Multiple Monitors, HDMI signal loss when GPU is stressed

[cinderman]

Hey Guys

I’m worried there may be an issue with my 4090 I purchased from Overclockers. So I’m wondering if anyone has a solution or is this common problem with Nvidia on multiple monitor setups.

Issue

When the GPU is stressed on certain games the HDMI connected monitor disconnects & reconnects constantly.

When I tried the Early access Diablo 4, whenever I had a busy screen, the HDMI started to disconnect & reconnect . I had Aida64 running at the time and could see the 4090 with a 374w power draw 100% usage when running the game on max settings at 3440x1440p 175hz.

At the time I had 2 monitors on displayport and Aida64 monitor panel on the HDMI port.

Recently I started playing The last of us and the same thing started happening again…

Resolutions attempted to address issue

1. Changed HDMI & power cables
2. Reseated GPU & cables
3. Changed 12VHPWR power cable
4. Reinstalled Windows 10
5. Updated all firmware on everything and have latest drivers
6. Replaced Aida64 panel
7. Purchased a Displayport to hdmi adapter and had no disconnects on Aida panel.
8. Bought another Dell S2721DGFA 2560x1440 and connected via HDMI and same thing happened in The Last of us (opening scenes)
9. Updated again to latest drivers, I haven’t tried windows 11 yet

Monitors
Displayport 1 - Alienware AW3423DW 3440x1440p 175hz
Displayport 2 - Dell S2721DGFA 2560x1440 165hz
Displayport 3 - Dell S2721DGFA 2560x1440 165hz
HDMI - Generic panel for AIDA64 1920x1080p 60hz.

PC Specs

CPU: AMD Ryzen 9 5950X
Motherboard: MSI MEG X570 ACE
RAM: Corsair Vengeance RGB Pro DDR4 3600Mhz 128Gb
M.2 1: WD Sn850x 4Tb
M.2 2: WD Sn850x 4Tb
M.2 3: WD Sn850 2Tb
PSU : Corsair RM1000x 80 PLUS Gold (latest model) connected via APC UPS
12VHPWR : Corsair Premium 600W PCIe 5.0 / Gen 5 12VHPWR PSU Cable
Case : Corsair 5000X
Fans: 10 x Corsair iCUE QL120 RGB

Typical Power & Temps on max load (nothing is overclocked)

GPU : 4090 @ 374w 60-62c (hotspot +10c) (msi afterburner)
CPU: 5950x @ 88w 80-84c (game profile ryzen master)
M.2s (all) : max 60c
North bridge : 69c
System : 48c

Power Draw from mains is 680-700w when running everything (sound,4 monitors 100% usage on both GPU/CPU, transferring large files on network & between m.2s.)

 

[Semple]

A good level of testing there, only thing extra I can think of is do you see the issue with:
- 1 DP + 1 HDMI
- 2 DP + 1 HDMI

That'll give some extra evidence around the card being unstable under load, or if there's just more of a generic HDMI issue.

 

[cinderman]

A good level of testing there, only thing extra I can think of is do you see the issue with:
- 1 DP + 1 HDMI
- 2 DP + 1 HDMI

That'll give some extra evidence around the card being unstable under load, or if there's just more of a generic HDMI issue.

It was 2DP+HDMI on Diablo 4, the extra DP is new

I’ll try the single DP + HDMI today and let you know.
I could also just try the main 3440x1440p on the HDMI too I guess (albeit not 175hz)

I’ll update some more of my specs like the temperatures in the mean time.

 

[cinderman]

I’ve just completely reassembled my whole pc and Will use this weekend’s Diablo IV server slam to see if I still get this issue. I’ll see if what I’ve done will fix it. I suspect that my card is having issues when I run >2 monitors

 

[cinderman]

Update


I didn’t experience those issues again on Diablo 4 this weekend nor on The last of us.

One thing to note though is both the games were patched and my 4090 is no longer running at 100% usage now (it’s more like 60% on max settings)

So in conclusion I’m not sure if it’s the games themselves causing the issue or that I‘m not stressing my GPU like before (tried path tracing on cyberpunk, and max settings on hogwarts too).

Does anyone have a suggestion on any games I should try to stress the 4090 on both memory and processor.

I’m not trying to jinx it, I just don’t want a repeat of what happened to gigabytes 3080s when they were failing on heavy loads.

 

[eeii]

Fur mark is a really good gpu stress test to use. I would not have it on for longer than 15 mins. If it does it during this test RMA the card.

 

[cinderman]

Fur mark is a really good gpu stress test to use. I would not have it on for longer than 15 mins. If it does it during this test RMA the card.

Ok thankyou I’ll give it a go tonight

 

[cinderman]

Fur mark is a really good gpu stress test to use. I would not have it on for longer than 15 mins. If it does it during this test RMA the card.

Ok I ran the Furmark test for an hour and nothing happened. I checked my Aida64 and saw the cpu and video memory wasn’t being used. So I tried cinebench with furmark, still no issue. Just now I tried heaven and furmark to try to use video. Interestingly heaven states my card is running at 3120mhz (2865 on aIda) and memory at 10701 with default settings on afterburner (but video memory usage is still low).

Anyways thanks for the help, I think the issued laid with those 2 games (the last of us previous update, Diablo IV previous beta) and probably the 2 month old nvidia drivers. Also I’m not sure if it’s nvidia themselves or afterburner but my gpu is running faster (stable) than the official nvidia spec now.

 

[eeii]

That’s good to hear.