7950 won't return to idle speeds

[guest]

 

[LtMatt]

Hi folks,

Got a MSI 7950 TFIII BE and it idles at 500mhz and fully clocked RAM (I'll have to double check the RAM speeds later).
The thing never returns to 300mhz even with 0% GPU load.

I am running a single Benq XL2411T at 144hz via DVI-D at 1920x1080.

Any ideas?

I hope it's not another dodgey MSI!
But I won't jump to conclusions.

What driver version are you using and are you using any overclocking tools?

 

[LtMatt]

Catalyst Software Suite 13.1 1/17/2013 from the AMD site and I had a look around in the overdrive function and MSI Afterburner but set every thing back to defaults.

Just found this old thread:

http://forums.overclockers.co.uk/showthread.php?t=18454165&highlight=7950+idle+500mhz



Slightly different card but suggestions a driver issue then. I'll install the beta driver that AMD have on their site to see if that helps - failing that I'll manually adjust it in CCC.

Try 13.3 beta 3 and see if that fixes the issue. If not we have an AMD rep on the forums here, so if needs be send him a trust message and report the issue.

http://forums.overclockers.co.uk/member.php?u=146051

 

[LtMatt]

Thank you LtMatt. Couldn't have asked for a better or quicker answer. I'll report back later after work.

Happy to help mate.

 

[seanpr92]

you could also install something like saphire trixx or afterburner

 

[anthony566]

Another option is to remove your current drivers and any addition gpu software r boot and do a fresh install with the current drivers.

 

[LtMatt]

Make sure you report that issue to Thracks Kieran.

 

[OptimaLnrg]

Maybe those extra 24hz that you aren't seeing are too much for the card to handle at the desktop clocks

 

[Thracks]

A single 144Hz panel requires the same GPU bandwidth as multiple 60Hz monitors. It's probably not coincidental that they have the same idle clockspeeds. I'm looking into it, but I would not be surprised to learn that 500/1250MHz is required to avoid display corruption via insufficient bandwidth on the display PHY.

 

[humbug]

A single 144Hz panel requires the same GPU bandwidth as multiple 60Hz monitors. It's probably not coincidental that they have the same idle clockspeeds. I'm looking into it, but I would not be surprised to learn that 500/1250MHz is required to avoid display corruption via insufficient bandwidth on the display PHY.

Its about time we have you on this forum

Welcome to the mad house

 

[Thracks]

An update:

We reached out to BenQ to obtain the EDID data for their monitor, and have discovered that the vblank interval for this refresh rate is extremely short. In many cases, the speed of the vblank is faster than GDDR5 memory could feasibly respond to any changes in clockspeed that might be required to avoid display corruption. Such a clockspeed change could occur when you switch from one power state to another, with a commensurate change in mClk.

To avoid display corruption in all scenarios, the only one-size-fits-all fix is to run the VRAM at max clock. Luckily this represents an extremely small percentage of overall graphics card power consumption.

So, in short, I was right that it was the expected behavior, but wrong as to why.