Hi I'm currently using EVGA PrecisionX as an overclock utility, but I'm having some problem with the memory clocks. Every time i change the slider to say +150MHz, hit apply and save the profile. But when I load the profile, it loads everything the way I set it EXCEPT the memory clock slider, which always goes back to zero. I'm not even sure it's applying the memory overclock in the first place, at this point. 
I have a GTX 970 G1 Gaming.
Anybody else having this issue or knows how to solve it?

Yes I have been having that problem with my GTX 970 FTW, in my case anytime I set a memory overclock and hit apply, the memory overclock resets to 0. It will not " take ".
 
I posted about this some time ago, thought I was the only one.
This happens to me with Precision X, even the legacy version , and it also happens with MSI Afterburner as well. No problem applying a core overclock, just the memory change will not take.  I finally just gave up on it.

same, tried both.. weird thing about mine is, it will work some times after restarting my computer, others it wont.

we are probably the only four people with this issue lol. so pretty much we are screwed. wont get support on this issue

same!! so thats 6 people
 

It's a problem for me too guys. It's getting frustrating. 

Indeed, would sure be nice to get this issue resolved.

I'll have to direct this further up.  I tried another test bench, another GPU and still haven't been able to reproduce the error but from the information posted so far I can't determine whether the source is hardware or software.\
 
For now, if you haven't wiped the drivers from the system, do so.  Then use the clean install option under the Custom selection when installing drivers.  Looks like some users have had success with DDU.

For me it's the other way around. I was folding and overclocking at the same time when I stumbled over this behavior. I paused the folding process and then the mem offset stuck.

I have this problem too, can't change the memory clock, resets to 0
 
Gtx 960 FTW

Still no progress on this? I'm on my 3rd 980TI SC that cannot perform a memory overclock. It can overclock on idle, but as soon as the GPU is loaded up, it returns to stock clocks, no matter the overclock. 

Just got a 960 4gb SC, same issue. Tried uninstalling drivers using DDU in safe mode + clean install option when reinstalling, reinstalled PrecisionX, I can only get the memory clock to set while the card isn't being utilized, then it gets "stuck" and the card reverts back to stock clocks when I put a full load on it. The slider is stuck until a reboot/uninstall at that point. Add me to the list!
Edit: Forgot to add, its also in a system with a 1060 6gb like the above poster.
 
double edit: Don't know what I did, it resolved itself.

Same with me, I just finished the thread going back to 2015, nobody has an answer. My memory clocks stay at 3304MHz when they should be at 3505MHz (2 980Ti Classified's, this effects both). I wonder if this has anything to do with that buggy Fall Creators Update.

Just saw this post by random chance, and have some idea of what's going on (possibly).
 
The problem is probably that the card is stuck in Performance State P2, and not shifting in to max performance, P0. It gets tied to the application you're using, so 3D games WILL unlock P0, which compute tasks may not. PrecisionX and Afterburner do not tell you which state you're in, so use GPU-Z or NVinspector to check, or the instructions below. Setting the clocks in one PState may or may not force the shift into P0.
 
For compute and games, set the card to Prefer Performance in the 3D settings.
If you're using the card for compute tasks, make sure to set the "Optimized for Compute" option in the control panel.
 
For 9xx cards, you can force P2 clocks to the same level as P0 using the application clocks.
 
Note first:

• This is not supported on 10x0 cards.
• I did NOT mess with voltages, as the process below is unpredictable enough. The worst that will happen is you'll get a display driver crash.
• The command to set the GPU clocks does not persist through reboots. You can easily make a quick bat file to run it, though.

 
To check your P-State:
 

• Fire up the program you're having trouble with or a similarly stressful application. You must be running something stressful.
• Open an admin console and enter:

"C:\Program Files\NVIDIA Corporation\NVSMI\nvidia-smi.exe" -q -d PERFORMANCE

 
 
If it says you're in P2, you can copy over the memory and GPU speeds from P0 by doing this:
 

• Check your supported clock speeds:

"C:\Program Files\NVIDIA Corporation\NVSMI\nvidia-smi.exe" -q -d SUPPORTED_CLOCKS

• This throws up a long list. Scroll to the top and note the first 2 entries. It should be something like:

Memory : 3505 MHz
  Graphics: 1622 MHz   

(DON'T USE THESE NUMBERS - USE YOUR OWN)

• If you have more than 1 GPU, you can specify which one by checking the order with:

"C:\Program Files\NVIDIA Corporation\NVSMI\nvidia-smi.exe" -q

         then adding -i (#-1) to any nvidia-smi commands, like

"C:\Program Files\NVIDIA Corporation\NVSMI\nvidia-smi.exe" -i 0 -q -d SUPPORTED_CLOCKS

        (This can vary. I have an Intel CPU with onboard graphics, so I think it lists that as GPU0 for general queries, but it's not counted for nvidia-specific ones.)

• Enter the command below:

"C:\Program Files\NVIDIA Corporation\NVSMI\nvidia-smi.exe" -ac Memory#,Graphics# 
(using your numbers from the last step, so ... -ac 3505,1622)

 
Now comes the questionable/I can't figure out the pattern part. Sometimes this will kick the card into P0. Other times it will be clocked as if in P0, but stay in P2. If you are overclocking from here, I THINK being stuck in P2 will cause the clocks to jump back to this new fake P0 speed if you try to change the memory clock.
 
If you do get into P0, overclocks should stick, but be aware that the overclock is on top of the the changes made above, depending on which software you use. What I mean is, you might see +300MHz on the memory clock after the tweak above, then overclock another 400MHz, so you'll see +700MHz. If you are in P2 and then somehow drop into P0, the software may change to show a drop to +400MHz, because the base in P0 is already +300MHz. It makes sense when you see it, but if you saved a profile/preset, it may try to apply that +700MHz to the new +300MHz baseline, causing your display driver to crash.
 
Where it gets weird(er), and speaking of display driver crashes, crashing the display driver has almost always pulled my 980Ti and 1080Ti out of P2 and into P0. No idea why. So, as weird as it sounds, you could try intentionally crashing the driver by setting the locks a little too high (don't go nuts with it).
 
Whatever the case, if you use profiles, you'll want to save 1 for if you're in P2 and 1 for if you're in P0 (for each card) and apply them as necessary.
 
Other things to try:
- Reboot and fully power down (unplug or switch off PSU).
- Overclocking from NVInspector (Show Overclocking then pull down P2 and unlock limits)
 
This is the edge of the rabbit hole. If you want to go further down it, you can look into WDDM 2.1 changes, CUDA 7 vs 8 vs 9, and the changes to the Nvidia display drivers around versions 375 and 350.
 
It is a royal mess and definitely not EVGA-specific.
 
Hope that helped someone.