Printable View
E8600 + CM V8Quote:
Originally Posted by cstkl1
RE
Team Xtreem 2x1GB 1600
ASUS 4870 1GB
Antec Sig 850
Just the relevant specs.
So how?
And could anyone PM me the numbers that i should key in my BIOS extreme tweaker menu? for 24/7 stability and 4ghz and abv speeds.
A picture would do too....
Thanks!
try thisQuote:
Originally Posted by GregSG
should work
but before that
u need to find out whats the default vcore voltage for 3.33ghz and whats the set vcore in the bios to achieve the exact same vcore with loadline enabled
once uve done that
do this
cpu configuration.. disable everything except executable bit and multi core
multi 8
fsb 500
strap 333
ram 1600
set 1T 7-7-7-20 tRFC 110 first
disable static read and dynamic write
ai clock booster medium
performance level 8
vcore = as above default vcore for 3.33ghz stock
vdimm = ure rated vdimm for 1600mhz
vtt(fsb) - 1.12
loadline - enable
pll,nb,sb,sb 1.05 - lowest voltage aka 1.51/1.25/1.51/1.05
cpu 0 gtl +40, the rest on auto
nb dram vtt +12.5
dram channel a/b +25
spectrum disable..
run memtest86 test 5 for atleast 20 rounds without any error
keep changing ure vdimm until u get exact
once perfect
reduce the tRFC by steps until u can pass around 20 rounds
and then only run the full memtest86 for a atleast 2 hours - 4 passes
load into windows diagnotic mode
and try running interl burn test for around 50-60 runs.
have fun
btw for 4.25 ghz same setting above except
multi 9
vcore = stock vcore for 3.33ghz +0.1v...so example if 4ghz is 1.25v set then u put 1.35v...
if ure cooling is sufficient u can run inter burn test for 50-60 runs
take note this is for e8600 only.. common trait with all the e8600 i tested
'New' CellShock 1600 CS3222271 2x1GB (D9JNL ?)
I now have slightly better CL7 clocking than I did, but 900mhz is the maximum I can find for Memtest stability.
Max vDIMM before errors start is 1.90V (set in BIOS). So 900mhz/CL7 and 1000mhz/CL8 seem the limit for 86+/HCI.
CL7 up to 950mhz is mostly ok for 32M with ~ 1.96V.
Can get to windows / run Everest at CL7/1000mhz but can fail at any time.
However even if the max, 900mhz/CL7 opens up more of the board than the Transcend could - 562FSB/5:8/899mhz isn't great at CL8/tRD9, but starting to get pretty quick with CL7/tRD8.
For Read and Latency, 560+FSB/5:8/CL7/tRD8 takes over nicely from where 475/1:2/CL7/tRD6 gets to, and the board takes the strain instead of the DIMMs.
CS3222271 works ok with 0206/0403. 1001 seems good enough to keep using it with this kit.
could work as 4x1GB @ 800-900mhz/CL7, but not enough CL7 headroom for 2x1 on this board.
Using CS3222271 @ 1.90V, I have made a bit more progress with Blend, Large FFT and IBT stability at 500fsb/1:2/CL8, adjusting GTL and DDRref in line with some of cstkl1's recent recommendations :up:
All Voltage set in BIOS 0403/1001 (not measured)
Memtest HCI 2000% / 900mhz (1800) / CL7 / vDIMM 1.89V / vNB 1.40V
http://i447.photobucket.com/albums/q...CI2000_CS1.jpg
475x9.5 / 950mhz / 266strap / CL7 / tRD 6 / CT Strong / vDIMM 1.96V / vNB 1.51V
http://i447.photobucket.com/albums/q..._SP32_CS_2.jpg
562x8 / 899mhz / 333strap / CL7 / tRD 8 / CT Mod / vDIMM 1.90V / vNB 1.59V
http://i447.photobucket.com/albums/q..._SP32_CS_1.jpg
I've included a quick IBT run aswell but more work needed for a longer test. vNB may go lower.
562x7 / 899mhz / 333strap / CL7 / tRD 8 / CT Mod / vDIMM 1.90V / vNB 1.59V
http://i447.photobucket.com/albums/q...IB10_CS_1s.jpg
nice clocks cheap
heck its a coincidence..
i just got my 2x2gb stick stable for the gtl at 1800 Cl8 tRFC 60 tRD 9 at 563fsb vdimm 1.9(trying to lower this since memtest86 does 4 passes at 1.65) vnb 1.59 vtt 1.21( trying to lower this )
testing now with IBT
and man a lot of skewing this took ...
will post in a minute the full settings..
this thread is getting better and better by the day. atleast no rubbish and all good clocks..
didnt try 1001 ??
i still cant understand y memtest86 doesnt seems to be a good indicator for vdimm voltage.
cause ibt says otherwise
and loading into windows also says otherwise especially when it comes to 4gbs.
Not many here with Quads. This board needs a better bios for Quad cores.
Once I get my case up and running with WC I will be using this board with my QX9600 so will be able to give some feedback. I decided to go down this route because my S2E won't accept 8gb (I have chosen to go down the Corsair route for the memory on the X48 rather than OCZ).Quote:
Originally Posted by Think
cheap do u have a issue with 600fsb
bios shows 1800mhz
but everest and cpuz shows 2000mhz.???
4 sticks on this mobo..Quote:
Originally Posted by SimonD
is a cause for concern
cause its very chipset orientated at that point
Quote:
Originally Posted by cstkl1
i have the same if 1800 set in bios on 400fsb strap, in cpuz shows 1000 |(ddr2000) , i am on bios 1001
http://i259.photobucket.com/albums/h...8GB1600Mhz.jpg
8GB @ 1600 7-6-6-16 1T
400 strap divider bug showing in cpuz. the ratio is 2:3 so BIOS is correct.Quote:
Originally Posted by cstkl1
DerekT's screen :) does show a 2:3 ratio in Memset, but I think this is using the 266 strap as shown by DRAM:FSB ratio in Everest.
going up from 563fsb, if the DIMMS can do +900mhz/CL7, then 333strap/CL7/5:8/tRD8/CT Mod is a faster combination for chipset performance, than using 400strap with same real memory clock (600fsb/2:3).
over 600fsb, I found 333strap/5:8/CL8 is still faster tRD for tRD than 400strap/2:3/CL7.
Also guys, which tRD are you using at 600fsb on 400 strap ?
Using tRD8, I couldn't POST with 600fsb/2:3/CL7 on 400, only tRD9.
on 333strap, tRD8 still 'OK' with CL7 at +630fsb/5:8 - with good enough DIMMs (not mine).
Quote:
Originally Posted by cheapseats
Auto gives tRD 9 which is fine, ttRD 8 does post with enough Vnb but is very unstable, i hope i can get it stable with better cooling on the NB.
That is correct.Quote:
Originally Posted by cheapseats
A good bandwith 620 PL8:
http://images4.hiboox.com/images/500...81ff6536f2.jpg
and a 32M PL9:
http://images4.hiboox.com/images/500...af0193a371.jpg
to set coeff 8x I need my cascade or more :D
They move far less air, thus you hear less air movement, thus, they are quieter.Quote:
Originally Posted by aln688
sorry guys
wanted to post the 563 full setting
but i wanted to solve one issue
the 4.5ghz multi 8 issue when ure running above 1600mhz
first of all to do that high fsb
u need to manually set every skew
from cpu/nb and all the dram skews.cmd/clk/ctl.. u cant leave them on auto..
and i found out already how it works
but for 4.5ghz the multi 7.5/7 settings 4ghz 4.2ghz doesnt work because of the skews
i needed a definite proof.it was very easy to see the skewing gone wrong
when u run 3dmark06..the sm3 scores will vary eventhough memtest86 already says ure nb/dram voltages are perfect.
3dmark06 sm3 test very good test for high cpu clocks and skewing affects
but just go give in inconclusive hint the cmd/clk skews depend on ctl skews..kekeke and the cmd skew is dependent on the cpu/nb skews
so i did the cpu/nb skews .. so do the cpu/nb skews first.. nothing in auto then cmd and then clk and then ctl
Morning Cstkl1 & Cheapseats - some great posts on straps and performance level settings :up:. I decided to lower my multi to see what I can do and managed to hit 480FSB which I think is the limit for my setup although the temps were reasonable. Pretty good for a quad though :). Funny thing is my memory peformance reduced from 12k to 11.7k read and from 10k to 9.4k write.
Is it simply that I have moved beyond the sweet spot of my memory or is there any strap & memory settings that you could recommend ?. Currently I am on the 400 strap with CT on AUTO and tRD of 6 also from AUTO.
475
Attachment 90586
480
Attachment 90587
sorry rob cant help much on the quad
cause dont know the quad limitations for the ram /gtl and skews
anything i say or recommend would be assumptions
but for c2d the high mem clocks with higher bus
theres a link between the cpu clock skews... also found leave the nb skew to normal.
and mess with cmd/clk/clt skews
afaik i can tell on intial test
cpu skews reduces fsb vtt with higher bus speeds
nb skews does nothing and best left at normal
cmd skews and clk skews are nb voltage/gtl related
and clt skews are are vdimm voltage/gtl related.
hope this would help but again no definite answer but it was the key to get 563 stable
but didnt work on 4.5ghz.. only up to 4.22
Cheers for the info cstkl1 - I need to have a play to explore how the different settings effect one another. Might just bump the multi up another half and try to tighten the timings and leave it at 480. May even find some time fot a game or two then :rofl:
or another thing
if u had to raise the nb skews
be prepared to raise the pll voltages..
this is what happend whenever i tried for e8600 on 2-4dimms
last test
cheap
do u get inconsistent sm2/sm3 scores with 563fsb.. difference can be 1-2k for both sm2/sm3 scores at different runs
even though prime/memtest/ibt/memtest hci stable...
any idea whats causing this??
I would love to run 8gb ram on my Rampage Extreme, but $1k for ram is a little crazy. Currently operating with 4gb of the Kingston Hyper X DDR3. Will probably drop in another 4gb of the same HyperX very soon. Running a quad here (qx9770) with default settings (no OC) for past few months, it's a new build. Planning to crank up some OC soon, but not very far with the HyperX ram ? 4ghz air cooled seems very easy on this setup.
good runs guys I like to see 7-7-7 at 1900+ mhz ive pushed my Gskills about as far as they will go im sure 9-9-9 and 2t I can get well over 2000 but I dont want to use training wheels to get their stick with cas8 and 1t
you just need to pump up the voltage check this thread out.. http://www.xtremesystems.org/forums/...d.php?t=199292Quote:
Originally Posted by Ozzfest05
i tested those sticks with my evga 790i board a while back and they overclock pretty similarly with my rampage extreme too :yepp:
Trying to reach a good Super Pi 32M under auto-cascade.
Everest gave me this:
http://images0.hiboox.com/images/500...6245455d6a.jpg
Super Pi1M:
http://images2.hiboox.com/images/500...d4c679e95e.jpg
Max fsb is:
http://images4.hiboox.com/images/500...ead1089f96.bmp
And max fsb valid:
http://valid.canardpc.com/show_oc.php?id=462328
Stil can't finish a decent 32M, system crashes during the bench. Voltages are already at high value :D :(
4 hours :) my insulation isn't that bad actually... ('was doubting about it...)
Edit: Well I finished the 32M 620x9 PL8 as desired... that was... fun... :D :D
http://images4.hiboox.com/images/500...c78ed0dda9.jpg