GPU mem clock

[Woahduds]

does it affect FPS in games?

[CoolaxGaming]

Yes but the effect is minimal at lower resolutions

[Enderman]

yeah it does, sometimes more than the GPU clock, which I dont understand...

but yeah if youre overclocking, push the mem clock as far as it can go

[Woahduds]

yeah it does, sometimes more than the GPU clock, which I dont understand...

but yeah if youre overclocking, push the mem clock as far as it can go

how would i benchmark that?

[Enderman]

how would i benchmark that?

just run a benchmark and look at the fps

then OC the memory, and run the benchmark again and compare

[Brandon Cassie]

yeah it does, sometimes more than the GPU clock, which I dont understand...

but yeah if youre overclocking, push the mem clock as far as it can go

It helps as it increases the GPU's memory bandwidth which in turn will hasten post/texture processing, anti aliasing etc

[Brandon Cassie]

how would i benchmark that?

Use FurMark 

[Enderman]

It helps as it increases the GPU's memory bandwidth which in turn will hasten post/texture processing, anti aliasing etc

yeah but i wouldnt expect it to give a huge 5% or 10% fps difference...

[Brandon Cassie]

yeah but i wouldnt expect it to give a huge 5% or 10% fps difference...

It really does lol

[TheElt]

Use FurMark 

Don't use FurMark for benching, it's an awful program for benchmarking since it puts unrealistic stress on the GPU.

 

OP, for benchmarking I'd reccommend using Unigine Valley, ramp up the memory until you see artifacting, then dial it back little by little until it gets stable. But the performance gain should be taken with a pinch of salt, since Valley loves memory clocks and scales well with them.

[JoyboyP]

Use GPU-Z to see what kind of GDDR5 memory you have.

 

Hynix - you can get a average 200 MHz increase.

Samsung - can go over 500 MHz.

 

With my 780Ti GHz Edition I can only get a +200MHz increase as it has Hynix memory.

My friend has a 980 Strix and got +525MHz increase as his has Samsung memory.

This should help you see what sort of OC you can get.

 

@ TheElt Brohoof /)

[Brandon Cassie]

Don't use FurMark for benching, it's an awful program for benchmarking since it puts unrealistic stress on the GPU.

 

OP, for benchmarking I'd reccommend using Unigine Valley, ramp up the memory until you see artifacting, then dial it back little by little until it gets stable. But the performance gain should be taken with a pinch of salt, since Valley loves memory clocks and scales well with them.

Furmark is good for testing the stability of the GPU as it maxes out most of the resources (framebuffer, pixel & texture fillrates etc) not sure about these other programs, they don't seem to stress the memory controller as much

 

But yeah I guess I'm a bit of an old fart when it comes to overclocking GPU's...been ages