6600XT possible problems!

[elbubsio]

Hi all,

I've just treated myself to a shiny new ASUS 6600XT, which seems like a great card despite all the negative press it's gotten, but I've noticed a couple of things that are worrying me a little. In blender, it's very quick at doing simple renders like a single figure, but give it something more complicated and it starts great, with the GPU and VRAM going at high speed, but within 3 seconds it seems to hit a 'barrier' and slows right down with the GPU and VRAM dropping to double digits. The clocks will occasionally jump back up, then straight back down again. In the end, it can take 50% longer than my old card. This is not caused by heat, as the card is only about 45c at all times, unless the radeon software is lying, which brings me to the second point. According to the radeon software the fans are always at 0rpm, even when they're audibly spinning at high speed.

So, the question is, is this likely just down to buggy drivers for such a new product (there's only one non WHQL driver available ATM), or does it sound like I may have a duff card?

Thanks.

P.S. - The rendering is being done using cycles if that matters. Eevee doesn't seem too bad.

 

[RJARRRPCGP]

For goodness sake, turn off the 0-RPM thing! I know that in the case of the RX Radeons, 1,700 RPM is a sweet spot, if the temps don't go too high. At least, looks true for XFX and MSI.

 

[Bomby569]

For goodness sake, turn off the 0-RPM thing! I know that in the case of the RX Radeons, 1,700 RPM is a sweet spot, if the temps don't go too high. At least, looks true for XFX and MSI.

0rpm never did no warm to a gpu, what matter is temp, and every card have different fans and different rpm values, so i don't know how is anyone supposed to follow the 1700 advice.

Hi all,

I've just treated myself to a shiny new ASUS 6600XT, which seems like a great card despite all the negative press it's gotten, but I've noticed a couple of things that are worrying me a little. In blender, it's very quick at doing simple renders like a single figure, but give it something more complicated and it starts great, with the GPU and VRAM going at high speed, but within 3 seconds it seems to hit a 'barrier' and slows right down with the GPU and VRAM dropping to double digits. The clocks will occasionally jump back up, then straight back down again. In the end, it can take 50% longer than my old card. This is not caused by heat, as the card is only about 45c at all times, unless the radeon software is lying, which brings me to the second point. According to the radeon software the fans are always at 0rpm, even when they're audibly spinning at high speed.

So, the question is, is this likely just down to buggy drivers for such a new product (there's only one non WHQL driver available ATM), or does it sound like I may have a duff card?

Thanks.

P.S. - The rendering is being done using cycles if that matters. Eevee doesn't seem too bad.

if it downclocks that badly temps are not fine, look into junction temperatures. If the fans are spinning it really doesn't matter what the software says, probably needs an update to latest version to support newer cards or something or some bug

 

[eidairaman1]

For goodness sake, turn off the 0-RPM thing! I know that in the case of the RX Radeons, 1,700 RPM is a sweet spot, if the temps don't go too high. At least, looks true for XFX and MSI.

0rpm never did no warm to a gpu, what matter is temp, and every card have different fans and different rpm values, so i don't know how is anyone supposed to follow the 1700 advice.



if it downclocks that badly temps are not fine, look into junction temperatures. If the fans are spinning it really doesn't matter what the software says, probably needs an update to latest version to support newer cards or something or some bug

There aint enough data to know wtf is going on, vague info doesn't help.

COMPLETE SYSTEM SPECS?

ROOM TEMPERATURES?

Blowing hot air on parts doesnt help at all.

Fyi the card is bus limited like the RX560 and lower.

 

[RJARRRPCGP]

I found a sweet spot for fan RPMs, which seem to be in that range for likely a lot of dual-fan video cards. How to find out? If familiar with Adrenalin, disable "0 RPM" and manually define a fan curve and try different RPM ranges and watch the GPU core temps.

 

[eidairaman1]

0rpm never did no warm to a gpu, what matter is temp, and every card have different fans and different rpm values, so i don't know how is anyone supposed to follow the 1700 advice.



if it downclocks that badly temps are not fine, look into junction temperatures. If the fans are spinning it really doesn't matter what the software says, probably needs an update to latest version to support newer cards or something or some bug

Room temperatures need to be given...

I found a sweet spot for fan RPMs, which seem to be in that range for likely a lot of dual-fan video cards. How to find out? If familiar with Adrenalin, disable "0 RPM" and manually define a fan curve and try different RPM ranges and watch the GPU core temps.

 

[pavle]

If core and memory clocks drop to 2D range during rendering, it's either a driver issue or you'd have to make a driver profile for blender, forcing 3D clocks.

 

[Bomby569]

If core and memory clocks drop to 2D range during rendering, it's either a driver issue or you'd have to make a driver profile for blender, forcing 3D clocks.

i was trying to remember where that avatar comes from. It's from Transport Tycoon right? i love that game.

 

[eidairaman1]

If core and memory clocks drop to 2D range during rendering, it's either a driver issue or you'd have to make a driver profile for blender, forcing 3D clocks.

Heat from the room can cause this, which has not been given.

So until @elbubsio comes back we cant do anymore

 

[pavle]

i was trying to remember where that avatar comes from. It's from Transport Tycoon right? i love that game.

Yes! It's my favourite game of them all; a room-full of trains without actually having to have a room for trains.

Heat from the room can cause this, which has not been given.

So until @elbubsio comes back we cant do anymore

I agreed on second part, but how much heat are we talking about? I mean he did mention the card is @45°C - that's nothing; even my GTX980 isn't that cool, and believe me that Maxwell card is really cool.

 

[elbubsio]

Hi all,

Thanks for the replies and apologies for not getting back sooner but I've had a busy weekend

An update: I've had some success with the blender issue. I found that if I turned both the CPU and GPU on for rendering instead of just the GPU, the GPU seems to work fine, with the clocks shooting up and remaining there till all tiles are rendered. Weirdly though, if I turn the CPU back off now, it still seems to work fine. Blender is weird

For those asking for specs:

Ryzen 2600 with Cooler Master 212BE Fan
RX6600XT graphics (of course)
16GB Crucial Ballistix RAM (with slight o/c from stock 3000 to 3200)
Fractal Edison-M 650w gold PSU
ASRock B450M Pro-F Mobo
Gamemax Kamikaze Case

Moving on to temps, the room temp is about 16-18c, so I doubt that's a problem. The card is idling at about 40-45c on both core and junc and under load (Metro Exodus at max settings and no R/T) it's at about 65c for the core and 80-90c for the junc (which seems a little high to me).

Does that all sound reasonably okay?

 

[pavle]

Looks like everything is in order, I found that a Powercolor RX 6600 XT also can get its "junction temperature" (=hottest spot on gpu die) up to 90°C (stock clocks) so no surprises there; glad it was only a blender issue.

 

[eidairaman1]

Yes! It's my favourite game of them all; a room-full of trains without actually having to have a room for trains.

I agreed on second part, but how much heat are we talking about? I mean he did mention the card is @45°C - that's nothing; even my GTX980 isn't that cool, and believe me that Maxwell card is really cool.

Blowing 80+ degree farenheit air on already hot parts just makes them hotter and can't remove heat faster.