[tom]

One thing that may be hurting you is having only a single stick of ram. Having four matched 512 sticks has lower latency than having 1 2gb stick since all eight cores are fighting to use the same stick of ram. Adding another 2gb stick should help.

Greg is right here. It's not about the amount of RAM, it's about dual channel memory configuration. You computer is not faster now, it was before slower than it should.

ftp://download.intel.com/support/mother ... rev_15.pdf

Also see Page 14, Paragraph 2 >>> ftp://download.intel.com/support/mother ... 960004.pdf

The Intel® 5000 MCH supports a burst length of four in either single-channel mode or dual channel mode. In dual-channel mode this results in eight 64-bit chunks (64-byte cache line) from a single read or write. In single-channel mode, two reads or writes are required to access a cache line of data.

[deadalvs]

i guess there has to be a rule about priority settings in the task manager.

i posted once my «best time» of 24:05 min with a high priority.

which way to go ?

lowest priority?
highest priority ?

[tom]

You're right, deadavls. It should be in high priority and you can switch to high from Task Manager. Benchmarking while running tens of other services, tasks makes no sense in accuracy.

[deadalvs]

that was my feeling too.

i think the best (and most accurate) is to eliminate near all cycles for other tasks running and see the FULL speed buttered into maxwell.

it is so much dependent on how much other soft is installed by each user that this must be filtered out a little - agree too with You, Tom !

killing other tasks is risky, so just cutting them off seems the real deal.

[ivox3]

dealdvs, ...

I remember your 24m time with high priority, .. but that's what I'm investigating now, ... I think there's something suspect with those type readings maxwell's reporting ------ for instance, .. I just ran a test in high priority(with as many services killed off as possible) and the maxwell clock is indicating 12m !

This can't be right ... so, .. back to testing...

btw: Thanks Tom for grabbing that pdf ..

[sandykoufax]

chris, congratulation.

At finally, you can do it.

[beppeg]

So, .. I ask that everyone disregard that last score for now until I have some more accurate testing to leverage it with.

Great ivox3
I'm waiting for your final test for changing the first place in the db

[ivox3]

Hi Beppeg, ...yes, ... please wait. The results will have a higher degree of accuracy.

[GM5]

Chris,

Glad the extra ram helped you out. Now you have to wait for the applications to catch up with your system

-Greg

[ivox3]

Okay folks ..

First off, .. I apologize for rocking the boat the other day .. , ..I got kind of excited, ..but I was innocently duped and I hope I can show how that happened. ... I beg forgiveness.


Now, ..on to the facts.

First off, .. the extra 2GB's of ram did indeed bring about a performance boost --- roughly a 20% gain. ...so that's good, ..not earth shattering, but I'll take it.

Secondly, ..the question of " Does setting CPU priority to anything above Normal bring a performance gain ? " Basically, ...as you will see, ... the answer is essentially no. From the chart you can see that from comparing individual times that sometimes there is and sometimes there isn't -- but through compiling the average scores a slight gain is revealed , ..but it's basically negligble.



So that brings me to this -- What can be said about altering the CPU setting is that it definitely interferes with the Maxwell clock cycle -- The only setting that matched the stop watch readings was the Normal setting, ..all others yielded inaccurate readings, ..although generally consistant. The chart should illustrate this fact.

So, ...to our friend Dealdalvs, ..the 24 m 05s time you received and my 17 m time are out. Gone. ...to be discarded. The only way to get a good time is through the use of a stopwatch if the setting is altered beyond the Normal setting.

That's about it, ... sorry for the mix up folks. I hope this helps and let's us all know a little more on where we're at in terms of cores vs. performance.

BTW: ...everyone from now on should have to perform this same level of testing when posting a new bench mark ! ...

chris

BTW: @ Beppeg , ... I'll let you choose which score to use for the master page.
___________________________________________

specs: Intel s5000XVN 8 cores [ 2.33GHz]
4GB's FB Ram

MT = maxwell clock time
SW = stop watch
Orange box = best time

[deadalvs]

thanks for this nice insight, ivox3 !

so we know about another «bug» in maxwell that has to be ironed out !

* * *

it is nice to be called «friend» !

[ivox3]

... you bet.

..get your stopwatch out.

[deadalvs]

i'd have killed You if Your rendering really had taken 45 seconds ...

[Mihai]

Still it's not a bad score. Looking at the chart, the dual dualcore Xeon machine, also at 2.33ghz, gets about 44 minutes. So 26 minutes is not bad...

[Frances]

thanks for this nice insight, ivox3 !

so we know about another «bug» in maxwell that has to be ironed out !

* * *

it is nice to be called «friend» !

I mentioned inconsistancies with the maxwell clock a year ago. Nobody cared.