Question about GPU and 144Hz

[Caliph]

So here is the story :

 

I got MSI R9 390 and Benq Xl2411Z 144 hz conected with DVI

 

When i set my monitor on 144hz memory and core clock goes to max settings ( 3d mode)

When i set 120hz it is 2d mode on idle

 

So my question is :Can i somehow make the GPU stay on 2d mode on 144hz and how.
If not ill stay on 120 hz cause its not a big deal, but i dont want 144hz 3d mode on idle cause my gpu temp is 60 then and it warms up the room...

 

+ Should I use a DVI to display port adapter , will that allow me to get 144hz in 2d mode ( i read somewhere it will maybe)

 

[Sauron]

what do you mean with "memory clock" and "core clock"? Monitors don't have those

[Caliph]

what do you mean with "memory clock" and "core clock"? Monitors don't have those

I frazed it wrong, GPU core and mem clock goes to max settings

[Sauron]

I frazed it wrong, GPU core and mem clock goes to max settings

 

So if you set 144hz the gpu goes to max usage for no reason?

[Caliph]

So if you set 144hz the gpu goes to max usage for no reason?

Not usage, usage is 0% ,

My memory clock and gpu core clock goes to max settings ( 1500 and 1040)

[Sauron]

Not usage, usage is 0% ,

My memory clock and gpu core clock goes to max settings ( 1500 and 1040)

 

It happens because the gpu needs to keep up with the refresh rate, and is therefore unavoidable.

[Caliph]

It happens because the gpu needs to keep up with the refresh rate, and is therefore unavoidable.

Thought so, well ill keep it on 120hz, then its normal

[Caliph]

It happens because the gpu needs to keep up with the refresh rate, and is therefore unavoidable.

1 Extra question, my friend got Asus 144 hz monitor and GTX 970 and his gpu is acts normaln on 144hz

[Falkentyne]

It doesn't have to do with the GPU keeping up with the refresh rate.

 

It has to do with the memory clocks.

144 hz uses reduced vertical blanking.

 

Here is the real reason why usually the memory clocks run at full speeds at 144hz (and sometimes the GPU is also running faster as it detects a load on the memory):

 

http://www.monitortests.com/forum/Thread-Custom-Resolution-Utility-CRU?page=167

 

 

The memory needs to be retrained after changing clock speeds. This is done between screen refreshes during the vertical blanking period to avoid screen flicker. If the vertical blanking period is not long enough, there won't be enough time to refresh the memory contents before the next screen refresh.

 

So there's your answer.

Since 144 hz require a reduced vertical total (vertical blanking period), the memory runs at full speed instead of switching speed states from idle to load.

[Caliph]

It doesn't have to do with the GPU keeping up with the refresh rate.

 

It has to do with the memory clocks.

144 hz uses reduced vertical blanking.

 

Here is the real reason why usually the memory clocks run at full speeds at 144hz (and sometimes the GPU is also running faster as it detects a load on the memory):

 

http://www.monitortests.com/forum/Thread-Custom-Resolution-Utility-CRU?page=167

 

 

 

So there's your answer.

Since 144 hz require a reduced vertical total (vertical blanking period), the memory runs at full speed instead of switching speed states from idle to load.

Do u know why is this happening on r9 390 and not on GTX 970?

[BrightCandle]

I actually reported bugs in and around this issue with the old AMD 7970's. One of the problems then was the GPU's did not increase their memory clocks to compensate for the increased pixels necessary to run either a high refresh rate or eyefinity with 3 screens. Desktop performance really suffered and I actually wrote a program that sat in the toolbar and tricked the GPU in expecting a compute task soon that made it run at full speed with no usage that solve the problem. AMD in response to that bug changed the power profile to do the same thing. Its actually a really old problem that is still being solved the same way due to how AMDs power management works.

 

The issue is that AMD's GPUs don't determine clock speeds based on usage, they just use profiles to determine what to use. So if you have a so called "power vampire" like fur mark then AMD uses a profile to reduce the GPU core clock speed from the norm. On the desktop they use the power saving clockspeed unless it exceeds a certain pixel or refresh rate. They have profiles kick off when programs are run or in certain circumstances. Its basically all software defined. Nvidia works differently and a lot more like Intel's CPUs, its driven by actual usage so the GPU will clock higher when using high refresh rate screens but it wont max out unless it needs to and it will fluctuate as you do things on the desktop and indeed it will protect itself from furmark not by setting a maximum frequency but by balancing TDP/clock speed and its actual temperature. So I can for example run furmark with watercooling at maximum clocks with the TDP set much higher and the card doesn't throttle because I can cool it. Its just one of the interesting ways in which the two companies approach the same problem differently.

[Caliph]

I actually reported bugs in and around this issue with the old AMD 7970's. One of the problems then was the GPU's did not increase their memory clocks to compensate for the increased pixels necessary to run either a high refresh rate or eyefinity with 3 screens. Desktop performance really suffered and I actually wrote a program that sat in the toolbar and tricked the GPU in expecting a compute task soon that made it run at full speed with no usage that solve the problem. AMD in response to that bug changed the power profile to do the same thing. Its actually a really old problem that is still being solved the same way due to how AMDs power management works.

 

The issue is that AMD's GPUs don't determine clock speeds based on usage, they just use profiles to determine what to use. So if you have a so called "power vampire" like fur mark then AMD uses a profile to reduce the GPU core clock speed from the norm. On the desktop they use the power saving clockspeed unless it exceeds a certain pixel or refresh rate. They have profiles kick off when programs are run or in certain circumstances. Its basically all software defined. Nvidia works differently and a lot more like Intel's CPUs, its driven by actual usage so the GPU will clock higher when using high refresh rate screens but it wont max out unless it needs to and it will fluctuate as you do things on the desktop and indeed it will protect itself from furmark not by setting a maximum frequency but by balancing TDP/clock speed and its actual temperature. So I can for example run furmark with watercooling at maximum clocks with the TDP set much higher and the card doesn't throttle because I can cool it. Its just one of the interesting ways in which the two companies approach the same problem differently.

Okay, thanks for the answer.

Can i edit those profiles somehow so they use similar thing as nvidia, or im stuck with GPU on 60C on idle ?

[BrightCandle]

Okay, thanks for the answer.

Can i edit those profiles somehow so they use similar thing as nvidia, or im stuck with GPU on 60C on idle ?

What Nvidia does is a hardware feature, it protects the GPU from overheating, overcurrent and tries to achieve the maximum performance possible given its cooling and TDP limitations. AMD is using a software solution however and the GPU could in theory burn itself up with a program it doesn't know.

 

The solution used to be to edit some XML files that were on the drive and the clock speeds were in there, but the problem was they applied kind of globally there wasn't one for the desktop specifically. Now I am reading today its all moved into the registry so rather than butcher a process I have never personally tested I recommend you have a look into it yourself now you know how it works. I don't know if what you want is possible, it might be I just don't have an AMD card I can test it with.

[Caliph]

What Nvidia does is a hardware feature, it protects the GPU from overheating, overcurrent and tries to achieve the maximum performance possible given its cooling and TDP limitations. AMD is using a software solution however and the GPU could in theory burn itself up with a program it doesn't know.

 

The solution used to be to edit some XML files that were on the drive and the clock speeds were in there, but the problem was they applied kind of globally there wasn't one for the desktop specifically. Now I am reading today its all moved into the registry so rather than butcher a process I have never personally tested I recommend you have a look into it yourself now you know how it works. I don't know if what you want is possible, it might be I just don't have an AMD card I can test it with.

Well that sound complicated.

I can create profiles in MSi afterburned for everything, can i set 2d for everything other than games?