Tech benchmarking
Miscellaneous Forums/Blitz Showcase/Tech benchmarking
Greetings, I've been working on a new system for blitz to compliment my blitzTree system.. This new system handles grass and weeds. My goal is to be able to supply unlimited amounts of entity locations via colormap or simple x,y,z placement towards the grass structure.
Anyways, this little demo runs several loop tests inside the exe and compiles a total score at completion.. If you would download the demo and post back your bench scores I would be very happy. At this time the only score im interested in is the Test 2 score (med avg)..
thanks!
http://www.aliencodec.com/temp/sys_grass_alpha.zipBare in mind this is a very early demo, the grass isnt 'pretty' yet but polish comes last. :)
Your link doesn't work.
can't download
To start things off, my score was:
19318.3 / 87 avg fps, GF5950,amd64fx51
May need to refresh the page, i had the wrong addy in there...
http://www.aliencodec.com/temp/sys_grass_alpha.zip
looks pretty good, although it doens't quite look like grass.
It also doesn't run very well on a GF2MX (~10fps)
Perturbatio, I suspect you may have gotten the wrong zip.. My bad - i had some foulups on uploading.. The zip should be sys_grass_alpha.zip
But, for this system running on a GF2MX i'd suspect 10fps is rather fabulous as for the looks - I agree, it's not ment to pretty just yet (as mentioned above)... but by the time of release it'll be looking right :). Thanks for the test and feedback!
7112.6 and 32 fps, P4 1.8 GF3 Ti200
20026.6 - 91fps, P4, 2.4ghz, 512mb, FX5900+128mb
score: 3359.4
avg: 15 fps
System specs in me sig
Nice score zawran!
So far it looks like theres to much flux in the way its distributing performance... Theres a catch in the system that attmepts to smooth out performace based on scale.. Ie more handling done on systems that show more fps - quads being released when outside the view spectrum etc.. Theres still several opt's to be done.
30129.0 & 136fps, P4 3GHz & 512Mb & Radeon 9800Pro
Marg
Test 3: 23179 - 105 fps AMD Barton XP 2800+, 1 gig ram, Radeon 9700 Pro... (from the center of the map looking around)
high flux in performance tho.
Also noticed that the scoring/linearity of the demo isn't propper, due to the terrain ending if I walk around I can easily hit 40000 points or more. (without getting close to the edge)
Just some tech notes about the demo:
test 1: 24,000 quads @ 48,000 tris/poly ( X2 bcull enabled? )
test 2: 36,000 quads @ 72,000 tris/poly ( X2 bcull enabled? )
test 3: 48,000 quads @ 96,000 tris/poly ( X2 bcull enabled? )
Yah retro, avg scores should be based on a stationary pivot... I'll compile a new test later, as soon as i add the final system, which will yeild more accurate readings.
test3:
16508.8
75fps avg.
specs:
AMDXP2600+, 512MB, RAD9600XT
Test 2: Score 4754.2/avg 21 ...Nice demo :)
fps: 132
score: 29117.0
amd 3000+ radeon 9800pro 512mb ram
looks pretty cool
ps. test 3:
fps: 96
score: 21195.0
test 2: 19764.8 avg fps 89
test 3: 16022.6 avg fps 72
AMD 2600+, Radeon 9600XT, 1GB RAM
As a benchmark used in comparison to other machines this is (sorry) useless. Depending on where you look, the score changes. Depending on where you stand the score changes.
I got 17615.4 @ 80fps on a p4-2.0 with a pos ibm video card that I believe shares system ram.
In order to have a good benchmark you have to have perfectly repeatable conditions. YOU move the camera and look at things. The user has no interaction with the program. Otherwise you'll get results like mine above. I ran to the edge of the map and stared at blackness for most of the "benchmark". That inflated my score badly.
Nice grass though... Just no use as a benchmark.
This was mentioned earlier...
It's not intended for a blitz3d - computer benchmark score.. its a test for a grass system im making.
I want to get an idea if my speed is in the right direction before moving on...
The score provides me with information which in turn helps me determine what area i could imporve in.
will this be a free upgrade for blitztree3d owners or a separet product and at what cost if that is so
ah. I see. I'll see what I get on.... the beast.
...
Well... just to see what it'd do, I ran to the edge and my machine renders blackness so fast I think it must have a flux capacitor in it. I got negative exponential values for both numbers. :)
Lemme try it looking in a circle from just the center of the map. :)
...
38086.4 @ 173 fps
lowfps was 125 and high was 250 I believe.
Athlon64 3000+ OC'd to 2.2Ghz / 1.5Gb DDR400 / Radeon 9800XT
Several new opts' went in, including some better sorting of the quads.. results compared to my pic posted earlier! much higher!

Smurfpass, undecided as of now..
Kanati, thats some insane FPS! hmmm does radeon own geforce that much?? I may invest in a radeon soon.. or hold out for a gf 6800... hmmm
Well... actually yours seems pretty low for an FX51 and 5950... I would expect you to be faster than me. But I *am* oc'd on my cpu which brings it pretty near the FX51's performance...
Lemme try the new one... Which isn't linked... :)
I suspect it's the geforce as Sandra benchmark reports a huge bottleneck from agp to my cpu and Geforce has been known to have massive bottleneck problems.... I havent linked the new demo yet as im working out some other issues... but will soon.
Impressive.
Kanati for the test to be even i suggest you stand still in the middle of the map at your starting position until the test runs its course. thats what i did
Standing still where I started:
20770.2
94
Moving around:
13921.6
63
specs in sig
Out of all three tests, the average one had the most vertical tearing. Maybe because the average fps was nearer to my refresh rate? Also, some z-order problems - I used to get that on my stuff. Something to do with the chosen entity fx.
Cannot get it to work. After selecting any of the three options Low/Med/Extreme my PC just sits there doing what seems to be nothing for a very long time so I have to CTRL-ALT-DEL out of it. Probably my machine is too low spec:
AMD K6-500 Mhz, Geforce 2 MX. 256 Mb Ram.
Kantai, whats your card settings at?
Im using ffa x4, also I benchmarked during windowed mode in the blitz editor (debug mode)..
I've got the last of the optimizations in so I'll post a demo today that sits the camera in a hot spot and slowly rotates... this will render more accurate results across-the-board..
Thanks for the early alpha test guys! It'll be neat seeing you compare the results you got yesterday with the new version today... I look forward to seeing improved scores!
More later.
Kanati for the test to be even i suggest you stand still in the middle of the map at your starting position until the test runs its course. thats what i did
That's what I did. I expected a bit lower score based on other's, but I did stand in the middle of the map... Even spun in a circle a few times to look at the loverly grass. :)
Alienhead: I have full screen antialiasing set to "application preference", so if your program uses it, it's turned on (and I did notice it looks a little like AA was on, but didn't think much about it). The card isn't overclocked (don't like the possibility of blowing a 500 dollar+ video card.) but the CPU is (220x10@ 2.2ghz).
I'm interested in what you get with running the exe. You should get much better results while not windowed and in debug mode. :)
11785.4 and avarage of 54fps. Specs below.
Well I feel like i've finished the core routines and made all the improvements needed. The end results are very very fast while supporting insane amounts of poly and textures.
This screenshot isnt pretty but it shows more than 878,000 tris being processed and yet maintaining a very high fps rate.
pass 3 screenshotI've moved into the last part of the systems development stage.. Grass placemnet options and effects... After this i'll cmpile a package of high-quality photo-real grass and weed textures..
Heres a shot of a simple colormap being used to position grass on my map. I plan on a few other techniques to allow easy grass placement as well.
colormap placementSome filters are being worked on as well, in this shot a soft-filter is appiled to the grass system.
soft-filter soft-filter2 You may even mix filters amongest grass types, heres a soft and normal type mixed in the same scene.
soft-filter3 An ability to force masking or alpha channels is coming!
More later!
well i must say that this is impressive i can't wait until this comes out well deppends on with price it is going to have and if it work good together with blitztree3d
Score = 3476
AVG 15 fps
AMD 1.6Ghz Geforce4 MX 64MB WinXP Home
Score = 17770
AvG = 62
(Highest Detail 3 & Moving around)
Spec = AMD XP 2100, 512 DDR, Ati Radeon 9600xt w. 256 MB
On highest detail:
Score = 8958.4 and AVG = 40
On lowest detail:
Score = 24063.6 and AVG = 109
test3, stationary: 16152.4/73fps (using my new GFX Card).
Ty guys, these scores have given me an overall idea where i stand. Ive been hard at work the last 3 days furthering this engine and it's starting to look close to finished. I plan to release this new demo sometime this weekend.
I've had several emails with concerns about mixing the grass system with blitztree system. Rest assured the systems will interchange with each other perfectly.
As matter of fact I've already made sure of this and plan to demostrate this in the next demo. Heres a few pics
Blitzgrass+blitztree 1Blitzgrass+blitztree 2
Excellent. Where's the BUY NOW button ? :)
ooh really cool !!!
Like John-R said...where's the BUY button.
(oh and make a blitztree/blitzgrass bonuspack with a discount for TreemahgikPro users will ya :-)