@ earlygrey

_________________
____________________________
c64-2sids, A1000, A1200T-060@50(finally working!),A4000-CSMKIII
! My Master Miggies- Amiga 1000 & AmigaOne X1000 !
mancave-ramblings
X1000 I BELIEVE

@Lokaty
Quote:

How it is possible almost three times faster graphics card returns result only 51 fps more than weaker gpu? It's problem of Cow3D benchamrk, CPU or something else??

Performance is affected by multiple factors. There are multiple potential bottlenecks: CPU power, PCIe transfer rates, GPU power, GPU texel-rate, etc. With Cow3D PCIe transfer rates appear to be the major bottleneck. I've already commented on this a few times, so scroll back to my previous posts.

The GPU's speed can limit Cow3D's performance if you use a rather low-end GPU. As you move to faster GPUs, eventually the GPU is faster than the CPU can feed it commands. At that point, doubling the GPU's performance only means that the GPU is twiddling its thumbs more while it awaits instructions.

Hans

_________________
http://hdrlab.org.nz/ - Amiga OS 4 projects, programming articles and more. Home of the RadeonHD driver for Amiga OS 4.x project.
https://keasigmadelta.com/ - More of my work.

@all
Just discovered that Cow3D's fps counter measures *all* operations, including rendering the line of text at the top containing the fps counter. Text is currently rendered in software, and so involves RAM=>VRAM transfers (CPU-based because it's line segments which are usually too small to use DMA on). This could explain some of the results.

So the fps counter does *not* measure purely the Warp3D/Nova performance; it's skewed by the relatively-slow text rendering done afterward (and all other operations).

Hans

Last edited by Hans on 17-Jan-2017 at 09:17 AM.

_________________
http://hdrlab.org.nz/ - Amiga OS 4 projects, programming articles and more. Home of the RadeonHD driver for Amiga OS 4.x project.
https://keasigmadelta.com/ - More of my work.

AmigaOne X1000 Radeon R7 250X AmigaOS 4.1 FE Upd 1

coW3D6-os4 170 fps
coW3D6-os4-Nova 470 fps

Hey!

So cool! Glad to see Nova is on the very good way, Warp3D working quite well on moderns GPU, etc... etc... etc...!

EDIT: See you already answered the question (about important performances differences between X1000/X5000 on same GPU). Still curious, the informations to send every frames should be *very* limited for any PCI-E

With a much more detailed object, the difference may be less significant.

Very great job! You did it!

Last edited by Crisot on 17-Jan-2017 at 08:11 PM.
Last edited by Crisot on 17-Jan-2017 at 06:29 PM.

_________________
AmigaOne XE --- 7455A@1400 --- Radeon 9000 Pro --- RAM 2GB --- SSD 40GB

@Crisot
Quote:

EDIT: See you already answered the question (about important performances differences between X1000/X5000 on same GPU). Still curious, the informations to send every frames should be *very* limited for any PCI-E

The length of the command sequences sent to the GPU is longer than you think. However, as I discovered yesterday the fps calculations include the time taken to render the fps counter and info bar at the top of the window (plus whatever else the computer is doing). That increases the amount of PCIe bandwidth used per frame.

Hans

_________________
http://hdrlab.org.nz/ - Amiga OS 4 projects, programming articles and more. Home of the RadeonHD driver for Amiga OS 4.x project.
https://keasigmadelta.com/ - More of my work.

@Hans

Thank you for clarifying. Hope after implementation in AmigaOS4 second core support you can extend driver performance in W3DNova in the future. Meantime I think there is no sense buying something more powerful than R7 265 (or maybe is but Cow3D need some improvements to prove it - hope Alain do this... ).

Cheers Guys!

What is the histoty of the cow program I wonder? Nothing in relation to the cow object as made famous by Imagine? Inspired no?

XE , 9250 Showoff, 40fps

Now can we buy the usefull proggies.

Fast pc 60. How is that possible..

Last edited by PR on 19-Jan-2017 at 12:08 PM.

@Amigo1

OS 4.1.7 was never released, so...

Quote:

AmigaOS 4.1 FE (no update 1, so 4.1.7)

..."OS 4.1 FE" is really 4.1.8, and...

Quote:

AmigaOS 4.1 FEUpd1 (with update 1, so "4.1.8")

..."4.1 FEUpd1" is really 4.1.9.

Just so you know!

@Hans

Quote:

The GPU's speed can limit Cow3D's performance if you use a rather low-end GPU. As you move to faster GPUs, eventually the GPU is faster than the CPU can feed it commands. At that point, doubling the GPU's performance only means that the GPU is twiddling its thumbs more while it awaits instructions.

A fact that I think some people around here should meditate a bit over.

@Lokaty
Quote:

Thank you for clarifying. Hope after implementation in AmigaOS4 second core support you can extend driver performance in W3DNova in the future. Meantime I think there is no sense buying something more powerful than R7 265 (or maybe is but Cow3D need some improvements to prove it - hope Alain do this... ).

Not necessarily. Cow3D is but one test under specific conditions. Plus, I've discovered that rendering the fps counter & info-bar is skewing the results significantly. Have a look here for my comments and a modified test: https://keasigmadelta.com/blog/warp3d-novas-performance-boost-partially-hidden-by-lazy-cow

Hans

_________________
http://hdrlab.org.nz/ - Amiga OS 4 projects, programming articles and more. Home of the RadeonHD driver for Amiga OS 4.x project.
https://keasigmadelta.com/ - More of my work.

"Good" that we do not have games like oolite yet for AOS4, I might be forced sell my car and upgrade my Amiga HW.

_________________
- KimmoK
// For freedom, for honor, for AMIGA
//
// Thing that I should find more time for: CC64 - 64bit Community Computer?

@Hans
As the GPU goes faster the CPU at one time would not be fast enough and the GPU will wait for commands. If the CPU is slower than GPU but is say bi-core, if AmigaOS support multicore and if Nova driver use this feature then the slower CPU will be able to send more commands and so the GPU will work more and so you'll get better performance, right?
Kamelito
Ps it could be the case for the A1222 maybe...

How about OpenGL Elite 2 and MElite?
Kamelito

@Hans
First of all thank you for all the work you put into Warp3DNova and the great tutorials you have written on Kea Sigma Delta.

I'll test the new(your version) Cow3d and post the results later today when I'm in front of my X5000.

_________________

Thanks Hans for Nova, your support and enhancing my little prog cow3D

>What is the histoty of the cow program I wonder? Nothing in relation to the cow object as made famous by Imagine? Inspired no?

In fact it all began with a small GL program I made "starship" that show how to use glDrawElements to draw simply & efficiently a 3D object
This program was ported to Warp3D
But as the object was small it couldnt serve as benchmark so I found in my A4040 harddrive this cow object that was coming from a lightwawe viewer i think

Then I added a simple texture to make it looks like an "Aubrac cow" a very old & rustic race that come from montaneous France where my family come from
https://en.wikipedia.org/wiki/Aubrac_cattle

Note that I textured it with female cow colors as the more impressive bull dont have the same shape so will have need an other 3d model :


https://www.google.fr/search?q=taureau+aubrac+prim%C3%A9&biw=1185&bih=811&source=lnms&tbm=isch&sa=X&ved=0ahUKEwju8pvXrdDRAhUGWRQKHaXHAioQ_AUIBigB#tbm=isch&q=taureau+aubrac

Alain Thellier

AmigaOne X5000,R9 280X
Hans version of Cow3d

CoW3D6-NoInfoBar-os4
391fps

CoW3D6-NoInfoBar-os4-Nova
1399fps

_________________

@Hans
Its totaly out of cow3d-topic, but still w3d related: maybe you know, is there anyone who works on the minigl.library stub, so it will use w3d nova ? I mean, just to boost all the minigl apps we have already on os4depot. Maybe aeon, or daniel ?:)

_________________
Join us to improve dopus5!
zerohero's mirror of os4/os3 crosscompiler suites

@Hans
>"because the CPU writes the data to VRAM rather than getting the GPU to fetch it using DMA (would love to fix this, but it'll take time)"

Any idea of what gain will we get from doing it through DMA?
I guess that you've no timeframe for that to happen, what are the difficulties to use DMA Instead of CPU. In "classic" chipset using DMA was easy :).
Kamelito