FPS Fix

[tin]

Hey, I haven't seen any appropriate slash working threads regarding this.

I bought a new laptop and I'm stuck at around 20fps, but laptop should be able to get more fps.

AMD Quad Core A8-4500M 1.90GHz (4MB cache)
AMD Radeon HD8750M 2GB
RAM 4GB DDR3 1600MHz

Thanks in advance! :)

[VladVP]

Are you playing in fullscreen or windowed mode?

[tin]

Fullscreen.

[VladVP]

That migth be a problem on some CPUs... have you tried playing in windowed mode, and noticed anything different regarding FPS?

[HoboHob]

What's your CPU temps when you play? CPU throttling can kill your FPS. Though I somewhat doubt it's your CPU overheating.

Even why my CPU hits its T.J. Max I can still play AoS ~40 FPS.

[Gingerninjaperson]

Someone told me (have no clue on the legitness of this) that this games engine is optimized for older computers, don't know why that would impact FPS on better ones though. I have an i5 3570k stock clock and a 650, and I usually get ~60fps.

[learn_more]

Someone told me (have no clue on the legitness of this) that this games engine is optimized for older computers, don't know why that would impact FPS on better ones though. I have an i5 3570k stock clock and a 650, and I usually get ~60fps.

it doesnt use the gpu for rendering so a good video card doesnt do anything.

[VladVP]

Someone told me (have no clue on the legitness of this) that this games engine is optimized for older computers, don't know why that would impact FPS on better ones though. I have an i5 3570k stock clock and a 650, and I usually get ~60fps.

it doesnt use the gpu for rendering so a good video card doesnt do anything.

It does use MMX and/or SSE though, which on most, if not all, modern systems is contained in the GPU.

[learn_more]

It does use MMX and/or SSE though, which on most, if not all, modern systems is contained in the GPU.

and again, you are full of shit.
please don't comment at all if you don't have a clue what you are talking about.
http://en.wikipedia.org/wiki/Streaming_SIMD_Extensions
http://en.wikipedia.org/wiki/MMX_(instruction_set)


and to stay ontopic:
i have found amd processors to perform worse on cpu-heavy games than intel processors.

[VladVP]

I meant that MMX registers and instructions on most modern systems were located in, and handled by the GPU. The fact that MMX and SIMD are CPU extensions, doesn't mean they have to be used on a CPU.

[learn_more]

that would be a totally retarded implementation, as performance would degrade severely.

[tin]

Alright so... any actual fix for this?

[VladVP]

Alright so... any actual fix for this?

Did you feel any difference when you tried my suggestion?

that would be a totally retarded implementation, as performance would degrade severely.

And once again, you are full of excrement.
It's a common and well-known fact that most GPUs which are not from the 80's have a significantly better performance than CPUs. Both in terms of calculation speed (Both integer ops and flops), but also graphics calculations or read/write speed. That last thing is exactly what MMX and SSE is designed to deal with. How can performance "degrade severely" if the technology in question is moved to a platform which makes it operate faster?

[learn_more]

And once again, you are full of excrement.
It's a common and well-known fact that most GPUs which are not from the 80's have a significantly better performance than CPUs. Both in terms of calculation speed (Both integer ops and flops), but also graphics calculations or read/write speed. That last thing is exactly what MMX and SSE is designed to deal with. How can performance "degrade severely" if the technology in question is moved to a platform which makes it operate faster?

do you know ANYTHING about pc architecture at all?

what is the processor executing a program?
the cpu
now imagine (lol) someone wanted mmx instructions to occur on the gpu
first, the cpu determines it's a gpu instruction
then it sends the instruction + data needing processing to the gpu over whatever the gpu is in, (so could be anything including pci) to the gpu
the gpu executes
the gpu sends the result back


would that be faster than the cpu handling the instruction itself?

not by a long shot.

[VladVP]

And once again, you are full of excrement.
It's a common and well-known fact that most GPUs which are not from the 80's have a significantly better performance than CPUs. Both in terms of calculation speed (Both integer ops and flops), but also graphics calculations or read/write speed. That last thing is exactly what MMX and SSE is designed to deal with. How can performance "degrade severely" if the technology in question is moved to a platform which makes it operate faster?

do you know ANYTHING about pc architecture at all?

what is the processor executing a program?
the cpu
now imagine (lol) someone wanted mmx instructions to occur on the gpu
first, the cpu determines it's a gpu instruction
then it sends the instruction + data needing processing to the gpu over whatever the gpu is in, (so could be anything including pci) to the gpu
the gpu executes
the gpu sends the result back


would that be faster than the cpu handling the instruction itself?

not by a long shot.

Come on, now you're just pulling this out of your ass.

How is

Spoiler:

Code: Select all

mov rax, [src]
mov rbx, [dst]
shr rax, 48
shr rbx, 48
add rbx, rax
and rbx, FFFFh
shl rbx, 48
or rcx, rbx

mov rax, [src]
mov rbx, [dst]
shr rax, 32
shr rbx, 32
and rax, FFFFh
and rbx, FFFFh
add rbx, rax
and rbx, FFFFh
shl rbx, 32
or rcx, rbx

mov rax, [src]
mov rbx, [dst]
shr rax, 16
shr rbx, 16
and rax, FFFFh
and rbx, FFFFh
add rbx, rax
and rbx, FFFFh
shl rbx, 16
or rcx, rbx

mov rax, [src]
mov rbx, [dst]
and rax, FFFFh
and rbx, FFFFh
add rbx, rax
and rbx, FFFFh
or rcx, rbx

mov [dst], rcx

processed on the CPU, faster than

Code: Select all

movq mm0, [src]
movq mm1, [dst]
paddw mm0, mm1
movq [dst], mm0

processed on the GPU?


Hint: It's not!
Even with the 0.5-tick delay that CPU-GPU communication might generate, you still don't get close to the performance capability of MMX/SSE.
Do you think those GPU-based 10-minute WEP password crackers would be possible with a delay as huge as you're assuming?
We had some hefty discussions about this very topic when I was working on the Voxlap port with Sonar and Cajun.