Vega 64: VR SOC / VR Mem temps >100C

[Sumanji]

Hi All,

Put together my new machine on the weekend, and I'm a little concerned by the temps reported in GPU-Z...

While playing PUBG, VR SOC and VR Mem both going up to ~100C. GPU core is 78C and HBM is 85C ish (all numbers are max reported temps under load). Card is an Asus ROG Strix Vega 64.

The fan is also getting pretty loud (2400rpm / 60%), which I think might be caused by the VRM temps?

Case airflow pic: https://imgur.com/a/XUrhRR5

Would be great if you guys could share load temps so I can compare please!

Cheers,

Su

 

[Sumanji]

Try taking the side panel off of your case, might drop temps a bit.

I'm sure it definitely will, but that sort of defeats the point of having a case right...

My 56 & 64 bios and heavy clocks never seemed to get that warm.
Maybe some cards dont have good contact with the cooler.

Roughly what temps do you get on your 64?

The Stryx isn't the best cooler for Vega, I'd only go with Sapphire personally (and possibly Powercolor).

Yeah I wanted the Sapphire Pulse 56, but no stock

 

[Sumanji]

Those are PCB temperatures so Can add 5c easy.
Remember when People said Vega can set Yours house on fire ??
I would not get one without WC

Urgh, the Asus card temps are a complete joke versus the Powercolor... wonder if I should exchange it?

Cheers,

Su

 

[Sumanji]

Like I said in this post back here, you just answered with stock clocks and voltages.
But what are they? 1.15v at a throttling 1440-1480mhz ?

Apologies buddy, please see below:





Getting differing readings from GPU-Z and the Strix tool.

Strix is reporting 1.2V which seems insanely high for stock?! Whereas GPU-Z shows between 1.00-1.05V

I guess I should try manually setting the "stock" Vega 64 voltages and speeds and seeing if that makes a difference?

Cheers,

Su

 

[Sumanji]

Thanks for looking Dave.

I added the 2nd pic with AMD Wattman and HWMonitor also. Yes, I just loaded up PUBG (which puts 99% GPU load at the title screen). But maybe it's not boosting all the way up to 1600MHz because it doesn't need to? I have AMD frame rate limited to 100fps, so that could be why.

Apparently the Asus GPU Tweak software is a pile of garbage, so I am going to uninstall and use AMD Wattman.

 

[Sumanji]

Huh weird, I think it may not have been the frame rate cap after all. Played around with Wattman:

1.10V 1537MHz +50% power limit:




1.05V 1537MHz +10% power limit:




Something feels wrong and I'm not sure I can fix it by tweaking in Wattman?

Considering just returning this and getting the Sapphire Pulse Vega 56 - ComputerBase.de review shows it about the same noise & temp levels as a GTX1060 under load.

Happy to bite a restocking fee just to avoid all this faff...

Cheers,

Su

 

[Sumanji]

...as long as in realtime gaming you are seeing it hold around 1460mhz and 945 on the hbm then it should be working properly.

Did some further playing around last night... at stock volts/clocks, the card is NOT holding "full" speeds under load. It hovers between 1200-1260MHz on the core, but 945MHz on the HBM.

In Wattman, I played around to give it 1.050V on the core, 1.000V on the HBM and max core clock of 1537MHz.

At stock power limit, it would not go past 1200MHz on the core:



Raising the power limit to +30% it initially stayed at ~1400MHz for a while, but then the temperature got crazy and the card started throttling... (109C on the VRM!!):



I guess I could try giving it even fewer volts on the core (<1.000V?), but really it feels like I shouldn't have to be fannying around so much to get the card to run properly at stock speeds...

You can have a card pulling 1000W and still run cool if the cooling is designed properly.

Exactly... I think this Asus cooler is total pants.

I'd personally go with the Sapphire if they were my £££

Sapphire is always my first choice... if you can find me one in stock over here, I'll buy you a cookie

The Sapphire Pulse Vega 56 is in stock for ~$500 so I am just considering pulling the trigger on that and getting rid of the Asus tbh...

Use the iChill software in your raedon settings. It'll downclock the GPU watts without affecting performance.

Have you undervolted the card?

Chill is a good shout - is it always a good idea to leave that on? No effects on frame drops etc?

See my reply to Dave above re undervolting... seems to have made the temps worse!

Are you using stock case fans?

Nope, replaced the stock 120mm with a pair of FD Venturi HF-14 (using the low speed adapters to run them at ~900rpm, which feels like a good balance between airflow and noise... not that it matters because the GPU is so blood loud).


Thanks all!

Su

 

[Sumanji]

Edit just to clarify that testing was in a game and not a torture test ?

Yup, was just playing Dawn of War 3 in campaign mode (borderless window) with the GPU-Z displaying on top.