tFAW=30

if ur playing below 4k uncore no problem i guess .
max i think the board could handle with no FB is around 1.5v VTT ?
above that is considered too hot already for me .
now running 1.41v for 4k uncore .
still no time to test . gotta play around with CUDA cause its my FYP >.<"

ic... burnt SLI micro at 1.55 vtt last round

now the limitation to go 2k dram is my ram chip, if want get fully stable need to run cl10 i guess... voltage requirement for 2k cl9 is too high for this chip, around 1.8 +/-. Too high for x58.

CL10 2k dram 4k uncore seems a bit slow... now still priming on 1866@cl9. cant pass the 3-4 hours mark.. keep getting freezing problem. Will try adjust TFAW as owikh suggest after work

Just an update of the RTL combi stated by cstkl1 for try out

DRAM 2000mhz 7-8-7-20 T1 @ 1.61Vdimm and 1.425VTT

54-56-58 = 1 worker error after 45 minutes. Ran this RTL twice, and failure timing were consistent

54-57-58 = 1 worker error in 1 minute

54-55-58 = Only 2Gb memory detected

53-56-58 = 1 worker error in 1 minute

53-55-58 = 1 worker error in 1 minute

53-57-58 = 1 worker error in 1 minute

54-56-59 = 1 worker error in 1 minute

55-57-58 = 1 worker error in 1 minute


so it seems like RTL 54-56-58 is the one that could bring me the furthest. Gonna try 55-56-58 later.


Attached thumbnail(s)

its a insufficient vdimm.

no use that same rtl 54,56,58 first.. ( later tweak it to 54,56,57 (bios 74).. ( then try again to 53,55,56)
but change ure vdimm to 1.62v set

u guys seriously need to understand fft's of prime.

do this
prime 95
set custom
set small fft to 128 and large fft to 4096
test atleast 5400mb


eh now for 2150 i am running 60,62,63 C8
and 2000 C7 at 53,55,56

also all hypers that i have tested always are in even number vdimm's.. for 2k's
only those that i have seen on classy's that are not is on the e759.. the first batch classy with nf200 triple sli

btw to me ure first 45 prime.. is a 15 mins fail prime..
try to understand fft's 8,10 is not what u want now
blend custom's are to get overall switches in phases, voltages, stressing different components

u want uncore and ram with cpu
so do that custom i posted above
u just need around 4 hours of that

This post has been edited by cstkl1: Sep 3 2010, 01:00 AM

just playing with my 950..

stable on 4.0ghz 200x20 on 1.235vcore(set in bios)..






Attached thumbnail(s)

wow, this 950 is not bad.
but pls show us the other CPU-Z tabs ie. mobo & dram tabs.
or use CPU Tweaker.

good proc, time to test the IMC with uncore stress...

Ok bro.. Later i will post some other pic..

~dah lame x oc, byk sudah lupe.. Lol..

I got a problem. I tried to fix my sound card in the 3rd PCIe slot and my GC in the 1st PCIe slot but then my display went wrong. I can't set it to High resolution anymore. (I think the bandwidth become 8x)

I remove my sound card from the 3rd PCIe slot - Display became normal.

This mean I can't fix 2 cards in different PCIe slot if I'm not doing CF.

Is there any way, I can use both PCIe for different purpose?

By the way, my mobo is DFI DK, Thanks~

What about another PCIe slot? Try swapping between all those 3 slots.
Reinstall gc driver after installing sound card?
The board should be able to run gc+sc at any slot w/o issues.

ok just to make sure the correct slots are numbered as follows

PCIE 1 (16x)
PCIE (4x)
PCIE 3 (8x/16x)
PCI
PCIE 2 (8x/16x)
PCI


plug in either the 4x or PCIE 3.
and not pcie 2.

This post has been edited by cstkl1: Sep 8 2010, 12:11 PM

1. You can check the bandwidth using softwares like GPU-Z
2. This looks to be a motherboard specific problems. PCI-E 1x, 4x, 9x, 11x or whatever "x" will work on PCI-E slots as they are using the same interface
3. I have my 5870 on the 3rd PCI- E slot and sound card on 2nd PCI-E slot and it works fine, no problems. Even my R2E didn't gave me problems moving the PCI-E sound cards to different slots

i7 950 + Patriot Value Ram DDR3 1333

3.8Ghz @ DDR3 2000 @ 9-9-9-27


Not so efficient IMC, but maybe its also depends on ram chipset ???

the rams were using micron D9KPV, which based on old tech ddr3. Not as good as new ddr3 chips.

But not bad for a value ram to deliver this speed

@Owikh, I tried every slot with every possible combination but hardly found a working one except for the one Cstkl suggested. Iinm, this is a default PCIe bandwidth setting for DFI mobos.

Anyway, the pickle is, I'm planning on fixing a T-Rad2 for my HD5870, it's a tri slot cooler and this board just made it hard for me to fix the GC and the SC in the order I desired.

My initial plan is to fix in the GC+T-Rad2 in 1st slot and my SC will go into the 2nd slot (slot numbers are followed as referred by bro Cstkl).

But the only configuration left for me is fixing my SC in PCIe (4x) and my GC in the 3rd slot. Since I'm using CM690, fixing my GC in the 2nd slot is quite impossible since the space is not enough for my T-Rad2.

Is there any chance for me to Mod the Bandwidth settings?
Anybody, have any solution for this.. Thanks~

I think this is a common bandwidth setting for DFI mobos, iinm, DFI UT have the same configuration of bandwidth between it's PCIe slot.

i7 950 + Patriot Value Series DDR3 1333@ DDR3 1874 Prime95 Custom Test 5 hours


i7 950 + Patriot Value Series DDR3 1333@ DDR3 1900 Prime95 Custom Test 5 hours

Spend some time with R3e Last two days

hmm ok this is the big difference on bandwidth r3e vs classy

basic..
classy main bandwidth on reduction of Cl on equal ram speed is the RTL's
rather than the subtimings of Read to Write
r3e gains bandwidth from lower Read to write than classy

uncore 4k dram 2000

R3e - RTL for any CL on R3e is best at 58,60,61
Classy - CL7 is at 53,56,58,CL8 is 58,60,61 CL9 is 61,63,64 etc


on CL8's the R3e is faster.. but when it comes to CL7 classy bandwidth is higher with lower latency


uncore 4k dram 1600
this is where the classy starts kicking r3e butt on latency
R3e - RTL 64,60,61.. on any CL's is optimum
EVGA - CL7= 2k C9 rtl's, CL6 = 2k C8 rtl's , CL5 = 2k C7's RTL's,

Vdimm Scaling
I am assuming this is the reason for r3e.. its because of the Read to Writes

r3e = 2k C8 = 1.52v , 2k C7 1.6v
classy = 2k C8 1.5v , 2k C7 1.6v
r3e = 1600 C6 1.5v.. and overclocking on this fixed preset is the same classy on vdimm
classy = 1600 C6 = 1.5v
ok this where i cheesed of with r3e
normally in any other mobo ure 2k C8 = 1866 C7 = 1600 C6 same vdimm.

Conclusion
And here is actually the cool part about r3e
Pro's of R3e
Its easier to clock 2000 compared to classy as the rtl's are more or less the same all out. So just the vdimm and ure qpi voltage for ure 2k rams.
Cons of R3e
Above CL8.. on 2k .. not going to make much of a difference.

and owikh is king of bantai voltage

This post has been edited by cstkl1: Sep 17 2010, 05:20 PM

cough cough..
i love to feed my sticks (more than) sufficiently hehe..
easier and quicker for me lahh.
Now studying on Channel Interleave & Rank Interleave Settings

How come my clocked i7 920@ 3.6Ghz have lower score compare to i7 930@ 2.9Ghz (Turbo boost) in Vantage score? Both same running Win 7 x64 bit..

This post has been edited by gengstapo: Sep 18 2010, 01:05 PM

vantage first benefit of scoring from overclocking comes in i7 920

1. GPU
2. Mem bandwidth ( take note i didnt say mem speed.. but bandwidth)
3. CPU uncore
4. Cpu speed.

i still havent found a way to measure the benefit of high qpi