Question Low GPU clock on 7900 XTX ?

[hayze]

Hey Everyone,

As mentioned I have a 7900 XTX, driver version 22.12.2, If I use the latest version 23.7.1 it causes me to have a memory leak, same with v23.5.2, first version I downloaded for this GPU was v23.5.1 which may have been stable for me, but still had low FPS in all versions mentioned and crashing when using any overclocking/tuning presets, so this isn't a versioning issue.
i5-13600k (turbo boost turned on, so typically at 5ghz in game)
EK-Mana MSI MAG Z690 TORPEDO EK X D-RGB
Monitor resolution 3440 x 1440 120hz

I have been noticing very low FPS in mw2 (or all game for that matter) about 60 FPS avg on ultra settings (even lowest settings I get maybe 110-140). I notice my GPU Clock avgs about 600-700mhz.
I went into Adrenaline to try and mess with frequencies, I can't even set the minimum frequency to 800mhz without it crashing, I was able to set it to 600mhz that ran fine, but only tested for about 10 min.
Also I can't even use the tuning presets or automatic tuning (besides default) without it crashing any game.
I see most people can get between 2300mhz (which is the game frequency of this card) and 3000mhz, which finally made me realize why my FPS is so low.

I'm not really sure how to fix this, is my card defective?

I am using this EK-Loop Vertical GPU Holder EVO - Gen4 Riser, do you think the riser cable could be causing some sort of performance degradation? I also can't really test if it is easily or not, full liquid cooled setup with hard tubing.

 

[thestryker]

Whenever someone is having issues like this and there's a riser cable involved the first thing to check is indeed the riser. Try setting your PCIe slot to PCIe 3.0 as sometimes that will work even if the riser is the problem.

There's an outside possibility it's a power delivery issue, but this doesn't seem particularly likely to me.

 

[hayze]

So in my BIOS, PCIe was set to auto, tried Gen3 first then Gen4, both performed the same, but I did see a slight improvement when specifying the gen.
Instead of 60fps with about 600-700mhs, I was around 90fps 900mhz.

So I may just drain my PC and put the card in the slot and get rid of the Riser.

 

[hayze]

I finally got around to removing the Riser, and unfortunately my problem still persists.

Only fix is that I do not have a memory leak any more, a few driver versions ago fixed it.

I'm stumped on my issue.

My core clock averages about 1100-1200 mhz, which is higher than what I originally reported.

When playing the latest CoD at max settings I'm still just at 1100-1200mhz, or any game for that matter.

Another issue I'm having is my PC just powers off with certain graphic settings on games.

CoD multiplayer: any setting above balanced graphics, eventually when playing my PC will just shut off instantly.
CoD Zombies: anything above the lowest settings, pc shuts off
Black Desert Online: Low settings I can play for about 5 minutes, max instantly shuts off.
Overwatch: I have not had power off issues even at max settings, but still do suffer low clock speed

It's not a power supply issue either, I originally had a Season Focus 1000w Gold, but I bought a Corsair HX1500i Plat to replace it, and I'm using two PCIe power cables for my card, and not the y splitter.
Also heat is not an issue, my GPU doesn't exceed 60c and my CPU doesn't exceed 70c

At this point I'm thinking I have a bad card, can a bad card also cause PC to turn off?

 

[drivinfast247]

I finally got around to removing the Riser, and unfortunately my problem still persists.

Only fix is that I do not have a memory leak any more, a few driver versions ago fixed it.

I'm stumped on my issue.

My core clock averages about 1100-1200 mhz, which is higher than what I originally reported.

When playing the latest CoD at max settings I'm still just at 1100-1200mhz, or any game for that matter.

Another issue I'm having is my PC just powers off with certain graphic settings on games.

CoD multiplayer: any setting above balanced graphics, eventually when playing my PC will just shut off instantly.
CoD Zombies: anything above the lowest settings, pc shuts off
Black Desert Online: Low settings I can play for about 5 minutes, max instantly shuts off.
Overwatch: I have not had power off issues even at max settings, but still do suffer low clock speed

It's not a power supply issue either, I originally had a Season Focus 1000w Gold, but I bought a Corsair HX1500i Plat to replace it, and I'm using two PCIe power cables for my card, and not the y splitter.
Also heat is not an issue, my GPU doesn't exceed 60c and my CPU doesn't exceed 70c

At this point I'm thinking I have a bad card, can a bad card also cause PC to turn off?

You didn't reused any power cables after the PSU swap, correct?

And, when you say "shuts off" you mean the entire PC turns off, or just the monitor?

What symptoms lead you to the assumption that you had memory leak problems?

 

[hayze]

Edit: Well nevermind, my entire system just instantly powered off, plugging straight into the outlet did help though.


I just resolved my pc powering off issue, and yes it was the whole system just turning off.
This post actually helped me think of it https://forums.tomshardware.com/threads/faulty-graphics-card.3833965/
My computer is plugged into a surge protector, that also powers my 3 monitors and work laptop.. I looked at it yesterday and thought I don't think all of that should go to one outlet lol. Now my pc is plugged directly into the wall, which was the fix.

I'm still suffering from low clock speeds. I have Black Desert Online currently at maxed settings, custom clock settings on card, min 2500mhz - max 2900mhz (also with just default settings).
At the start of the game I was getting about 1500mhz, but then it dropped off hard to 600-900mhz .

I also saw a post about disabling ULPS in MSI afterburner that helped someone, I disabled that, but didn't help me.

and for the memory leak, I have 64GB of RAM, and it was a brand new build so not a lot of stuff running on it yet and after 36 hours my RAM would be maxed out. Just through troubleshooting, using process of elimination I found it was my GPU/GPU driver

 

[Avro Arrow]

Hey Everyone,

As mentioned I have a 7900 XTX, driver version 22.12.2, If I use the latest version 23.7.1 it causes me to have a memory leak, same with v23.5.2, first version I downloaded for this GPU was v23.5.1 which may have been stable for me, but still had low FPS in all versions mentioned and crashing when using any overclocking/tuning presets, so this isn't a versioning issue.
i5-13600k (turbo boost turned on, so typically at 5ghz in game)
EK-Mana MSI MAG Z690 TORPEDO EK X D-RGB
Monitor resolution 3440 x 1440 120hz

I have been noticing very low FPS in mw2 (or all game for that matter) about 60 FPS avg on ultra settings (even lowest settings I get maybe 110-140). I notice my GPU Clock avgs about 600-700mhz.
I went into Adrenaline to try and mess with frequencies, I can't even set the minimum frequency to 800mhz without it crashing, I was able to set it to 600mhz that ran fine, but only tested for about 10 min.
Also I can't even use the tuning presets or automatic tuning (besides default) without it crashing any game.
I see most people can get between 2300mhz (which is the game frequency of this card) and 3000mhz, which finally made me realize why my FPS is so low.

I'm not really sure how to fix this, is my card defective?

I am using this EK-Loop Vertical GPU Holder EVO - Gen4 Riser, do you think the riser cable could be causing some sort of performance degradation? I also can't really test if it is easily or not, full liquid cooled setup with hard tubing.

I don't think that it's the drivers because I'm using the same drivers as you with my XTX and I've been running games at 2160p60Hz with max settings without issue.

 

[kira-faye]

Bad Windows install. Make new media and try again. Latest drivers refusing to install was a symptom I only cleared up with a new install, tried everything but nothing showed corrupt.

 

[hayze]

I don't think that it's the drivers because I'm using the same drivers as you with my XTX and I've been running games at 2160p60Hz with max settings without issue.

at the time 23.7.1 was the latest version, we are now on 23.12.1. When I completely removed my gpu drivers, I let my pc run for 24 hours, and during that period it remained stable. When I installed the drivers, memory leak was back. But this is no longer an issue, as it's fixed now.

Bad Windows install. Make new media and try again. Latest drivers refusing to install was a symptom I only cleared up with a new install, tried everything but nothing showed corrupt.

When I first had this issue, I thought it was a windows problem, I actually went back to windows 10 and still had the same issues, then reverted back to 11. All clean installs.

My only problems as of now are low clock speeds and pc shutting off instantly on too high of graphics. varies per game.

Something interesting I noticed while playing New World, when I play on low settings my clock speed is avg 1200MHz, but when on medium settings my clock speeds drop to about 650MHz avg, weird it would drop, should increase since more load. Max settings I can only play for maybe 10 minutes before PC shuts off.

I may have found another reason for my pc shutting off, I didn't realize my GPU has two heat sensors, my hotspot at medium settings in New World sits at a consistent 110c, sometimes spikes to 116c, while my avg temp is only 37c. I was reading these two shouldn't vary more than about 10c and hotspot shouldn't really exceed 110c. Possibly due to bad/not enough thermal paste?
Anyone know much about this?

My next step is to put my 7900 into another rig and see how it performs, and then I'll take my old 1080ti and put it in this machine see if it has any issues. Maybe a week or so until I have results, hard tubed liquid cooling definitely not fun to work with while troubleshooting lol.

 

[thestryker]

Hotspot temps can vary quite a bit, but generally speaking you should be looking at 10-30C difference maximum. Usually the difference gets higher as the temps go up (ex at 55C hotspot might be 65C, but at 75C hotspot might be 95C) though that isn't always the case.

If the disparity is that high then it sounds like the cooler isn't properly mounted. It's possible the wrong thickness thermal pads were used on memory/VRM and it's causing mounting pressure issues or the cooler might just not be tightened correctly. Thermal paste alone wouldn't cause that unless you used none or there was a protective film not removed from the cooler plate.

 

[Avro Arrow]

at the time 23.7.1 was the latest version, we are now on 23.12.1. When I completely removed my gpu drivers, I let my pc run for 24 hours, and during that period it remained stable. When I installed the drivers, memory leak was back. But this is no longer an issue, as it's fixed now.

How did you fix it (or did it just fix itself)?

 

[hayze]

Hotspot temps can vary quite a bit, but generally speaking you should be looking at 10-30C difference maximum. Usually the difference gets higher as the temps go up (ex at 55C hotspot might be 65C, but at 75C hotspot might be 95C) though that isn't always the case.

If the disparity is that high then it sounds like the cooler isn't properly mounted. It's possible the wrong thickness thermal pads were used on memory/VRM and it's causing mounting pressure issues or the cooler might just not be tightened correctly. Thermal paste alone wouldn't cause that unless you used none or there was a protective film not removed from the cooler plate.

it's an EKWB block directly from them, so one would hope/think they would not give the incorrect thickness.
Do you know how thick they need to be? Or yeah maybe I didn't tighten it down properly lol.

How did you fix it (or did it just fix itself)?

One of the patches fixed it, it was about 2 or 3 patches after 23.7.1. Once this was solved I didn't mind playing at low graphic settings, which is why I'm back 6 months later, finally drained my system, cleaned it, and got rid of the riser cable lol.

 

[thestryker]

it's an EKWB block directly from them, so one would hope/think they would not give the incorrect thickness.
Do you know how thick they need to be? Or yeah maybe I didn't tighten it down properly lol.

Well the pads can be varied thickness so something like the wrong pad in the wrong place etc. It's not a guarantee either way just a potential point of failure. As for what they should be you'd have to look up the specific model of card and see if anyone's taken measurements though with a custom block these might not apply. Definitely a circumstance where you are going to need to take the block off to check everything to make sure the pads are in the right place, there's good thermal paste coverage and then tighten it up.

 

[hayze]

Well the pads can be varied thickness so something like the wrong pad in the wrong place etc. It's not a guarantee either way just a potential point of failure. As for what they should be you'd have to look up the specific model of card and see if anyone's taken measurements though with a custom block these might not apply. Definitely a circumstance where you are going to need to take the block off to check everything to make sure the pads are in the right place, there's good thermal paste coverage and then tighten it up.

I finally fixed the issue and it was definitely because of the thermal pads, I guess when I first put it together I didn't pay attention to thickness. Thank goodness I didn't damage my card.

Everything is 100% fixed now, PC doesn't just shut off any more, my clock speeds are correct now, and hot spot temp is only ~10c higher than avg now.