuck at maximum voltage the entire time, even at an idle desktop.

That is pretty much insane IMO and I had to go back to the normal BIOS. Don't like idling at 85C on the desktop.....

You should try Naennon bios, should drop your voltage to 0.875v at desktop 2D as long as your card its in idle state under 620mhz
 
I used this on one of my Asus Titan's. Proceed with extreme caution. When using this I was stuck at maximum voltage the entire time, even at an idle desktop.

That is pretty much insane IMO and I had to go back to the normal BIOS. Don't like idling at 85C on the desktop.....

Do you have multiple monitors?
 
Do you have multiple monitors?

I have dual monitors. With dual/triple monitors power states on the GPU flip a lot more frequently, but - The GPUs normally still go in idle states when all applications are closed and/or idling even with multi monitor.

I didn't get any vdroop ever with the BIOS listed in this thread, I was 1.25V 24/7. Since flashing back I can idle at idle state speeds now. Is there another BIOS that is better in this respect? I can't go full voltage like that 24/7...
 
I don't know, but that's the danger with a modded BIOS--whoever hacked it up clearly didn't test with a multi monitor setup. What other contingencies are untested? You're dicking around with a $1000 piece of hardware and trusting some random stranger on the Internet to not brick or otherwise cause permanent damage to it. More power to you if you're cool with that, but that's way above my risk tolerance threshold. I'm really waiting on an official fix for Titan before I consider a purchase.
 
I'm using multi-monitors (3X 30'') and my voltage goes down to 0.875v no problems at desktop/2D, using a modified marteen-naennon BIOS.

No reason to fear that one. No contingencies. 3X 30''. Working fine.
 
I flashed both my Titans and now I get rock solid clocks. So far just sitting at 1163mhz pegged the entire run of Valley.
 
I don't know, but that's the danger with a modded BIOS--whoever hacked it up clearly didn't test with a multi monitor setup. What other contingencies are untested? You're dicking around with a $1000 piece of hardware and trusting some random stranger on the Internet to not brick or otherwise cause permanent damage to it. More power to you if you're cool with that, but that's way above my risk tolerance threshold. I'm really waiting on an official fix for Titan before I consider a purchase.

Nvidia just released a statement saying that the throttling is intended.

There will be no fix. The bug was that the power % was misreported, it was being reported lower than actual power. Some people were confused because they were being throttled at 90% power - apparently the bug causes cards to report a lower than actual power percentage.

But the throttling is apparently working as intended, so no fix.
 
If that's the case, then isn't it reasonable to assume that Nvidia's engineers implemented the throttling for a reason?

Forgive my ignorance... I've seen conflicting information on this. Isn't it true that the throttling only affects boost clocks, so the cards don't throttle below their base clock speed? Is the point of contention that people want to maintain full boost speed 100% of the time, or are the cards actually throttling below their intended minimum speed?

If it's the case that the throttling only affects boost clocks and the card really is working as intended, I'm inclined to trust Nvidia's judgement for implementing this as they have more knowledge of the power delivery and thermal specifications of their hardware. Is it possible that there are other factors at play that aren't exposed to the user, such as VRM temperatures or power rippling? Flashing a BIOS to force the card to run above its safe tolerances may be [H] but it may be a good way to waste $1000 over a period of months or a year.

If the card is actually failing to maintain it's minimum advertised clock speeds, then it's just broken.

Sorry, I really just want to know what's going on. There's so much conflicting info flying around out there right now on this issue. I'm only interested in purchasing if the hardware is actually working as intended without significant firmware modification.
 
Basically, the crux of the matter is that nvidia is prioritizing efficiency over performance. Titan certainly could draw more than 265 watts from the pci express slot, WAY more, but 265 watts is the hard limit. Once you hit 265, you're done and you will start to throttle, and apparently that is intended behavior.
 
But what I want to know is whether the power delivery circuitry on the card would be overtaxed by drawing more than 265 Watts. Similar to how pushing an original i5 / i7 chip over 1.3625v would cause electromigration and eventual failure (I wore out an i7 860 by running "exactly" 1.3625v for a year). Just because it's theoretically possible for the Titan to draw more power doesn't mean the circuitry on the card is able to safely handle it ... ?

Or is Nvidia really just being bitches about this whole thing?

It just doesn't make sense to sell a $1000 video card and then gimp it in this way unless there is a really solid reason for it.
 
Basically, the crux of the matter is that nvidia is prioritizing efficiency over performance. Titan certainly could draw more than 265 watts from the pci express slot, WAY more, but 265 watts is the hard limit. Once you hit 265, you're done and you will start to throttle, and apparently that is intended behavior.

On my todo list is to use my Kill-A-Watt before and after the vmod just for my own curiosity, but not this week, maybe early next week. I'm just curious how much extra people are approximately burning in their Titans with the 1.21v hack

I'm guessing that if someone enabled full DP in the nv control panel with 1.21 @ 1202 the card would probably go poof.
 
But what I want to know is whether the power delivery circuitry on the card would be overtaxed by drawing more than 265 Watts. Similar to how pushing an original i5 / i7 chip over 1.3625v would cause electromigration and eventual failure (I wore out an i7 860 by running "exactly" 1.3625v for a year). Just because it's theoretically possible for the Titan to draw more power doesn't mean the circuitry on the card is able to safely handle it ... ?

Or is Nvidia really just being bitches about this whole thing?

It just doesn't make sense to sell a $1000 video card and then gimp it in this way unless there is a really solid reason for it.

How did you manage to fail a i7 this easily?

I bought a i7 920 when it first came out and has been running that voltage for years...

:confused:
 
I don't know. I've had bad luck. That chip was just a dud, it ended up slowly failing (originally was running fine at 4.0ghz / 1.3625v for ~ 1 year and then little by little had to drop the speed down because of BSoD's until it couldn't even run stable at stock speed).

I had problems flashing modded BIOS to try to unlock my 6950 Crossfire setup.

I couldn't unlock a Phenom 2 X3, it would just fail to POST.

I've just had really bad luck with BIOS mods so I'm just a little hesitant on the idea of flashing a mod BIOS to a card like the Titan, but I'm afraid I wouldn't be happy with the performance with the throttling. I think I might just sit this one out.
 
I really feel like the issue is being blown way out of proportion. I run a reasonable overclock on mine and seldom see much throttling at all. Mine is boosting to 1125 core with only occasional drops to 1110. Even when there is that throttling happening, how much do you think 15mhz is affecting performance? Some people just don't like to see the clocks fluctuate.
 
Shame we can't just increase the power limit reasonably, say to 115% without breaking other features like voltage downclocking.
 
The modified BIOS files are very safe. I have had my Titans for less than a week and have flashed then more than a dozen times. Start at http://1pcent.com/ and see how deep the rabbit hole goes. My biggest complaint is that my cards are not game stable at 1202 mhz. I had to use Kepler BIOS tweaker to reduce the max boost to 1176mhz. FWIW, having a UPS can make the process seem less risky with a $1000 card, but its otherwise nearly fool proof.
 
For those of you who are interested the most recent BIOS posted at Tech Inferno is the new sweetness. It removes all the Kepler non-sense and gives us back control like we had with Fermi cards. All the power and efficiency benefits of the stock BIOS with the ability to hand pick frequency and voltage best for your card.

This is the BIOS the cards should have come with.

Features include:
– ‘Boost 2.0′ disabled
– voltage adjustable up to 1.2125V
– idles normally
– default power target set to 350W
– fan adjustable up to 100%
– normal fan behavior
– default 3d speed 928 MHz (EVGA SC clocks)

I have also found that it removes the 80C automatic throttle feature. Now you can run your overclocked overvolted card at 80-94C and keep it quiet if you choose.

http://www.techinferno.com/downloads/?did=61

:
 
if you are going to double some specs then double them all which would mean the 690 actually has more memory bandwidth. 384 GB/sec for 690 and 288 for Titan.

That's not how it works. And considering the 690 is 2 cards/gpus vs 1. Sorta apples vs oranges wouldn't you say? A more equal comparison would be a 690 vs sli titans. But we know how that would turn out :p
 
Back
Top