R9 290 auto-underclocking NOT due to thermal throttling

[JeffMan1212]

So I recently picked up a Sapphire Tri-x r9 290 4GB UBER MEGA SUPER-COOL EDITION, (Just kidding about that last part, but come on guys, GPU names can be silly) and let me tell you, it's been no walk in the park. When the card runs at it's full clock rate (1 Ghz), everything works like a charm. Unfortunately, the damn thing decides to underclock itself automatically IN GAME. This is no slight underclock either. The card will literally half it's speed in an instant. As you can imagine, this drops the frame rate quite significantly. I'm not going to stand for 20fps on a $300 card playing Far Cry F***ing 3. I've tracked the issues in MSI afterburner, and the screenshot attached to the post illustrates the problem pretty well. Is there some way to diasable this excuse for a feature, even if temporarily? I understand why my card underclocks on the desktop, but in game? Please help! Oh, and ignore the 130 memory clock. it's really 1300, the last zero just got cut off. the top sector is te GPU temp. 

 

System Specs: 

sapphire tri-x r9 290

AMD fx 8320 factory clock

biostar ta970 mobo

8gb g.skill ripjaws ram

tb seagate hard drive

windows 7 64 bit

asus optical drive

corsair CX750M PSU

Driver version: Omega (14.12)

[QueenDemetria]

So I recently picked up a Sapphire Tri-x r9 290 4GB UBER MEGA SUPER-COOL EDITION, (Just kidding about that last part, but come on guys, GPU names can be silly) and let me tell you, it's been no walk in the park. When the card runs at it's full clock rate (1 Ghz), everything works like a charm. Unfortunately, the damn thing decides to underclock itself automatically IN GAME. This is no slight underclock either. The card will literally half it's speed in an instant. As you can imagine, this drops the frame rate quite significantly. I'm not going to stand for 20fps on a $300 card playing Far Cry F***ing 3. I've tracked the issues in MSI afterburner, and the screenshot attached to the post illustrates the problem pretty well. Is there some way to diasable this excuse for a feature, even if temporarily? I understand why my card underclocks on the desktop, but in game? Please help! Oh, and ignore the 130 memory clock. it's really 1300, the last zero just got cut off. the top sector is te GPU temp.

Is it only doing this in Far Cry, or is it doing it with all games? If you only have tested FC3, then I recommend testing other games just to isolate variables. Most Ubisoft games run like garbage and cause many problems, so it could just be a problem with FC3.

[JeffMan1212]

Is it only doing this in Far Cry, or is it doing it with all games? If you only have tested FC3, then I recommend testing other games just to isolate variables. Most Ubisoft games run like garbage and cause many problems, so it could just be a problem with FC3.

I've tried it in far cry 3, and crysis 3. Both show the same results.

[QueenDemetria]

I've tried it in far cry 3, and crysis 3. Both show the same results.

Okay, and I forgot but what driver are you using?

[JeffMan1212]

Okay, and I forgot but what driver are you using?

Oh, sorry, I'll add that to the post. I'm using the latest omega drivers. I've also tested the problem with the previous version.

[Majestic]

Could also be TDP throttling. Or Vsync going to 30fps, try out that new framerate locking feature from the drivers instead of using vsync.

[JeffMan1212]

Could also be TDP throttling. 

Yeah, but I don't run into this problem when the GPU usage is constantly maxed. Only when it fluctuates. I think that the card thinks it's idling when the GPU usage is low, and lowers the clock rate to save power. 

[Majestic]

Yeah, but I don't run into this problem when the GPU usage is constantly maxed. Only when it fluctuates. I think that the card thinks it's idling when the GPU usage is low, and lowers the clock rate to save power.

 

Which could happen if vsync goes to 30fps. It's too hard to tell, due to the amount of information on the graphs. Try increasing the polling frequency and only log coreclock and framerate, see which drops first, the coreclock or the framerate.

The phaselag then tells us what is wrong.

 

A quicker assessment would ofcourse be to turn off vsync.

[JeffMan1212]

Which could happen if vsync goes to 30fps. It's too hard to tell, due to the amount of information on the graphs. Try increasing the polling frequency and only log coreclock and framerate, see which drops first, the coreclock or the framerate.

The phaselag then tells us what is wrong.

I'm a newbie, what do you want me to do?

 

[Majestic]

I'm a newbie, what do you want me to do?

 

 

Well, I was trying to figure out what happens and in what order. But an easy way to tell if it's vsync, is to disable it ingame.

 

Elaboration;

Vsync locks the framerate to your monitor's refreshrate, but when it cannot attain that framerate, it halfs it so it remains smooth. So if your graphics card is hovering around the 60fps mark, it is constantly going from 30 to 60.

What I asked was that you disabled all the monitoring values of afterburner except for coreclock and framerate, and increased the "polling rate", so we could see in which order it occured.

[JeffMan1212]

Well, I was trying to figure out what happens and in what order. But an easy way to tell if it's vsync, is to disable it ingame.

 

Elaboration;

Vsync locks the framerate to your monitor's refreshrate, but when it cannot attain that framerate, it halfs it so it remains smooth. So if your graphics card is hovering around the 60fps mark, it is constantly going from 30 to 60.

What I asked was that you disabled all the monitoring values of afterburner except for coreclock and framerate, and increased the "polling rate", so we could see in which order it occured.

Vsync was off during the gameplay. I feel like an idiot at this point, but could you explain the polling rate?

 

[Majestic]

Vsync was off during the gameplay. I feel like an idiot at this point, but could you explain the polling rate?

 

 

And it's not forced in the CCC?

 

It's on top of the "monitoring" tab. Lower the value to like 100ms, so you have more interpolling during the assessment of the problem. It's the time it waits for "sampling". So a value of 200 means every 200ms it checks the values. 

The higher this number is, the less accurate the data is, especially if you want to catch something like downclocking, which happens in milliseconds. Remove the checkmarks in the "monitoring tab" except for framerate and coreclock and enlarge the screen so you can see timedelays.

 

Could also be that Biostar board throttling your 8320. If the CPU throttles back, the framerate drops, causeing the 290 to clockback for powersaving. But both are confirmed when you do what i said above.

[JeffMan1212]

And it's not forced in the CCC?

 

It's on top of the "monitoring" tab. Lower the value to like 100ms, so you have more interpolling during the assessment of the problem. It's the time it waits for "sampling". So a value of 200 means every 200ms it checks the values. 

The higher this number is, the less accurate the data is, especially if you want to catch something like downclocking, which happens in milliseconds. Remove the checkmarks in the "monitoring tab" except for framerate and coreclock and enlarge the screen so you can see timedelays.

 

 

 

 

[Majestic]

 

 

The framerate drops before it downclocks. Meaning your CPU is likely throttling.

[JeffMan1212]

From heat?

[Majestic]

From heat?

 

Possibly, but could also be from the VRM's. Trying to find a utility to monitor throttling for AMD CPU's. I normally just use XTU, but ofc. that won't work on your machine.

 

EDIT: @JeffMan1212, Here you go; http://www.techpowerup.com/downloads/2288/throttlestop-6-00/

[JeffMan1212]

You are such a beast... My problems are solved! My CPU temps were far too high. All I had to do was blow out the CPU heatsync.