There are two on Zoson's thread for the FTW.  The first, the one that's says "-AIR" at the end of the name, raises the voltage to 1.281...which the card can't even use.  1.275 is the limits on the FTW.  The other one doesn't look like it does anything.  /shrug.  

If they're working for you, great, but.....neither one of them are set up right.

mdlawson

kwkrnu72
I never did get r0ach's bios versions to work.  I made some improvements upon the 1.262v bios from the LTT forums.  Upped the voltage to 1.275v, improved the scaling in the voltage tables, moved the TDP and power limits up to 300 watts (105 from each 6 pin and 80 from the PCIe slot), disabled the boost state and set the TDP base clock to 1531.5 and memory to 3705.  Works well for my cards, but....I tailored it specifically for them.  Other than the voltage, which I wish would go higher.  = (  

It didn't allow me to push too much further, but it did help.  FS against 2 x 970s I'm now in 2nd (21193), FS Extreme against 2 x 970s I'm in 1st (11414), and I have yet to find 2 x 970s with a higher Valley score than 5555.  I just wish they could have run this well straight out of the box....wouldn't have been as much fun getting there, but....that kind of performance is definitely there.

I've been running the bios you've linked, working great so far and better than Zoson's one. Seems to hit the full 1531MHz frequently, but I'd be interested in trying out your latest version with a bit more voltage/power headroom to see if I can get it more consistent. Is disabling boost/upping the memory clock giving any benefits? 
 
Thanks for your work so far!

This is the one I'm using right now.  I upped the memory to 3900 (1950), and raised the power limit base to 295w and 310w with the slider all the way to max.  Voltage is still set to 1.275.  I still use Precision X to increase clock / memory speed for benchmarking, but other than that, I just use the bios settings for my daily gaming clocks.

https://drive.google.com/...JwdlE/view?usp=sharing



  Updated to that last week, and was pushing above 1600 in SLI in Heaven and above 2000 on the base memory clock.  Still working, but...it's pretty much maxed out now I think.

Should work now.  I forgot to set it to "public"

Muahahaha

Single card at 1629 / 4200



SLI run at 1586 / 4205


 

Haze88

kwkrnu72
Muahahaha

Single card at 1629 / 4200



SLI run at 1586 / 4205


 

You have water blocks on both you 970's correct?  Would the settings you made be too much for an air cooled card?  Thanks!  Edit: tried it anyways my card can't handle it xD.  I'll try your other BIOS.  Looks like it can't handle the other BIOS either strange.  My ASIC quality is 75% so in theory it should be fine?  

Yes, my gpus are under water.  Try this.  I lowered the base boost clock to 1506, and the voltage to 1.262v.  I moved the power limits and TDP to 310w straight up.  No more power slider.  Shouldn't need it.  If this doesn't work for you, let me know, and I'll give you a link to follow to the magic man.

https://drive.google.com/...Z0qTCB5ZkhIZlhVX1NxaTg

Have you run your game with GPUz open to the sensors tab?  Try that, and stretch it across the screen, like shown below, then after gaming for a while, take a look at the 'perf cap reason' line.  If you're seeing green in there, you're up against your power limits, and the card is throttling.

nextgencfl
Is this needed for the 970 FTW+?

Are you overclocking?  Have you run any of your games or benchmarks with GPUz sensors tab open to see if your card is throttling due to power limits when it's under load?

veizelman

flexy123
Just my comment here...yes, late to the party:
 
I see what the OP is claiming and saying, but I want to see real evidence that jumping up/down in clocks/voltages and throttling translates into problems such as stutters in games.
 
I simply have a hard time to believe that Nvidia implemented something (GPU Boost, that is) where the throttling that we see would cause "tearing and stuttering". A more objective test here would maybe be measuring frame times. I personally really do not believe that throttling, even fast in sequence is even noticeable. You dont get tearing because your card goes -13 mhz slower or faster and throttles one bin down. Sorry that I have a problem buying this. Watching the graph in Afterburner doesn't count. Please provide real numbers (frame times) to back this up. Not trolling, just want to see evidence for this.

Ihave massive issues while trying to play games, the clock will be fine but utilization is low. with kboost gpu clock fluctuates between 1200-1450. Changing the graphical settings in game or nvidia panel increase no fps. Many gamers cannot believe i would need more than 120 minimum fps, but the input lag and consistency of graphics quality is very different when trying to play a competetive game. There is no reason the 970 could not get better frames on low settings with obly 30-45% at best gpu utilization. cpu stays at 70-80%, so there is no bottle neck in my opinion, only gpu ramping down because it does not think I need the extra frames. no software seems to be able to force the card to work harder.. Tried to force power state 0, witch is supposed to be the boost clock, but it only runs at gpu clock 600 ish.
Point being, we all have our own preferances for stuttering/fps, and the card should not hold back power because it thinks i dont need 300fps. But I do. I do think I've heard of more tearing from fps varying too much, but lets not dwell on the term too long as the tearing is an minor issue. Finally, would never use v-synch for anything playable. It adds so much input lag. Plz help me shake the change out of this card

Are you running with any AA?  

Also, what kind of monitor are you playing on?  Because, even a 144mhz monitor, it's only going to display like what...a max of 280ish?fps.  On a 60mhz monitor, the most you'll see is maybe 100?fps.  And if you're playing on a TV....well, you've found the source of your input lag.  Input lag doesn't really come from frames per second.....it comes from wireless keyboards, and monitors with a slow response time.  Like most televisions....


 

realmister

kwkrnu72
Yes, my gpus are under water.  Try this.  I lowered the base boost clock to 1506, and the voltage to 1.262v.  I moved the power limits and TDP to 310w straight up.  No more power slider.  Shouldn't need it.  If this doesn't work for you, let me know, and I'll give you a link to follow to the magic man.

Hey Green man, thanks for this second nerfed BIOS you posted, it tried it and had some issues. Some red flares/dots would appear during Wither 3. So I thinks it's still a bit too high for me. So can you "give me the link to follow to the magic man"? I want my BIOS without those pci-e handcaps, and experience the full potential of my card, within its silicon limits, of course. I have a very high quality PSU and ventilated case, although still running on air, and full intention of making use of it.

Are you hitting any perf cap limit PWR?  If not, we can just redo the bios to lower the base clock a bit more.  How are temps?

realmister

kwkrnu72

realmister

kwkrnu72
Yes, my gpus are under water.  Try this.  I lowered the base boost clock to 1506, and the voltage to 1.262v.  I moved the power limits and TDP to 310w straight up.  No more power slider.  Shouldn't need it.  If this doesn't work for you, let me know, and I'll give you a link to follow to the magic man.

Hey Green man, thanks for this second nerfed BIOS you posted, it tried it and had some issues. Some red flares/dots would appear during Wither 3. So I thinks it's still a bit too high for me. So can you "give me the link to follow to the magic man"? I want my BIOS without those pci-e handcaps, and experience the full potential of my card, within its silicon limits, of course. I have a very high quality PSU and ventilated case, although still running on air, and full intention of making use of it.

Are you hitting any perf cap limit PWR?  If not, we can just redo the bios to lower the base clock a bit more.  How are temps?

Thanks for your answer, sire, and sorry for the delay. I was using Roach #2 for the past couples of weeks. Yesterday I flashed your nerfed BIOS and only had to drop 13MHz (1506 -> 1493) in order to stabilize. I was playing Dying Light 2k with everything on Ultra with decent FPS, around 50-60. I use an agressive fan curve (the one from EVGA Precision) on Afterburner, so despite the ACX2.0 jet propulsion noise and summer here in my region, temps at 98% usage were at 70C tops. I also use Afterbuner to monitor my limit caps, and I didn't hit any of them using your BIOS. Impressive.
 
The first reboot was fine, and thats how I was able to play and get the above data. But, when I tried to up VRAM clock a little, the crashes appeared. And even reflashing wouldn't help....when I reboot my system, when I reach windows workspace, I get BSOD and the infamous nvlddmkm error and system reboots. Idk what it could be. Perhaps my mobo can't handle the stress (GA-Z97-D3H) but I find it unlikely...
 
It seems my sweetspot is 1493/3900Mhz but I suspect something power related with your BIOS isn't welcome in my system.

If you've updated to the newest drivers, they aren't liking overclocks very much.  Unless you need them for a game, roll back to 355.82.  Also....what PSU are you using?  Is it 6 pins, or are they 6+2 pins that come out of the PSU?

Try this:  https://drive.google.com/file/d/0B6zqzZ0qTCB5Wm9vNkV4dEJmdWc/view?usp=sharing

It's good, but not outstanding.  

As for the driver version I'm using, no....I'm using one even older than that.  I'm using 353.62.  = )  They've been done making "performance improvements" to the 9xx series drivers for a while now, now they're tweaking things for new release games...hot fix driver releases for "xx game".  I haven't bought any of the new release games yet, because I know any more, a "release date" actually means, "here's a game that we kinda finished, please beta test this for us".  lol

kwkrnu72

realmister

kwkrnu72

realmister

kwkrnu72
Yes, my gpus are under water.  Try this.  I lowered the base boost clock to 1506, and the voltage to 1.262v.  I moved the power limits and TDP to 310w straight up.  No more power slider.  Shouldn't need it.  If this doesn't work for you, let me know, and I'll give you a link to follow to the magic man.

Hey Green man, thanks for this second nerfed BIOS you posted, it tried it and had some issues. Some red flares/dots would appear during Wither 3. So I thinks it's still a bit too high for me. So can you "give me the link to follow to the magic man"? I want my BIOS without those pci-e handcaps, and experience the full potential of my card, within its silicon limits, of course. I have a very high quality PSU and ventilated case, although still running on air, and full intention of making use of it.

Are you hitting any perf cap limit PWR?  If not, we can just redo the bios to lower the base clock a bit more.  How are temps?

Thanks for your answer, sire, and sorry for the delay. I was using Roach #2 for the past couples of weeks. Yesterday I flashed your nerfed BIOS and only had to drop 13MHz (1506 -> 1493) in order to stabilize. I was playing Dying Light 2k with everything on Ultra with decent FPS, around 50-60. I use an agressive fan curve (the one from EVGA Precision) on Afterburner, so despite the ACX2.0 jet propulsion noise and summer here in my region, temps at 98% usage were at 70C tops. I also use Afterbuner to monitor my limit caps, and I didn't hit any of them using your BIOS. Impressive.
 
The first reboot was fine, and thats how I was able to play and get the above data. But, when I tried to up VRAM clock a little, the crashes appeared. And even reflashing wouldn't help....when I reboot my system, when I reach windows workspace, I get BSOD and the infamous nvlddmkm error and system reboots. Idk what it could be. Perhaps my mobo can't handle the stress (GA-Z97-D3H) but I find it unlikely...
 
It seems my sweetspot is 1493/3900Mhz but I suspect something power related with your BIOS isn't welcome in my system.

If you've updated to the newest drivers, they aren't liking overclocks very much.  Unless you need them for a game, roll back to 355.82.  Also....what PSU are you using?  Is it 6 pins, or are they 6+2 pins that come out of the PSU?

Try this:  

Damn dude, I've been testing the new BIOS for a couple of days now and I think it's damn near to perfect! Windows now starts normally and I only had one random crash with Dying Light. My most demanding game (The Witcher 3), now is running like a charm, without artifacts and stuff. I don't hit any perf caps and temp is around 65C, even at 1531Mhz (1506 normal) 99% usage and 70% power. With almost everything on ultra w/o vsync I'm always above 60FPS at fullHD! Idk if the driver downgrade helped a great deal but now I'm pretty satisfied. I'll keep you posted about any occurences. Thanks again!

Hi guys... I have an GTX 970 ftw too and was kind of paranoid with this issue...
Did a lot of research and i think i have some answers.

First of all is: Is this an issue? Sure, if it wasn't than EVGA would not have released the "+" versions, wich has more power phases and a higher TDP.
But... this is a deal breaker? No. This cause tearing or stutter? No. It only stops the card from being a better overclocker.
We have 3 brickwalls on turbo boost 2.0.
1)Temps
2)Reliable Voltage
3)TDP

If any of these is on its limits, than the gpu would not boost anymore.
Tearing can't be caused by a lower clock, tearing is caused by not matching the refresh from the gpu with the monitor, as you guys know. I can't think of a scenario where a gpu clock not boosting any higher would cause tearing. What may have caused the extra tearing that the OP related is the extra fps the FTW may have when compared with the Gigabyte due to higher stock clocks. Actually, a framerate CLOSER to the monitor refresh rate can cause a lot more tearing, due to the frame "tear" appearing in close areas of the screen in each refresh. When in a lower framerate, that does not happens. This can be seen when you lock your fps to let's say... 58fps in a 60hz monitor. You will see LOTS of tearing. Now try it at 50 or 45. No tearing anymore. So, thats my guess for the extra tearing the OP related.
And stutter? Well, same thing. 60hz is when our eyes stop seeing frames, and get a feeling of perfect fluid motion. if we oscilate between 60, and NEAR 60. you see a lot of stutter due to contrast.
I'm not saying that i am 100% right, but i think these are the logic conclusions at the time.

Now, let's go the the bios fix... A lot of people claimed that when checking gpuZ even with a higher TDP modded bios they are still getting pwr perfcap limiting right?
Thats why: overclocking.guide/increase-the-nvidia-power-limit-all-cards/

As you can se in the guide, TDP is measured in hardware level. What you can control in the bios is the software interpretation. So, if you set your SOFTWARE TDP higher, you will still get limited, but on let's say... 80%.
The software thinks the board has a higher TDP. But the hardware sensor will still limit the board.
If you guys really think this is a serious problem, follow the guide above, is the only real solution. it will not fix tearing nor stutter, these are caused by other reasons, but it will let you overclock your card more than we can get at stock.

PS. Sorry for any mistakes, english is not me primary language.

Nope, it's the 170 at 100% and 187 watts at 110% on the power limit slider that's the problem, at least initially, and not the 250 watt TDP that's set in the bios in a completely different table.  Different things, man.

Most of the ACX 2.0 cards, depending on ASIC quality of the card, and quality of the PSU, overall build of that specific card, will start to see PWR limit perf caps when they reach anywhere from 1450 to 1550.  The power limit perf cap, will throttle the GPU core and the voltage, to get the card back under said power limit, making the, in some cases, VERY mild overclock effectively useless.

So, we modify the bios to increase the power limit to avoid the power limit throttling, and...viola, fixed.  A card that will properly overclock, and not throttle itself because of artificially low power limits. 

In fact, Maxwell pulls a LOT more power than NVIDIA wanted to admit.  They just do it in microsecond spikes. As outlined in the article linked below.

Will it affect MOST of the ACX 2.0 970s with stock clocks, or even very mild overclocks?  No....but it is a HUGE problem for some, and WILL be a problem if you plan to overclock the card to help it reach it's maximum potential.  These cards are AMAZING overclockers.  My 187 watt at 110% 970 FTW, needs a 300 watt power limit to keep it from throttling at the overclock I'm running.  I've created a couple of hundred bios versions for my own card, experimenting with different settings, and...anything under 300 would hit a power limit perf cap.  And then, of course, like above, a few hundred custom bios versions for a whole array of 970s, for a lot of 970 owners that were tired of seeing the exact same thing I was tired of seeing.  Bouncing core clocks with mild overclocks.

If you're happy with LOWERING your core clocks?  Good for you....most of us don't want 960 performance out of the GPU we paid 970 type money for.  

I could show you some pictures of the bios file, and maybe even draw you some pictures, but....I'm just really not in the mood.  Enjoy the read.  /wink

http://www.tomshardware.com/reviews/nvidia-geforce-gtx-980-970-maxwell,3941-11.html 
 

It's easy to make assumptions unless you experience the power limit throttling. It has nothing to do with the TDP in the bios. As Vellinious stated it relates to wattage and percentages of the power limits in the bios. I tried for weeks to squash this with my 970 SC (even though Best Buy called it a SSC). Lowering voltage to 1.212v, power to 201W-250W, and getting stuck at 1506.5MHz. Finally got tired of the card with the crappy 4+2 power and got a better card. Now, I oc'd it to 1506.5/8000MHz by choice, game-stable, rather than benchmark-stable. I have no throttling period, with the new card...6+2 power phase, 1506.5/8000MHz, 1.243v and 288W power limit. Never had screen tearing or stuttering either.. 60FPS with v-sync and triple buffering on in nvidia control panel, with the witcher 3, SW BF, Batman AK all on Ultra. I repeat...no lag...even with the power hitting 117%, because my power limit is 124%. Same old argument...this is getting old. Vellinious knows what he's talking about, he tried to help me when I needed. It WAS the card, not user error.

Table 5 is higher because I have 1x6pin and 1x8pin.