Post your 3dmark fire strike score.

OK i will post 3dmark scores soon...in the mean time, I am getting a really odd directx low video memory crash error. After playing a game for 5 min, the game crashes to desktop with the error saying I have ran out of memory. I have this issue in ME catalyst, Andromeda, Wildlands. Witcher 3 and gta v runs fine.
 
Also, none of my gpus are reaching 100% usage in any of the settings.

I am not using the 950. On gta v i really want to get 120 fps but my fps hovers around 110 with my gpus at 60-70% and my cpu at 50-60%. It seems like something is holding my cards back. Also the low vram/low system memory crash is really not helping my case either.

Yeah, the 950 is too far of a performance gap for 1080 ti in SLI. I know, because I had a 970 paired with Titan X SLI (all Maxwell) and the 970 became the bottleneck as the PhysX card. So imagine a stronger SLI pair with an even weaker PhysX card (950 has less than half the CUDA cores of the 970)... 
 
If you run MSI AfterBurner, to make it noticeable whether the CPU+GPU are both taking a dive at the same time, overlap the graphs in the hardware monitoring portion (right click -> attach). Better with the individual CPU cores (all of them) as it'll give more data to use.
 
Here's the test where I found this out.
 
And I have to make an addendum to my post at GeForce... now that I've thought about it, I don't expect the 950 to hit 100% GPU load as a dedicated PhysX card... as CUDA cores (that do the PhysX) don't account for the whole GPU. Example, my 970 pegged 80% when it became the bottleneck and the main GPUs + CPU took a dive in usage. The 950 will probably see an even lower percentage.
 
Of course, this only accounts for games that use PhysX on GPU, not non-PhysX or CPU-only PhysX.
 
~~~~~
 
On a tangent (tirade?)
 
I personally wished people would understand that saying "my cpu hovers around 60-70%" isn't very useful with multi-core processors. Primarily because if a single core pegs 100%, the average can still be low and "look" like it's not a CPU bottleneck to people reading the post. "My cpu hovers around 60-70%" is vastly different from "none of the cores went past 70%."

I have never experienced this issue with sli before. Usually the game itself might not work, but never ever seen multiple games give the exact same error.

https://ibb.co/bZRf9v
 
Heres my 3dmark fire strike score. I know my cpu is old and is probably holding my gpus back, but I want to know if these scores are acceptable or not. My gpus were running at 2038 mhz. 
 
and yes to the previous poster, I did DDU every time in installed a driver.

the 3dmark that i downloaded wont let me make a link. Also i think i ran this benchmark at 1440p. I am just not sure whats holding my cards back....

shafat
the 3dmark that i downloaded wont let me make a link. Also i think i ran this benchmark at 1440p. I am just not sure whats holding my cards back....

You ran the correct test. Fire strike extreme is the 1440p test. Looks like you've ran fire strike before and were able to get a link. http://www.3dmark.com/fs/3971717
 
Are you using a hb sli bridge?
 

Run cinebench and tell me your cpu single core score.

shafat
My 4820k at 4.5ghz ht on gave me a score of 791.

Do the single core test.

I highly dobut your cpu is the bottleneck. I'd suggest re-installing windows onto a separate partition for testing.

The difference between x16 vs x8 3.0 is like 1-3%.

Set the paging file to auto manage size and you should be fine.

shafat
Thnx for the suggestion sajin. But is that a proper solution? I thought the reason I had 16gb ram is so that I dont need to use pf. Also, doesnt pf slow you down?

Yes, it is a proper solution. Disabling the paging file is a bad idea. Looks like you found that out the hard way. The only time the paging file slows you down is when your computer has to constantly swap to the paging file.

shafat
 Also, doesnt pf slow you down?

Yes, it causes stutter.