Hello, today i've found strange thing:
660 ti and 680 clock is only 705 MHz instead of nvidia's claim 905 and 1003 MHz
respectively.
i had try different 3xx.xx drivers, but results are the same: 705 MHz
maybe it is the answer of low performance of 6xx series compared to 5xx series?
and how can we fix it? searching the web gives no distinct recipe.
Copyright © 2024 Einstein@Home. All rights reserved.
Nvidia drivers underclock 6xx GPU?
)
I dont know anything about Linux, but on windows every utility Ive used to check the clocks gave me the expected frequencies on all my keppler GPUs...
What is that "adaptive" thing? Usually adaptive means that the card will be underused to save power and to be more "green"... Isn't there a setting saying "max performance" or something alike?
Also, Keppler GPUs are designed to adjust themselves to fullfill a target power utilization, and they will get less "boost" with the default settings, so ussually adjusting the target power to the maximun allowed will avoid those downrated clocks needed to keep the power consumption in the selected range...
To set this power target, in Windows, you need a 3rd party app (those ussually used for OCing)...
Not sure if anything of this will be usefull for you but by comparisson you may get an idea of what is going on...
"Adaptive" as "Maximum
)
"Adaptive" as "Maximum performance" doesn't affect. maximum frequency is still 705 MHz (
My system reports 705 MHz
)
My system reports 705 MHz also and is running via performance level 3. One possibility could be that nvidia-settings is not reporting the boost frequency.
according to wiki GTX 680 has
)
according to wiki GTX 680 has 3 frequency: base (1006), average boost (1058) and max bust (1110).
our 705 MHz is still far away even from the base 1006 MHz.
Maybe one should ask this on
)
Maybe one should ask this on the NVIDIA forum....yes, the forum is back, after it was shut down after a hacker attack for several months.
Cheers
HB
already asked. 7 views for a
)
already asked. 7 views for a 12 hours and 0 reply. i think it is pointless )
RE: already asked. 7 views
)
Be patient...give those guys in the US a chance who are just now waking up :-)
CU
HB
You need to first take into
)
You need to first take into consideration that Kepler was designed from the start to be a variable power consumption card. That consumption is throttled by varying the clock speed.
You need to also check with several different programs as some won't report correctly or consistently. GPU-Z for example, is bad about underreporting load and adjusted speeds.
You also need to be looking at the clock speed under load. When mine is not under load, it goes as low as 162mhz, up to 1146mhz, and everything in between. Have maximum performance set, put the card under load, then check the reported clock speed with more than 1 application. If you're still at 705 then it's time to start looking at other things.
If you're running a 680M in a notebook on a battery, you won't go above 705.
There are not only 3 set clock speeds for the 680. It's capable of multiple speeds, and can go as low as 300mhz.
Low performance with regard to what? 6xx cards are not clocked slower than 5xx cards. Furthermore, Kepler smokes Fermi cards with regard to gaming. Fermi is better at crunching. We already talked about this on another thread here where I outlined the differences with links. The difference between performance with different types of applications has to do with the GPU type, and nothing to do with clockspeeds. Besides, the GTX 580 runs at 772mhz.
RE: RE: already asked. 7
)
34 views, 0 replies for 3 days. i told it was pointless )
I still run 2 first
)
I still run 2 first generation of FERMI GPUs, GTX470 & 480, the first with
the latest drivers and the latter with 275.33.
The GTX470, has to be downclocked by f.i. Smartdoctor or EVGA Precision,
otherwise they only downclock when no load exists.
My 2 AMD/ATI HD5870 GPUs, also only downclock if no load is present.
From 750MHz, via 550 to 157MHz.
Probably much todo with the change in architecture and market..........?!