Help with Intel XTU efficiency

PolRoger · Nov 19, 2014

I'm getting crappy XTU score(s) with my 2500K compared a few others posted for this hardware bench. :shrug:

What can I do to improve my score to get closer to the more efficient runs posted (higher score/lower speed) HWBOT ranking #3 thru #7?

http://hwbot.org/benchmark/xtu/rankings?hardwareTypeId=processor_2289&cores=4#start=0#interval=20

EDIT: An improvement but I still need ~10 to 15 marks greater efficiency...

Johan45 · Nov 19, 2014

Not much to say more speed faster mem/ tighter timings. Make sure you're using Win7 with SP1

ATMINSIDE · Nov 19, 2014

Realtime priority?

QuickFast · Nov 19, 2014

PolRoger you got a PM since this isn't in the team lounge -someone give this man a benching team badge

Woomack · Nov 20, 2014

yep, would be nice as he already made 100+ points for the team

PolRoger · Nov 20, 2014

Well I messed around with this bench some more yesterday and I'm still kind of surprised at how variable the results can be... I had results from the lower to mid 900's right up to ~1000.

Sometimes the score would go down even while running with tighter timings? I did manage to increase my score some but it is still not quite what I was hoping for. :shrug:

C8 @ ~2240MHz (5.249):

C7 @ ~2200MHz (5.263):

ATMINSIDE · Nov 20, 2014

Are you throttling at all?

PolRoger · Nov 20, 2014

I not certain but I don't think I was trottling... I guess I could re-test and check for that.

ATMINSIDE · Nov 20, 2014

Just watch the bar for any blips

You shouldn't be at an 82°C max, but you never know.

That's the only thing I've seen besides core speeds that cause fluctuations of 40+ points in XTU.
Do you have all your power-saving stuff disabled?

QuickFast · Nov 20, 2014

PolRoger- Only time I saw score differences like that was when using software to tweak the memory in windows don't know if any one else seen the same thing but when I stoped using it and tweak memory in bios everything seemed fine and only saw a margin of error of 1 mark\point + or -

Woomack · Nov 20, 2014

Sometimes scores are +/- 50 without special reason. Last time when I was testing dual xeons I had +/- 150 difference and scores between ~1500 and ~1650. It's often the same as with hwbot prime or some other tests. On Haswells I haven't seen differences like that. +/- 5 points maybe but not much more.

ATMINSIDE · Nov 20, 2014

Woomack said:
Sometimes scores are +/- 50 without special reason. Last time when I was testing dual xeons I had +/- 150 difference and scores between ~1500 and ~1650. It's often the same as with hwbot prime or some other tests. On Haswells I haven't seen differences like that. +/- 5 points maybe but not much more.

Good note, all my experience was with Haswell.

PolRoger · Nov 20, 2014

Yeah.... This bench on this setup has been kind of bugging me. :screwy:

Here are a couple more screen shots. I finally decided to keep pushing on MHZ and I finally took Bronze cup for hardware :thup:

... But I'm still ~ 20+ Marks from Silver.

C9 2206MHz (5.170):

C9 2206MHz (5.377):

Woomack · Nov 20, 2014

New hardware has always more points in XTU. I just think it's the way to show users that new generations are so much better and they should upgrade the hardware.
Some older series are not even supported. Like older Xeons won't run in XTU. I was surprised that I was able to make it run on dual IB based Xeons on Server 2012. Usually I get info that this platform is not supported.

Try to play with sub timings to get couple of points more. Generally final score is basing on CPU and memory performance. CPU is higher clock = better. Memory is balance between high clock and timings to get highest bandwidth ( I didn't really check if it look at read, write or copy ).

What memory you have there ?

PolRoger · Nov 20, 2014

Woomack said:
What memory you have there ?

Its an older G.Skill triple kit originally for i7/x58 with Elpida ic.

http://www.newegg.com/Product/Product.aspx?Item=N82E16820231388

http://www.xtremesystems.org/forums/showthread.php?255797-GSKILL-TRIDENT-F3-16000CL8T-12GBTDD-3x4GB

Woomack · Nov 21, 2014

Then you can set much tighter sub timings. I won't tell you now how much you can lower them but looking at your last screenshot I can say that a lot.
tRC to 80-110, tRRD 5-6 and all other try 1-2 lower, from tertiary try everything that is 4 lower to 3 ( check that one by one )

PolRoger · Nov 21, 2014

Woomack said:
tRC to 80-110

Did you mean tRC or something else?

I don't think I can adjust tRC in BIOS for my board? Isn't minimum setting for tRC... tRC=tRAS+tRP.... So ideally for 9-9-9-24 it should be 33 and any higher number would be a more relaxed setting?

Woomack · Nov 21, 2014

tRFC

PolRoger · Nov 22, 2014

Two more runs with a 2x4GB RipjawsX 2133C8 (BBSE) kit. First run is with default XMP and 1T. The Second run is also with some of the Secondary timings tighten up.

2206MHzC8 (5.170):

2200MHzC8 (5.160)

Help with Intel XTU efficiency

PolRoger

Member

Johan45

Benching Team Leader Super Moderator

ATMINSIDE

Sim Racing Aficionado Co-Owner

QuickFast

Member

Woomack

Benching Team Leader

PolRoger

Member

ATMINSIDE

Sim Racing Aficionado Co-Owner

PolRoger

Member

ATMINSIDE

Sim Racing Aficionado Co-Owner

QuickFast

Member

Woomack

Benching Team Leader

ATMINSIDE

Sim Racing Aficionado Co-Owner

PolRoger

Member

Woomack

Benching Team Leader

PolRoger

Member

Woomack

Benching Team Leader

PolRoger

Member

Woomack

Benching Team Leader

PolRoger

Member

Similar threads