RX580 Undervolt Temps and Fan noise for gaming

[Meddling-Monk]

Not sure if this should be in Overclocking or here, but feel this is probably just as viable.

So back in March, when I my old 7950 bit the dust, I obtained a Powercolor RX580 Red Devil as a replacement. It's cool and quiet when not AAA gaming, but the fan noise was always audible, even back before I replaced the case + AIO fans for Noctuas, just it was much more silent than the old 7950.

Now I've silenced the system where there's really only airflow noise and not much of any case fan noise, the RX580 I've discovered is very noticable in noise once it hits 1500+ RPM for fan speed (roughly 50% or more). Which is often the case when doing any heavy gaming.

So to cut to the chase, I've now done the following, and wondering if anyone can see if there's anything else I can do to make things better (reducing temps further, reduced power, further clocks, etc), but primarily to keep the noise level down (below the GPU fans 1500RPM).

1. GPU Bios selection is on "Silent", so 1355Mhz Core.
2. I've increased Power Limit by +20 in Wattman.
3. I've undervolted PStates 3 to 7. All now reside at 1.06V (1060mV).
4. Overclocking core beyond 1355Mhz was found to crash the entire system in games. Therefore maximum core is 1355Mhz only where it never crashes anything.
5. Memory voltage was set to manual and reduced from 950mV to 800m in Wattman.
6. Memory was increased from 2000Mhz to 2150Mhz.
7. Memory can reach 2200Mhz at the above voltages but will very rarely throw an odd Memory error (displayed via hwinfo sensors). 2250Mhz at above voltages will produce a lot more errors per minute. (These can be reduced from increased power, but that increases temperatures further)
8. Fan speed is set to Manual and to run from 1300RPM and 1500RPM. Switches off entirely at below 60C.
9. Room temperature is around 24C.
10. GPU idle temps are around 20-34C (20 or less from cold boot, 34 after heavy gaming and it idles down)
11. GPU load temps when AAA gaming are around 80-82C after extended session with the above voltages and clocks.
12. No throttling in games.
13. Can't hear the GPU fans over the airflow of case fans unless I take off case and listen to the card.

Anyone able to spot anywhere where I can try and improve things further? As I'm aware the tha GPU limit is roughly 90C, and letting temps reach that is unwise. Or have I taken things as far as I can without doing some custom (water?) cooling on it?

Thanks all.

 

[Ross Thomson]

with the AIO how much air flow do you actually have to the graphics card?

 

[AKAK]

You have undervolted a lot more than me. I went as far as 1080 core at 1375mhz and left it there as it stopped the crazy loud episodes. Have too much idle noise in my system to deal with first which I think may be the Corsair aio fan.

 

[Meddling-Monk]

with the AIO how much air flow do you actually have to the graphics card?

The AIO shouldn't be an issue in that regard. The AIO is for the CPU only, GPU remains stock Powercolor cooler for the Red Devil version.

Spoiler: Fan layout on the 650D

The area marked in Green is the H100i AIO.
The two fans above is marked A and B are Noctua A12x25 PWMs running at 60-70% for the temperature ranges of 20C to 75C. Both average 122m3/h for exhaust for their fan rating. (Although should be less as these are in Pull config)
The fan marked C, is also a Noctua A12x25 PWM, also running at a set 60%, averaging 61.26m3/h for exhausting air at the back of the case.
The fan attached to the PCI slot area marked E, is a basic Akasa fan on a resistor line, pushing only 11m3/h at best as an exhaust, to help keep airflow between the front and back of the case (Although, as the back slots are what it is attached to, most of that will be blocked)
The fan at D is a Noctua A14 PWM running as an intake, at a set 60%, averaging 84.12m3/h.
The fan at F, is a Noctua A20 PWM, running as an intake, set at 80%, averaging 117.52m3/h. (Although there will be some reduction from the case mesh at the front of the case and the grille, which I have not cut out)
Overall, Positive Pressure from roughly 201.64m3/h intake, and 193.78m3/h exhaust.
All of the removable 3.5" drive bays have been removed, so there is no obstruction between the fan at F to the fans at the back of the case. No 5.25" devices installed so there is a clear path from the Fan at D to the fans at the back.

Everything in between the front intakes and the exhaust fans gets plenty of sufficient airflow. The RX580 sits between fans C and E. The fan marked C at the back of the case near the top, still throws largely cool air out when gaming for lengthy periods of time. Whilst the fan at E (The PCI slots) is where most of it is slightly warmer air, but not overly so. So I don't think I can get the card more airflow without increasing noise from airflow from fans or the fans themselves.

You have undervolted a lot more than me. I went as far as 1080 core at 1375mhz and left it there as it stopped the crazy loud episodes. Have too much idle noise in my system to deal with first which I think may be the Corsair aio fan.

Thanks for giving me a rough guideline where another user is at for their card when undervolting. Appreciate it.

 

[Ross Thomson]

I'd play about with your fans a bit more. The fan on the PCI slots isn't needed nor should the other rear fans. Your CPU will be running hotter with the AIO set to exhaust. have a look at the actual airflow in the case, you might find the cold air is being dragged up before it gets to the GPU.

 

[Meddling-Monk]

I'd play about with your fans a bit more. The fan on the PCI slots isn't needed nor should the other rear fans. Your CPU will be running hotter with the AIO set to exhaust. have a look at the actual airflow in the case, you might find the cold air is being dragged up before it gets to the GPU.

Hmmm. Interesting.

Regarding the CPU, I'm not as concerned about the temps. The 4930k at sits at 4.3Ghz (offset -0.02v) and putting it under prime for stressing in this exhaust config, it barely reaches 60C. It only just about hits 70C if I OC it from 43x100 for 4.3Ghz, to 36x125 for 4.5Ghz (also in Pull config). But unfortunately it needs a lot more juice to get running, 1.4v nearly, so I'm not keen on switching to that until I really need that extra cpu power later down the line. So that's why I'm not that overly concerned about the temps for the AIO and CPU as they're well below what's considered the norm even when pulling air through them from the case. But what you've pointed out, the fans pulling more cold air out might be something to look into again. My previous smoke test was mostly external so checking that out again internally this time would definitely be something to do.

As for the PCI slots fan, prior to installing it, there was a massive heat trap in the area where the fan was (below the GPU, PCI slot 1, and the PSU), so much so that you can feel the side panel and window heating up from the GPU during gamiing sessions. So that's why that was added to better pull the incoming cooler air across from the front and to vent some of the heated air from the GPU at the same time. Will see how the removal of that PCI fan impacts the GPU when I flip the AIO fans around.

 

[Meddling-Monk]

Well, swapped the AIO around, and unfortunately the temps went up. Both for the CPU and the GPU.

CPU went up 10C degrees, and as it's now an intake, I had to replaced the old Demciflex dust filter on or risk dust everywhere, taking it from 60C to 70C during prime stress testing. Took the dust filter off, and it still went up by a few degrees.
The GPU went up maximum of 82C to 85C.

I'm thinking it's because by turning the AIO fans to an intake, the cool air it's supposed to bring, is mixing with the warm air from the GPU, but the preivous 180m3/h exhaust is now only taking the warm air out by something like 61m3/h instead at the back. So the warm temps are stuck being slowly removed and stacking up before the cool air can do any more to it. At least that's what I think has happened. So I'll be switching the top AIO fans back to Exhaust. (It had great amounts of Positive pressure though , but a shame about the temps ).

So I think unless someone else can see something I've overlooked, I "think", this is probably the best combination I can ask for with regards to a low(er) power GPU, low(ish) temp GPU, (virtually) silent (compared to airflow sound) GPU that doesn't throttle in games, with a relatively cool CPU as well that I can OC later to 4.5Ghz.

Oh, I didn't test switching off the PCI slot fan at the moment, since the GPU went to 85C with AIO intake, so I didn't want to take the risk of it hitting 90C or more (just in case). Will do that test a bit later once I switch the fans back around.

 

[Ross Thomson]

pretty odd the CPU temp went up give you turned it to intake and still went up when you removed the dust filter. were the fans in pull or push for the intake?

you could try switching the AIO to the front as an intake.

 

[Meddling-Monk]

Whilst is it no longer as bad as before, the GPU does make the side panels window a "teeny tiny" bit warm without the PCI slot fan running. Nothing amazing (in this cold weather), but something I'll probably keep attached and as is, just to keep as much of the GPU as cool as possible (especially during hotter weather).

pretty odd the CPU temp went up give you turned it to intake and still went up when you removed the dust filter. were the fans in pull or push for the intake?

The fans were in pull Exhaust and switched to pull Intake during the test just now. Incidentally, this wound up slightly louder than when in Exhaust configuration, and I think it has something to do with the screws for the H100i being flat against the fans below in pull Intake instead of being flat on the anti-vibration pad. Whereas the anti-vibration pads actied as a barrier between the AIO and the case during pull Exhaust.

As for the the possible reasons of the increased temps, I'm guessing:
1. The 4930k is overclocked, and fully loaded with RAM. So voltages were added to make sure it is stable. At 4.3Ghz, the CPU needs 1.264-1.288v. Not a lot by any margin, but certainly not very little. Checking the fluid temp of the H100i, it appears during the Exhaust phase (I forgot to check the Intake test), the temps of the fluid inside the AIO max out at 39.9C (40C). So I'm guessing when it was in pull Intake, the cool air going through the AIO was being heated up by the AIO during the CPU stress test, and the reduced Exhaust had more air being pushing in overall, but not enough of where the warm air was located was being extracted out quickly enough and it slowly built up a hot spot and that spread out to the CPU and AIO.
2. The pull Intake also possibly broke the air steam. Since it was pushing air down and is two of the strongest fans in the case. This probably broke the cool air going from front to the back, and caused the air to gather more easily and warm up. And with the Exhaust now only 1 fan, it wasn't enough to remove the now trapped warm/hot air hot spot before new cooler air was heated up from mixing with the warm air from the AIO and just stacked up on itself.

To be honest, I was just as surprised as you are after testing it. But nothing to do about it I'm afraid in this particular case.

you could try switching the AIO to the front as an intake.

Unfortuanately, this won't be possible. The front of the Corsair 650D was never designed with front mounting AIO's in the first place. Only if you did some custom modding of the case would that even be viable. And the H100i (original, not v2), has some short tubes. I doubt it would even be possible to mount it at the front and reach the CPU.

:: edit ::

Just checked, the distance between the front of the case and the location of the CPU is greater than the 11.5 inches that the tubes of the H100i are (assuming mounting the AIO with tubes up so it is closer to the CPU location). So yeah, no chance that is possible even if I were to start modding the case further.

:: further edit ::

And an open side case test had the GPU temps go from 80-82 to 79-81. So I think from what I'm reading that it means the airflow and cooling is probably going to get as good as it can if I'm not mistaken.

 

[Ross Thomson]

sadly your VRM's are to the left of the GPU so it would be inadvisable to fit a kraken G12 (my GPU runs at 45C with one MSI gaming X 580)

 

[Meddling-Monk]

Tiny bit of an update.

Not satisfied with the state of things, I've been reading around, looking at articles, etc. And have managed to help cool the RX580 down by a few degress by putting a few fans on top of the RX580 in Exhaust (pushing upwards). Accomplished via putting two 80mm fans on top of two empty ethernet sheathes atop the GPU on the backplate. This has helped keep the general temperature down by a few degress (1-3C) at full GPU usage for extended periods of time during gaming, and an average of between 4-7C whilst at benchmarking full GPU usage.

This has consquently brought back up the heat trap section that I discovered in earlier testing in the 650D in the top left corner. The primary troublemakers were the AIO, where the extra heat also heated up the case and kept the heat trap ongoing, so I have reversed that again so it is now an intake, which incidentally has no impact on the GPU at this time (as it's in a Push config instead of Pull like last time, which fully disrupted the airflow cooling involved) and changed the rear fan back into an Exhaust again. Currently, almost four fifths of the case when touched is chilly/cold. Only the window section slight above the GPU remains minorly luke warm. With the heat trap noticable in the top left corner when you run your fingers along the case there. But it is an improvement.

Will be trying to fit a fan roughly where the CPU is that exhausts to the rear (dunno how yet, but will see), and hopefully this helps dissipate and control the stuck heat trap section there. But for those who are with the Red Devil and want it more silent whilst running at full/near full speeds, you may want to consider some of what I've done to keep the GPU cool if you can't get an AIO onto it.