NVIDIA GeForce RTX 3070 Crashes With All Unreal Engine Games - HELP!

[Chrispy_]

Abbreviating numerous times you've stated concerns this is a counterfeit.

BB and all their vendors, buyers, etc on down the line do not produce or sell products you can buy anywhere else. Kitchen appliances, accessory electronic cords, tv, computer, even Apple products are... it is all SKU specific to their line of stores. Internet searches on manufacturer pages will turn up less than nothing.

LOL, no - your assertion that the PSU is unique-to-Best-Buy-PSU is mad. Nothing else in that CyberPower PC is exclusive to Best Buy, and apart from the CyberPower decals on the fan hubs and pump top, it's not even CyberPower-exclusive. It's all off-the-shelf parts available globally with webpages for info/drivers/firmware/support easily findable.

Best Buy do not have exclusive sales rights to these parts:

• AMD Ryzen 5 5700
• Gigabyte RTX 3070 Eagle OC Revision 2.0
• Western Digital SN570 SSD
• Asus PRIME B550-PLUS AC
• G.Skill Flare DDR4-3600 CL18 kit

What on earth makes you think CyberPower would contract EVGA to make a special PSU just for them, and then only use it in their Best Buy SKUs when nothing else in the system is bespoke? Even if that happened, why do you think it' would be hidden from web searches with no results whatsoever, no support pages or specs on the EVGA website, and no mention of it by either Best Buy or CyberPower?

I can't prove that it's a counterfeit PSU, but it sure bears all the hallmarks of one.

This is what OCCT shows me, does that show the AIO and pump info you wanted?

Yes! That proves the pump is moving. I don't use many AIOs but the RPMs seem low to me; the Corsair and BeQuiet AOIs I've occasionally used have had pump speeds in the 4000-5500RPM range.
Maybe someone else has experience with Coolermaster AIOs and can comment on whether 2400RPM is normal for their pump.

It wouldn't hurt to go into the BIOS hardware monitor / fan control page and confirm that the pump is set to full speed. You should be able to see the BIOS version on the home screen of the UEFI BIOS too, and if it's too far behind the current version (3404) then grab a new version from here:

PRIME B550-PLUS AC-HES｜Motherboards｜ASUS Global

PRIME B550-PLUS AC-HES AMD B550 (Ryzen AM4) ATX motherboard with dual M.2, PCIe 4.0, WIFI 5, 1 Gb Ethernet, DisplayPort/HDMI, SATA 6 Gbps, USB 3.2 Gen 2 Type-C®, front USB 3.2 Gen 1 Type-C®, Aura Sync ARGB headers, FAN Xpert 2+, Armoury Crate, and 5X PROTECTION III. PRIME B550-PLUS AC-HES...

www.asus.com

edit:
Check you're not using BitLocker in Windows if you want to use the 3404 BIOS as it will reset your security key. Download BIOS version 3205 if you're unsure.

 

[HumongousGigantic]

I believe I've had the fans on a BIOS setting of Optimize All. I just went and checked and regardless this is what I found. Looks like it is already set at max all time?

 

[Chrispy_]

Yeah, that's what it looks like. Perhaps CoolerMaster just use larger, slower pumps than I'm used to, so I wouldn't worry about the pump.

It's only a 120mm radiator and those are never great but I'm still a little surprised your CPU is getting so hot with PBO disabled and a liquid cooler. The fan on that radiator is definitely spinning when the system is on, right?

Depending on how much you want to investigate, I'd say you might want to unmount the cooler, check it doesn't still have the plastic peel on it, check the paste has been making good contact etc. The 5800X in my living room PC is using 50% more power than your 5700 and is never hotter than about 86-87C using a cheap, low-profile air cooler that is pretty wimpy in a cramped HTPC case that doesn't have a lot of airflow.

 

[Dragam1337]

Gigabyte, and in particular their eagle models, sadly often have very questionable build quality.

There is a quite high probability that the crashing is due to this.

As others have said, you extra paid alot of money for a prebuilt system... so why are you using time troubleshooting it on your own? Then you might aswell have saved yourself that money, and gotten a better system by building it yourself. But you didn't, so contact the seller and have them fix it (aka replace gpu)...

Yeah, that's what it looks like. Perhaps CoolerMaster just use larger, slower pumps than I'm used to, so I wouldn't worry about the pump.

It's only a 120mm radiator and those are never great but I'm still a little surprised your CPU is getting so hot with PBO disabled and a liquid cooler. The fan on that radiator is definitely spinning when the system is on, right?

Depending on how much you want to investigate, I'd say you might want to unmount the cooler, check it doesn't still have the plastic peel on it, check the paste has been making good contact etc. The 5800X in my living room PC is using 50% more power than your 5700 and is never hotter than about 86-87C using a cheap, low-profile air cooler that is pretty wimpy in a cramped HTPC case that doesn't have a lot of airflow.

Cpu temps wouldn't make a game crash though... if it was a cpu instability issue, then the entire system should hardlock. But it shouldn't become unstable from high temps in the first place, it would just throttle...

 

[HumongousGigantic]

Gigabyte, and in particular their eagle models, sadly often have very questionable build quality.

There is a quite high probability that the crashing is due to this.

As others have said, you extra paid alot of money for a prebuilt system... so why are you using time troubleshooting it on your own? Then you might aswell have saved yourself that money, and gotten a better system by building it yourself. But you didn't, so contact the seller and have them fix it (aka replace gpu)...



Cpu temps wouldn't make a game crash though... if it was a cpu instability issue, then the entire system should hardlock. But it shouldn't become unstable from high temps in the first place, it would just throttle...

 

[Chrispy_]

Cpu temps wouldn't make a game crash though... if it was a cpu instability issue, then the entire system should hardlock. But it shouldn't become unstable from high temps in the first place, it would just throttle...

I've already stated exactly that in earlier posts.

Sorting the cooling problem is separate to the crashes; there's no way in my experience that a 65W CPU like the 5700 with PBO disabled should be hitting the thermal limit with an AIO - even a weak 120mm AIO. They're easy-to-cool chips that even the free, bundled little Wraith Stealth has absolutely no problem keeping at under 90C.

Gigabyte, and in particular their eagle models, sadly often have very questionable build quality.

There is a quite high probability that the crashing is due to this.

I often wonder where statements like this come from.

Gigabyte aren't my first choice of GPU vendor but I sure have bought hundreds of them. I think I bought close to 500 Gigabyte RX570 Gaming OC cards when the mining boom collapsed in 2019 and everyone was selling brand-new RX 570 8GB cards for dirt cheap. I took that opportunity to upgrade every workstation across three offices with 2GB, 3GB, 4GB GPUs in them to make 8GB the minimum worstation GPU spec we had to worry about. To date, I don't think I've had a single one come back faulty through anything other than abuse - I've definitely had one with the HDMI socket ripped off the PCB by a careless cleaner aggressively vacuuming under-desk cables, and I've binned at least a couple that weren't worth replacing worn fans on - but statistically I should have at least half a dozen failures under warranty, based on average RMA rates of 1-2% for GPUs, and I can safely say that 5 years later, no such thing happened under warranty and no such thing has happened now that all the warranties are long-expired.

I've only bought 18 Gigabyte Eagle cards to date, and I don't like how they're so plastic, but all of them are still running fine, crash-free with many of them running around the clock as architects and engineers are rendering animations on them in Enscape or Unreal Engine (TwinMotion) overnight on a regular basis. My own HTPC is using a Gigabyte 4060Ti and I don't think I've experienced a single crash of any type on this PC yet. I personally had a Gigabyte GTX 1080 for a couple of years and it was a perfectly well-behaved, sturdy-feeling, cool, quiet, and reliable GPU during that time, even when my ancient PSU died and needed replacing.

There have been many high-publicity GPU failures of late affecting several vendors but the only one that seems to affect Gigabyte was some cracked PCBs on extremely heavy 4090s and their RMA department handled it poorly. I don't see how that one-model incident which even Louis Rossman weighed in on has anything to do with all other Gigabyte GPUs. I can't call it FUD because I haven't given you a chance to validate your claims or point me to further evidence educating me on Gigabyte atrocities I've missed.

I'm just curious why I keep seeing Gigabyte's name dragged through the mud and blamed for crashes with zero evidence or reasoning behind the accusations. Either they're genuinely terrible and I'm just incredibly luck to beat the odds with the dozen or so different models I've had zero issues with across hundreds of samples, or more likely (IMO) the competition is astroturfing hard, and successfully. Feel free to change my mind, but it'll take either hard evidence, independent journalism on the subject (not an upset child on Reddit) or so many google results that it can't possibly be an anomaly.

 

[Dragam1337]

I've already stated exactly that in earlier posts.

Sorting the cooling problem is separate to the crashes; there's no way in my experience that a 65W CPU like the 5700 with PBO disabled should be hitting the thermal limit with an AIO - even a weak 120mm AIO. They're easy-to-cool chips that even the free, bundled little Wraith Stealth has absolutely no problem keeping at under 90C.


I often wonder where statements like this come from.

Gigabyte aren't my first choice of GPU vendor but I sure have bought hundreds of them. I think I bought close to 500 Gigabyte RX570 Gaming OC cards when the mining boom collapsed in 2019 and everyone was selling brand-new RX 570 8GB cards for dirt cheap. I took that opportunity to ditch every single workstation GPU with less than 8GB. To date, I don't think I've had a single one come back faulty. NOT ONE.

I've only bought 18 Gigabyte Eagle cards to date, and I don't like how they're so plastic, but all of them are still running fine, crash-free with many of them running around the clock as architects and engineers are rendering animations on them in Enscape or Unreal Engine (TwinMotion) overnight on a regular basis.

There have been many high-publicity GPU failures of late affecting several vendors but the only one that seems to affect Gigabyte was some cracked PCBs on extremely heavy 4090s and their RMA department handled it poorly. I don't see how that one-model incident which even Louis Rossman weighed in on has anything to do with all other Gigabyte GPUs. I can't call it FUD because I haven't given you a chance to validate your claims or point me to further evidence educating me on Gigabyte atrocities I've missed.

I'm just curious why I keep seeing Gigabyte's name dragged through the mud and blamed for crashes with little or zero reason for the accusations. Either they're genuinely terrible and I'm just beating the odds with the dozen or so different models I've had zero issues with across hundreds of samples, or more likely (IMO) the competition is astroturfing hard, and successfully. Change my mind!

First part, fair - i missed that.

2nd, it comes from personal experience, including the 2 gigabyte 4090s i bought... while the cards won't be outright faulty, the built quality and quality control is extremely lackluster compared to other vendors. It is very common to see gigabyte cards have very bad tim applications (mine did, half of die having no tim) which leads to extremely high hotspot temps and... potential instability during very high load. Add to that that the fans are rubbish quality. None of this has been an issue with the 4090s i've bought from other brands.

And in regards to the eagle, check any review of any gigabyte eagle card... they are always THE worst cards you can get... running at very high fan rpm and still getting very hot.

Then why dafuq am i still using a gigabyte 4090 in my private gaming pc ? Because ironically the low quality power delievery made it the only 4090 without any coilwhine of all the ones i tested... the higher quality power delievery, the worse the coilwhine. And i absolutely detest coilwhine, so here i am with the technically worst built 4090... lol.
But at least the chip itself is quite good - does 2700 mhz at only 925 mv, which allows me to drop those poop fans to 800 rpm and still remain below 70c.

 

[Chrispy_]

First part, fair - i missed that.

2nd, it comes from personal experience, including the 2 gigabyte 4090s i bought... while the cards won't be outright faulty, the built quality and quality control is extremely lackluster compared to other vendors. It is very common to see gigabyte cards have very bad tim applications (mine did, half of die having no tim) which leads to extremely high hotspot temps and... potential instability during very high load. Add to that that the fans are rubbish quality. None of this has been an issue with the 4090s i've bought from other brands.

And in regards to the eagle, check any review of any gigabyte eagle card... they are always THE worst cards you can get... running at very high fan rpm and still getting very hot.

Then why dafuq am i still using a gigabyte 4090 in my private gaming pc ? Because ironically the low quality power delievery made it the only 4090 without any coilwhine of all the ones i tested... the higher quality power delievery, the worse the coilwhine. And i absolutely detest coilwhine, so here i am with the technically worst built 4090... lol.
But at least the chip itself is quite good - does 2700 mhz at only 925 mv, which allows me to drop those poop fans to 800 rpm and still remain below 70c.

I think the Eagles are the worst-reviewed cards of the ones that are reviewed. Being hot and loud isn't desirable, but don't confuse that for defective.

Trust me when I say there are far worse cards out there from brands like KFA2, Palit, PNY, and Colorful that never see mainstream reviews. Even hotter, even cheaper build, even louder fans etc. It's also super important to not make sweeping generalisations of a whole brand based on a single GPU. One of the worst GPUs designs I have ever encountered in my long career was from Asus, the supposed king of GPU build quality, and honestly most of their GPUs are good (if pricey).

PNY have the highest failure rate for NVIDIA for me, based on actual returns of in-house builds. Inno3D is ranked second on my shit list of RMAs needed, but that's just my small sample size of a few hundred at most - and I'd buy an Inno3D over a Palit based solely on all the Palit 20-series and 30-series GPUs I encountered suffering from both extreme PCB sag and glaring fan control issues that were swept under the rug and dismissed as "not a problem" by their customer service reps. I'd still look at a PNY, Inno3D or a Palit GPU in the future because you care rarely apply the result of one SKU to another, even in the same name range (eg Asus DUAL, or Gigabyte Eagle). It's unlikely that the same people were even involved with the design and manufacture of two GPUs more than a few years apart.

 

[Dragam1337]

I think the Eagles are the worst-reviewed cards of the ones that are reviewed. Being hot and loud isn't desirable, but don't confuse that for defective.

Trust me when I say there are far worse cards out there from brands like KFA2, Palit, PNY, and Colorful that never see mainstream reviews. It's also super important to not make sweeping generalisations of a whole brand based on a single GPU. One of the worst GPUs designs I have ever encountered in my long career was from Asus, the supposed king of GPU build quality, and honestly most of their GPUs are good (if pricey). PNY have the highest failure rate for NVIDIA for me, based on actual returns of in-house builds. Inno3D is the second worst on my shit list, but that's just my small sample size and I'd buy an Inno3D over a Palit based solely on my sample of Palit 20-series and 30-series GPUs with extreme PCB sag and/or glaring fan control issues that were never addressed with vBIOS updates. I'd still look at a PNY, Inno3D or Palit GPU in the future because you care rarely apply the result of one SKU to another, even in the same name range (eg Asus DUAL, or Gigabyte Eagle).

For the first thing you write, i mentioned that as the first thing in my previous reply :

"while the cards won't be outright faulty, the built quality and quality control is extremely lackluster compared to other vendors."

And here is a fun little trial for you - have a sweep on this forum, looking up threads where the gpu has very high hotspot temps and how big a percentage gigabyte cards account for in that... it's gonna be high.

Gigabyte are as said notorious for their terrible quality control, and it's especially true for their tim application on gpus. Which again won't make them faulty per say, but it will make them run alot worse, with throttling and speeding up the degradation of the chip.

While you are correct that you can't compare one sku with another in most regards, you can compare company trends - and lack of quality control is certainly one for gigabyte. You could say that having to redo paste is a minor thing with gigabyte cards (you nearly always need to do that with even with brand new gigabyte cards if you want the hotspot delta to be good), but most people aren't gonna do that, and it will as said hurt performance... and in the case of the eagle in this thread, if we try to stay just a bit on topic, if his core temp is 80+ and his hotspot as a consequence is 100+ that could deffo cause instability during heavy load. With that said, when it's only some games that crash, it seems alot more likely to a software issue - most likely gpu driver.

In regards to the other brands you name, i personally wouldn't even consider them - but they don't really sell them in my part of the world anyways, so that's a moot point.

 

[nomdeplume]

I can't prove that it's a counterfeit PSU, but it sure bears all the hallmarks of one.

We could pursue this to the detriment of any useful conversation or leave the unknown variables important to us lie dormant. It is of suspect quality.

My concern after reading through all responses was failing to be convinced money spent on new parts would fix what appears to be a rather large hiccup UE games trigger more than other loads. A lot wasn't adding up. Prebuilt are not immune to poor assembly. Selling it or returning it in favor of a better suited option may be worth the headache it saves.

 

[HumongousGigantic]

Gigabyte, and in particular their eagle models, sadly often have very questionable build quality.

There is a quite high probability that the crashing is due to this.

As others have said, you extra paid alot of money for a prebuilt system... so why are you using time troubleshooting it on your own? Then you might aswell have saved yourself that money, and gotten a better system by building it yourself. But you didn't, so contact the seller and have them fix it (aka replace gpu)...



Cpu temps wouldn't make a game crash though... if it was a cpu instability issue, then the entire system should hardlock. But it shouldn't become unstable from high temps in the first place, it would just throttle...

While I see where you are coming from this doesn't sound constructive. "You got scammed, don't buy prebuilds, build your own" comes off as elitist. I am sorry I am trying to troubleshoot through this forum. Clearly I am in the wrong place.

 

[Chrispy_]

Gigabyte are as said notorious for their terrible quality control, and it's especially true for their tim application on gpus. Which again won't make them faulty per say, but it will make them run alot worse, with throttling and speeding up the degradation of the chip.

While you are correct that you can't compare one sku with another in most regards, you can compare company trends - and lack of quality control is certainly one for gigabyte. You could say that having to redo paste is a minor thing with gigabyte cards (you nearly always need to do that with even with brand new gigabyte cards if you want the hotspot delta to be good), but most people aren't gonna do that, and it will as said hurt performance... and in the case of the eagle in this thread, if we try to stay just a bit on topic, if his core temp is 80+ and his hotspot as a consequence is 100+ that could deffo cause instability during heavy load. With that said, when it's only some games that crash, it seems alot more likely to a software issue - most likely gpu driver.

Where the hell are you getting this FUD from?
Show me any compelling evidence of this bad TIM Gigabyte application because I can't recall seeing any photos in forum posts or heard reviewers/Youtubers bring it up. Only FUD comments that blanket-statement condemn all Gigabyte GPUs with no evidence. That's a pure, dictionary definition of an astroturfing campaign, AKA FUD.

By contrast, in addition to already telling you my approximate sample size of Gigabyte GPUs and failure rates, I went back through @W1zzard 's reviews for all the 40, 30, and 20-series Gigabyte GPUs as he has some of the best and most reliable teardown photos on the net.

• Every single one of them shows flawless TIM application.
• Not one of them failed quality control.
• Repasting was clearly not necessary.
• Plenty of Editor's Choice and Highly Recommended awards.
• Low temperatures and hotspot deltas that seem normal compared to other brands in the cooler comparison page of each review.

Don't take my word for it, look at the proof for yourself - it's all here on this very site.

TechPowerUp

www.techpowerup.com

Change my mind, prove me wrong, show me that this isn't just FUD.

 

[TumbleGeorge]

Crashes With All Unreal Engine Games​

How play games with other engines?

 

[Dragam1337]

Where the hell are you getting this FUD from?
Show me any compelling evidence of this bad TIM Gigabyte application because I can't recall seeing any photos in forum posts or heard reviewers/Youtubers bring it up. Only FUD comments that blanket-statement condemn all Gigabyte GPUs with no evidence. That's a pure, dictionary definition of an astroturfing campaign, AKA FUD.

By contrast, in addition to already telling you my approximate sample size of Gigabyte GPUs and failure rates, I went back through @W1zzard 's reviews for all the 40, 30, and 20-series Gigabyte GPUs as he has some of the best and most reliable teardown photos on the net.

• Every single one of them shows flawless TIM application.
• Not one of them failed quality control.
• Repasting was clearly not necessary.
• Plenty of Editor's Choice and Highly Recommended awards.
• Low temperatures and hotspot deltas that seem normal compared to other brands in the cooler comparison page of each review.

Don't take my word for it, look at the proof for yourself - it's all here on this very site.

TechPowerUp

www.techpowerup.com

Change my mind, prove me wrong, show me that this isn't just FUD.

"Change my mind, prove me wrong" - what are you, an 18 year old edgelord ?

As i've said, it's based on personal experience, and lots of other people having the exact same experience. And yes, i've posted pics of the teardown on this forum - look it up yourself...

 

[Chrispy_]

I'm calling you out on it because you're the one making the ridiculous claim that contradicts my experience, all of the reviews here, and other reputable review sites that perform independent testing and teardowns.

The burden of proof is on you, and this is your actual posted gigabyte 4090 OC as you claim, which looks fine to me:

https://tpucdn.com/forums/data/attachments/325/325199-60915e5fe565c9ed11ecc435274fb132.jpg

It's quite clear that you're talking rubbish. All four corners of your GPU die are pasted properly, and you can see that Gigabyte's paste application was accurately positioned and neatly applied with a nice clean-edged application template. Why lie? Why bluff? Forum search is easy and quick.

You say temps dropped 10C but all you've likely done is replaced lifespan-optimised paste for performance-optimised paste and that 10C delta will not be permanent unless you repaste every year or so.

I hate FUD. It's bad for everyone except the manufacturer if they're the ones who spread it. Unless you're being paid by the manufacturer, engage critical thinking and back up your claims with evidence, links or whatever.

 

[nomdeplume]

While I see where you are coming from this doesn't sound constructive. "You got scammed, don't buy prebuilds, build your own" comes off as elitist. I am sorry I am trying to troubleshoot through this forum. Clearly I am in the wrong place.

You might be taking the shift in direction personally. Don't worry, the mods here are very good and will clean up the other nonsense.

Nobody is questioning that your first action was seeking help here. That your second isn't following through with either BB, who operate their own repair shops, or the company selling this prebuilt is worrisome. The upcharge you paid was for guarantee of a working product or they fix it. Make them troubleshoot while we discuss this further because...

Offsite internet help is actually quite limited. You were very cooperative about probing for cause of issues. Actual testing with a meter and other equipment will solve if the psu corrupted other parts or whatever the issue(s) impacting it really are. That includes poor assembly or manufacture defects they didn't pick up on.

 

[Dragam1337]

I'm calling you out on it because you're the one making the ridiculous claim that contradicts my experience, all of the reviews here, and other reputable review sites that perform independent testing and teardowns.

The burden of proof is on you, and this is your actual posted gigabyte 4090 OC as you claim, which looks fine to me:

https://tpucdn.com/forums/data/attachments/325/325199-60915e5fe565c9ed11ecc435274fb132.jpg

View attachment 337299

It's quite clear that you're talking rubbish. All four corners of your GPU die are pasted properly, and you can see that Gigabyte's paste application was accurately positioned and neatly applied with a nice clean-edged application template. Why lie? Why bluff? Forum search is easy and quick.

You say temps dropped 10C but all you've likely done is replaced lifespan-optimised paste for performance-optimised paste and that 10C delta will not be permanent unless you repaste every year or so.

I hate FUD. It's bad for everyone, including the people spouting it.

If that looks fine to you, then that really says everything about your "knowledge". There should not be a large spot entirely devoid of tim.

And no, i didn't say that temps got reduces by 10c - i said that delta got reduced to 10c. Prior to repaste the hot spot would get as high as 100c - now it rarely goes above 80c.

Same story for alot of other people with gigabyte cards, including the thread you got that pic from.

It will likely be the same with alot of your gigabyte cards, but judging by your comments here, you couldn't tell a good paste job from a bad one even if your life depended on it...

 

[Chrispy_]

You might be taking the shift in direction personally. Don't worry, the mods here are very good and will clean up the other nonsense.

Nobody is questioning that your first action was seeking help here. That your second isn't following through with either BB, who operate their own repair shops, or the company selling this prebuilt is worrisome. The upcharge you paid was for guarantee of a working product or they fix it. Make them troubleshoot while we discuss this further because...

Offsite internet help is actually quite limited. You were very cooperative about probing for cause of issues. Actual testing with a meter and other equipment will solve if the psu corrupted other parts or whatever the issue(s) impacting it really are. That includes poor assembly or manufacture defects they didn't pick up on.

The problem here is that it's an inconsistent fault too - The PC initially hit 95C and crashed in OCCT power testing, now - having made no changes - it's fine in that test.

High temperatures might be a red herring, as they shouldn't affect stability.
The PSU may be good, may be bad, but we don't know for sure without swapping it for testing or contacting EVGA to ask if it's a genuine model they make.

Unfortunately I think we're at the point where further steps are disruptive enough to invalidate the warranty. Unreal games might be crashing from registry corruption if the problem is persisting between uninstall and reinstall of the game, so a full wipe of the OS and clean OS install is one angle of attack. Hardware swap of the PSU and/or cooler are other options.

It really does now depend on whether OP wants to abandon warranty and start poking around at his own expense, or take advantage of the warranty and send it back for testing and repair.

If that looks fine to you, then that really says everything about your "knowledge". There should not be a large spot entirely devoid of tim.

And no, i didn't say that temps got reduces by 10c - i said that delta got reduced to 10c. Prior to repaste the hot spot would get as high as 100c - now it rarely goes above 80c.

Same story for alot of other people with gigabyte cards, including the thread you got that pic from.

It will likely be the same with alot of your gigabyte cards, but judging by your comments here, you couldn't tell a good paste job from a bad one even if your life depended on it...

That void on the left edge is caused by you. It has nothing to do with the quality of the TIM application, which you can clearly see was applied correctly from the neat edges of the TIM around the edge of where the die made contact. Your high-res macro shot beautifully shows the perfect even square of TIM that gigabyte initially applied, as well as the thickness of the layer of TIM. Nothing about it screams "poor QC" to me.

There's every chance (you can see the flow marks in the accompanying die shot) that the TIM flowed like that as you broke the seal in your unnecessary teardown to repaste it. This is the corner you broke the seal on first as you lifted the cooler off the PCB and it sucked air in causing this pattern:

 

[Dragam1337]

The problem here is that it's an inconsistent fault too - The PC initially hit 95C and crashed in OCCT power testing, now - having made no changes - it's fine in that test.

High temperatures might be a red herring, as they shouldn't affect stability.
The PSU may be good, may be bad, but we don't know for sure without swapping it for testing or contacting EVGA to ask if it's a genuine model they make.

Unfortunately I think we're at the point where further steps are disruptive enough to invalidate the warranty. Unreal games might be crashing from registry corruption if the problem is persisting between uninstall and reinstall of the game, so a full wipe of the OS and clean OS install is one angle of attack. Hardware swap of the PSU and/or cooler are other options.

It really does now depend on whether OP wants to abandon warranty and start poking around at his own expense, or take advantage of the warranty and send it back for testing and repair.


That void on the left edge is pump-out. It has nothing to do with the quality of the TIM application, which you can clearly see was applied correctly from the neat edges of the TIM around the edge of where the die made contact. Your high-res macro shot beautifully shows the perfect even square of TIM that gigabyte initially applied, as well as the thickness of the layer of TIM. Nothing about it screams "poor QC" to me.

There's every chance (you can see the flow marks in the accompanying die shot) that the TIM flowed like that as you broke the seal in your unnecessary teardown to repaste it.

You're so full of it, it's crazy... yeah, absolutely unncessary repaste with a hotspot delta of 20+ c and reaches 100c, which funnily enough got instantly fixed with a repaste...

Tell us what IT firm you are working for, dying to know who'd have such a clown on the payroll...