TENTATIVELY SOLVED: Afterburner?! Continuing with the SLI Power-bug issue

[BonzaiDuck]

UPDATING THIS INITIAL POST:

If you read my "Call to Arms" thread, you will have to concede that there was some sort of problem with SLI configurations of GTX 9x0 cards -- that I wasn't the only person who noticed it. I can't say whether the others "figured it out," but I did, after putting in an NVidia support ticket. I stumbled onto the solution before receiving any sort of guidance from NVidia.

The "fixes" in this initial post here were not the real McCoy, and only led me to more drastic measures. Those follow through the length of the thread as of March 12.

=====================

My request with this thread arises because I can't understand how my problem solution worked.

The solution to the higher-than-normal power consumption and idle clock speeds was to simply select the AfterBurner checkbox that makes it load at Windows startup, putting its little fighter-plane icon in the system tray.

I needed Afterburner to implement custom fan-control to my satisfaction, and when I deselected "Apply OC settings at startup," fan control would revert to the default -- with the fans not running until temperatures reached 50 or 60C -- maybe higher.

Solving the power consumption problem only required making AfterBurner load at startup. Of course, this means that my SLI graphics configuration is joined at the hip to a piece of software -- never mind the software was bundled with the GTX 970s.

So I'm wondering how Afterburner manages graphics card settings -- including OC choices. Does it save some sort of "profile" to be loaded at startup? Does it make changes to the cards' BIOS'? Why is fan-control dependent on "Apply OC settings at startup," when the Ai Suite fan-control for my motherboard remains implemented at the BIOS level, and I don't need to run AI Suite to keep the fan settings?

Maybe someone knows. I've done some web-searches today, and all I find are guides and instructions for using Afterburner, but no explanation of how and why I need it running at startup.

 

[Alpha0mega]

Just a heads up, but for me, applying the fan profile does not require selecting "Apply OC settings at startup". I can leave that off, and as long as Afterburner is running, the fans are controlled by it, irrespective of whether AB was launched at startup or not.

However, the aforementioned checkbox must be selected for changes to power/temp target and frequency changes to take effect.

Also, my Zotac cards didn't come bundled with Afterburner, but I use it anyway, since it's probably the best one there is, and I have been using it for sometime. So I don't imagine that the downclock problem is specifically because AB isn't running at startup. Rather (just spit-balling here) it's because something is stopping the cards from properly doing so until AB steps in and corrects it . Why that happens only if it's started at Windows boot, I can't say.

 

[digitaldurandal]

My fan settings also work without apply OC settings at startup. (using r9 290 cf)

 

[Burpo]

1. Does it save some sort of "profile" to be loaded at startup?
Yes, it's a graph..

2. Does it make changes to the cards' BIOS'?
No, that's why the software & graph have to be loaded..

3. Why is fan-control dependent on "Apply OC settings at startup?
It needs instructions from the graph..

This is what I gathered from a quick read of the Manual..
I could be full of sh*t..

4. "the Ai Suite fan-control for my motherboard remains implemented at the BIOS level, and I don't need to run AI Suite to keep the fan settings?"

Asus bios allows AI Suite to make/save changes..

 

[Alpha0mega]

Just tested another thing. I stopped Afterburner from starting with Windows (by unchecking "Start with Windows" in Settings), and rebooted. My cards are still downclocking to 135 MHz without Afterburner (checked via GPU-Z). Then I started AB manually, and checked again. Same thing, cards downclock with or without AB running, or being run in any number of ways.

Could be the Windows difference we talked about it the other thread, but possibly more likely it's something specific to your setup that's causing the cards to run at higher clocks. Maybe some software draw calling in the background? Though why having AB load at start changes, I don't know.

 

[BonzaiDuck]

Just tested another thing. I stopped Afterburner from starting with Windows (by unchecking "Start with Windows" in Settings), and rebooted. My cards are still downclocking to 135 MHz without Afterburner (checked via GPU-Z). Then I started AB manually, and checked again. Same thing, cards downclock with or without AB running, or being run in any number of ways.

Could be the Windows difference we talked about it the other thread, but possibly more likely it's something specific to your setup that's causing the cards to run at higher clocks. Maybe some software draw calling in the background? Though why having AB load at start changes, I don't know.

I had sent PM to Stringjam just now with the latest developments about this.

First, let me assure that my CPU OC settings were rock-stable before I added my second card. As far as I can tell from running stress-tests again, they were still tip-top. Take that out of the equation.

After adding my second GTX 970, I would very occasionally have a BSOD stop-code 124. I thought I got rid of it: system was running 24/7 nearly 2 weeks. It came BACK four hours ago. Stop-Code 124 may have been associated with hardware errors and graphics card troubles, but it can also -- just as frequently -- occur because of a malfunctioning driver. Personally -- I strongly suspect a driver issue. When I thought I "got rid of it," I did what I THOUGHT was a clean NV driver installation.

HOWEVER! "Clean driver installation" will not replace ALL THE OTHER NVIDIA DRIVER S*** -- SHIT! -- IF THE EXISTING 3D/Controller/driver/PhysX/Audio/GEForce Experience-etc. are STILL THE NEWEST VERSIONS.

What I finally did : I Completely uninstalled Afterburner. I then DISABLED SLI and rebooted. Then I completely uninstalled (from Ctrl-Panel->Programs/Features) EVERY SINGLE NVIDIA SOFTWARE PROGRAM/DRIVER.

WITH SLI STILL DISABLED, I reinstalled the entire enchilada. Then I re-enabled SLI. On reboot, power-saving idle clocks come back to 135/162 (or howsoever it is reported by different monitoring SW).

I think this sort of thing happens when you have one NVidia card successfully installed and working with all the drivers, then add a second card without any back-steps. Only time will tell. People are either going to install them all at once and together, or they'll put them in one at a time.

So I think -- whether or not my problem is really solved at this point (but HOPE!) -- moving from single-card to SLI, best practice would be to completely uninstall all NVidia driver/software/SHEE-it! And then reinstall.

And I wouldn't be fiddling around with AFterburner -- I wouldn't even install it -- unless SLI, the clocks, everything -- is working properly.

I'm going to reboot again, and then update this post. JUST TO BE SURE.

I feel like a member of a brotherhood/sisterhood here. People take an INTEREST in the misery of their fellows. I just want to say how much I appreciate it.

 

[BonzaiDuck]

I had sent PM to Stringjam just now with the latest developments about this.

First, let me assure that my CPU OC settings were rock-stable before I added my second card. As far as I can tell from running stress-tests again, they were still tip-top. Take that out of the equation.

After adding my second GTX 970, I would very occasionally have a BSOD stop-code 124. I thought I got rid of it: system was running 24/7 nearly 2 weeks. It came BACK four hours ago. Stop-Code 124 may have been associated with hardware errors and graphics card troubles, but it can also -- just as frequently -- occur because of a malfunctioning driver. Personally -- I strongly suspect a driver issue. When I thought I "got rid of it," I did what I THOUGHT was a clean NV driver installation.

HOWEVER! "Clean driver installation" will not replace ALL THE OTHER NVIDIA DRIVER S*** -- SHIT! -- IF THE EXISTING 3D/Controller/driver/PhysX/Audio/GEForce Experience-etc. are STILL THE NEWEST VERSIONS.

What I finally did : I Completely uninstalled Afterburner. I then DISABLED SLI and rebooted. Then I completely uninstalled (from Ctrl-Panel->Programs/Features) EVERY SINGLE NVIDIA SOFTWARE PROGRAM/DRIVER.

WITH SLI STILL DISABLED, I reinstalled the entire enchilada. Then I re-enabled SLI. On reboot, power-saving idle clocks come back to 135/162 (or howsoever it is reported by different monitoring SW).

I think this sort of thing happens when you have one NVidia card successfully installed and working with all the drivers, then add a second card without any back-steps. Only time will tell. People are either going to install them all at once and together, or they'll put them in one at a time.

So I think -- whether or not my problem is really solved at this point (but HOPE!) -- moving from single-card to SLI, best practice would be to completely uninstall all NVidia driver/software/SHEE-it! And then reinstall.

And I wouldn't be fiddling around with AFterburner -- I wouldn't even install it -- unless SLI, the clocks, everything -- is working properly.

I'm going to reboot again, and then update this post. JUST TO BE SURE.

I feel like a member of a brotherhood/sisterhood here. People take an INTEREST in the misery of their fellows. I just want to say how much I appreciate it.

YES! Yes-yes-yes-yes!! Second reboot of the system confirms it! Idle clocks still come back to 135/162.

This is a "digital" realm. I think for that reason, "small sample" statistics carry greater significance. It's either "0" or "1," and a second reboot to "normal behavior" confirms this has been fixed.

Afterburner was just the sugar-coating on the M&M. The NVIDIA drivers are the chocolate core of the real problem.

 

[BonzaiDuck]

Just reinstalled Afterburner.

The bug is back.

If I uninstall it, the bug is still there.

It was working fine until then, with only the NV drivers and (S***) installed.

I don't know if I should find another version of AB, or . . . look for something else . . .

 

[BonzaiDuck]

Just reinstalled Afterburner.

The bug is back.

If I uninstall it, the bug is still there.

It was working fine until then, with only the NV drivers and (S***) installed.

I don't know if I should find another version of AB, or . . . look for something else . . .

Still there? It was. Then I disabled and re-enabled SLI, rebooted, and the clocks settle to 135/162 again. This was after uninstalling AB a second time, already reported in my quote above.

Now I'm wondering what happens if I DISABLE SLI before installing Afterburner. Then -- re-enable SLI. I've been through this so many times this morning, I can't remember if I'd done it one way, or the other.

 

[Alpha0mega]

I think they gods have it in for you. Remember, "Whom the gods would destroy, they first make mad".

Kidding aside, I installed by second 970 about a month after getting the first. I don't remember if I did the "uninstall driver before installing" thing, but that's my standard MO for driver updates anyhow, so more than likely I did.

You could try some other fan control software, like EVGA's. I honestly can't begin to guess why Afterburner would, for you, stop the downclock, and that too partially, but not do so if it was started at boot (is that still the case?).

You could either try another software, or try upgrading to Windows 8, or just leave AB to start with Windows (since you are going to need it running to control the fans, why add the extra step of manually starting it?), at least until Windows 10 comes out.


EDIT: when I said that my standard MO is uninstall-then-reinstall, that was only for the drivers. I don't uninstall AB for driver changes, and didn't for when I got SLI.

 

[BonzaiDuck]

I think they gods have it in for you. Remember, "Whom the gods would destroy, they first make mad".

Kidding aside, I installed by second 970 about a month after getting the first. I don't remember if I did the "uninstall driver before installing" thing, but that's my standard MO for driver updates anyhow, so more than likely I did.

You could try some other fan control software, like EVGA's. I honestly can't begin to guess why Afterburner would, for you, stop the downclock, and that too partially, but not do so if it was started at boot (is that still the case?).

You could either try another software, or try upgrading to Windows 8, or just leave AB to start with Windows (since you are going to need it running to control the fans, why add the extra step of manually starting it?), at least until Windows 10 comes out.


EDIT: when I said that my standard MO is uninstall-then-reinstall, that was only for the drivers. I don't uninstall AB for driver changes, and didn't for when I got SLI.

Last time I might have spoken too soon, but things were fine after the first driver update until I re-installed Afterburner. Then it would boot to the 912/1500 clocks again.

So I uninstalled Afterburner, took a quick inventory of my system tray, uninstalled two annoying Startups I'd rather live without, burned the old drivers from Programs and Features, C-Cleaned the registry, downloaded a new copy of the same (damn) driver package, and reinstalled everything but GeForce Experience. [I'll try that again later, but you don't need it, even for the game profiles and I'll explain in a minute. You can always monitor the NV web-site, and get the updates yourself without complicating your system tray and background processes.]

At this point, GPU_Z is at least a reliable monitoring program; it gives you the clocks and the temperature, even if I don't see the fan speed. The odd part about it is this: GPU_Z won't display my GPU fan speed, but the temperature is indicative of the fan settings I had made with Afterburner (which was uninstalled.) That is, the temperature with 135/162 clocks now is the same as it was before expunging Afterburner when I had to force the low clocks for a windows session by cycling the SLI Enable. At the time, the fan profile gave the same temperature.

This means that my fans are at the 30% floor where I had wanted them at idle.

So that leaves the overclocking. I may check for newer versions than the Afterburner I have, but in meantime I'm going to try it with NVidia Inspector.

NVidia implicitly recommends TechPowerUp GPU-Z, because they told me to use it for generating log files to send to them about this overall problem. To get the tech-support ticket, I had to give them all my specs, so they know I have the MSI cards. Instead, I was directed to use GPU-Z.

Any idea of cites, "warranties," "guarantees," "positive reviews" -- recommendations or just good press for Nvidia Inspector? Is Inspector a product of NVidia or an affiliate? Or does Nvidia lend any positive assessment of the software?

I think I'll install that program and attempt to overclock again with the settings I discovered with Afterburner.

Too bad about AB, though. That was a nice interface -- well-designed.

Anyway, this is a long screed about my difficulty, but I should say I've tweaked my monitor for game settings, got the clocks behaving again through another two boot cycles, and ran some races with GRID2 and Assetto-Corsa. Something "happened." It was like being released from Purgatory.

I never saw anything like this before -- adding up the monitor profile and the game settings. It's like Butthead in the cartoon with Beavis: "Ah . . . huh-huh . . . Ah have been to the mountain . . . huh-huh . . ah-h-h-"

 

[xthetenth]

Last night I downloaded nVidia Inspector to deal with a high clock state from multiple monitors. Getting it working was quick and easy. Seems nice and straight forward, and the interface is pretty nice.

Haven't poked around fan control or overclocking options actively yet though.

 

[BonzaiDuck]

Last night I downloaded nVidia Inspector to deal with a high clock state from multiple monitors. Getting it working was quick and easy. Seems nice and straight forward, and the interface is pretty nice.

Haven't poked around fan control or overclocking options actively yet though.

The moral of this story is the same as it would be for the motherboard, RAM or anything else: Do whatever you need to do to get it working properly at the stock speeds and voltage. Get all the snags out of the key components and aspects, like SLI. Clean up the system tray of all the updaters and other garbage.

It could be that the real culprit was something I deleted from the system tray, but I'd noticed too many anomalies running Afterburner. I'll have to check the version number and see what's available now. I'm just not that eager to try someone's "Beta" . . . . Like I said -- you could go from the frying pan into the fire.

One more to your remark. I also had a multi-monitor setup with the HDTV on a DVI-to-HDMI connection. Right now, the 2x 970 have a DVI-to-DVI connection to my BenQ. I expect a DP cable to arrive early next week.

But I disconnected the TV, and the anomaly apparently returned afterward. So with only the BenQ, the clocks are now behaving properly.

I still can't be sure whether this was a problem with Afterburner, the systray garbage, or bad driver installation. But when things improve like this, it's likely leading to progress . . . of some sort . . . anyway . . .

 

[kasakka]

Based on your problems you'd be better off just doing a BIOS mod for fans and overclock and forget Afterburner. Personally haven't had issues with it though.

 

[BonzaiDuck]

Based on your problems you'd be better off just doing a BIOS mod for fans and overclock and forget Afterburner. Personally haven't had issues with it though.

Yeah -- It's MSI's own damn software. You could speculate about a "lag" in NVidia driver versus Afterburner versions; there's the different (possible) experiences between Win 7 and Win 8.

All I can say now, though, is that it's behaving itself -- I think I've done about three restarts so far to settle at idle clocks of 135/162.

You can't create fan curves in Inspector. While you can access fan settings for each card individually, you can choose either "Auto" or a fixed % duty-cycle and speed. But I see the fans ramp up under "Auto" to between 50 and 70% with a tweaked two-GPU full load setting for Heaven Benchmark (force alternate frame rendering 1). Both GPUs will go to 98% load for Heaven, but the power-consumption is only 75% TDP (approximately). Temperatures max out somewhere around 68C. I think they'd be just over 70 with Kombustor with the "alternate" setting. But now, Power consumption is only about 11% with WMC running -- clocks between 135-270/162.

I know I'm headed in the right direction, because I'd even had trouble with Windows Media Center under its default "single GPU" SLI rendering mode, and I'd chosen the "alternate" for WMC previously. Now it all seems to work perfectly in "single GPU." [Knock on wood.]

Next thing I should do I suppose is to change the refresh rate toward its full extreme. Maybe I'll start with 120 Hz.

 

[Alpha0mega]

Try other software as well, if you are having problems with this one. Most, like Afterburner, are brand agnostic. I have used AB with MSI, ASUS and Zotac cards, and it works perfectly well with all of them. I just like AB because it has advanced features, no bling and I am familiar with it.

 

[BonzaiDuck]

Try other software as well, if you are having problems with this one. Most, like Afterburner, are brand agnostic. I have used AB with MSI, ASUS and Zotac cards, and it works perfectly well with all of them. I just like AB because it has advanced features, no bling and I am familiar with it.

I won't disagree but to say the only thing that Inspector lacks is a custom-fan-profile feature with the graphical presentation. It's either "Auto" or a fixed setting.

But the essential monitoring items are there, and all the clock and voltage tweaks. You can say "It even has slider controls." It doesn't have the Afterburner bling.

The MSI Afterburner was released in December, and the only NVidia driver version posted on the MSI downloads page is 347.25 and one earlier. 347.52 was released last month. Throw into that mix the possibility of OS differences, and I'm going to be happy without the blingy interface until they fix Afterburner, or GeForce driver, or both.

Back to the fan profile. I was going to insist on running those fans at a minimum 30% duty-cycle. Now that I see what the temperatures are like with these well-behaved idle clocks and how the fans ramp up under gaming or "Heaven" loads, I've concluded I really don't need to do anything but let them run at 0% with the 135/162 clocks. At worst, they settle down to maybe 42C before more time under idle puts them at (lemme-see here . . ) -- HAH! -- 34 and 36C respectively! [Oh! Oh! Ah'm afraid I'm going to burn up my Gaming Golden graphics cards!! I must tell Chicken-Little! :biggrin: ]

SCREW the fan-profiles!! AND the 500MB of differently-allocated RAM!! I'm beginning to LIKE these cards MORE! 7.2% of TDP power!!

Everybody wants to worry about how these cards perform at the top-end!! Of course they're pretty good at the top-end! But the low-end is just as important!

Anyway!! Inspector actually does have a "blingy" feature like Afterburner. It's got the little graphs! And it doesn't present all the extra stuff you can get by without, or which makes you scroll up and down to see the measures you need to see.

We just want to overclock effectively. We don't care about the skins and slider-controls and the endless monitoring presentations!

By the way, AlphaOmega!! Do you think that ASUS Tweak would work with the MSI cards? On the surface, that looks like it could be a risky leap. Then again -- well -- maybe somebody knows something before I put myself through more misery and frustration.

 

[Z15CAM]

I "HATE" USB GUI Software to control CPU Temps - Just use your MB BIOS CPU Temp Settings.

And as for as far as GPU temps - Use a simple Manual Restate Control.

Whether you're on Air or under Water

 

[Alpha0mega]

I actually meant that Afterburner does not have "bling" i.e. interfaces that are too cool for their own good, and become less readable. Look at EVGA's PrecisionX, at least its default skin (I don't know if it has other skins that are better).

I don't know about ASUS's own software specifically, but most card vendors' offerings aren't limited to their own cards, simply because there is no need. They (usually) don't add any hardware unique to their cards, so the software just talks to the firmware, and that's standard, despite any tweaks. MSI and EVGA's software are two that I am sure about working with other brands. Zotac does too, but not sure about that.

Not familiar with your situation, but I personally can't live without fan control. My cards get fairly toasty when gaming, and I need to have the fans spin up faster then. At idle, they spin at their slowest, and are silent. If your ambients are good enough to get by without ramping up the speeds, at load, then awesome.

Wasn't your problem also solved by having Afterburner start with Windows boot?

 

[BonzaiDuck]

I actually meant that Afterburner does not have "bling" i.e. interfaces that are too cool for their own good, and become less readable. Look at EVGA's PrecisionX, at least its default skin (I don't know if it has other skins that are better).

I don't know about ASUS's own software specifically, but most card vendors' offerings aren't limited to their own cards, simply because there is no need. They (usually) don't add any hardware unique to their cards, so the software just talks to the firmware, and that's standard, despite any tweaks. MSI and EVGA's software are two that I am sure about working with other brands. Zotac does too, but not sure about that.

Not familiar with your situation, but I personally can't live without fan control. My cards get fairly toasty when gaming, and I need to have the fans spin up faster then. At idle, they spin at their slowest, and are silent. If your ambients are good enough to get by without ramping up the speeds, at load, then awesome.

Wasn't your problem also solved by having Afterburner start with Windows boot?

I thought it was, but it didn't seem to be the case. I didn't change anything, or if I did, I can't remember what borked it. Like I said, in the "digital realm" you expect things to either work or not work, as opposed to having "random changes." All I can say is the second re-install of the NVidia drivers and excluding Geforce Experience (a possible factor) and some system tray updater garbage -- left it tip-top and consistent through several reboots, running WMC, playing Assetto Corsa for a half hour and GRID2 for another half hour -- and I did that two or three times.

Frankly, I wish I felt confident about Afterburner, but installing Afterburner seemed to be the threshold two times in a row. If I say it's "blingy," it's because Inspector and GPU_Z aren't -- with their little red time-series graphs. But those graphs do as much for you as Afterburner with its excellent concept and design.

What I'd also noticed about Afterburner -- the version I installed, anyway -- is that I'd see inconsistencies. I'd see AB reporting 40% power-consumption on both cores, then I'd terminate Afterburner and look at GPU-Z -- which has consistently shown about 23.5% of TDP. I also noticed that I'd get inconsistent clocks when cycling SLI, depending on whether AB was running or put to rest for the cycling.

And now that I think of it, what borked the promising status-quo this last time was a stop-code 124 BSOD -- possibly driven by the driver or a conflict with the driver. Time will tell, but . . . Dat's da bidnis we're in! Patience, frustration, endurance -- troubleshooting. The longer the time between events, the harder it is to pinpoint the cause or replicate them.

As for the heat -- the case has 2x 200mm intake fans rated above 144 CFM. All the air goes into the carefully sealed case and only goes out through the CPU cooler, the PSU's vents, and the graphics cards' vents.

It just takes longer for the cards to cool to (now it's) 34 and 37C at this likely 77F ambient (and farewell So-Cal winter -- it's warmin' up.) But it doesn't take too long for them to drop from 72C to 50C.

Put it another way. I'm a fanatic for thermal-fan-control. And those particular fans running at maybe 60% on their way to 70C or at a fixed 70% set in Inspector can't be heard any more than if they're at 0%. That's not because my system is noisy; it's noteworthy because it isn't noisy. It sounds like a top-end OEM sitting under some bureacrat's desk at lunchtime. If you hear some turbulence during momentary bursts or CPU speed, or at stress conditions, it just makes you think the system is just a little alive. I don't really recall hearing anything during gaming, but I also don't have the speakers turned up to a point of distortion.

I think I could just run the fans at 60% all the time, if I wanted. Or I could work some magic with some foam art board channeling some sidepanel intake down the back sides of the two cards: when you think about it, it would just be an odd-looking assembly but it would be simple, even so.


This is a plateau. It's like the McGee Creek trail in the Sierras, up those giant staircase cliffs on the way to McGee Lake. Halfway or "almost" there -- to the top. It all works like it's supposed to at stock settings, and I can still attempt to boost the clocks a little -- but with a little less abandon while I wait, watch for updates for AB and contemplate other similar software options.

How do you make a BIOS mod to a graphics card unless you program it yourself? Has anybody got such a thing? What are the risks? Do you have to do one card at a time?

I bet they fix this through a driver revision and a new AB version, though.

 

[digitaldurandal]

How do you make a BIOS mod to a graphics card unless you program it yourself? Has anybody got such a thing? What are the risks? Do you have to do one card at a time?

When I made changes to my GPU BIOS it was done via an XML editor basically. It was for the AMD 5870. Not sure if the same is true for Nvidia or not.

 

[kasakka]

How do you make a BIOS mod to a graphics card unless you program it yourself? Has anybody got such a thing? What are the risks? Do you have to do one card at a time?

http://www.overclock.net/f/71/nvidia-drivers-and-overclocking-software

Basically you need Maxwell BIOS Tweaker and your cards BIOS files plus nvflash to flash them back in. Yes, you do one card at a time preferably. No need to remove the cards, just disable them individually.

I did a fan speed mod for my Gigabyte GTX 970 cards and now they're very quiet on idle.

Of course you could always brick your card but if you're careful it should work out allright.

 

[BonzaiDuck]

http://www.overclock.net/f/71/nvidia-drivers-and-overclocking-software

Basically you need Maxwell BIOS Tweaker and your cards BIOS files plus nvflash to flash them back in. Yes, you do one card at a time preferably. No need to remove the cards, just disable them individually.

I did a fan speed mod for my Gigabyte GTX 970 cards and now they're very quiet on idle.

Of course you could always brick your card but if you're careful it should work out allright.

Thanks for the leads and guidance on this.

Just my perception that flashing the BIOS on a motherboard is a fairly mainstream and recommended practice when it comes to resolving some anomaly or shortcoming. BIOS mods to graphics cards are likely suggested only within our enthusiast community, but in forum sites of which Anandtech is only one.

One thing in the back of my mind while contemplating this as a possible next-step alternative is a comparison of the costs (with risks) and benefits.

What I see with these cards is that they run with idle temperatures in the mid-30's -- in my well-ventilated case. As I said, I have 2x 200mm fans running at half-speed under idle, but rated at 140+ CFM each at top-end. so my idle temperatures may be lower than many.

At default, when the cards are pushing beyond the 50C range, the fans turn on and I'd seen them running at ~60%. The GPU temperatures never exceed the temperatures I'd observed under equally-loaded SLI cards running under a custom AfterBurner fan-curve and using Kombustor for the comparison. So I have to ask myself if I really want to mod the fan-speed or its range on these cards, set at least one card to a fixed fan-speed or both with Inspector, or just leave it alone.

They seem quiet at 60% duty-cycle!

Really, the key to this is the "power bug" that may derive from interacting causes like Windows version, driver version and Afterburner version. If the cards are constantly running at between 24% (Inspector) and 40% (afterburner) TDP, then you want more fan RPMs at idle. If they run at 7.5 to 8% (Inspector or Afterburner, respectively), then it seems less of a concern.

At this point, my trust for the numbers AfterBurner reports (for my system) has fallen a bit. But I can verify the power draw at the wall with my UPS software -- with or without Afterburner's "power-consumption" monitor -- and it amounts to 60 to 80W.

And like I said -- there are some low-tech fixes of a tedious nature that could actually resolve the entire enchilada without introducing fan curves or BIOS mods. I can even imagine cooling two gfx cards perpetually with a single fan which also pulls in a pile of Intake CFMs, and that sort of prospect could actually work better than the cards' own dual fans.

Just a lot of trouble. Main thing: the idle wattage draw, and the idle temperatures. Somebody with less airflow in their case would be more eager to mod the BIOS of their gfx cards.

But the ultimate fix for this seems to derive from software: OS, system-tray clutter, and Afterburner. And Afterburner seems to be a common thread for those reporting the problem, even as its also a common thread with others like AlphaOmega who are using Win 8.

 

[PhIlLy ChEeSe]

I didn't read the whole thread but most of the first few posts, I use EVGA percision in SLI with 780TI. I've not had many issue's, on occasion I do had black screen crashes or hard locks. I attribute it to other stuff such as my CPU overclock. Always run my fans at full tilt, I'm just funny like that.
As I have the Asus DCii version of these cards the factory fans break easily and I run my system without a case, so I used a few intel fans I had sitting around as I cant wait a month to RMA the card then the same thing happens all over again..........

 

[BonzaiDuck]

I didn't read the whole thread but most of the first few posts, I use EVGA percision in SLI with 780TI. I've not had many issue's, on occasion I do had black screen crashes or hard locks. I attribute it to other stuff such as my CPU overclock. Always run my fans at full tilt, I'm just funny like that.
As I have the Asus DCii version of these cards the factory fans break easily and I run my system without a case, so I used a few intel fans I had sitting around as I cant wait a month to RMA the card then the same thing happens all over again..........

Too bad about those 780TI fans. Never thought about using Intel cooler fans, but they'd seem about right.

For running the fans at full tilt, maybe people would worry about how long they might last. I really never thought about it. and like I said -- on these MSI 970's -- higher RPM can't generate enough noise for me to hear them at all so it doesn't seem different than when the fans are at 0% at their coolest idle.

I have to put it into perspective, having become a nitpicker for perfection-with-simplicity. I want all the fans thermally controlled from either the mobo or graphics cards. You'd like to do it with a single piece of software like ASUS Suite Fan Xpert, but it's obvious why you can't, and obvious that someone like me would prefer the software that ships with a mobo as opposed to "Speed Fan" or some alternative.

Your own problem could well be derived from your CPU overclock, but I went through that aspect of troubleshooting last year, and learned a few things. As I said, a stop-code 124 BSOD can be caused by a bad driver or a good driver interacting with buggy software. The problem always becomes more difficult if the crashes/resets/BSODs are very infrequent.

We also have "forum-member-veteran" confirmation (with my own) that monitoring software can also be a source of the trouble, but most specifically when the user is running two or more monitoring softwares at the same time. In my case, there could be a problem between ASUS Suite which starts at boot time, and Afterburner -- which is polling some of the same sensors.

The designers of AFterburner ambitiously chose to add CPU-related sensors in their creation. That could seem nice from the aspect of making the software look "comprehensive." But it overlaps other proprietary software like AI Suite that users feel compelled to use for the rest of their fan-control or other aspects. I could imagine that this was a miscalculation ignoring users who have those needs.

On the one hand, they want a software program with so many features that it will increase the customer-base -- especially when so many folks use AB who don't even have MSI cards. On the other, the prudent thing to do would be to provide functions specific to the graphics cards, and let the mobo maker provide software with features specific to that product.