LinkBack URL
About LinkBacks
Overclocking nVidia and Folding
This is The method I used for finding the ceiling at which my cards will fold......
Normally you want to shoot for the highest possible stable clocks for your card for core, shader, and memory if you want the most bang for your buck. But not just any clock increase will show improvement. First, unlink the core and shaders so you can maximize both. Try to max each singly to get an idea of the limits you will be reaching.
Use Rivatuner hardware monitoring to monitor your clocks in real-time. This has the benefit of showing you that your nVidia card will increase your clock speeds in steps. For example, shader clocks increase in 54MHz steps only and nothing in between. For example, it goes from 1674 to 1728 to 1782, nothing in between. No matter what MHz you set it at in that range, it'll only be one of those numbers. When you get close enough to the next, it'll switch, and you can SEE this in Rivatuner monitor.
The same is true for core and memory, although the steps are smaller.
Have a look, and you'll see what I mean.
When overclocking for the purposes of folding then Shaders is king so this is where we will start.
Winding the core down to 600 with the shaders linked gives a starting point
core core shader shader mem 1% 1% 1% 1% 1% 1% Fahmon set actual set actual ppm 601 1545 999 1m26 601 1626 1620 999 1m24 601 1641 1620 999 1m24 601 1655 1674 999 601 1700 1674 999 601 1707 1728 999 601 1766 1782 999 601 1811 1836 999 601 1869 1890 999 601 1920 1944 999 1m14 1m13 1m14 1m14 1m14 1m13 5681 624 1920 1944 999 1m13 1m13 1m13 1m13 1m13 1m13 5681 642 648 1920 1944 999 1m13 1m13 1m13 1m13 1m13 1m12 5681 660 663 1920 1944 999 1m13 1m13 1m13 1m13 1m13 1m12 5681 680 676 1920 1944 999 1m12 1m12 1m13 1m12 1m12 1m13 5760 701 702 1920 1944 999 1m12 1m13 1m12 1m12 1m12 1m12 5760 722 720 1920 1944 999 1m11 1m12 1m12 1m12 1m12 1m12 5760 740 738 1920 1944 999 1m12 1m11 1m12 1m12 1m11 1m12 5760 750 756 1920 1944 999 1m12 1m11 1m12 1m12 1m11 1m11 5841 766 771 1920 1944 999 1m11 1m12 1m12 1m11 1m11 1m11 5841 775 777 1920 1944 999 1m11 1m11 1m11 1m11 1m12 1m11 5841 781 783 1920 1944 999 1m11 1m11 1m11 1m11 1m11 1m12 5841 784 783 1920 1944 999 1m11 1m11 1m11 1m11 1m12 1m11 5841 787 786 1920 1944 999 1m11 1m11 1m11 1m11 1m12 1m11 5841 790 792 1920 1944 999 1m11 1m11 1m12 1m11 1m11 1m11 5841 790 792 1920 1944 1053 1m11 1m11 1m10 1m11 1m11 1m10 5924 790 792 1920 1944 1107 1m10 1m10 1m10 1m10 1m10 1m10 5924
The working area is denoted in yellow
Setting core shader and memory to 601,1545,999 gives a time for 1% of 1min 26 seconds
Next selecting shaders and setting to 1626 produces an actual setting of 1620…changing the setting to 1641 does NOT actually change the setting it is still 1620.
Setting 1655 however means that the actual setting snaps to 1674 and again increasing the shaders to 1700 (increase is <54) the actual setting remains at 1674
Adding a further 7 making the new setting 1707 (>54) the actual speed snaps to 1728 (54> 1674)
Continuing down the shader set column increases of not less than54 make the actual setting increase in steps of 54
For this particular card increases above 1944 cause Folding to fail
Next Adjustments to Core
With the core set to 601 the columns show the time taken to complete each percent and Fahmon’s reported ppd of 5681 these are where the differences are seen when increasing core speed.
Again we can see that setting the core speed produces an actual core speed that can be either higher or lower as the setting snaps to the nearest fixed amount.
Again for this particular card going above 792 causes Fah to fail.
Finally; Memory settings
There is not a great increase to be had from increasing memory clocks it is however possible to eke a little more performance from your card as can be seen from the results in the table (again in yellow)
So guys I hope this is useful or saves time and gets you to where you want to be when you overclock your CPU
Reply With Quote
(I couldn't paste the chart in so here's a picture)
Originally Posted by Skyguy
