Help tuning for 3dmark 11

uksoldierboy · 6 Dec 2013 at 17:46

Hiya 8Pack,

Having a little trouble in regards to benchmarks so i'm here seeking your expert knowledge.

My issue is with my Physics scores, both on 3DMark 11 and firestrike, i seem to be getting much lower Physics scores than i should be.

SPECS
-------

4770K @4.6|1.3vcore|Cache ratio 44|1.3v
Asus Maximus Hero z87
8gb Corsair Vengeance Pro series 1866mhz@2133mhz|9-11-11-30 1T tRFC 320.
Asus 290X
Corsair 120gb LS SSD.

Temps are all fine and well within thresholds.

I'll start with firestrike first as i think i've cured that but would like an opinion on why/how it happened.

basically if i run the benchmark at stock on my gpu(happened on my last gpu too which was a 6970) my physics score is always much lower than what it should be for my overclock, i normally score around 11500 @4.6 with the gpu at stock.
http://www.3dmark.com/3dm/1504194?

but after a lot of digging i've found what seems to be causing it, if i go into afterburner's settings and select "Unofficial overclocking-without powerplay support" so my GPU never downclocks, my Physics scores jump up to about 13100, in-line with what they should be i believe.
http://www.3dmark.com/3dm/1670572?
Cpu @4.7 in that run, 4.6 gets me 13100.

I have no idea why this setting is effecting my CPU score but would very much like to know.

In regards to 3DMark 11, my physics score is again around 11200, but even after trying the fix for 3dmark firestrike, it doesn't change, other's seem to be getting around 12800-13000.

I know your a busy bloke so any help or advice you can give would be greatly appreciated.

Mtom · 26 Mar 2014 at 22:36

Old thread, but im bringing it up, it would be great if AMD fix this, i just tried, and really with powerplay off, i get 12500-13000 physics score, if its on (normal mode) i get 11200-11600.

LtMatt · 27 Mar 2014 at 07:38

Yeah theres something up with it. I've directed Warsam to the thread. Hopefully they can start testing this. Its a powerplay issue for sure.

8 Pack · 27 Mar 2014 at 10:19

Physics is a CPU only test. I have not noticed Power Play effecting Physics at all. Combined yes for sure and driver version yes for sure in the combined test but not physics.

With 11 you can improve that a lot by using version 1.01 of the benchmark instead of the latest.

Mtom · 27 Mar 2014 at 17:18

8 Pack said:
Physics is a CPU only test. I have not noticed Power Play effecting Physics at all. Combined yes for sure and driver version yes for sure in the combined test but not physics.

With 11 you can improve that a lot by using version 1.01 of the benchmark instead of the latest.

You are right but still it is a fact, it works as uksoldierboy wrote above. if you switch AB to "unofficial overclocking without powerplay", then 3DMark gives the physics score it should. I say it should, because i immidiately noticed this in january, when changed my 7870 to 290X, and lost 1500 physics score point depending on OC.

Heres a comparison: http://www.3dmark.com/compare/fs/1574435/fs/1586661
january 20 - 7870 - [email protected] -12622 physics point
january 22 - 290X - [email protected] -11248 physics point

If i switch off powerplay, im getting the higher score again.

Mtom · 27 Mar 2014 at 17:25

I just made a qucik comparison:

with powerplay - 11151:

without powerplay - 12529:

LtMatt · 27 Mar 2014 at 18:20

Spot on Mtom. I can replicate it exactly as you and Soliderboy describe. I can also confirm it does not affect crossfire. This is a single card issue only. When i run crossfire i get just over 40fps on Physics on firestrike. When i run single card that drops to the low 30's. Its such a weird bug. Its been present since i got my 290's in every driver.

LtMatt · 27 Mar 2014 at 20:44

Got an update for you.

wazza300 · 27 Mar 2014 at 20:47

isn't it something to do with lower cpu power state? think you have to disable c7

that or use high performance profile in windows

(on haswell atleast)

8 Pack · 28 Mar 2014 at 10:04

Oh OK I never use power play anyway. Pointless!!!!

Kaapstad · 12 Apr 2014 at 09:42

8 Pack said:
Oh OK I never use power play anyway. Pointless!!!!

Have you tried 337.5 drivers for 4 cards on the P setting in 3dmark11, my Titans gained about 2k on the graphics score using them.

For all other benches 337.5 does not seem to do much.

8 Pack · 12 Apr 2014 at 09:52

Not tested. Been doing single card stuff.

You get frame cap around 220 on GT2 in 11???

Kaapstad · 12 Apr 2014 at 10:25

8 Pack said:
Not tested. Been doing single card stuff.

You get frame cap around 220 on GT2 in 11???

Here is the link to my run on 337.5

http://www.3dmark.com/3dm11/8213662

GT2 was 284 fps

Mtom · 12 Apr 2014 at 11:10

I can confirm that on 14.4 beta the physics score is fixed.
Now its the only thing needs a fix, is the fan control.

8 Pack · 12 Apr 2014 at 14:57

Kaap did you see GT2 being capped on previous drivers?? What version of teh bench you using??

To fix fan problem. Rip off stock cooler add water cooling!!! Fixed!!!

Kaapstad · 12 Apr 2014 at 21:27

8 Pack said:
Kaap did you see GT2 being capped on previous drivers?? What version of teh bench you using??

To fix fan problem. Rip off stock cooler add water cooling!!! Fixed!!!

Here is a comparison with the latest version of 3dmark11 on the left and an old version on the right.

http://www.3dmark.com/compare/3dm11/8213714/3dm11/6516720

By the looks of it GT2 is much the same but GT1 is noticeably higher.

I was running the CPU @5.0 in the latest run compared to 5.1 in the old run yet the scores were better with the new run.