Lower Memory Clock = Better Performance?

[ruffel]

So I was running Time Spy on my RTX2070 Super (Inno3D Gaming OC x3), and I got around 9.7k on the graphics score. Then, I lowered the memory clock (from stock) by 502 in MSI afterburner, and the score went up to 9.9k. I ran it a few more times and it;s the same. I read that it could be the memory was having error because the clock is too fast, so is that the case here? If so, is this something I can get my card replaced for? And should I keep the memory clock at stock, or drop it to -502 for now? I think the card was manufactured at the 31st week of 2019, if that means anything. It also uses Micron memory, but I thought that was fine now. 

Oh, and while we're here, the card often boost itself to around 1965 MHz, and sometimes 2010 MHz out of the box. It says boost clock is 1815 MHz on the box, and I'm pretty sure I didn't overclock it to that. So do these cards just boost as high as they can, or did I actually do something?

 

CPU: Ryzen 5 1600 Stock

MB: Gigabyte ab350-gaming 3

RAM: Kingston 2400MHz 8GB x 2 

GPU: Inno3D Gaming OC x3 RTX2070 Super

[Skiiwee29]

There is a diminishing return to overclocking your GPU memory and at a point you do actually start to hurt it. I remember it was all explained by Kingpin in an overclocking battle live stream last year between Jayztwocents and Steve from Gamers Nexus. 

[ruffel]

1 minute ago, Skiiwee29 said:

There is a diminishing return to overclocking your GPU memory and at a point you do actually start to hurt it. I remember it was all explained by Kingpin in an overclocking battle live stream last year between Jayztwocents and Steve from Gamers Nexus. 

Yes, except it got faster when I lowered it from stock (7000MHz to 6498MHz). (sorry i didn't made it clear)

[Eigenvektor]

How often did you repeat those tests and are you certain nothing else has changed? The difference doesn't seem too big (~2%), so that could just be margin of error and/or test results influenced by other stuff like background programs.

[ruffel]

2 minutes ago, Eigenvektor said:

How often did you repeat those tests and are you certain nothing else has changed? The difference doesn't seem too big (~2%), so that could just be margin of error and/or test results influenced by other stuff like background programs.

I ran them in both configs like, 3 times in a row each, back to back. I did have some background programs running, but it's the same for both cases. But I probably should run them again now, and maybe on more programs too. I'll come back with some new results.