My processor's frequencies are fluctuating wildly!

A 2700X outperforms an i7-7700K in gaming, so you can say it really packs a punch. And those 8 SMT cores won't hold back the 2080TI.
On paper, the setup is solid.

Something bothers me: just one exhaust fan? Sure, I overclock, but I have 6, personally!
Try to max out the ventilation possible. Given the price of the setup, we can afford that ;-)

As for overclocking, +0.204V is nearly overkill; increasing the voltage boosts stability but also increases consumption and thus... heat. If you can lower it as long as the system doesn’t crash, do so. Either your CPU will work, or the system will crash. It’s that simple. In any case, it’s better to deal with a BSOD because of "undervolting" than to fry your CPU due to "overvolting."

The MIN framerate is what’s bugging you. That’s typical of a CPU bottleneck. I wouldn’t be surprised if it’s overheating... Ryzen (and Ryzen 2 for sure) made significant strides in heat management, even teaching Intel a lesson, but it’s still high-end... it doesn’t cool like an Atom.

If you want advice, go into detail, provide the specs of your case and your budget, and I’ll recommend what to buy. Photos would also be helpful ;-)

As for advice on Ryzen overclocking... sorry. I haven't touched an AMD CPU motherboard in 6 years, and I didn't overclock it either :-D You’ll need to dig into the tutorials, like the one I sent you.
But once again, regarding cooling, I can give you some pertinent answers.

Okay

First of all, the CPU should not exceed 85°C, otherwise it will reduce its frequency to limit overheating, which is known as "thermal throttling." See here the CPU specs.

The graphics card does not like to heat up to that extent at all. Mine never exceeds 50°C while gaming with the overclocking I've applied. For that, you just need to adjust the fan curve in Afterburner...
Personally, my curve is (was) simple: 0% below 30°, then a straight line up to 100% at 50°. If the GPU hits 50°, I want 100% of the available fan speed of the card :-D

For CPU overclocking, feel free to experiment, in any case, if you don't put 2V to your CPU it should be fine :-) Take it step by step, and check the MAX voltages from different Ryzen overclocking tutorials you’ll find, it will give you an idea of the MAX voltage you can afford to reach.
Example: My CPU is not supposed to exceed 1.3V (4.4GHz for me) but some have attempted a temporary OC at 1.46V. I knew I had some headroom, so! I pushed to 1.37V, I gained 300MHz. And it is stable 24/7. Of course, the CPU cooling is keeping up behind it

Now, let's talk about cooling!

- completely revisit your cable management! Here’s what mine looks like: Day, night.
The picture may look bad, but apart from the watercooling pipes and the PCI-e cables from the graphics card, NO cable is wandering around. They are ALL placed against the case.
There's just one you see "going out" from the GPU, it manages a fan on the case door. I put it back in place when I close the case ;-)
- replace the front fans with NF-A12x25 (I don’t know where your 2x120mm are pulling air from, but it doesn’t seem large, prioritize static pressure over airflow in this case)
- set all your other fans (those already present here) to exhaust (2 on top, 1 at the back)

And don't forget to
- revisit your GPU fan curve in Afterburner
- revisit the case fan curve in the BIOS
- follow overclocking tutorials. It takes some time to learn, it’s not just "putting in larger values," but the gain is worth it: +42% performance for me :-D

So

First of all, the cable management is spot on for a start.
It's not perfect, but it's sufficient for now so you don't need to worry too much about it.
When you start overclocking seriously, trying to gain every MHz, trying for that extra 0.01V... you will be hunting for every lost Celsius degree :-D Then it'll be interesting to optimize everything.

Placing the rear fan at the top is a good thing. Hot air rises, and the ATX standard is designed for that (no joke, but I can't remember where I saw that)

If the graphics card fans are too loud at 100%, set them to 70% at 40° and leave it at that. That would give you a curve that goes from 0% at 30°, then a straight line up to 70% at 40°, followed by a horizontal line at 70%. Still, let a line start at 60° (still 70%, then) up to 70° for 100%. At 70°, the card shouldn't exceed that. If your GPU is at 70°, your VRMs are at least at 90°!

From there, keep the power limit at 100%. No need to go higher, the Pascal architecture doesn't like exceeding the 100% power limit. It should be the same with Turing.

You don't need to change your fans.
In intake mode, given the little airflow coming from the front, you MUST switch to effective hardware. But your current three fans, which provide plenty of airflow since there are three of them, don't need to be replaced. You will set them up for exhaust at the back (1) and at the top (2). Over-pressure or under-pressure are generally to be avoided, but two very good fans will almost be worth three average or even good fans.
Check this out, you'll like it ;-) Noctua 120mm, gray, highly efficient in static pressure, silent, PWM, and at less than €19 each, I don't know what more you need!
Take two, that will work. Those will go in the front, just a reminder ;-)

Normally, the airflow will change everything. There will be more air, the fans will spin slower (silence!) but it'll move enough to cool the graphics card (silence!)
In short, the more fans you have, the quieter it is.
My PC is in my bedroom and sometimes it stays on overnight... I actually have 11 case fans! And 3 for the graphics card! I can't hear them from 3m away from my bed.

From there... well, you just have to enjoy yourself ;-)

Okay, you've followed EXACTLY everything I told you, so it's good :-)
The cable management is clean now! It's perfect. Going forward, use zip ties; that's the secret to perfect cable management. That way, even at the back of the case, it's clean.

The fan curve is as I told you, perfect.

You still need to improve the ventilation. Adding 2 fans, which are also suited and sized to your needs, will really make a difference. That's the only reason your card heats up so much.
Some even go as far as replacing the thermal paste of the graphics card with Kryonaut; I did it with my previous cards. You really gain from it.

OKAY

This is problematic.
As for the assembly, nothing to say, it's clean, everything is fine.
You frankly shouldn't be hitting 70° without overclocking at 70% fan speed! Is the fan curve still the same?

Let's assume that's the case.
First of all, don't touch your power limit. If there's a problem, it doesn't come from there. Trust me, my GPU doesn't have such temperatures, and its power limit is at 100%.

Do a test. AT IDLE, set the fan speed of the card to 100%. Watch the temperature drop, and once it's stabilized, note the obtained temperature.
From there, launch Far Cry, and note the maximum temperature.
This is just to see how your card behaves at full fan speed.
Clearly, overclocked, it has NEVER exceeded 52° at 100% fan speed in-game. And I don't think I've ever exceeded it in benchmarks, that's saying something! If it goes above 52°, there's a cooling problem with the card.

During the test in-game, check if all the fans are working. Especially the graphics card fans.

How do you control the case fan speeds? If it's done through the BIOS, take a look in there and adjust the speeds. Personally, at my place, my fans are set like this:
0% until 30°
flat line from 30° to 50°
If my CPU exceeds 50°, I want 100% of the available ventilation. For my case, it's different, my CPU cannot exceed 64°. But nothing prevents you from testing this curve, it's still fine for a CPU handling 85°. It won't hurt it, and you can be sure you'll have good airflow in the case while gaming; your CPU should definitely exceed 50°.

Replacing the thermal paste on the GPU does not void any warranty if there is no warranty sticker on the mounting screws. To remove the GPU from its cooling, normally 4 screws are enough. They are the ones that form a square closest to the GPU. CPU thermal paste will work very well; I used NT-H1 on my GPUs.
Again, Kryonaut is the most suitable: it's very effective plus its extremely high viscosity. The application is a bit more delicate, you feel like you're applying glue, but it works a ton.

As for the procedure to apply a ventilation profile in Afterburner, here is the sequence to follow to ensure it is properly applied.
Assuming that your current profile is profile 1
I imagine you already know it, but if it’s coming from that, then that’s something.

Okay

First of all, your fan curve dropping from 39~40° to 38° while you're supposed to have the fan speeds going from 70% (barely) to 100% is not normal.
Mine goes down to 23°, the room temperature! It takes some time, sure, but it gradually approaches that value.

Make sure all your fans are spinning, and that all the custom curves in the BIOS are set correctly.
The recorded speeds are consistent with the curve. The temperatures are not.
We can test another GPU curve:
a straight line at 40% from 0° to 30°, then a straight line from 30° (40%) to 65° (100%).

The first part of the curve will provide a constant airflow in the card while keeping noise levels very low. This will prevent the card from reaching high temperatures when the system starts.
Since you’re not overclocking heavily on your card, you can apply this same profile at Windows startup after saving it in a profile. If you start hitting the red zone (and right now, conditions don't allow it so I wouldn't recommend it), think about removing this setting ;-)

We have a lead to resolve your custom fan curve issue. Apparently, there was a problem on that side. 31° is the temperature you should have with the fans at 40%.

Now, another problem.
Fan management in the BIOS.

The options in front of you are very simple:
- The curve, we're in agreement on that

- CPU FAN Speed Control: custom management in "Manual", that's your case, we're good

- Fan Control Use Temperature Input: source, thermometer used to measure the temperature to correlate with your curve: CPU
We're good, I don't think there are other parameters than "CPU" and "Chipset". From the two, choose the CPU.
THIS MEANS that your CPU is the source. If it doesn't heat up too much, the fans won't spin fast, so the airflow in the case won't be great, meaning the GPU will struggle to cool down.
There's your "other problem."
You should, at the very least, be more aggressive in the curve. Already, keeping them at 50% (or 40% if you can) until 30° and then ramping up to 100% at 50° will make them faster on average, this curve never crossing the current curve until their maximum at 50°. And it will provide a constant airflow in your case. Your GPU will still be cooled down, even in the absence of tasks.

- Temperature interval: 3. This is a time in seconds between the different measurements, you can lower it to 1. Don't go below that, it will drain performance otherwise.

- CPU FAN Control mode: PWM (perfect)
PWM is a way to manage the fan.
There are two types:
- voltage variation (classic)
- pulse width modulation (PWM)
The first will simply deliver more or less voltage. When you ask your fan to run at 70%, your motherboard will supply 8.4V.
The problem is that the fan motor is not optimal at 8.4V; it will start losing efficiency and generating unwanted vibrations. Clearly to be avoided when you can use PWM.
The second, PWM, is a management via modulation. When you ask your fan to run at 70%, for a given period (short generally), the fan will run at 12V 70% of the time. For the remaining 30%, it won't be powered.
This technique allows the motor to run consistently at its cruising voltage, thereby avoiding too much loss in efficiency and limiting vibrations.
The only check to make is to ensure that the fans, as well as the motherboard connectors, support PWM.
If you have 3 connectors, it's voltage variation. If you have 4 connectors, it's PWM.

- CPU Fan Stop: Allows the fans to stop for silence. To be avoided in your case since we're facing a temperature issue.


Make sure all the BIOS curves are the same and correct, check that the fans are all spinning (if you consider the profiles above, they should ALL be running constantly) and look at the results.

If all these changes do not resolve anything, try returning your card for RMA. It will save you from disassembly manipulations that could unnecessarily void the warranty. And go directly to the store for that.

I also forgot to mention that I did a new benchmark with Userbenchmark and my graphics card literally crushes the average scores: 230%. Whereas before I was at 203% (with overclocking).

However, there are no results for "DirectX11", probably because I haven't installed the appropriate software (?). Is it serious? Is it because of this that the graphics card behaves better? Do I have to install DX11 and 12 even though my GPU seems to be working better without it?

\o/
Wow! There have been a lot of proposals around here ^^' Not a single one was about the graphics driver settings. I'm still surprised that it comes from a setting, but it clearly is the case
By the way, aside from MSI Afterburner, I don't touch any other graphic settings ;-) G-Sync in Full-screen Only at most

when I switch from my game to another window, for example Chrome, the graphics card starts to be heavily utilized, its temperature skyrockets to over 70 °C and the coil whine kicks back in. As soon as I switch back to the game, it instantly calms down
I have that too. However, my fan profile remains active so even if the card gets excited, the fans cool it down.

maybe Far Cry 5 is simply not demanding enough [...] I am re-downloading Assassin's Creed Odyssey and GTA5 (to) test the "new" behavior [...] with these 2 other games
If your problem has disappeared in Far Cry 5 (which is demanding! Let's not kid ourselves, it's still AAA), it should disappear in the others as well

However, there are no results for "DirectX11", probably because I haven't installed the appropriate software (?).
DX11 is integrated into Windows and/or the games.
That’s not the reason for performance loss!
The last time I had that, it was because the DX11 benchmark test crashed due to excessive greed in the OC frequencies.

Do I have to install DX11 and 12 even though my GPU seems to work better without them?
There, on the other hand, if you download a "pseudo-DX", give up. DX is integrated in the games at worst. Not for direct download.

Certainly because your card is not being utilized to its full potential in FC5/GTAV with a max fps of 60 ;-)
I run GTA V on ultra at 3440x1440 (1080ti), it fluctuates between 90 and 110 fps. With scenes hitting 160.
In the end, these games may not be the most demanding on the list...

Why do temperatures vary so much from one game to another?!
A more demanding game will push your card to 100%, thus making it heat up.

I just checked. All the tests of the 2080ti Zotac concern the triple-fan, or the AMP!. But not the dual-fan.
It smells like a discount 2080ti without being rude. And if that's the case, you're going to have to choose between silence and performance ;-) The right compromise is only possible with an adequate cooling system.
I suggested a warranty return on 02/22, but it might be better to proceed with an exchange for a more expensive, but better-built model.

I would like to see an official test of the 2080ti twin fan; I haven't found one.

Sorry, this is for me the only logical explanation.

I'll let you read:
Like previous GeForce GTX models, the GTX 1080 uses a technology for turbo and monitoring its vital parameters: GPU Boost. It operates as follows as soon as an application is launched:
- The GPU is pushed to its maximum voltage and frequency
- A monitoring system kicks in and checks the GPU every 100ms
- If the total power consumption of the card reaches the limit, voltage and frequency are reduced by 6.25 mV/13MHz
- The fan gradually speeds up until it reaches 82 °C
- If the GPU reaches 83 °C, voltage and frequency are reduced until it returns to 82 °C
- If the GPU drops to its base frequency, the temperature can rise above 82 °C
- If the GPU reaches 92 °C, the frequency is significantly reduced and the fan ramps up to 100%
- If the GPU reaches 94 °C, the card shuts down
Source

With this, you should have the necessary information to make LDLC's warranty work ;-)

Thank you for your response,

at the risk of coming across as slow, I don't really understand how this could compel them to exchange the card for me... What you highlighted in bold is just a security system, isn't it?

Well, it's not supposed to activate without overclocking
In itself, your card can't deliver 100% of its performance "out of the box"

In comparison, I have... never reached the temperature limits of my card, and if I ever did, in any case, I had already started to push the limits with "interesting" overclocking.

Alright, but the problem is that when letting the motherboard manage the fans, it doesn't work, since it automatically increases them to around 65/70%. So in terms of warranty, that won't pass...

Well, I tried a little experiment: instead of positioning the 2 Noctua fans at the front of the case, I decided to attach them together with zip ties so that they are side by side, and, with a bit of tinkering, I managed to secure them just underneath the graphics card.

I was so happy when I managed to get them to stay, already believing in a miracle. And believe it or not, but that doesn't change the temperatures at all! It's really starting to piss me off.

Normally with 2 Noctua fans running at full speed at 1700 RPM just 2 cm from the graphics card, it should at least cool it down by 1 or 2 degrees, right? Can someone explain the logic here?



Furthermore, I have the feeling there is a resolution issue that might be causing unnecessary calculations by the graphics card (pure hypothesis, especially since I don't know much lol)

Earlier, I launched a game (Monster Hunter World) at 1600*900, the graphics card starts to rise to 80 °C and the fan ramps up to 100% automatically until it drops back to 79 °C. Uh, what's with this madness?!

So to test, I decided to switch to 800*600 in Windows display settings and in-game, to see how it went. No overheating issues (thankfully!). I then exited the game, switched back to 1600*900 in Windows, relaunched the game and adjusted the resolution (1600*900), and now, no overheating issues... Except that now the image is absolutely disgusting, it's super pixelated, with huge aliasing effects everywhere, despite having activated the most powerful Antialiasing in the game. It's still the resolution of my screen. And I've noticed this on all my games: as soon as I'm in native resolution, the image is so pixelated and blurry that it feels like I'm in 800*600 (I'm not exaggerating).

Could activating DSR in the Nvidia control panel have something to do with this (the resolution scaling wasn't activated in the game, and I believe the option doesn't even exist)?

How is it possible to reach such high temperatures without even being in full HD? I thought I had seen it all...

P.S.: the card peaks at over 95% usage when activating DSR x4 (so in 3200*1800). At 900p, we're at about 50%. What will it be in 4k? An explosion?

"Okay, but the problem is that by letting the card manage the fans, it doesn't happen because it automatically increases them to about 65/70%. So in terms of warranty, that's not going to work...
When do you leave your fans at that speed? It overheats, that's a fact. You just have to deal with it. I don't remember my suggested fan profile forbidding reaching that speed...
"straight line at 40% from 0° to 30°, then straight line from 30° (40%) to 65° (100%)"

In any case, it's too late to initiate the exchange now. The legal 15 days have expired quite some time ago.

"I would like to see an official test of the 2080ti twin fan; I haven't found any": Have you found one since then?

Earlier, I launched a game (Monster Hunter World) at 1600*900, the graphics card started to reach 80 °C and the fan ramped to 100% automatically until it cooled back down to 79 °C. What the hell is this?!
I completely agree, something's wrong. At that resolution, the GPU is so underutilized that a 2070 or even a 2060 could produce as many FPS as your 2080ti, as lower resolutions are much more CPU intensive rather than GPU intensive.

"So to run a test, I decided to switch to 800*600 in the Windows display settings and in the game to see how it goes. No overheating issues (thankfully!). I then quit the game, switched back to 1600*900 in Windows, restarted the game, and adjusted the resolution (1600*900, of course), and now, no overheating issues... Except that now the image is absolutely terrible, it’s super pixelated, with huge staircase effects everywhere, despite having enabled the most powerful Antialiasing in the game. Yet, it's the resolution of my screen. And I notice this in all my games: as soon as I'm at native resolution, the image is so pixelated and blurry that I feel like I’m at 800*600 (I’m not exaggerating much).

Could the fact that I activated DSR in the Nvidia control panel be the cause (the resolution scaling wasn’t activated in the game; by the way, I think that option doesn’t even exist)?"
Are you sure you uninstalled and reinstalled the nVidia drivers properly?

"Normally, with 2 Noctua running at full speed at 1700 RPM just 2 cm from the graphics card, it should at least cool it by 1 or 2 degrees, right? Can someone explain the logic?"
Let me explain, and you won’t like it.
If I made you install these Noctua from the start, it's because I don't know where your 2x120mm are getting their air from, but it doesn't seem like a lot, prioritize static pressure over airflow in this case. In other words, your case doesn’t seem great for cooling, to put it mildly. I have no idea how air is getting into the case.
If you can open the front panel, do it. And put those Noctua back in their place :-)
From there, that should fix all your temperature issues. A case with good airflow is the key to effective cooling."

I admit, it’s not normal
Your 53° is the temperature I have at home, so it matches, it’s even pretty good

Make sure to check the processes that are running. You can sort the processes by GPU usage in the task manager on Windows 10 for some time now

That was the last advice of the evening :-)

Hello!

Not bad this feedback. I had already looked at this but I didn't think it was that effective...

I decided to keep the fan speed constantly at 100%
Same for me. I went through the same phase as you, except that I have a Kraken G12 with a 360mm on top. Maxed out at 57° yesterday, during a heatwave with 37° in the room. The VRMs didn't suffer too much since there were 8 hours of gaming without artifacts or anything. Just 2 or 3 crashes but that was the CPU overheating
Just goes to show...

Your noise might be coming from the PSU fan ^^ There’s no reason for your fans to speed up if ALL your fans are manually set to 100%. By the way, that’s also what I had to do this week to make sure the system would stay stable.

Uh no.
Honestly, don't take it apart.
There are capacitors and coils in there! If you disassemble it and by bad luck touch a component on the primary side, you'll get a shock like never before. If only that was all... if you reassemble it poorly, you create a short circuit... your PC catches fire.
I've experienced that, without disassembling anything, the power supply at the end of its life delivered 12v on all lines... the HDDs burned. If I hadn't been there, the apartment would have gone up in flames.

Instead, buy a good power supply! Lepa, Seasonic... If your power supply is really an 80+ Bronze or lower, if it's over 3 years old, etc... and especially if the noise is unbearable.

That said, be sure of this. Put your ear close to the power supply before and during charging.

No.

First, make sure the problem comes from there

And if you insist on changing it (because there are no other options), well
here it is

Your power supply should be at least 550W according to this calculator. I went with a safe margin.

A RM750x is a very good choice. You're getting into the high-end range from Corsair. According to Cowcotland, the voltages are very stable and the consumption is reasonable (thanks to the 80+ Gold certification, or how to pay off your power supply on your EDF bill)

Perfect then :-)

So is everything settled now?

Sorry for the duplication.

I saw your PM, I'm going to respond here just in case someone else has the faith to read this whole post

Seasonic is excellent. It's often Seasonic rebranded or another specialized brand hiding under power supplies like "Cooler Master" or "Corsair." I'm even sure that Seasonic is hiding in the RM range.

If you can get an official modular Seasonic Gold 750W, go for it

I hope so too, honestly changing the cooler on my 1080ti Strix OC has made a difference. And it's not bad at all by default. The one on your GPU can't be better, but I didn't think the Arctic Accelero would be so good :-) so much the better, I'll know

Looking forward to helping out :-) I've learned more here by helping people than on my own, and I like that my advice is useful