• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

290 throttling.

fanboy

[H]ard|Gawd
Joined
Jul 4, 2009
Messages
1,057
After playing around with my 290 for afew weeks I really don't think the card is throttling because of heat like most think. it's more to do with gpu load then anything I have recorded so far no matter what cooler is on the card.

Here is a Valley 1.0 run with the target set to 90c and fan speed set to 100% @stock clocks... if the gpu load drops off well so does the core clock and I let it run 2 loops before starting benchmark to get temp up.









2013-12-01 11:40:54 , 947.0 clock , 1250.0 memory , 89.0 temp , 46 fan , 2411 fan speed , 98 gpu load , 0 , 1110 , 11.75 , 1.141 , 1.000 , 15.1 , 117.0 , 165.3 , 147.5 , 56 vrm 1 temp , 74 vrm 2 temp ,

2013-12-01 11:40:55 , 947.0 clock , 1250.0 , 89.0 temp, 46 , 2412 , 98 gpu load , 0 , 1110 , 11.75 , 1.133 , 1.000 , 14.1 , 124.0 , 159.8 , 158.0 , 56 , 74 ,

2013-12-01 11:40:56 , 947.0 clock , 1250.0 , 89.0 temp, 46 , 2414 , 98 gpu load , 0 , 1110 , 11.75 , 1.141 , 1.000 , 14.8 , 141.5 , 200.3 , 158.0 , 56 , 74 ,

2013-12-01 11:40:57 , 928.3 clock , 1250.0 , 88.0 temp , 46 , 2414 , 82 gpu load , 0 , 1133 , 11.75 , 1.141 , 1.000 , 14.8 , 130.5 , 188.5 , 153.3 , 56 , 74 ,

2013-12-01 11:40:58 , 947.0 clock , 1250.0 , 89.0 temp , 46 , 2413 , 98 gpu load , 0 , 1116 , 11.75 , 1.141 , 1.000 , 16.4 , 132.0 , 186.3 , 150.8 , 56 , 74 ,

2013-12-01 11:40:59 , 947.0 clock , 1250.0 , 89.0 temp , 46 , 2413 , 98 gpu load , 0 , 1100 , 11.75 , 1.141 , 1.000 , 15.3 , 134.0 , 180.8 , 149.5 , 56 , 74 ,

2013-12-01 11:41:00 , 947.0 clock , 1250.0 , 89.0 temp , 46 , 2414 , 97 gpu load

00000.png
[/URL][/IMG]
 
It can throttle with heat and power target.
Although it happens in many games with the GPU maxing out around 50C (using Xtreme III cooler), ie Crysis Warhead.
In these, if I increase core clock, the throttling increases and performance hardly changes (with vsync off, power target maxed).

Other games it hardly throttles. ie COD Ghosts.
In these if I increase core clocks, performance increases.

This is accounting for CPU bottlenecks.
If the CPU is maxing out, there will be more throttling.
 
Last edited:
Is there a way to force constant voltage to the gpu as to keep the gpu from throttling?
 
If you take Trixx last build and only use the force constant voltage part and run at def setting of 95c and 47% it does peg it to 947Mhz even in 2d clocks but once it get it 94c it down clocks some but not much.. if you disable ULPS it may disable temp down clocking as I don't know..
 
Afterburner has an option to do that but it has no effect.
The voltage looks to be as dynamic as the clock speed, it will change depending on need.
I'm not sure that a constant voltage would help, the voltage level should be a response mechanism rather than a controller.
ie the voltage changes to meet the conditions, the conditions arent set depending on the voltage.

We cross posted.
 
the new power scheme does not use a constant voltage. its dynamic.

if you force it, it will completely disable all power features, and force the card to run in a mode for which it was not designed,

expect shortended life span.

instead, create a user fan profile.
 
the new power scheme does not use a constant voltage. its dynamic.

if you force it, it will completely disable all power features, and force the card to run in a mode for which it was not designed,

expect shortended life span.

instead, create a user fan profile.

What about those of us that are on full water loops? I haven't had a chance to do anything with my 290 since my EK block was defective and it seems like its taking forever and day for the new one to arrive..I did install it and make sure it wasn't DOA, and I was curious about the ASIC score so I checked that, but that about sums up my experience with it so far..

On my Golden 7950, I kept the voltage @ 1.23V with 3D clocks of 1325Mhz/7Ghz..I usually run my cards @ 100% 24/7, so the voltage wouldn't "change" very much for my usage situation no?
 
I dont think its possible to lock the voltage unless there is a feature not revealed yet.
Using the method provided in Afterburner, nothing changes, its still variable.
 
I dont think its possible to lock the voltage unless there is a feature not revealed yet.
Using the method provided in Afterburner, nothing changes, its still variable.

I had read that early on when I first got my card, but many seemed to think at the time that this was just because AB needed to be updated to work with the new cards..I know that TriXX hasn't been updated in forever...A lot of people using water seemed to recommend flashing over the Asus 290X bios, since that allowed for higher voltages (and clocks) but I am not sure if it will (or is designed to) keep the voltage @ a constant value.
 
Yea, I cam in here to ask just this. I notice when playing BF4 the core clock fluctuates anywhere from 850s to 900mhz. Cant get stock with the stock fan profile. 70% is actually quite loud but it keeps it from throttling a lot for me. Looks like I need to buy a aftermarket solution.
 
The reason I ask about (force constant voltage) is that it was the answer to the HD 7950 Boost which had a dynamic clock speed that really didn't work well. . if you forced it then the clock rate would stay at 925Mhz all the time.
 
I had read that early on when I first got my card, but many seemed to think at the time that this was just because AB needed to be updated to work with the new cards..I know that TriXX hasn't been updated in forever...A lot of people using water seemed to recommend flashing over the Asus 290X bios, since that allowed for higher voltages (and clocks) but I am not sure if it will (or is designed to) keep the voltage @ a constant value.

I'm running my card with a 290x BIOS, using the AC Extreme III cooler, load temps around 50 to 55C at 1000/1250 clocks.
Boosting the voltage works in the same way, just with higher voltages that allow higher clocks, the voltage doesnt remain at a set value.

Changing the GPU voltage allows a much higher memory clock as well as core clock, but beware:
When testing memory overclocking with Afterburner, I got over 1650MHz and then I finally had a crash (black screen).
On reboot, when the desktop loaded, the screen went black and the machine was locked up.
Stopping CCC loading on boot made it work ok again so I tested a high overclock with Afterburner to see if the card was ok still and it was perfect.
But I couldnt stop CCC immediately crashing the PC when run.
The remedy was to reinstall the gfx driver.
 
I'm running my card with a 290x BIOS, using the AC Extreme III cooler, load temps around 50 to 55C at 1000/1250 clocks.
Boosting the voltage works in the same way, just with higher voltages that allow higher clocks, the voltage doesnt remain at a set value.

Changing the GPU voltage allows a much higher memory clock as well as core clock, but beware:


What do you get without touching the voltage and what do you get with it?
 
I'm running my card with a 290x BIOS, using the AC Extreme III cooler, load temps around 50 to 55C at 1000/1250 clocks.
Boosting the voltage works in the same way, just with higher voltages that allow higher clocks, the voltage doesnt remain at a set value.

Changing the GPU voltage allows a much higher memory clock as well as core clock, but beware:
When testing memory overclocking with Afterburner, I got over 1650MHz and then I finally had a crash (black screen).
On reboot, when the desktop loaded, the screen went black and the machine was locked up.
Stopping CCC loading on boot made it work ok again so I tested a high overclock with Afterburner to see if the card was ok still and it was perfect.
But I couldnt stop CCC immediately crashing the PC when run.
The remedy was to reinstall the gfx driver.

Thats why people say dont use CCC to overclock, use MSI instead.
 
I never had any throttling at all, I've been running mine at 1000/1300 since they came out of the box. On water now at 1125 on both cards and I still don't see any throttling, even after 4 or 5 hours of BF4. How are VRM temps?
 
What do you get without touching the voltage and what do you get with it?

Running Unigine heaven 4.0 until it gets well warmed up:
Stock clock 1000MHz, power target +50%

Stock clock, stock voltage = 53C
Stock clock, +100mV = 57C
+100MHz, +100mV = 60C

Heaven gets it close to max temps, but not quite.
COD Ghosts gets a few C higher when at stock clocks, potentially more when clocked, will check some time.
 
Sounds like CCC tried to take over the OC settings from Afterburner. Weird.

As you change settings with AB, they change in CCC too if it is running, immediately.

I think what happened was this.
High overclock was set in AB, also gets set in CCC.
System crashed.
Rebooted, system crashed.
Booted safe mode, disabled CCC auto start.
Rebooted, system stable.
Tested slightly lower overclock to make sure the hardware was ok.
Reduced clocks back to normal.
Ran CCC, system crashed. <- Tested a few more times after rebooting.
Reinstalled driver, all ok again.

Whichever was run last (AB or CCC) gets control of the clocks.
CCC was set to use high clocks and cannot reset its clocks after being clocked too far if it then crashes on load.
This is why a driver re-install was needed to reset them.
 
just disable AB load on windows start temporarily.

Should fix the problem till you turn off Overdrive.
 
AB doesnt load on Windows boot up.
I have never allowed it to.
The crash continued to happen without AB running at all.
edit: Windows desktop would load, then a moment later as CCC loaded, the screen went black and the machine locked up. This was repeatable every boot until CCC was stopped from auto loading. Once on the desktop, running CCC caused an immediate black screen lockup as well.

I forgot to mention:
I increased VGPU by 100mV with AB so the clocks would have been stable at idle but not under load if the voltage is set +100mV.
CCC when it loads doesnt add 100mV so the memory overclock was likely way too high even for idle.
edit: memory clocks are affected a great deal by VGPU. I can get almost an extra 300MHz on the memory clocks with +100mV VGPU.

The problem is fixed now as I re-installed the driver, this is a headsup for others.
 
Last edited:
yes, i had this problem as well.

disable overdive in CCC, and this will stop.

It does not need to be enabled for AB to funciton,

in fact its better to leave it disabled so they wont interfere.
 
What i meant was difference from voltage vs stock oc.

I wondered after I made the post.
Here are the results running Unigine Heaven 4.0.

* Stock voltage will get me to 1070MHz core, 1350MHz ish memory.
* +100mV will get me 1155MHz core. Max load temp was 64C. I ran memory at 1500MHz so it didnt cause a problem (see other posts), it will go over 1600MHz with +100mV VGPU.
* -69mV allows me to run stock 290X clocks (1000/1250) stable.

This card is a Powercolor 290 OC (original clocks 975/1250) with an XFX 290X BIOS, AC Extreme III cooler.

edit: tested max memory speed with +100mV VGPU and it was stable at 1620MHz.
Max I can set to with AB is 1625MHz, but this gave me the black screen crash earlier so didnt bother going that far.
 
Last edited:
yes, i had this problem as well.

disable overdive in CCC, and this will stop.

It does not need to be enabled for AB to funciton,

in fact its better to leave it disabled so they wont interfere.

I see what you are saying.
It wont recover the crash I had, but it will prevent CCC from being the issue in the first place.
I also did this after the event :)

For those using the non beta Afterburner, you will need the newer beta version to change the power target without using CCC.
 
Last edited:
correct,

unfortunately, if youve already enabled OD and are getting the black screens, the only way to get rid of it is to uninstall CCC in safe mode.

im sure there is a registry key or an xml setting as well, i just dont know what it is off the top of my head
 
I wondered after I made the post.
Here are the results running Unigine Heaven 4.0.

* Stock voltage will get me to 1070MHz core, 1350MHz ish memory.
* +100mV will get me 1155MHz core. Max load temp was 64C. I ran memory at 1500MHz so it didnt cause a problem (see other posts), it will go over 1600MHz with +100mV VGPU.
* -69mV allows me to run stock 290X clocks (1000/1250) stable.

This card is a Powercolor 290 OC (original clocks 975/1250) with an XFX 290X BIOS, AC Extreme III cooler.

edit: tested max memory speed with +100mV VGPU and it was stable at 1620MHz.
Max I can set to with AB is 1625MHz, but this gave me the black screen crash earlier so didnt bother going that far.


Thanks man. Did you mod the vrm cooling? what vrm temps you get with max oc/voltage?
 
Yes, I stuck the angled heatsinks on the VRMs that come in the Xtreme III kit (see pics, VRM heatsinks are the rightmost and leftmost in both)
The fan is running at 12V permanently, I cant hear it above my PC case fans that are on quiet mode (Antec 900 case).
I read that 7V heatsink fans arent a good idea if increasing the voltage as the VRMs get much hotter, so I havent tried 7V.

The hottest VRM temp stabilises somewhere around 64C with +100mV at 1155MHz when running Unigine Heaven 4.0 for over 1/2hr.
They may get a bit higher running other things, I havent checked.


Quick edit:
Previously I could get down to -69mV and run stock clocks.
I went a bit lower even, pushed a bit more and crashed, on reboot, -69mV worked straight away so stuck with it. Load temp dropped 3C compared to stock volts.
Now, its only stable down to -25mV, a very big change.
Not sure why yet.

213g86d.jpg

2n7lj5d.jpg
 
Last edited:
Back
Top