GZDoom benchmarks - next round

[Gez]

The shaders menu is only enabled if your card is up to SM3. If you have SM4 or higher, it's disabled because there's no reason to turn them off. If you have SM2 or lower, it's disabled because there's no reason to turn them on.

[Graf Zahl]

I'd really like to see some tests on an ATI X1xxx, ATI HD 4xxx, ATI HD5xxx and Geforce 7xxx card. Especially without the modern ATIs it doesn't help me that much.

[Nuxius]

I'll try to get you some results from a 7900GS when I'm over at my parents later today. Unfortunately, I don't have any ATI cards to test with.

Not going to bother with mine, since it's a GeForce 8 series (8800GTS) and you obviously already know how it will act.

I wonder what these tests would do to my ancient GeForce 3?

[Graf Zahl]

I'll try to get you some results from a 7900GS when I'm over at my parents later today.

Thanks. That one would help a lot.

Not going to bother with mine, since it's a GeForce 8 series (8800GTS) and you obviously already know how it will act.

Correct. I've seen enough data of those. But there's actually one WAD you could check. That's Super Sonic Doom. I got major discrepancies with these tests that I can't really explain. They are considerably slower on my system in relation to the 3 NVidia results I got from Gez and Enjay so seeing another one might shed some light.

I wonder what these tests would do to my ancient GeForce 3?

Me, too. Keep in mind that these levels were all picked for their demanding content so most will probably grind it to a halt. Back in the day when I had such a card the lowest was the P:AR scene which was around 30-35 fps. Most of the others didn't even exist back then.

[Edward850]

Since it took me awhile to do the benchmark test last time, I might as well ask first if there is a point in me doing this one? (Nvidia 9400 GT).

[Graf Zahl]

No. For NVidia I'd like to see a Geforce 7xxx and a 9800. I don't need any others.

[Nash]

I plan to participate this to provide data for the ATI HD 4870 X2... but there seems to be a lot of WADs that I need to download and I'm kind of short on time currently... argh...

I really, really want to help though...

[Rachael]

Here are the results for mine. I hope you don't mind that I didn't actually write all the numbers down and only took screenshots instead.

The most important info:
CPU 1.8 GHz AMD Sempron Dual-Core (pretty much a Celeron equivalent)
GPU ATI Radaeon X2400 HD w/ OEM Catalyst drivers version 9.4 on Windows XP

[Graf Zahl]

Wtf???

That thing is really slow. Now the question is, is that card really not better or is it the driver? I really, really need more ATI data, especially from newer cards.

I won't do a final 1.3 release until I know how the engine performs on modern ATI hardware so it's up to you ATI users to make it happen.

[Graf Zahl]

I just renamed tutnt.pk3 into tutnt-v105.pk3 and it worked and it liked it!

Yup, works for me too.

My machine:




However, my daughter's laptop (ATI) is significantly slower:




Full screenshots here:

http://files.drdteam.org/index.php/file ... hs/njt.zip

I was finally able to cross-check this on my old GF 6800 and if shaders are off the same problem appears. After that I ran with '-sm2' (disabling all shader features) on my current system and the same problem appears. For some reason the software texture warping is not working right anymore (or the map uses too many too large warped textures - I have to check.)

[Graf Zahl]

Wtf???

That thing is really slow. Now the question is, is that card really not better or is it the driver? I really, really need more ATI data, especially from newer cards.

I won't do a final 1.3 release until I know how the engine performs on modern ATI hardware so it's up to you ATI users to make it happen.

UPDATE:

I did some research about common benchmark results and so far it looks like this is a good representation. The Geforces tested so far sort themselves in the order that is to be expected and the 2 ATI results I got also seem to fit in to that order. Now if only I got some results of other ATI users...

[playerlin]

This machine is my brother's, and it's older than mine.XD

CPU : Intel Pentium D 805 @ 2.66Ghz
Memory : Kingston DDR 333 512MB*2
Display card : MSI R4350 D512H (ATi HD 4350, BIOS ver.:011.015.000.012) Core: 600Mhz ; Memory: 500Mhz. 512MB Display memory.
OS : Windows XP SP3
Display driver version: Catalyst(TM) 9.8

GZDoom v1.3.08 beta(r539M) @ 800x600 32bpp.

The startup log are fucked up on that machine...
the console give me a lot of 'something-shader failed' message.

Results :
Screenshots @ my googlepage.com:
http://playerlin.googlepages.com/gzdoom ... _shots.rar
(If this link failed, please tell me!)

EPIC.WAD (EPIC.PNG) 22-25fps

njbgpabh.pk3 (NJBGPABH.PNG) 16-17fps

DarkHour.wad (DARKHOUR.PNG) 36-39fps

HC20.WAD (HC20.PNG) 15-16fps

kdizd_11.pk3 (KDIZD11.PNG) 29-30fps

njol.wad (NJOL.PNG) 36-38fps

PAR.WAD (PAR.PNG) 32-34fps

PAR.WAD with lights.pk3 (PAR2.PNG) 22-23fps

phobiata.wad (PHOBIATA.PNG) 12-13fps

Sonic.wad -1 (SONIC1.PNG) 37-38fps

Sonic.wad -2 (SONIC2.PNG) 26-27fps

tutnt.pk3 (TUTNT.PNG) 8fps (DAMN! XD)

zpack.pk3 (ZPACK.PNG) 18-19fps


Hmm...according to other posts, I think it doesn't needed my 8600GT's data anymore... :p

[Graf Zahl]

I have to admit that from that card I would have expected a bit more. According to benchmarks it is almost twice as powerful than the HD2400 and yet the results are nearly identical.

It's by no means a great card (half the benchmark performance than a Geforce 8600) but these values are even worse than that - even in relation to other ATI hardware.

[InsanityBringer]

I just tried it on my ATI Raedon x1950 pro, but the results were.. interesting

The numbers were flailing all over the place, from 10 to 100, and even dipping down into the negatives! I'm not sure how reliable my numbers are due to that.

EDIT: The Raedon 4830 HD's numbers are consistant. I'm grabbing those right now

edit

Code: Select all

par no lights
63 fps
w: render=5.78, split = 0, setup=1.50, clip=1.62
f: render=5.56, setup=0.23
s: render=2.83, setup=0.43
all: all=22.15, render=15.39, setup=5.18, portal=0, finish=0.94

par lights
26 fps (would get higher sometimes)
w: render=10.51, split = 0, setup=0.98, clip=1.07
f: render=18.88, setup=0.23
s: render=2.71, setup=0.22
all: all=97.00, render=32.69, setup=3.85, portal=0, finish=0.26

epic
39 fps
w: render=6.66, split = 0, setup=1.90, clip=1.73
f: render=5.01, setup=0.30
s: render=2.57, setup=0.39
all: all=23.52, render=16.47, setup=6.59, portal=0.41, finish=0.20

zpack
37 fps
w: render=6.78, split = 0.67, setup=1.20, clip=1.24
f: render=9.01, setup=0.43
s: render=0.39, setup=0.12
all: all=25.58, render=18.99, setup=5.49, portal=0.86, finish=0.41

burghead
34 fps
w: render=7.73, split = 1.72, setup=1.80, clip=1.10
f: render=10.31, setup=0.55
s: render=1.65, setup=0.25
all: all=28.89, render=21.56, setup=6.63, portal=0.23, finish=0.31

sonic1
43 fps
w: render=1.42, split = 0, setup=.67, clip=.81
f: render=2, setup=0.13
s: render=3.21, setup=0.49
all: all=18.49, render=9.25, setup=9.09, portal=5.60, finish=0.35

sonic2
42 fps
w: render=4.41, split = 0, setup=1.35, clip=2.66
f: render=1.24, setup=0.17
s: render=3.90, setup=0.60
all: all=22.98, render=11.67, setup=6.37, portal=4.58, finish=0.24

hc20
33 fps
w: render=7.59, split = 0, setup=2.18, clip=2.37
f: render=10.90, setup=0.50
s: render=0.52, setup=0.18
all: all=29.25, render=20.92, setup=7.85, portal=0.00, finish=0.28

edit2:crap, forgot a few. I'll have to get them later

[Graf Zahl]

I just tried it on my ATI Raedon x1950 pro, but the results were.. interesting

The numbers were flailing all over the place, from 10 to 100, and even dipping down into the negatives! I'm not sure how reliable my numbers are due to that.

Such numbers are relatively useless.
This means that the timer on your computer does not work as expected.
Does it list a proper CPU frequency in the startup log?


REgarding the HD 4830, those numbers are a massive disappointment. According to all benchmarks I have seen, that card should be at least as fast as my GF 8600 but it barely manages half of that...

What's interesting though is that I got approximately the same performance on my current system with last year's drivers and older versions of GZDoom.

Seeing that the loss comes almost exclusively from the flat rendering I think I'll add a few checks in there to see if it can be sped up. I'd appreciate if you could run the test again (2 saves should be enough for that though) once I upload the new version.