High DX11 CPU overhead, very low performance.

Any chance to try with the 1018.1 drivers?

EDIT:Meanwhile.

 

I was going to but it gave me BSODs for some reason. It said said thread stuck in driver. Tbh I've never installed a modified driver before, but I did follow the instructions closely and no idea what might have gone wrong.

 

That's the difference between an optimised DX 11 driver and one who is not optimised.

Near 100% difference.

 

Nice find. Bad to see a GTX 970 can do twice the draw calls than r9 290x.

So my work pc is running a r9 270x with i5-3330 and I can install windows 10 to test. Anything I can do to test for you guys? I will only be able to test on Monday or Tuesday

I will do a 3dmark test from my side now and post results


My 3dmark api overhead test results

dx11 sungle thread = 1 099 658
dx11 mt = 1 053 303
mantle = 18 773 929

Even on my 4790k @ 4.4GHz it's not enough to beat the 750ti draw calls/Second.

 

What's amazing is that my 7950 in DX11 is faster than your 290X.. Wonder how that happened....




Hello MonarchX

 

****

 

Got no issues with GTAV ... maybe saying "game that a lot of AMD users have problems with" is a bit bold. There are some but not 100% entirely sure if it's AMDs fault or what (some 290x users).

 

What happened is that you are using the 1018.1 drivers, while he is not. You know, the ones you ****trolled the other thread about them not working.

 

It should've installed nicely had you followed all the instructions clearly.

disable driver signature enforcement
uninstall
safemode, ddu
boot into windows
device mgr, update display adapter

And it should work.. you'll go through the installation process.. the screen may go black as well and stuff... but it should be ready soon.. then you can restart and it should be done.

 

You're saying drivers made my 7950 draw nearly twice as fast as a 290X? Those would have to be some incredibly tweaked drivers. And btw I'm running W10.

I've seen the before and after shots from Death already. The drivers do increase draws, just not to point of insanity.

Death had a modest increase, around 1500 iirc.


And yes I'm sure it was all just a big misunderstanding when MonarchX was finally banned for trolling.
If u don't agree then I suggest you talk to Hilbert since it was Hilbert who banned him. Not a minute too soon either.


Edit: So we're all happy I wlll install the set MonarchX used and post the results after testing.

 

where does one get this driver to test it out?

 

Here.

 

thanks.

 

SIMD is Single Instruction Multiple Data and is not related in any way to command lists or AMD's CPU differences from Intel's.

Not sure what you're linking there?

 

my 7950 gets about 1,000,000 draw counts in DirectX 11, old drivers or new

new windows 10 WDDM 2.0 drivers were slightly faster, very slightly

I have a hard time believing a 290x gets about half as many draw counts,

sounds like something is ****y

 

You failed to realize that this overhead test is more affected by the CPU performance than a higher GPU, unless you CPU is so powerful that you are getting GPU Bound which can happen

Ofc changing brands have a diferent impact.
Altough i'm waiting for your results Pill with the 15.4B and for resuts of the monarch with the 1018.1.

I already posted mine

 

MonarchX i5, 290X 14.xxx dribvers

643 143







My 8320, 7950 14x.xx

664 779


\







@Yasamoka, you're right about the terminoligy but wrong about the architecture. Some GPU's have SIMD processors on the ASIC which get instructions from another Control Processor. A list of commands can be stored in the Controller which reduces repetative single instruction API calls to the CPU.....


That's about all I know.

 

How this does not confirm that his numbers are real? Your 8320 is at 4.7 GHz and it has eight threads.

 

-.-
Can I send you a dictionary and a pair of glasses?

What I need is...
Your results with the 15.4b and with the 1018.1(done).
And for monach I need the results on 15.4b (done) and with the 1018.1.

 

Guys stop the bickering, check out my results