Author Topic: Nvidia vs AMD brown townysis in Blockland  (Read 3890 times)

I got my hands on both an MSI R9 390X and a EVGA SC GTX 970. the R9 390X is typically regarded as performing about ~10% better on average than the GTX 970 at 1080p, but I was wondering how well it would perform in blockland, an "indie" game (or whatever the hell you want to call it)

My test consisted of starting up a singleplayer "freebuild" server and loading up the bedroom brick build. I then climbed on top of the table/drawer with the lamp, and did a timed walk/look motion. I used FRAPS to do a frametime graph and used FRAFS to decipher and save it as a picture.

I thought Nvidia would generally perform better due to me thinking that it had superior OpenGL performance - apparently I was wrong!

(note! they are not to scale! the most relevant information is in the top left corner, with the avg/1%/0.1% times/framerates. right click -> open in new tab to see full size picture)


GTX 970 Tests



AVG - 183 FPS
1% LOW - 130 FPS
0.1% LOW - 115 FPS



AVG - 182 FPS
1% LOW - 130 FPS
0.1% LOW - 114 FPS

R9 390X Tests



AVG - 265 FPS
1% LOW - 202 FPS
0.1% LOW - 91 FPS



AVG - 264 FPS
1% LOW - 201 FPS
0.1% LOW - 99 FPS

Results

AVG - R9 390X performed ~45% better than the GTX 970
1% LOW - R9 390X performed ~55% better than the GTX 970
0.1% LOW - GTX 970 performed ~20% better than the R9 390X

Verdict

This is based off two single cards and could not be completely representative of overall GPU manufacturer performance and my testing methodology could be completely off or I might have some sort of driver issues/etc, but it seems that from my testing, AMD has better average performance in Blockland compared to comparable Nvidia cards. If you want to play Blockland with the highest average framerates, get a comparable AMD card over an Nvidia model. Something interesting to note is that the R9 390X had fairly bad 0.1% LOW results, however they are still high enough that I do not consider it to be a large problem, although it could be a problem for some. In short, AMD wins in this game.


specs of build in case any of you want to know: http://pcpartpicker.com/p/zY64FT
picture 1
picture 2
« Last Edit: April 27, 2016, 08:20:20 PM by communistsquared »

also forgot to say, all settings and shaders were at max, with trilinear filtering in advanced options. VBO is disabled enabled
« Last Edit: April 22, 2016, 12:52:23 PM by communistsquared »

You need to have vbos enabled for a realistic test. So leave that option unchecked



Blockland seems to be cpu limited though. Loading a huge build (say 250000 bricks) it'll run at less than 60 fps for me with 30% gpu usage (i7 6700k and 980 ti)

alright, ill enable VBO and try again. I remember getting a performance hit from it for some reason, so I disabled it

alright, ill enable VBO and try again. I remember getting a performance hit from it for some reason, so I disabled it

The option is inverted for whatever reason, make sure you actually have them on

The option is inverted for whatever reason, make sure you actually have them on
if the option is inverted in the menu, then I had them on this whole time.

uh i have a 970 and get better fps than this

uh i have a 970 and get better fps than this
it depends on the amount of bricks. this was a ~30k build

Rendering 262k~ ( 64 x 64 x 64 ) 1x1f bricks with a 970, I get 34 fps at max shaders.

Rendering 262k~ ( 64 x 64 x 64 ) 1x1f bricks with a 970, I get 34 fps at max shaders.

could you post the map? I wonder what results I would get.

Well I used a function to generate the 1x1f's, and well I didn't save it.
I'll rewrite it in just a second.

also, what is your CPU, and resolution (and game should be run in fullscreen for best results)

also, what is your CPU, and resolution (and game should be run in fullscreen for best results)
My CPU is an i5-6600, not the K model.
And I was running at 1280x720, I might do fullscreen later.
Here's my exact build, with CPU, motherboard, RAM, and GPU.
My GPU's clock is at stock, but if I remember correctly I overclocked it a bit and it didn't affect my performance at all.

Code: [Select]
function generateBricks(%s) {
%cl = clientGroup.getObject(0);
%blid = %cl.getBLID();
for(%x = 0; %x < %s; %x++) {
for(%y = 0; %y < %s; %y++) {
for(%z = 0; %z < %s; %z++) {
%br = new fxDTSBrick(_T) {
position = %x SPC %y SPC %z*0.4+10;
dataBlock = brick1x1fData;
angleID = 0;
isPlanted = true;
client = %cl;
stackBL_ID = %BLID;
};
%br.setTrusted(1);
%br.plant();
%cl.brickgroup.add(%br);
}
}
}
}
Just copy this, and type eval(getClipboard()); into your console.
Don't run the eval code with the eval function copied.
After that's done, just type generateBricks(64); into your console, it'll take a little while for it to build all of the bricks.

getting ~70FPS with the R9 390X. at fullscreen 1920x1080p.



(bottom right corner)

What's your CPU usage while running this?