Experimenting with Overclocking

[Guest]

I have an XFX 390. It is a decent card, but I feel like it wasn't really made for overclocking (was just purchasing the cheapest 390 that wasn't a gigabyte or asus card). And my opinion of overclocking has changed pretty drastically with my experience with this card. So, like most of us do, I pushed my card to it's limit. My particular limit, sans BIOS hacking, etc., just standard overclocking was:

 

+100 mV

+50% power limit

core clock: 1165 mhz

memory: 1650

fans: 80%

temps: ~82C at full load

 

This made the card LOUD, hot, but fast. Lately I have been looking to upgrade, and I was likely going to start with my CPU. My goals for the new build were to draw less power, be quiet, and have a small form factor. This led me to experiment with the other end of the spectrum - undervolting. 

 

With the power limit all the way to -50% I was unable to get the card to ramp up over ~550 mhz on the core clock (memory had no issue staying at 1625). So I went for a reasonable overclock at (as close to standard power limit as I could) 5% increase.

 

Results:

-60 mV

+5% power limit

core clock: 1065 mhz

memory: 1625

fans: 52%

temps ~75c at full load

 

The games I play will use the extra mhz to push further than the physical limit of my monitors refresh rate (60 Hz), but at 60 frames I am more limited by my CPU than I am by my GPU - so pushing this extra 100 mHz on the core has net me little but, likely, the degradation of my chip, higher temps and more power draw. 

 

So, I am mainly posting because I always thought, "push to the absolute limit". But, this isn't necessarily the right choice for everyone. I am much happier with the new overclock and didn't lose much of anything (mainly gained, actually). The card runs at about the same noise level as the system as a whole, meaning I can't hear the GPU from the other room anymore. I'm not sure what reducing the mV does in conjunction with increasing the power limit, so any insight on that would be great.  

 

 

[Morgan MLGman]

I did the exact same thing with my R9 290X. It had a stock clock of 1030 MHz core and 1250 MHz memory, I was using it at -75mV voltage, 1060MHz core and 1260MHz memory which resulted in much lower temps and the performance benefit usually wasn't worth it. I ran the card at 1150/1460MHz with +44mV for benchmarking though.

[jools]

reduce mV to lowest stable under full load reduces temp

VxA=W proportional to heat output

[Guest]

1 minute ago, jools said:

reduce mV to lowest stable under full load reduces temp

VxA=W proportional to heat output

power increase seemed to have a much more drastic impact on the temps. 

 

For instance, having the power limit at the +5% and the mV at 0 compared to -50 mV had little impact on the temps (maybe a few degrees C at most). When increasing the power limit, however - I immediately saw changes in the temps. 

[jools]

I always thought power limit was when throttling occurred ie oh my card is drawing my power limit lets reduce it...power algorithm

I could be wrong cpu clocking is my thing

[Guest]

1 hour ago, jools said:

I always thought power limit was when throttling occurred ie oh my card is drawing my power limit lets reduce it...power algorithm

I could be wrong cpu clocking is my thing

I'm not sure. Basically - the mV seems like a secondary control to that of the power limit. Like a way to finer tune it, but the power limit being the main point of control. 

[jools]

gfx cards come with a wattage power usage rating eg.150w card . power limit either increases or decreases this number. while mV is the supply to thje gfx chip.

 

I can understand , most gfx cards now days have built in algorithms power and clock speeds automatically.

 

have u looked at this

 

http://www.hardocp.com/article/2015/09/21/xfx_r9_390_double_dissipation_8gb_video_card_review/10#.WBdi6Wd75ZQ

[jools]

also

http://www.hardocp.com/article/2015/09/21/xfx_r9_390_double_dissipation_8gb_video_card_review/3#.WBdksmd75ZQ

mite give some comparative voltages and tweaks