You should check gpu load too. Maybe clock drops only when there is no job for gpu and it is waiting for cpu. 

You should try enabling boost lock or locking your clocks with msi afterburner along with enabling max performance mode inside the nvidia control panel.

the_Scarlet_one
is this only happening in RDR2? The game sounds like it is having major issues. What is your system setup and how is your COU reacting during these drops?

This is not a “normal” GPU boost 4.0 behavior as mentioned above. The core frequency spikes down and then back up rather than settling down one bin at a time.. 15mhz is not the same as a 100+ mhz plummet for >1 sec and then right back to level.

Please use a good monitoring software to monitor the CPU and see if it is having any adverse or strange drops in clocks or utilization issues.

Here's my setup. The issue seems to happen in every game I play. If you meant CPU, it's stable at 5MHz even when the GPU core clock drops.

 

kevinc313
That's crazy, you're running right into thermal limit at 88C and even before that the card is reducing boost clock by massive amounts.  To be running +115 on a FTW3 and only be getting 1940mhz means Boost 4.0 is taking out like 200-300mhz of clock due to the high temps.  Are you sure the radiator fan and pump are actually running? Is it getting air?  Is the fan curve actually working or is the rad fan sitting at 800rpm under load - try clicking the blue auto fan button on the main page?   Something is seriously wrong.

Fan and rad are the first things I checked when I realized there's clock drop. The fan works in both manual mode and auto mode with the fan curve. One of the rad hose is hotter than the other so I assume the rad is working as well. 
 
I ran the Kombuster at 2k for around 1 hour yesterday. The core clock stayed at 2010Mhz for maybe 1 or 2 mins and then gradually dropped to between 1850 and 1900Mhz but no huge drops as I showed in the graph. However, the temp never exceed 70 degree. Then I ran RDR2 again with the boost clock enabled for half hour. Same core clock drop, same 3-4 mins interval between drops and same temp spike.
 
Do you think it's something to do with the silicon or it's simply just a bad card/rad/fan? 

Or maybe your fan was on low speeds...

That's because your GPU is hitting the Temperature wall. That window you show in there, where the line goes across and drops to the floor. Compare that with your GPU peak temp which is paired with your clock drops..
 
You're basically running that thing up to max temp, clock dropping, temp drops, then you run up to max, and the clock drops again. 
 
It's saving your card
 
I'm not sure what your system setup is, but you might have a serious problem with cooling, or you might be pushing it too hard with an OC. You shouldn't be hitting 90+ or 85+. Looks like your wall is around 87
 
 
What you should do, is default the card in X1, so it goes back to stock, (wipe any profiles), and reboot, then see what kind of temps you are running in windows. Check your room ambient temp so you have something to compare against. If you have a WC loop on the GPU, (and iirc this is a hybrid, so it's air and water), then make sure there are no kinks or tight bends in the tubes, check your rad to see if it's clear of dust, etc... 
 
You can also check to make sure your system is getting good airflow, and that goes for Rads as well, and then compare the temp between the GPU itself, and the memory on the card. Those cards are I think air cooled on memory, and WC on GPU, so while you might see a difference anyway, it probably won't be anything major. If your GPU is a hot spot and the memory is cool, it might indicate a problem with your Pump or Rad. 
 
Last thing I would do is Scan for an OC, and let the card find something using Precision, then see how that performs. 
 
 
Actually, seeing your build there above my post, I'd say pull the fan on top that's right in front of your Rad. It's pulling and your Rad is pulling from the same area. So you've basically got 2 fans pushing out through the front Rad, or pulling if you switched them, (I did), 2 fans exhausting out the top, and your GPU fan exhausting out through your Rad. It's a negative pressure system as far as I can see, and the top fans are not pushing air through a Rad, while your GPU fan is.
 
Anyway, I'd fix that. It's a very close fit for everything in that chassis, and whatever air is entering is going to be entering through every crack and hole in it, sucking dust in at the same time. Mostly, that would be the bottom if it's ventilated, which will cool your GPUs memory at least. I would consider swapping your CLC fans to pull, intaking air instead of exhaust, leave your GPU on Push, and get rid of that top fan in front of it. If you've got the holes on the top of your chassis, swap the top front fan to a larger fan and make it an exhaust, then put some mesh or something over the old top rear fan hole to restrict airflow there. Tape it down from the top under the grill, or something, (even a bit of crazy glue dabbed on the mesh at the corners and put on the top over the fan hole would work).
 
And when you can, I'd suggest moving to a new chassis with some more room in it and better airflow. I have the Corsair Air 540 and it works almost perfectly for my system. I won't say I love it, but I like it. Keeps the PSU and SSDs out of the main component area, and gives me good airflow.
 
I use a Pull configuration on the CLC because it brings air in from the room, rather than pulling it from the inside of the chassis, and my CPU rarely hits 50C Most of the time it's 32, so the air going through it is barely adjustment from room temp. Exhaust on two top fans, (I got rid of all the original case fans except the rear fan), and exhaust on the rear fan. PSU is in the second compartment and pretty much draws from outside and exhausts independent of the rest of the system. 

iLLuser
Think I found the cause. Seems like the rad fan is pulling to much power from the card. the harder the fan blows, the higher the temp. As soon as I connect the rad fan to the mb, the temp spike's gone and it stays below 60 degree, which I believe is a reasonable temp for the hybrid card? Correct me if I'm wrong. But still the core clock is not staying above the 2000Mhz line with a +125 overclock. Is that just how the Boost 4.0 works?

 
Here's my PC setup if you are interested. I know it's a really tight fit cause I built it for my old 1080. Didn't realize 2080ti's that big till I get it lol. Big thanks to everyone.

 
 

 
Your card should clock to about 2100mhz at a sub 30C temp with the right type of light load, on just the stock clocks.  Try the splash screen on this website:
 
https://www.ontvtonight.com/
 
With temps in the 50's and a reasonably high load and stock clocks, the card should clock to around 2055mhz (three 15mhz steps, one step per 10C) if everything is ok, so something else is going on for you to only get 1995mhz (four more steps down).  I think the boost clock and OC settings you're adding in X1 either isn't actually being applied, or somehow it is kicking out and defaulting to stock clocks/settings without crashing the application you're trying to run.  +125 (actually +120, 15mhz increments) should give a clock of 2220mhz below 30C with moderate load, which might be enough to make the card reset OC settings. Under full load with +120 and a temp in the 50's your clocks should be about 2175mhz, if its actually stable, it easily may not be. So you're like 12 steps down if your OC is actually being applied.
 
Your temps look good for a card operating at stock clocks and power levels with your fan config, and unrealistically low for max power and an OC.  I'm pretty sure X1 isn't doing anything.  However, the fact that you're hitting precisely the 88C OC temp limit doesn't support that theory.
 
It may be possible that without the bump to the power limit and core voltage limits that you would normally apply taking effect, the card can't maintain the target clocks when under load, hence the extra steps of downclock?  Just a guess.
 
I'd wipe X1 and install MSI Afterburner.  The single fan slider will control both fans simultaneously as a percentage.
 
You'll get better cooling regardless if you move the hybrid rad to a front middle intake push fan and the cpu aio to top mount exhaust.  Push pull fans on the hybrid rad if you can fit them.
 
You're also on the hairy edge of PSU use with that 650w unit especially if you're pushing the CPU and GPU hard simultaneously with 5 fans, 2 pumps, 2 HDD's, 2 SDD's.  Try unplugging your HDD's and use a GPU load that has low CPU load.
 
 

Bump - dying to know what was causing this.

Well, TS said that he just connected rad to motherboard and problem is gone.
My guess that he used manual fan curve and faced same bug that I did: every time time fan speed updates - fans actually turns off and start up again from 0% to needed. That means that average fan RPM is very low and gpu is overheating like hell.

btw I can see proof that TS faced that bug: 1st of all - manual fan curve is enabled, 2nd - rad fan is set to 65%, but actual RPM is 800, and that is 40%: it must be about 1300.
There is X1 screenshots in 1st message I missed before :)
 
I recorded that behaviour once for evga support and used mic to check actual fan RPM:
https://www.youtube.com/watch?v=h602HsYw8sI
 

Timeline:
00:00 - 03:35: I am running fan on auto and warming up card with OC test - no issues here, fan never stops or something, you can confirm that by RPM values and fan sound. So fan react to temperature and spin up or slow down correctly. 
03:35 - 06:38: I am running manual (fixed RPM mode) - and there is issue: every time I press "apply" to change fan speed (even it is VRM fan) - radiator fan (FAN2) actually turns off and drops RPM to zero and only after that it starts slowly speed up. For example, at 5:10 I am changing fan from 100% to 80%: but it actually turns off and it takes 25 seconds to reaches 80% after I pressed "apply". Same for next test from 80% to 90%.
06:38 - 11:10: I am using fan curve, but there is an issue: every time fan speed updates (based on "fan speed update period" - FAN2 actually turns off so on default settings it turns off every second (7:23-7:50). I changing update period - and that affect how often fan turns off. At 9:30 I am starting another OC test to heat up GPU with next settings: update period - 30s, fan curve is flat: under 50C - 40%, above 50C - 70%. 10:00 - GPU reached 50C and fan settings changed, but FAN2 turning off every 30 seconds.
11:10 - 14:10: auto (bios controlled) again - and again there is no issues like it were on 1st part of video.