3dmark11 Rank thread

Gooberman

Gooberman

Active Member
GPUZ says stock is 835MHz but it's actually 790 so it's 1090 MHz on the core. I'm so close to 5k :cool: I know it's not going on the board, but i'm happy with my laptop lol.
KOrN1vi.png

Larger Image
 
G80FTW

G80FTW

Active Member
Heku said:
Think i scored best! :p
Click to expand...

Why would a 550ti have worse pixel and texture fill rates than an 8800? If I remember correctly, I could get 28GT/s texture fillrate overclocking my GTS and the GT that came later was over 30GT/s. I know the 550 is midrange, but I would think even a midrange 4 generations later would wipe the floor with a 4 generation old flagship GPU. Was there really not much performance increases from the 200 series through the 500 series?
 
Last edited:
8

87dtna

Active Member
I'm guessing there's just power saving features on the 550 Ti that have the Vram clockspeed really low. The clock shows 600mhz when it should be at 1800 and have 98.5gb/s bandwidth.
 
8

87dtna

Active Member
Not really. Most of the score is based on the graphics portion. The physics and CPU score is a smaller percentage of the total score.

If he had an I7 4770k at the same clockspeed, he *might* break 14k overall score.
 
Okedokey

Okedokey

Well-Known Member
87dtna said:
Not really. Most of the score is based on the graphics portion. The physics and CPU score is a smaller percentage of the total score.

If he had an I7 4770k at the same clockspeed, he *might* break 14k overall score.
Click to expand...

3DMark uses a weighted harmonic mean of the graphics, physics and combined scores, each have a coefficient for FPS and scaling constant (look up standardisation or z test). The scaling makes the scores fit within a 'normal probability curve' or 'bell curve' - central limit theorem.

What this actually means is the end score is a function of the mathematics being:

Graphics score is worked out by multiplying a scaling constant by:

4/ (reciprocal of the fps for each graphics test) i.e.

4/ ((1/FPS1) + (1/FPS2) + (1/FPS3) + (1/FP4)).

This is known as a harmonic mean test.

Physics testing is simpler; it allocates threads and multiplies this scaling constant by the FPS.

Combined tests do both of the above, and again, scale to fit to a normal distribution curve.

Your final score is a harmonic mean of all of the above but with an additional 'weighting' based on the pre-set chosen when you run the benchmark.

This means essentially that you

3Dmark Score = ((Weighting for graphics)/ (Weighting for graphics/Score for graphics)) + ((Weighting for physics)/ (Weighting for physics/score for physics)) + ((Weighting for combined/ (Weighting for combined/score for combined))

In simple terms, it’s the sum of the averages of the standardised and weighted parts of the test. Makes sense?

In terms of the actual amount physics plays a role (the scalar), it depends on the pre-set chosen (weighting) and ranges from 20%, 15%, 10% for 'entry', ''performance' and 'extreme' respectively. Since the constant is a constant, and we're all using the same preset (Performance @ 720p), it can be said that physics accounts for 15% of your score.
 
Last edited:
You must log in or register to reply here.
Top