Real-Time Ray-Tracing: Speed Test

SLH

SLH

Newbie
Joined
Sep 5, 2003
Messages
1,563
Reaction score
0
Firstly, sorry if this isn't an appropriate forum.

I'm developing a realtime ray tracer for larks and would appreciate some feedback on how it ran.

If your run the exe, having extracted it and the earth.bmp file a window'll open showing an image, with time per frame in a console window.

If you could post what the rough average is, plus your system specs (CPU clock speed and GPU important).

Thanks!

NOTE: You need DX9 to run it (not a DX9 card though), but i guess you all have that!

P.S. If you want to replace the earth texture for a nicer resolution one be my guest (silly upload size limit!). It needs to be twice as wide as it is tall for it to texture map correctly.
 
I get some 15s, some 47s but 90% of them are either 31 or 32.

System:

AthlonXP 2800+ (2.1 GHz)
X800 XT Platinum (overdrive enabled)
asus a7n8x-deluxe mobo (nForce2)
1024mb dual-bandwidth memory
 
63 and 78

AthlonXP 2500+
Geforce 5900Ultra
asus a7n8x-deluxe mobo (nForce2)
1.6gigabyte dual-bandwidth memory

I am running a lot of things in the background though.
 
I'm afraid to run it... :(

My step moms shitty laptop might explode and send shrapnel into my groin....
 
Mine does not do anything
it just opens up a window with a sphear that has the earth bmp as a texture. But it does not move or anything.
 
AgentXen said:
Mine does not do anything
it just opens up a window with a sphear that has the earth bmp as a texture. But it does not move or anything.
Click to expand...
It's not supposed to, just tell him what numbers you see in the window behind that one...
 
AgentXen said:
Mine does not do anything
it just opens up a window with a sphear that has the earth bmp as a texture. But it does not move or anything.
Click to expand...


A Cmdprompt window should open up with a bunch of numbers appearing, just look at the numbers for a little while and post in here what numbers come up the most.
 
47ish on my crap computer at home:
amd xp 2800+
Geforce4 ti4200
gig pc 2700
some asus nforce2 mobo
 
31 mostly, some 70s

Intel P4 2.5Ghz
512 DDRAM
ATI Radeon 9700pro 128mb
 
i got 60s and 70s, But what is the point of the window with the picture?
 
wtf i get 30 average and some 15s :O

P4 3GHz prescott
512mb ddr pc3200
Radeon X800XT

wtf is wrong :(
 
ZeeM said:
wtf i get 30 average and some 15s :O

P4 3GHz prescott
512mb ddr pc3200
Radeon X800XT

wtf is wrong :(
Click to expand...

maybe you've got too much memory being eaten by random apps in background ?
 
Are the lower numbers better or the higher ones, it seems that the lower the number, the better it is?
 
I get 15/16 some 31s

3.2Ghz Intel P$ with Hyperthreading
Radeon X300
1GB RAM
 
the number is the time taken to render the frame, not the number of frames per second (hence a lower number is better).

I think.
 
Thanks a lot everyone, the results have been useful. Before i wasn't sure if my program was CPU limited or GPU limited, and it seems that for now at least its GPU limited. I couldn't really test this since i've got a 1.5ghz Athlon with a GeForce2 Ti200. :D

Razor said:
Are the lower numbers better or the higher ones, it seems that the lower the number, the better it is?
Click to expand...
I'm measuring how long it takes to do one 'game loop' in milliseconds

Basically do 1000 / Number and you get frames per second. So yeah, the lower the better.

EDIT: hehe, you just beat me!
 
I'd say the results show that the program uses CPU horsepower and isn't affected by the GPU.
 
lePobz said:
I'd say the results show that the program uses CPU horsepower and isn't affected by the GPU.
Click to expand...
The other way around. Look at the numbers I got. It must be because of the mobility radeon. I'm using a laptop after all.
 
Bunch of 31's and 16's. Alternating.

2.6 GHZ Pentium 4
9800 Pro
512mb RAM
 
lePobz said:
I'd say the results show that the program uses CPU horsepower and isn't affected by the GPU.
Click to expand...
Not too sure about that. Although it uses the CPU to do the collision detection (between rays and the sphere), as well as a few other things, the actual rendering of the image is just the graphics card.

I've been doing a few more tests, and it looks like its the GPU that's slowing things down, especially in my case, but it's pretty hard to tell due to the fact that the GPU can do stuff while your preparing the next frame etc.

BTW, if anyone cares i've made a slightly different version to this post, where it renders lines instead of triangles, so you can sort of see what's going on.

Linky: http://www.vbforums.com/attachment.php?attachmentid=36738
 
Yes! I'm the slowest so far! :D
100-120. w00t.
 
lePobz said:
I'd say the results show that the program uses CPU horsepower and isn't affected by the GPU.
Click to expand...
Certiainly seems that way. I have the exact same machine as you (nforce 2 board, 2800+ running at 2.17, 1 gig dual chan) except for video card (9800pro) but my figures are identical. Some 46/47, a few 15/16, but mostly 31/32
 
ship said:
Yes! I'm the slowest so far! :D
100-120. w00t.
Click to expand...

:eek: 265 on some really terrible pc (P4 1.9GHz, GeForce MX 200 32mb, 512mb ram). I'll run it on mine later.
 
Wow, that's definatly the GPU that's causing it to be that slow.
 
f|uke said:
Certiainly seems that way. I have the exact same machine as you (nforce 2 board, 2800+ running at 2.17, 1 gig dual chan) except for video card (9800pro) but my figures are identical. Some 46/47, a few 15/16, but mostly 31/32
Click to expand...
That's what I mean, the same pattern is repeated with other peoples' results too.

There's a bit of a big difference in GPU horsepower between a 9800 and an X800... If you go through the results, similar specced CPUs give similar results, regardless of the GPU.

If I remember correctly, Ray-tracing doesnt touch the GPU unless its told specifically to do so via Direct3Ds code paths.
 
As i wrote the code i know exactly when it uses the GPU and when it uses the CPU. :)

It looks like its very system dependent, since the results can be read both ways (you can pick pairs of computers that show the CPU having a big impact, and others where its the GPU.

I wouldn't be suprised if user/driver settings are having a major effect, since my program just uses the defaults.
 
I'm just doing it to see how far i get really, although it'd be cool if i could get it good enough to make a game out of.

It's only in the very early stages at the moment, but i've built it with bigger things in mind. That's one of the nice things about ray tracing, you can start simple and make things as complex (and better) as you like!
 
That's the general idea, but i'm trying to do real-time ray tracing as a challenge.
 
I get like 204, as my lowest and then average of like 220. : p

P4 2.2 ghz
1 gig ram
6600 gt

So its deffinatly the CPU... good thing i have a new mobo, that can support up to a 3.2 p4.. i guess i will be needing to upgrade soon.

-NSF
 
How is it real time if it's measured in time per frame :LOL:


EDIT: I'm getting 70-100...
 
i got this:


ERROR

AppName: ray tracing engine.exe AppVer: 0.0.0.0 ModName: ray tracing engine.exe
ModVer: 0.0.0.0 Offset: 00003ba4
Click to expand...
 
You must log in or register to reply here.

Similar threads

-smash-
Using Unity at Valve - Vision Summit 2016
Replies
5
Views
6K
Ennui
Ennui
Higlac
Benchmarking, otherwise known as "How big is your computer's dangly parts."
Replies
22
Views
4K
Bakurei
Bakurei
VirusType2
x1900xt -> 4850
Replies
19
Views
1K
xcellerate
xcellerate
dream431ca
Ray tracing to revolutionise gaming graphics
Replies
20
Views
1K
Gargantou
Gargantou
Asus
Installing Folding@Home
Replies
2
Views
1K
Asus
Asus
Back
Top