I agree it should be completely fine at those temps. On a slightly related note, I recently moved to water-cooling my CPU for the first time, primarily to make it quieter. I was surprised by how easy it was. My PC is now silent when not gaming and still quiet under heavy load (GPU is still air cooled). And it dropped my CPU temp by about 20 degrees. I’m never going back.
I’m not sure what, in relation to my post, you’re agreeing to-- that the GPU is fine getting up to mid 80s C when I’m gaming? It peaked at 86 briefly, but spent an uncomfortable amount of time at 85. According to what I’ve read online, that’s technically just within spec, but higher than it should be. Especially for long periods of time.
That sounds like it’s slotted into the secondary PCIe slot, which would be suboptimal. Is there another slot higher up?
No, unfortunately, it doesn’t look like there’s another, higher up slot to move it into.
From a quick search, 85c peak is not crazy high for a 3060 (at least several people reported it in reddit threads about temperature) but the thermal throttle appears to be around 93c, which is closer than I’d like. Your artifacting / hiccups / does sound like it could be thermal throttling, and it sounds like your GPU is probably a more likely candidate than your CPU.
That said, unlike CPUs, the cooling in GPUs is not meant to be user servicible. It can be more complicated than it seems. Often there are thermal pads that are sized for a particular part of the heatsink and aren’t interchangable or movable. You have to be careful to disassemble it and reassemble it exactly the same way. I generally would not recommend it.
Better things to try first are changing the fan curves, as you mentioned, running the fan at high speeds at lower temperatures. You can also try undervolting it which reduces temperature generation in general. Most cards have some room to undervolt before they become unstable. Both of these things you can do without physically modifying or disassembling the card.
And this may be overkill or something you wouldn’t want to do, but you could always just remove the side of the case and take a box fan or other large sized fan and have it blow directly into the case to overcome airflow issues. You don’t want to put it directly up against the case - fans have a powerful magnet - but blowing into it is going to give you a much more powerful airflow than the case fans. I don’t know if that’s practical based on where you keep your case, and it’s a bit of an ugly workaround, but it’s better than having a PC that can’t do what you want it to do.
You might want to post in the buildapc subreddit - that’ll give you access to a much bigger and more current group of hobbyists than the SDMB will.
Thanks, some great advice.
But just to be clear, are you saying you wouldn’t recommend trying to separate the fan / heatsink from the GPU card and replace the thermal paste, per this video? The guy in the video did say it was kind of hard to separate them, to the point he was afraid he might break something, which did make me nervous to think of trying it. But when he did manage to get them separated, it seemed to be a simple matter to replace the paste and reassemble. And after doing a thorough internal cleaning and replacing the paste, it was running like 12-20C cooler in benchmark testing.
I suppose I could just remove the card and heatsink combo from the slot just to get better access to blow out the fan with some canned air, and try that first.
I personally would never do that. I consider it akin to soldering your motherboard. Way above my pay grade.
This isn’t something I’ve ever done personally, I’ve never opened a video card, so I can’t really give personal advice. I’ve read people advising against it on pc building forums on the basis that each cooling solution is unique to the manufacturer and some involve more complicated construction than others, like differently sized thermal pads for instance that need to be replaced correctly. Your video makes it look quite easy, so I wouldn’t be afraid of doing what he did in that video, but I’m not sure if your particular card has the same cooling setup. This one is beyond my familiarity to really make good recommendations. It might end up being viable and a good idea, but I did want to caution that I’ve read that it can be more complicated or finnicky than replacing the thermal material on a CPU.
Pulling the card out and blowing it out is a good first step. Cans of compressed air should come with a straw and electric air compressors should come with a small nozzle that ejects really high pressure gas - you can use that straw to get right up into the hard to reach places to really thoroughly dust out the card. Just make sure you aim horizontally. Tilting the can vertically can spray the refrigerant.
I’ve never really had thermal issues before and I’ve never used small form factor PCs so I’m not operating from personal experience here and just giving general sort of advice. Like I said, you might get better, more specific advice on a hobbyist PC forum
Thanks; I will abandon the idea of taking apart my GPU then and replacing the thermal paste / pad, or at least saving it as a last resort (and making sure any how-to video I follow is for my exact model of video card). Will try air blow-out and changing fan curves first, maybe lower voltage.
If you’re seeing glitches more frequently than before (2077 is a glitchy game and it’s easy to be paranoid when you’re looking for problems), the first steps should be to disable eco mode and revert any undervolting if you did that.
You really don’t need to regularly replace thermal paste. If you’re having thermal issues (I don’t think you are), my first suspect would be that 90mm exhaust fan. Even more so if the GPU isn’t blower-style. An easy test here is to run it with the side panel off for a bit.
But seriously, don’t do anything dramatic unless the PC is thermal throttling.
Yeah, I keep hearing this, not just here but from videos I’ve been watching. And honestly, if it makes a significant difference in cooling, heck, I’ll just take the panel off anytime I’m gaming (which is not that often for me), and maybe get a fan and have it blowing in as well, like @SenorBeef suggested. Maybe not the most elegant solution, but if it gets the job done, fine with me.
Running a gaming PC without the side panel and a fan pointed in is a classic. I’ve been there when my old Northwood system threw a heat alarm in the middle of a TF2 tournament.
But see if the case supports a second exhaust fan. You’d see the screw holes next to the current one. The Lian Li 170M does, but I’m not sure if that’s true for the G509 version. Probably a better long term solution if that’s the issue.
I was agreeing with others in the thread that brief spikes of CPU temps above 80 under heavy load isn’t a major concern. I missed that your GPU was also getting that hot.
I use an extremely small case, a little larger than a PS4, running a 6 core AMD CPU and 3060ti. I don’t believe my temps ever went that high on both my CPU and GPU at the same time, not even when it was entirely air-cooled.
How hot does it run at idle?
This is a good point and something I should’ve caught. If you weren’t getting errors before, and then you changed something, and now you are having issues, reverting that change to test if they go away is the first step.
I had mentioned to you earlier about hwinfo64 and thermal throttling your CPU, but I should mention you can do the same thing with your gpu. HWINFO64 has a GPU sensor readout too, you may have to do some scrolling. There’s a section called “GPU performance limiters” and performance limit - thermal is also a yes/no readout which will tell you if the GPU is being thermally throttled. Current tells you if it currently is, max being yes tells you at some point it throttled. Average, I assume, tells you the percentage of time that the answer is “yes.”
But… going back to the beginning of the thread, you weren’t actually experiencing any problems, right? You were just worried what your temp-range displaying case lights meant when they were red. The stuttering stuff in CP2077 came after adjusting settings and eco mode. If you weren’t actually having problems in the beginning, just concerns, I’d say just turn everything back the way it was and don’t worry about it.
There’s a good chance your system is running well enough especially given that small form factor cases tend to run hot.
First answer is the right answer.
As long as your CPU stays below the TDP max (it’s rated max temp…the manufacturer told you so and they know) it is fine. It is made to operate at those temps and will happily do so. Same as your car is made to run in a certain range. If you rev your engine really high temps rise and, if let to keep going, can damage the engine. In a CPU it can rev higher but if temps get too high it automatically slows down (called thermal throttling). This is a good thing. This is the CPU saving itself from wrecking itself.
I think the disconnect with @Askance (just IMO) is you really do not want to hit that thermal throttling barrier if you can avoid it. Your PC literally slows itself down to save itself. That is good but your performance now tanks. You don’t want to slow your PC but you REALLY do not want to wreck your CPU.
Leaving some headroom for spikes in CPU load and/or boost speeds is desirable so you want to keep things cool.
Don’t get me started on overclocking…all the above goes to 11.
This is why desktops are almost always better performers than laptops. Laptops simply cannot dump their heat fast enough for high performance. They’ve tried and some are better than others but either it becomes a big laptop or very loud fans or both…not to mention very expensive. A well built PC can easily get faster than a laptop for extended periods of time (like rendering a video or gaming) for this reason even if the specs are otherwise the same.
If all you do is browse the web, read email and Tweet (X) then I would not worry. It is mostly gaming and video editing that will push the limits (some other things too like engineering stuff, maybe crypto stuff).
Thanks for all the great advice, everybody. I made some good progress last night, but first I’ll summarize what’s happened so far, since a couple posters are recent to the thread:
- I started the OP a few weeks ago because I was concerned that my CPU was getting too hot when I played Cyberpunk 2077.
- Following advice here and stuff I read online, I lowered some graphics settings in CP2077, capped the frame rate at 60, and set my CPU to ‘eco mode’. This was successful in keeping my CPU under 80C when playing.
- But lately I’ve been noticing some glitching and stuttering in the game, especially during battles, so I started monitoring my GPU and noticed it was getting up to the mid 80s C. So I posted here again yesterday, this time concerned about my GPU.
So, on the more recent advice I got here on the GPU, and stuff I read online, I took off the side panel; I didn’t feel like going so far as to yank out the GPU unit yet so I could get to the fan to clean it better-- as mentioned, it’s mounted so the fan is facing down, barely an inch from the bottom of the box, so it’s hard to get to. I blasted it some more with canned air as well as I could. I figure if I need to yank the card out of the box to blow out the fan more thoroughly, that’s a weekend thing.
Next I downloaded MSI Afterburner and set the GPU fan curve to increase the speed more as temps went up. I only did a gentle change to the curve for now-- nothing too radical. I did not try to lower the GPU voltage as of yet.
Then, as recommended by a few of you, I got a tabletop fan, set it to high, pointed it into the box, started up my AMD Ryzen Master and NVIDIA apps to monitor CPU and GPU temps, and fired up CP2077. Big difference. Both CPU and GPU never got above 75C during game play, and often were in the high 60s to low 70s. What a relief, being able to play a couple solid hours of CP2077 and actually enjoy it worry-free from my components melting down. I’m pretty much fine with making that a permanent solution whenever I game-- takes two seconds to remove the side panel and turn the fan on.
The game is still stuttering though, so on advice from you guys, next I’ll take the CPU off of ‘eco mode’ and see if that helps. A couple questions:
-
If taking the CPU off ‘eco mode’ puts CPU temps back near the danger zone, is removing the CPU fan / heat sink and reapplying thermal paste a good idea? I know you guys warned me off trying to do that with the GPU, but I believe the procedure with the CPU is much more user-friendly, correct?
-
Could capping my frame rate at 60 be causing the stuttering, so should I increase that, do you think? Or should 60 be sufficient?
Answering these things can be a bit exhausting (so to speak) because there are many possibilities. I don’t mind but thought to ask my AI to make it easier. And it was. I know, I know…AI = bad but I DID read its entire response and it is reasonable advice and I do have experience and knowledge about this to judge that advice. So, here is what it said:
Good news first, and it might save you some work: your temps were never actually dangerous.
Modern Ryzen chips are designed to run right up to 95°C — that’s the throttle point, not a failure point. AMD explicitly says sustained boost at 95°C is normal operating behavior. And NVIDIA GPUs generally throttle somewhere in the low-to-mid 80s (GPU core), with memory junction rated to 105°C. So “CPU at 80, GPU at mid-80s” is a normal gaming load, not a meltdown. The mitigations you did aren’t wrong, they just weren’t solving an actual problem — which matters, because it means your stuttering was never a thermal issue. Cooling it further won’t fix it.
On eco mode: definitely take it off. Cyberpunk is genuinely CPU-hungry, particularly in combat with lots of NPCs and physics, and eco mode caps package power hard (typically 45W or 65W PPT depending on the chip). That’s a real performance cut in exactly the scenario where you’re stuttering. There’s no risk in disabling it — the CPU’s own thermal and power limits handle safety regardless of what you set.
On repasting the CPU: it is far more user-friendly than the GPU, you’re right. But don’t do it preemptively — do it only if you see something anomalous, like temps spiking to 95 within seconds of load starting, or a big gap between idle and light-load temps. If your paste is 4-5+ years old it’s worth doing as maintenance. One caveat depending on your socket: on AM4 (Ryzen 5000 and earlier) the CPU is pin-grid, and a cold cooler can yank the chip out of the socket with the pins bent. Run the machine for a few minutes to warm the paste, then twist the cooler gently before lifting. On AM5 (7000/9000 series) the CPU is held by a retention frame, so that risk doesn’t exist.
On the 60 fps cap: almost certainly not your problem, and raising it is more likely to make things worse than better. Frame caps generally reduce stutter. There’s one exception worth checking: if you have V-Sync on with a 60Hz display and no VRR, dropping below 60 can hard-snap you to 30, which feels like a violent stutter. If your monitor supports G-Sync/FreeSync, turn that on, and cap around 3-5 fps below your refresh rate.
What’s actually likely going on. Stuttering specifically during battles points at a few usual suspects:
- VRAM exhaustion. This is the number one cause in CP2077. Combat piles on effects, and if you’re over your card’s VRAM the game starts swapping textures and you get exactly the hitching you’re describing. Add “Memory Usage / VRAM” to your Afterburner overlay and watch it.
- CPU-bound in crowds. Turn Crowd Density down to Low — it’s one of the heaviest CPU settings in the game.
- Traversal/shader stutter. A known engine issue, worse if the game is installed on a hard drive rather than an SSD.
- Ray tracing or Path Tracing left on at a setting your card can’t sustain.
The single most useful diagnostic: put GPU utilization and a frametime graph on your Afterburner overlay. When it stutters, if GPU usage is pinned near 99%, you’re GPU-limited. If it drops to 60-70% while the framerate tanks, you’re CPU-limited or hitting a streaming/VRAM bottleneck. That tells you which direction to go instead of guessing.
Two last things — the desk fan and open panel work fine, but two decent case fans would be quieter, cheaper over time, and keep dust out. And an open side panel is a dust magnet, which will slowly undo the cleaning you just did.
What CPU and GPU do you have? The VRAM number in particular would let me narrow this down a lot.
Much obliged for the thorough and detailed info, @Whack-a-Mole !
Some follow-up comments, questions, and answers to your questions:
But, if NVIDIA GPUs throttle at low to mid 80s, isn’t that a little beyond a normal gaming load? I was getting pretty long stretches at 85C on my GPU the other night, with a peak at 86C. Maybe not meltdown temps, but it seems concerning.
Will do, for sure.
It is 5 years old, so I guess I should when I can get to it. And thanks for the tip on warming it up before trying to remove the heat sink!
I believe I did lower all those CP2077 settings, on advice of Google’s AI. Will double-check.
Adding case fans is a good long-term solution if I can fit them in my box config. For the near-term, I’d only remove the side panel when gaming; I wouldn’t leave it off permanently. For all other everyday stuff I do on my PC, temps run fine, so panel stays on.
CPU: AMD Ryzen 5 5600X 6-Core Processor (3.70 GHz)
Installed RAM: 32GB
GPU: NVIDIA GeForce RTX 3060 (12 GB)
I added this to the AI thread (it remembers what was already said) and here is what it had to say:
That changes a couple of things from what I said.
Scratch VRAM as the culprit. 12GB on a 3060 is generous — that card has more VRAM than the 3070 and 3080. You are not running out. That’s one suspect eliminated.
Your stuttering is almost certainly the 5600X hitting its limits in combat. Six cores is the current floor for Cyberpunk, especially post-2.0, which leans harder on the CPU than launch-era builds did. Dense battle scenes with lots of NPCs and physics are exactly where a 6-core chokes, and it shows up as frametime spikes rather than a lower average — which matches “stutters during battles” precisely.
And eco mode was making that materially worse. The 5600X is already a 65W TDP part; eco mode drops it to 45W. That’s roughly a 30% power cut on a chip that didn’t have much headroom to give. Take it off, and don’t be alarmed if it settles in the low-to-mid 80s afterward — that’s a completely healthy operating temperature for that chip.
Two things to check that are more likely than anything you’ve tried so far:
First, is DOCP/XMP enabled in your BIOS? If not, your RAM is running at the JEDEC default of 2133MHz instead of its rated 3200 or 3600. On Ryzen this is brutal, because memory speed is tied to Infinity Fabric clock, and it hammers 1% lows and frametime consistency specifically. Check the memory speed in Task Manager’s Performance tab. If it says 2133, that’s likely a big chunk of your problem, and it’s a one-toggle fix in BIOS.
Second, is the game on an SSD, and is the PC on carpet? You mentioned the GPU fans face down about an inch off the case floor. If that case is sitting on carpet with a bottom intake, the card is essentially breathing through a blanket. That alone would explain mid-80s on a 3060, which is a fairly cool-running card. Get it up on a hard surface or a stand.
Settings that will actually help on this hardware:
- Turn ray tracing completely off. Even RT Low is punishing on a 3060. Non-negotiable if you want stable frames.
- DLSS on Quality (or Balanced at 1440p). Substantial free performance. Note that Frame Generation is 40-series only — don’t go hunting for it.
- Crowd Density to Low. This is the single heaviest CPU setting in the game and it directly targets your bottleneck.
- Enable Resizable BAR in BIOS (Above 4G Decoding, then Re-Size BAR). Cyberpunk has a driver profile for it and the 3060 supports it.
Keep the 60 cap. It’s a realistic target for a 3060 at 1080p without RT, and it’s helping you, not hurting you.
On repasting: AM4, so the cold-pull warning applies — warm it up before twisting the cooler off. But here’s the thing worth considering instead: if you’re still on the stock Wraith Stealth that shipped with the 5600X, that cooler is genuinely weak, and a $35 tower cooler like a Thermalright Peerless Assassin would drop your load temps 15-20°C. That’s a better use of a Saturday than repasting a cooler that was never adequate to begin with.
Wow, I thought I was getting good AI advice, but clearly you’re a prompt ninja who knows the right questions to ask it.
So, it looks like my takeaways are:
- Turn off eco mode on CPU
- Do the recommended BIOS tweaks
- Double-check CP2077 settings
- Getting that $35 Thermalright Peerless Assassin tower cooler for the CPU sounds like a no-brainer; at least I will look into it to make sure it would fit my system, for sure.
- Extra case fans to keep the GPU cooler without having to remove the side panel?