MunkyMark 2006 benchmark

Toggle sidebar Toggle sidebar
M

Munky

Diamond Member
Feb 5, 2005
9,372
0
76
First, the theoretical numbers:

score = (core clock * (texture units + pixel shaders)/2 + mem clock * (bus width/16)) / 500

for example:
x1900xtx = (650 * (16+48)/2 + 775 * (256/16)) / 500 = 66.4
7800gtx512 = ((550 * 24) + 850 * (256/16)) / 500 = 52.4
x850xtpe = ((540 * 16) + 590 * (256/16)) / 500 = 36.16
6600gt = ((500 * 8) + 500 * (128/16)) / 500 = 16
For SLI, simply multiply single card score * 1.9 (edited from 2.0)

Then, run the FEAR video stress test at 1280x960, 4xAA, 8xAF, max settings, (*edit: disable soft shadows) and report your average fps. Compare the theoretical score with the actual result in FEAR.

My numbers:
x1900xtx = (650 * (16+48)/2 + 775 * (256/16)) / 500 = 66.4
Actual fps = 68

x800gto = (580 * 16 + 590 * (256/16)) / 500 = 37.44
Actual fps = 39
 
J

JAG87

Diamond Member
Jan 3, 2006
3,921
3
76
absolutely stunning!

now this is an excellent ranking system. that equation is going in my calculus notebook.

I guess you divide by 500 to get a reasonable number to compare with fps in fear, but may I add that if you do not divide by 500, that will give you a nice number in the thousands to be used as a mark in points.

So I would divide by 500, if you want a fear FPS comparison, but otherwise leave the number as it is, and keep it as a score in MunkyMarks!
 
M

Munky

Diamond Member
Feb 5, 2005
9,372
0
76
Originally posted by: Leper Messiah
The SLI scores aren't going to be accurate due to driver overhead.
Click to expand...

Correct. But FEAR just happens to be such a gpu stressful game that SLI should offer a bigger boost than in other games, so maybe i'll have to revise it to (score * 1.8), but it's gonna be somewhere close.
 
M

Munky

Diamond Member
Feb 5, 2005
9,372
0
76
The stress test in in the MP demo. I dont have the full game, but I believe it's there also.
 
L

Leper Messiah

Banned
Dec 13, 2004
7,973
8
0
Originally posted by: munky
Originally posted by: Leper Messiah
The SLI scores aren't going to be accurate due to driver overhead.
Click to expand...

Correct. But FEAR just happens to be such a gpu stressful game that SLI should offer a bigger boost than in other games, so maybe i'll have to revise it to (score * 1.8), but it's gonna be somewhere close.
Click to expand...

Too bad I don't have FEAR, I'd bench it my OC'd 7800GT...

Which would be:

(526*20)+605*(256/16))/500=40.4. So if anyone's got a 7800GT and FEAR bench it and see about what you get.
 
J

JAG87

Diamond Member
Jan 3, 2006
3,921
3
76
Who cares about SLI, it will obviously be faster then one card, but this system is awesome for single cards.

If you want SLI scores, I wouls say depending on applications you need to multiply the score by 150%.

But then again some games dont get any improvement at all from SLI while some get 200% performance increase.
 
J

JAG87

Diamond Member
Jan 3, 2006
3,921
3
76
munky, I am going to make a program that calculates munky marks for you.

I will send you the exe file after so you can evaluate it. If you want pm me a pic that I can put on the GUI of the program. Make a nice logo of your name in photoshop and send it to me.
 
M

Munky

Diamond Member
Feb 5, 2005
9,372
0
76
Thanks. You can just make some logo, I'm not too picky.

Also, I'll post the score of my bro's oc'd 7800gt shortly.
 
M

Munky

Diamond Member
Feb 5, 2005
9,372
0
76
Originally posted by: Leper Messiah
Originally posted by: munky
Originally posted by: Leper Messiah
The SLI scores aren't going to be accurate due to driver overhead.
Click to expand...

Correct. But FEAR just happens to be such a gpu stressful game that SLI should offer a bigger boost than in other games, so maybe i'll have to revise it to (score * 1.8), but it's gonna be somewhere close.
Click to expand...

Too bad I don't have FEAR, I'd bench it my OC'd 7800GT...

Which would be:

(526*20)+605*(256/16))/500=40.4. So if anyone's got a 7800GT and FEAR bench it and see about what you get.
Click to expand...


You can download the latest MP demo and run it.
 
V

Vinnybcfc

Senior member
Nov 9, 2005
216
0
0
Lol my 5900 didnt like FEAR very much on those settings and im having problems trying to set the game to 1280 x 960 so it was ran at 1152 x 864

But it only came out at 1 fps average

MunkyMarks:

((475 * 8) + 375 * (256/16)) / 500 = 16 = 19.6 MunkyMarks on a 5900XT core clocked at 475

Could it be a problem with the dodgy shaders on the FX series?

Edit: Calculation wrong
 
M

mwmorph

Diamond Member
Dec 27, 2004
8,877
1
81
by your calculations,
(435*(8+8)/2+375*(256/16))/500,
I get 18.96.

something is not right.

am I making a calculation mistake or is my oced 9800xt a true beast, beating a 6600gt like that?
 
S

Skott

Diamond Member
Oct 4, 2005
5,730
1
76
Originally posted by: munky
First, the theoretical numbers:

score = (core clock * (texture units + pixel shaders)/2 + mem clock * (bus width/16)) / 500

for example:
x1900xtx = (650 * (16+48)/2 + 775 * (256/16)) / 500 = 66.4
7800gtx512 = ((550 * 24) + 850 * (256/16)) / 500 = 52.4
x850xtpe = ((540 * 16) + 590 * (256/16)) / 500 = 36.16
6600gt = ((500 * 8) + 500 * (128/16)) / 500 = 16
For SLI, simply multiply single card score * 1.9 (edited from 2.0)

Then, run the FEAR video stress test at 1280x960, 4xAA, 8xAF, max settings, and report your average fps. Compare the theoretical score with the actual result in FEAR.

My numbers:
x1900xtx = (650 * (16+48)/2 + 775 * (256/16)) / 500 = 66.4
Actual fps = 68

x800gto = (580 * 16 + 590 * (256/16)) / 500 = 37.44
Actual fps = 39
Click to expand...


OC'd 7800gt:

(520 * 20) + (625 * 256/16) / 500 = 40.8
actual fps = 43

40.8 x 1.9 = 77.52
or
43 x 1.9 = 81.7




So, if this is all true two 7800GT in SLI beats a single x1900XTX? Is that correct?


Edit: Just realized its an OCed 7800GT score.
 
O

otispunkmeyer

Lifer
Jun 14, 2003
10,442
0
0
we need the muffin-mark

take some scales, take some speciality otis spunkmeyer muffins, and get weighing. the card that weighs the most muffins is the most badass. you know dats tru dawg!
 
M

Munky

Diamond Member
Feb 5, 2005
9,372
0
76
Originally posted by: Vinnybcfc
Lol my 5900 didnt like FEAR very much on those settings and im having problems trying to set the game to 1280 x 960 so it was ran at 1152 x 864

But it only came out at 1 fps average

MunkyMarks:

(475 * 8/2 + 375 * (256/16)) / 500 = 15.8 MunkyMarks on a 5900XT core clocked at 475

Could it be a problem with the dodgy shaders on the FX series?
Click to expand...

Yeah, the FX series are an exception due to poor sm2 shader performance.
 
M

Munky

Diamond Member
Feb 5, 2005
9,372
0
76
Originally posted by: mwmorph
by your calculations,
(435*(8+8)/2+375*(256/16))/500,
I get 18.96.

something is not right.

am I making a calculation mistake or is my oced 9800xt a true beast, beating a 6600gt like that?
Click to expand...

It's possible. You can download the MP demo and let me know your actual score, then we'll know for sure.
 
M

mwmorph

Diamond Member
Dec 27, 2004
8,877
1
81
Originally posted by: munky
Originally posted by: mwmorph
by your calculations,
(435*(8+8)/2+375*(256/16))/500,
I get 18.96.

something is not right.

am I making a calculation mistake or is my oced 9800xt a true beast, beating a 6600gt like that?
Click to expand...

It's possible. You can download the MP demo and let me know your actual score, then we'll know for sure.
Click to expand...

i have the game. ill run it in a few minutes and report back.
 
J

JAG87

Diamond Member
Jan 3, 2006
3,921
3
76
Originally posted by: mwmorph
by your calculations,
(435*(8+8)/2+375*(256/16))/500,
I get 18.96.

something is not right.

am I making a calculation mistake or is my oced 9800xt a true beast, beating a 6600gt like that?
Click to expand...

that sounds quite right actually...
 
You must log in or register to reply here.

TRENDING THREADS

COMPANY
RESOURCES
FOLLOW
Top Bottom