Nvidia GTX 7870 overheating issues, or what?

I’m running into a massive headache with my XFX Nvidia GTX 7870. Everything seems totally fine until you actually put a load on it for a while. Once it hits a certain point, the screen starts glitching out—colors shifting, flickering, just total visual chaos—and it stays like that for a few seconds. From what I can tell, after running 3DMark, this madness kicks in right when the temperature hits about 92-93°C. It happens like clockwork. After maybe 5 or 10 seconds, once the card cools down just a couple of degrees, the test picks up right where it left off like nothing ever happened, only for the whole cycle to repeat itself the moment it climbs back up to that same heat threshold. I actually took the card into a repair shop, and they told me that after running continuous stress tests including 3DMark for 24 hours straight, the temperature never even crossed 87°C and everything worked perfectly. Honestly, I'm about to lose my mind here, so I could really use some advice.
Not a single component in this build has ever been overclocked.
Here’s a video showing exactly what I mean: https://www.dropbox.com/s/2zygp7ff21...209_152319.mp4
I should also mention that when I swap in my old AMD Sapphire HD4850, everything runs flawlessly without a single hiccup.

The rest of my setup is:
CPU: i5 3570k
Motherboard: AsRock Z77 Extreme4
RAM: 8GB G.Skill DDR3 + 8GB Kingston DDR3
PSU: Corsair 600W

Thanks in advance!

Noah Smith23 said:I'm running into a real headache with my XFX Nvidia GTX 7870—everything works perfectly fine until the card stays under load for a while. Then, the image starts "glitching"—colors shifting, flickering... just for a few seconds at a time. From what I can tell, based on my 3DMark runs, this happens right when it hits about 92-93°C! It’s incredibly consistent. After about 5-10 seconds, once the card cools down by just a couple of degrees, the test carries on like nothing happened—until it heats back up to that specific threshold again. I actually took the card in for repair, and they told me that after running continuous stress tests (including 3DMark) for 24 hours, the temperature never even crossed 87°C and everything functioned normally! Honestly, I'm starting to lose my mind here—any advice would be greatly appreciated...
None of my components have ever been overclocked, by the way.
I've attached a video so you can see exactly what I mean: https://www.dropbox.com/s/2zygp7ff21...209_152319.mp4
I should also mention that when I use my old AMD Sapphire HD4850, everything runs flawlessly without any issues whatsoever!

The rest of my build is:
CPU: i5 3570k
Motherboard: AsRock Z77 Extreme4
RAM: 8GB G.Skill DDR3 + 8GB Kingston DDR3
PSU: Corsair 600W

Thanks in advance!

If I may ask a quick question... why are you running 3DMark in the first place? In my experience, that's primarily a tool for stressing a card to check stability during an overclock, but since you mentioned you haven't overclocked anything, I'm curious why you're putting it through that kind of stress.

3DMark stresses a graphics card more than any actual game; there is no need to keep running it.
Based on the temperatures, you can see them spike, drop, then slowly climb again—it’s overheating.
I doubt this will happen during actual gameplay, though.
It is possible your case isn't getting enough airflow, or perhaps the graphics card cooler is too weak or just clogged with dust.
I didn't watch the entire video, but I know when AMD cards overheat, they throttle the GPU and memory frequencies by about 200MHz. You notice the performance drop immediately.
I am not sure how Nvidia handles it. Perhaps they also lower clock speeds while they struggle.
If you continue to run the graphics card at these high temperatures, it will likely fail within two to four years of purchase.

Best regards,
B. W.

This is exactly why I always suggest looking for cards with at least a three-year warranty. My rule of thumb? Sell yours about a year before the coverage expires—say, somewhere between the 18-month and 6-month mark—toss in a little extra cash, and upgrade to a newer, more powerful model.
It’s a nice little safety net; if your graphics card happens to die while it's still under warranty, especially once that specific series is out of production, you can often just pay the difference for a superior replacement. Granted, that's an extreme edge case, but it's worth keeping in mind.
The real headache comes when you try to "stress test" a card to death by running intense benchmarks or even applying extra heat sources to it. Sometimes, despite your best efforts, the thing refuses to die. Then, in a stroke of pure irony, it decides to give up the ghost right in the middle of a game of Solitaire.

No, I have my desktop and a laptop. This laptop is almost exactly five years old, and I can tell you, it runs just as well as the day I bought it.
On average, I run this laptop for 12 hours every single day without any issues. However, I clean out the heatsink every three months, or at the absolute latest, every six months.

In short, what dishonest people won't tell you is this: never buy an older laptop under any circumstances. And stay away from graphics cards; they run incredibly hot, and that heat alone makes them extremely fragile.
There are very few honest people who actually know how to maintain their hardware, but how are you supposed to know which ones they are?
It is better not to buy anything that has been in use for a long time. If something has been used for more than, say, nine months, it likely isn't worth your money and could die tomorrow. Give or take a few months.

Regards,
B. W.

It’s the same kind of nonsense you see happening in video games, too; I’m running 3DMark right now because it gives you a much clearer picture of what those temperatures are actually doing!
I ended up dropping it off at the repair shop, and this is the ridiculous feedback I got from their online claim status:
Service Notes: The graphics card boots the PC without any issues. We ran it through 3DMark for 18 hours, and the peak temperature hit 87 degrees Celsius. We also ran the 3DMark benchmark for a full 24 hours, and there were no artifacts or color shifts detected.

If you want to monitor temperatures, there's OpenHardwareMonitor and several other software options available. You can actually configure them to display the temperature right on your Taskbar.

Tell me, are you running 3DMark or any actual games?

Does yours have one fan or two? To be honest, XFX isn't exactly known for top-tier cooling. It’s possible their models use a more ventilated case than yours does. Are you running a micro ATX setup where everything is packed in tight? Also, have you tried adjusting the fan speeds? Generally speaking, higher-end cards tend to run a bit warmer—around 70-80°C—but hitting 90°C+ is definitely pushing it. You might want to try adding a case fan, swapping out the thermal paste for a high-quality brand, or as a final resort, just grabbing an aftermarket cooler.

So, here’s the card I’m running:
http://cdn.overclock.net/d/de/de5639...1-580x435.jpeg
It’s got a dual-fan setup.
As for my case, I’ve got a Zalman Z11 Plus. It’s a massive chassis that can hold anywhere from five to six fans if you want to pack it out.
The weird part is that this stuff just started happening out of nowhere; everything was working perfectly fine until recently. Back during the summer heatwaves, temperatures would spike up to 100 degrees, but even then, things were manageable—I have no clue what changed. I even wiped everything and installed Windows 8.1, but it's the exact same nightmare I had on Windows 7 before that.
Meanwhile, my old Radeon HD 4850 is still chugging along without a single hiccup.

Hitting 100°C is definitely not okay. If it’s staying that hot for extended periods, something is clearly wrong—most likely, your thermal paste has finally given up the ghost and can't conduct heat like it used to.

Those extra 5 or 6 degrees really matter if the repair shop isn't cutting you any corners. They probably tested it on an open test bench where the ambient temperature was much lower than yours. (By the way, is your PC case sitting right next to a heater or something?)

Also, 87°C sounds way too high for a non-overclocked GPU with just dual fans. Based on reviews of the XFX core edition, it hits a max of 72°C, while the black edition goes up to 77°C—and that's under a full gaming load, not just some light benchmark.

I honestly have no clue what’s actually going on here, but I can feel myself getting absolutely hammered by this thing lately, and my boss is being a total pain about replacing it—he won't even consider a swap as long as the machine is technically still kicking, no matter how much it's struggling.
The PC is sitting right in my office, nowhere near any heaters or anything that could be messing with the temps. 😛
Does anyone have a smart way to tackle this or some actual advice on how to fix it?

There isn't much else you can try. Like I mentioned, your first move should be swapping out the thermal paste, and if that doesn't do the trick, you might need to look into an aftermarket cooler like—though I suspect you'll void the warranty if you pull the stock heatsink off (I know that happens with single-fan NVIDIA GeForce GTX 7870 cards).