Jump to content

Nvidia GTX Titan - modified VBIOS for more control and better overclocking


svl7

Recommended Posts

According to ManuelG @ NVIDIA

https://forums.geforce.com/default/topic/544882/geforce-drivers/official-nvidia-r320-18-whql-display-driver-feedback-thread-released-5-23-13-/post/3815631/#3815631

The GPU Power reporting mechanisms have been corrected in our latest drivers and is available for customers in new versions of 3rd party GPU monitoring applications like EVGA’s PrecisionX version 4.2 and latest MSI Afterburner BETA. In addition, we have added even more information about why a particular GPU is limited to a particular clock. This new information is also visualized though the same 3rd party applications. Last but not least, we have introduced a new fan speed algorithm for GK110 class GPUs which eliminates fan speed variations caused by fluctuating GPU temperatures.

I hope this does not mean algorithm change in power reporting and fan speed affects modified BIOS ?

Link to comment
Share on other sites

  • Founder
I just flashed both my Titan HC's with the OC bios last night and while both seem to be working fine, I'm encountering some problems in nvidia inspector. For my first titan, all of the options for increasing voltage, base clock..etc look the same as they do in the demo video and other pictures posted by users on this thread. When I select my second Titan to apply the same settings, my voltage slider has changed to an "over volt slider" ranging from 0 to 32.5 mV (for a max of 1.20 V). Also, I can only move my base clock slider for the second card 10-20 MHz, before the "estimated max" shoots above 1200 MHz. This is in contrast to my first card where I add +200 MHz on the slider and can reach a voltage of 1.212 V.

Has anyone encountered a similar problem when dealing with Titan's in SLI? Also NV Inspector shows the same bios ID for both cards, but the first card has "Modified" beside the ID. I've gone back and re-flashed the second card, but there were no changes. Any ideas?

Are you sure you flashed your second card correctly? Both my cards work the same and both show MODIFIED. Remember its nvflash -i1 -4 -5 -6 bios.rom<bios.rom></bios.rom>

Link to comment
Share on other sites

Also auto fan speed works perfectly fine with the vbios posted here.

Yeap, tested the 2nd (latest) version of the SC TI bios, the auto-fan works fine now, nice one svl7...

EDIT: silly question but i just have to...is there any way for a company/dealer to find out what bios is installed in a card if the card is "dead"? (meaning the card doesn't post, no signal, doesn't get recognized by any system in any PCI-E slot and so forth). Just trying to "calculate" the future risk for an 1K card just in case something goes wrong (yes, i hate Murphy's law, lol)

Link to comment
Share on other sites

Quote

Is there anyway to get clocks in something other than increments of 13?



I don't think that's possible, 13Mhz steps are predefined.

But i spent some time to "map" the frequency/steps so i don't have to guess every time, this is for the SC TI version so here goes, might be useful for someone:

gpu offset actual gpu freq
+0 928Mhz
+13 941Mhz
+26 954Mhz
+39 967Mhz
+52 980Mhz
+65 993Mhz
+78 1006Mhz
+92 1019Mhz
+105 1032Mhz
+118 1045Mhz
+131 1058Mhz
+144 1071Mhz
+157 1084Mhz
+170 1097Mhz
+183 1110Mhz
+196 1124Mhz
+209 1137Mhz
+222 1150Mhz
+235 1163Mhz
+248 1176Mhz
+261 1189Mhz
+274 1202Mhz
+287 1215Mhz
(didn't care for any higher frequency that's why my list stops at 1215Mhz)

Also these are all the voltage steps available in the TI SC version

1.137V
1.150V
1.162V
1.175V
1.187V
1.200V
1.212V
  • Thumbs Up 1
Link to comment
Share on other sites

  • Founder
Yeap, tested the 2nd (latest) version of the SC TI bios, the auto-fan works fine now, nice one svl7...

EDIT: silly question but i just have to...is there any way for a company/dealer to find out what bios is installed in a card if the card is "dead"? (meaning the card doesn't post, no signal, doesn't get recognized by any system in any PCI-E slot and so forth). Just trying to "calculate" the future risk for an 1K card just in case something goes wrong (yes, i hate Murphy's law, lol)

I don't think there is and if you're worried about EVGA, they don't go that far in checking. In fact I sent them a flashed 680 for RMA before and they didn't care.

Link to comment
Share on other sites

I don't think there is and if you're worried about EVGA, they don't go that far in checking. In fact I sent them a flashed 680 for RMA before and they didn't care.

Thank you Brian, no reason for me NOT using the TI bios then!!! ;)

Another question, anybody tried the new official Nvidia drivers with the TI bios's?

Also the Afterburner 3.0.0 beta 10 is out (new Precision too but i prefer AB's interface for some reason), anybody has tried it with the TI bios?

Thanks!

Link to comment
Share on other sites

  • Founder
Thank you Brian, no reason for me NOT using the TI bios then!!! ;)

Another question, anybody tried the new official Nvidia drivers with the TI bios's?

Also the Afterburner 3.0.0 beta 10 is out (new Precision too but i prefer AB's interface for some reason), anybody has tried it with the TI bios?

Thanks!

Sure no problem. I'm using the latest NVIDIA driver and its working great with the modded vbios. I don't use Afterburner though, just Precision + NV Inspector. With the new Precision, you can do everything and more with it that you can with NV Inspector. I just use NV Inspector because I like the shortcuts.

Link to comment
Share on other sites

Sure no problem. I'm using the latest NVIDIA driver and its working great with the modded vbios. I don't use Afterburner though, just Precision + NV Inspector. With the new Precision, you can do everything and more with it that you can with NV Inspector. I just use NV Inspector because I like the shortcuts.

Thanks again Brian, you are right.

Testing the new drivers as we speak, they are fine (Farcry 3, Metro LL etc). Yeah, the new Precision is what i use right now, still can't store the voltage per profile but no problem, i can set it manually in 3 seconds. ;)

Just for the record, AB beta 10 still can't change the voltage in TI bios, i am mentioning it just in case somebody think that the new version works.

Link to comment
Share on other sites

I just finished rebuilding my PC yesterday. After verifying everything was running properly I flashed my cards with the modified SC BIOS from here. Finally the clocks make sense again. Just like with previous generation cards before Kepler, they now logically increase with added increments of a GPU overclock. You no longer have to test and guess where the max GPU clock will eventually end up while boosting. Saves a lot of hassle and time. Very nice!

Another thing I noticed was that the GPU still down-clocks or throttles. Initially, I was a little disappointed, because it appeared there wasn't _that_ much of a difference compared to the original BIOS. However, after further testing I found out it was due to the simple fact that the GPU didn't have enough workload. As soon as I cranked up the AA I got full boost again. Not sure whether this is an intended feature, but it's a good thing. No point in running full clocks when the utilization is only ~35% and totally CPU bound. In that respect it works a little like the original BIOS and not completely wasting energy by just generating heat. Very nice again. Thanks svl7 for making this available!

  • Thumbs Up 1
Link to comment
Share on other sites

Another thing I noticed was that the GPU still down-clocks or throttles. Initially, I was a little disappointed, because it appeared there wasn't _that_ much of a difference compared to the original BIOS. However, after further testing I found out it was due to the simple fact that the GPU didn't have enough workload. As soon as I cranked up the AA I got full boost again. Not sure whether this is an intended feature, but it's a good thing. No point in running full clocks when the utilization is only ~35% and totally CPU bound. In that respect it works a little like the original BIOS and not completely wasting energy by just generating heat. Very nice again. Thanks svl7 for making this available!

I have spent hours and hours testing, benching, gaming at every possible freq/voltage/fan combination using this bios, i have kept tons of AB/Precision charts/graphs and i have never experienced any kind of throttling with this or any other high-TDP modded bios.

The only throttling you should ever get (with this or any other modded bios) is because of the low temp target/high temps if you forget to increase your fan rpms and/or forget to increase the temp target (which by default is 80C). Apart from that which is 101% normal you should not get any throttling.

For instance I spent about 2 hours yesterday playing Metro 2033 (every setting at max), 1 hour Farcry 3 (every setting at max, including MSAAx8) and half an hour Crysis 3 (every setting at max except AA which was SMAAx4) at 1150Mhz/1.175V. Power target to 100% (=350W), temp target 90C. Even with temps around 81-84C - as expected - i didn't get a single step of throttling (kept the Precision graphs) and i am mentioning this because i have read all over the net that we should keep the Titans below 80C no matter what the temp target is. I doubt it, has never happened to my card anyway.

Just my 2 cents...

Link to comment
Share on other sites

I haven't noticed this, but then again I never monitored the card when there was only a mild load on the GPU, I guess I could adjust it to always go full power even at such a low load, but as you said, it's probably not necessary.

Link to comment
Share on other sites

  • Founder

The card should always drop its clocks when utilization is low, its totally pointless to keep it at full clocks + heat for no reason.

Link to comment
Share on other sites

Just to elaborate, I'm running 3-Way SLI water cooled. The cards don't even reach 40C during gaming with an ambient of 23C. Clocks so far are set to 1150 and mem to +300 on all cards. Sync between cards is off.

In Crysis 3 single player when running 1920 x 1080 @ 130Hz with all details set to max, no motion blur, no Vsync, no AA or only FXAA, one or more of my cards will clock down to 1019MHz or something less. As soon as I set a higher AA or run the game on higher resolution monitor at 2560 x 1600 the cards will be at full clocks again.

I'll try more games once I get my system sorted out and stuff installed.

No need to change anything though. From what I've seen the BIOS works fine as it is.

Link to comment
Share on other sites

  • Founder

At 1080P with no AA I don't even think 2 of your cards are being stressed to the max let alone 3 which would explain the utilization + clock drops as you noted.

Link to comment
Share on other sites

  • Founder
Is there going to be a new version of this bios? With the new drivers (320.18) I think the power settings are changed as my previously stable clocks are behaving strangely.

That doesn't really tell us much does it? You need to provide better details and/or GPU logs + settings. I haven't noticed any changes in my OC with the latest nvidia driver.

Link to comment
Share on other sites

That doesn't really tell us much does it? You need to provide better details and/or GPU logs + settings. I haven't noticed any changes in my OC with the latest nvidia driver.

I have a lot of instabilities with my overclocks which some other users experienced too with TI bios and the latest 320.18 bios. We guess it is the way Nvidia changed the power limiting system.

Link to comment
Share on other sites

  • Founder
I have a lot of instabilities with my overclocks which some other users experienced too with TI bios and the latest 320.18 bios. We guess it is the way Nvidia changed the power limiting system.

And you'd be guessing wrong because if that was the case, everyone with the vbios and the latest driver would have the same problem. Additionally, you haven't provided an ounce of proof to support this. If you want to blame the vbios, you need to at least provide some GPU logs, hardware configuration, application used and what clocks you attempted. Otherwise it's completely meaningless and we're not going to go on OCN searching for that information.

Link to comment
Share on other sites

And you'd be guessing wrong because if that was the case, everyone with the vbios and the latest driver would have the same problem. Additionally, you haven't provided an ounce of proof to support this. If you want to blame the vbios, you need to at least provide some GPU logs, hardware configuration, application used and what clocks you attempted. Otherwise it's completely meaningless and we're not going to go on OCN searching for that information.

Well it's ok. I just flashed back to my stock bios. I guess I'll use a custom bios when I get my Titan on water.

Link to comment
Share on other sites

And you'd be guessing wrong because if that was the case, everyone with the vbios and the latest driver would have the same problem. Additionally, you haven't provided an ounce of proof to support this. If you want to blame the vbios, you need to at least provide some GPU logs, hardware configuration, application used and what clocks you attempted. Otherwise it's completely meaningless and we're not going to go on OCN searching for that information.

He might mean this , although my overclocks actually became more stable when i updated drivers

18th may ManuelG

Technically this is by design. We introduced a change in our overclocking policy where the clock management respects the VBIOS settings. VBIOS's GPCPLL maximum VCO frequency is 2000 MHz(GPC 999MHz) for Fermi that’s why we are capped it at 999MHz. However due to end user demand, we are reverting these changes and implementing the same policy we had pre-R295 drivers. You will see this in our WHQL driver that will be released soon. "

Link to comment
Share on other sites

Got to put in some time in Metro LL with my newly flashed HC Titan's in SLI over the weeked. Using the new 320.18 NVidia drivers and playing at 2600 x 1440 with all settings maxed out except for SSAA. Very stable, multiple hour, gaming sessions with the OC Bios running at 1163 MHz / 6600 MHz @ 1.212 V. svl7 has some beer coming his way later this week for the great work on the OC Bios. I can't even count the number of the hours I spent trying to get a stable overclock out of the factory Bios. Seeing big increases in my 3DMark 11 X scores as well. Cheers!

  • Thumbs Up 2
Link to comment
Share on other sites

Sometimes, overclocks can be too close to the bleeding edge. Without any breathing room to compensate, a different driver or BIOS may push your hardware too far. In my opinion, for everyday use, it's better to back off one or two notches from your max achievable. I do the same with my CPU and rarely experience issues with software or firmware updates.

  • Thumbs Up 1
Link to comment
Share on other sites

Sometimes, overclocks can be too close to the bleeding edge. Without any breathing room to compensate, a different driver or BIOS may push your hardware too far. In my opinion, for everyday use, it's better to back off one or two notches from your max achievable. I do the same with my CPU and rarely experience issues with software or firmware updates.

I agree, I run benchmarks a few steps higher on the base clock offset than the settings I use for extended gaming. I haven't run Valley 1.0 with the 320.18 and my current settings OC but will try it later tonight and see if I have a problem.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...

Important Information

By using this site, you agree to our Terms of Use. We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue.