By creating another topic I am requesting support from EVGA to a problem that I am experiencing since June 2022.
 
I bought EVGA RTX 3090 FTW3 Ultra in April 2022 on eu.evga.com and first encountered FPS/GPU power drops 2 months later. In next month I tested many different hardware and software configurations but I couldn't resolve the issue that was visible in different games. In July I created a post on this Forums describing the problem in detail (https://forums.evga.com/PowerFPS-drops-3090-FTW3-m3566315.aspx). I have also contacted SupportEU@EVGA.com at that time. The consensus between both sources was to RMA the GPU and so I did. After receiving different card I immediately found out that the issue with power drops still does exist. After contacting support using user panel on eu.evga.com website I was told I will get another, previously tested RTX 3090. I have received another GPU but the issue was still there.
 
I contacted support on eu.evga.com with information about receiving another GPU with power drops issues. After few weeks of almost meaningless conversation, after a request, created a video and I provided a link to recording of me running Death Stranding and swapping two GPUs: GTX 1080Ti and RTX 3090, with fps/power drops issue clearly visible, and additional information about my whole system, used hardware and software, serial number (so there will be no confusion about GPU I was using), stability tests, etc. The reply from SupportEU@EVGA.com was astonishing and I started to believe nobody even watched requested video of me reproducing this issue for half an hour or have seen screenshot I sent. Basically I was told that RTX 3090 is not faulty (even if it gets lower fps score that 8 year old 1080Ti and renders the game unplayable in certain locations)... and to check if I have all 3 PCI-E 8 pin cables connected... which is also clearly visible in the video. Since I have sent last message a week ago I assume above is their final word on this issue.
 
Currently I have a faulty RTX 3090 that I am afraid to use (such massive power drops won't go unnoticed by power delivery via PCI-E slot; PSU should be fine though). I have also witnessed another issues with artifacting and crashes in different games (more frequently in Death Stranding, Horizon, Doom 2016 and Eternal which I play mostly) and when connecting 3 displays using DPs.
 
I request reply from EVGA USA/TW support, since support from EU.EVGA is not working anymore and is clearly not able to resolve this issue.
 
I will provide link(s) to video(s) mentioned above as needed.
 
Best Thanksgiving wishes to anyone celebrating thanking, giving and everything in between.

https://youtu.be/8SlGsI9ZT-s

sheeler
https://youtu.be/8SlGsI9ZT-s

That is crazy. I'm thinking it's a driver issue myself as everything is working fine after a fresh driver install, but once you reboot things get out of whack. If you do a fresh driver install again after getting the bad fps results everything is fine correct? 

I am not going to repeat myself from another post, but to add and update the list of hardware I used, I will just copy information I have send to EU EVGA support in one of the recent emails:
"
I have tested this issue immensely and this is the same issue I had with other two GPUs you sent me. I have tested so far:
- 4 different CPUs: 10900K, 11900K, E5-1680v2, 10900KF
- 5 different motherboards: 2 different ASUS Z490 Maximus XII Formulas, ASUS RIVBE, ASUS Z590 Maximus XIII Apex, Gigabyte Z490 Aorus Xtreme
- 3 different PSUs: EVGA Supernova P2 1000W, Be Quiet! Dark Power Pro 11 1200W, Corsair AX1600i 
- 3 different set of RAMs: 4x DDR3 Corsair Dominator GT and 4x DDR4 Crucial Ballistix Max (in this set I have tested every stick individually), 2x8GB Patriot Viper Steel 4000 CL16
- 4 different disk: 2x M.2 SSDs (Samsung 980 Pro, WD_Black SN850) and 2x SATA SSDs (Samsung 860 EVO, Samsung 840 Pro)
- 5 different Windows installations: Windows 10, Windows 10 Pro, Windows 10 Pro (fresh installation, just for diagnosing this problem), Windows Server 2019, Windows 11 Pro (fresh installation)
- 7 different NVIDIA drivers: 496.76, 511.23, 512.95, 516.59, 516.94, 517.48, 522.25 (I believe this is the newest one)
- 2 different versions of Precision X1: v1.0.9, v1.3.6. I have also tested without installing PX1.
"
 
In addition to the above I have tested powering PC using another relay (16A@230V, EU style ;)). Not to mention different BIOS versions and some Windows Updates.
 
 

Sajin
That is crazy. I'm thinking it's a driver issue myself as everything is working fine after a fresh driver install, but once you reboot things get out of whack.

 
I will restrain myself from any conclusions about what can be observed in this video. I'd rather wait at least this holiday week for some response from EVGA.
 

Sajin
If you do a fresh driver install again after getting the bad fps results everything is fine correct? 

 
That is mostly correct but I have only found out about this last month when, by mistake, I didn't do the proper OS reboot after driver installation. RTX 3090 was not behaving that way with every NVIDIA driver I installed. I confirmed only 2 or 3 that do. WHQL/Windows Update drivers are worse.

sheeler
I am not going to repeat myself from another post, but to add and update the list of hardware I used, I will just copy information I have send to EU EVGA support in one of the recent emails:
"
I have tested this issue immensely and this is the same issue I had with other two GPUs you sent me. I have tested so far:
- 4 different CPUs: 10900K, 11900K, E5-1680v2, 10900KF
- 5 different motherboards: 2 different ASUS Z490 Maximus XII Formulas, ASUS RIVBE, ASUS Z590 Maximus XIII Apex, Gigabyte Z490 Aorus Xtreme
- 3 different PSUs: EVGA Supernova P2 1000W, Be Quiet! Dark Power Pro 11 1200W, Corsair AX1600i 
- 3 different set of RAMs: 4x DDR3 Corsair Dominator GT and 4x DDR4 Crucial Ballistix Max (in this set I have tested every stick individually), 2x8GB Patriot Viper Steel 4000 CL16
- 4 different disk: 2x M.2 SSDs (Samsung 980 Pro, WD_Black SN850) and 2x SATA SSDs (Samsung 860 EVO, Samsung 840 Pro)
- 5 different Windows installations: Windows 10, Windows 10 Pro, Windows 10 Pro (fresh installation, just for diagnosing this problem), Windows Server 2019, Windows 11 Pro (fresh installation)
- 7 different NVIDIA drivers: 496.76, 511.23, 512.95, 516.59, 516.94, 517.48, 522.25 (I believe this is the newest one)
- 2 different versions of Precision X1: v1.0.9, v1.3.6. I have also tested without installing PX1.
"
 
In addition to the above I have tested powering PC using another relay (16A@230V, EU style ;)). Not to mention different BIOS versions and some Windows Updates.
 

 

Sajin
That is crazy. I'm thinking it's a driver issue myself as everything is working fine after a fresh driver install, but once you reboot things get out of whack.

I will restrain myself from any conclusions about what can be observed in this video. I'd rather wait at least this holiday week for some response from EVGA.
 

Sajin
If you do a fresh driver install again after getting the bad fps results everything is fine correct? 

sheeler
That is mostly correct but I have only found out about this last month when, by mistake, I didn't do the proper OS reboot after driver installation. RTX 3090 was not behaving that way with every NVIDIA driver I installed. I confirmed only 2 or 3 that do. WHQL/Windows Update drivers are worse.

Looks like you have done everything man. Except for reporting your issue to nvidia to have them look into the issue. https://docs.google.com/f...GP9hMCN6gKssA/viewform

sheeler
I am not going to repeat myself from another post, but to add and update the list of hardware I used, I will just copy information I have send to EU EVGA support in one of the recent emails:
"
I have tested this issue immensely and this is the same issue I had with other two GPUs you sent me. I have tested so far:
- 4 different CPUs: 10900K, 11900K, E5-1680v2, 10900KF
- 5 different motherboards: 2 different ASUS Z490 Maximus XII Formulas, ASUS RIVBE, ASUS Z590 Maximus XIII Apex, Gigabyte Z490 Aorus Xtreme
- 3 different PSUs: EVGA Supernova P2 1000W, Be Quiet! Dark Power Pro 11 1200W, Corsair AX1600i 
- 3 different set of RAMs: 4x DDR3 Corsair Dominator GT and 4x DDR4 Crucial Ballistix Max (in this set I have tested every stick individually), 2x8GB Patriot Viper Steel 4000 CL16
- 4 different disk: 2x M.2 SSDs (Samsung 980 Pro, WD_Black SN850) and 2x SATA SSDs (Samsung 860 EVO, Samsung 840 Pro)
- 5 different Windows installations: Windows 10, Windows 10 Pro, Windows 10 Pro (fresh installation, just for diagnosing this problem), Windows Server 2019, Windows 11 Pro (fresh installation)
- 7 different NVIDIA drivers: 496.76, 511.23, 512.95, 516.59, 516.94, 517.48, 522.25 (I believe this is the newest one)
- 2 different versions of Precision X1: v1.0.9, v1.3.6. I have also tested without installing PX1.
"
 
 

This is by far and beyond the most complex troubleshooting done by user, ever.
Have no doubt that simply you got two bad cards.
 
Hope that EVGA still have resources to resolve this problem. 

I watched your video and had to replay a few sections several times to make sure I really saw how you set up DSR.  What made me curious is why you change your monitors native display which looks to be 1920 x 1200 (Recommended)?  You set up DSR correctly in the Nvidia CP, but your monitor resolution should remain at the native spec.  Then in the game is where you select the higher resolution.  The way DSR works is the game will render at the higher resolution you set (in-game) then it downsamples it to your native resolution.
 
You can verify if that's what I saw in your video and if so, don't change your native res for your monitor in Windows.  Only change it in the game settings and see if that helps things.

Thank you for your replies and sorry for taking me so long to respond. I wanted to provide more meaningful information but after few other tries I am not getting any different results.

EVGATech_MarkusK
do you have the chance to test your componentes on a different system as well

Well, yes I have:
"
I have tested this issue immensely and this is the same issue I had with other two GPUs you sent me. I have tested so far:
- 4 different CPUs: 10900K, 11900K, E5-1680v2, 10900KF
- 5 different motherboards: 2 different ASUS Z490 Maximus XII Formulas, ASUS RIVBE, ASUS Z590 Maximus XIII Apex, Gigabyte Z490 Aorus Xtreme
- 3 different PSUs: EVGA Supernova P2 1000W, Be Quiet! Dark Power Pro 11 1200W, Corsair AX1600i 
- 3 different set of RAMs: 4x DDR3 Corsair Dominator GT and 4x DDR4 Crucial Ballistix Max (in this set I have tested every stick individually), 2x8GB Patriot Viper Steel 4000 CL16
- 4 different disk: 2x M.2 SSDs (Samsung 980 Pro, WD_Black SN850) and 2x SATA SSDs (Samsung 860 EVO, Samsung 840 Pro)
- 5 different Windows installations: Windows 10, Windows 10 Pro, Windows 10 Pro (fresh installation, just for diagnosing this problem), Windows Server 2019, Windows 11 Pro (fresh installation)
- 7 different NVIDIA drivers: 496.76, 511.23, 512.95, 516.59, 516.94, 517.48, 522.25 (I believe this is the newest one)
- 2 different versions of Precision X1: v1.0.9, v1.3.6. I have also tested without installing PX1.
"
To update the list, I can also include MSI 2080Ti Gaming Trio, which I have borrowed from a friend. No power drop issues.

EVGATech_MarkusK
specially the PSU? 

Your colleagues from SupportEU had also suggested that many times. That was also my initial impression of this issue.
I have been using EVGA Supernova P2 1000W with 2x GTX 1080Ti and those cards can draw much more power that RTX 3090. I have also tested RTX 3090 with Corsair AX1600i and the power drop issue was still there. Luckily enough, my friend with 2080Ti also had EVGA 1600 T2 to drive his GTX 2080Ti SLI so I tested RTX 3090 I received via RMA with this PSU. And I made a video of that:
https://youtu.be/g5Gl4jBH29M
Just to summarize: this is not (external to the GPU) power issue and I hope I showed that in linked video. EVGA website suggests 850W PSU for this GPU and Corsair AX1600i and EVGA 1600 T2 are one of the best PSUs created for consumer PCs right?
In this video you can see the power draw reads that HWiNFO provides. Those reads are accurate enough and multimeter attached to 8-pin PCIE#1 connector proves that. You can also see drops in power usage and FPS I am observing since 6 months.
You can also see how unstable this GPU is regarding core load. Even when power drops are not occurring (at least at the beginning and end of first segment) GPU struggle to achieve 98+ core load. CPU is not a bottleneck. So is not memory. There is no vsync, gsync, frame caps. There are no limits on GPU other than set by GPU itself. There is a fan blowing air at backplate and all fans are set at 100%. So is GPU power slider in EVGA PX1.
This didn't happen when I was using GTX 1080Ti or GTX 2080Ti where I was getting constant 98-100% core load.
 
In the next segment where I changed screen resolution using DSR (core load is higher) this is not happening, core load is usually at 99%. Power drops still exist though as you can clearly see at the end of the video, when I put to results next to each other. I wanted also to prove there is no CPU bottlenecking. Screen resolution was 5120x2160 and Intel 10900K@5.6GHz is a 12900K/13900K territory.
 

bdary
I watched your video and had to replay a few sections several times to make sure I really saw how you set up DSR.  What made me curious is why you change your monitors native display which looks to be 1920 x 1200 (Recommended)?  You set up DSR correctly in the Nvidia CP, but your monitor resolution should remain at the native spec.  Then in the game is where you select the higher resolution.  The way DSR works is the game will render at the higher resolution you set (in-game) then it downsamples it to your native resolution.
 
You can verify if that's what I saw in your video and if so, don't change your native res for your monitor in Windows.  Only change it in the game settings and see if that helps things.

That is what I am doing usually, but DSR doesn't provide anything if application is not running in fullscreen mode. In this instance, setting upscale in DSR allowed me to use higher resolution for game engine which put additional stress on tested GPUs. I have applied the same reasoning to the second linked video and you can see GPU is being utilized much more.
 

Sajin
EVGA doesn’t make the drivers for the gpu. Nvidia does. If you truly believe this issue is with evga 3090’s why don’t you just buy another vendors 3090 to see if you have the same issue with a different brand other than evga?

After spending so much time testing 3090 GPUs from EVGA, I don't believe much in regard with this problem anymore, other than it is not an issue created by any of my hardware and I have used other GPUs that prove there is something wrong with this one. If it is with all 3090 FTW3 models, whole EVGA series, whole 30 series GPUs or NVIDIA drivers I don't know but I don't think so since this issue would be much more public.
Besides, suggesting to buy another GPU to diagnose/resolve this issue is not very helpful, you know, from the fellow user perspective. If you bought a car and it was behaving faulty and if you brought it to dealership that sold you this car would you like to hear to go but another car from another dealership to prove something is wrong with this one? I don't want to live in a world like that, do you?

Sajin
I wouldn't mind going to checkout the same car at a different dealership to see if it has the same issue or not.

I won't go to electronics store with all my other computer parts just to test this issue (if they would provide me with other 3090 just for that kind of test). I also do not have a source for borrowing 3090 from other manufacturer. But this is going well beyond any user should put up with, so not a very good advice after all.

Sajin
I'm guessing you just want a 3090 ti from evga?

I didn't know EVGA is replacing (assuming faulty) 3090s with 3090 Tis. It is kinda understandable when you think how 3090s, even 3090 FTW3, is getting hot on the back side and there is no way to cool it without aftermarket dedicated solution - you can see that even on my last wideo: 95+°C on just 80% core load even when cooler is blowing air on the back side. I would be good to know if 3090 Ti have the same power drop issues.
 
I am still waiting for a reply from EVGATech_MarkusK. Maybe he can shine some light on such a proposition. I hope he didn't go to early holiday break - I sent him an email almost a week ago.