dart@windeath:~/Downloads> vkmark
=======================================================
vkmark 2025.01
=======================================================
Vendor ID: 0x10DE
Device ID: 0x1380
Device Name: NVIDIA GeForce GTX 750 Ti
Driver Version: 2412773376
Device UUID: 83901a4578007e5e207dc911985d5d68
=======================================================
[vertex] device-local=true: FPS: 6603 FrameTime: 0.151 ms
[vertex] device-local=false: FPS: 6846 FrameTime: 0.146 ms
[texture] anisotropy=0: FPS: 6251 FrameTime: 0.160 ms
[texture] anisotropy=16: FPS: 6209 FrameTime: 0.161 ms
[shading] shading=gouraud: FPS: 5898 FrameTime: 0.170 ms
[shading] shading=blinn-phong-inf: FPS: 5912 FrameTime: 0.169 ms
[shading] shading=phong: FPS: 5767 FrameTime: 0.173 ms
[shading] shading=cel: FPS: 5754 FrameTime: 0.174 ms
[effect2d] kernel=edge: FPS: 9415 FrameTime: 0.106 ms
[effect2d] kernel=blur: FPS: 5818 FrameTime: 0.172 ms
[desktop] <default>: FPS: 6317 FrameTime: 0.158 ms
[cube] <default>: FPS: 10955 FrameTime: 0.091 ms
[clear] <default>: FPS: 9980 FrameTime: 0.100 ms
=======================================================
vkmark Score: 7055
=======================================================
Now I want to try Xlibre/OpenMandriva Cooker in a VM … but I want to boot it from Grub if possible so I can compare apples to apples with KDE/Xorg-Wayland without any “layers” in between and not have to repartition my drive to do it … what can I say … I only have 8G of RAM
Interesting that our scores are very similar but yours are so much more “consistent” where I have some lower and some much higher … must be an Nvidia/Radeon thing
I do find interesting all the talk about 8GB VRAM being too little and yet I never use more than 4GB even with Minecraft open. I don’t use shaders because my card already gets pegged playing Minecraft but the framerate is good enough. I have a NVIDIA Tesla K80 but the fan kit i bought is in shipping so I’ll probably be able to see what changes soon.
my plan is to use prime render offload (I believe @malcolmlewis uses this) and just do that for things like games, but then when i need to do AI/ML work I can use my K80 exclusively. (R9 390 support in ROCM got dropped years ago)
In my opinion if you’re serious about AI you either need an NVidia GPU (32G min) (not a gaming card)
or an AMD 5950 (powerhouse) or whatever supersedes it … or both
So this is what I have and as you can see there is no integrated cooling solution, so i bought a cooling kit. The interesting part is that a K80 is actually two GPU units on one PCB, and will likely be more than enough for my use case.
That’s WAAYY better than what I’m gonna get and yes…Cuda is the important part if you wanna crunch some numbers pair that with a ryzen9 5950x and 64G of RAM and DAMN …