finally i got my system stable at 427fsb, thats about as high as i can go.
Printable View
finally i got my system stable at 427fsb, thats about as high as i can go.
Well, i'm using the 0406 beta bios and trying to get this rig stable at 450x8 but found something strange in CPU PLL regulation.
I've started with a value of 1.62 and then lowered it step by step running every time 1 hr of prime95 blend until it get error. So, after many hours of testing, i've come down to 1.50 without an error, but i was expecting the CPU temp will drop a bit lowering CPU PLL but it remained at the same value (the hottest core got to 66°C with real temp).
So i've took a look in bios and found that even if i set a CPU PLL value of 1.50, the bios hardware monitor read 1.55!
Did someone else noticed this? It's normal? Or is it a BIOS issue?
Maybe someone else has already posted this, but i had problems with nb temp. I've reseated the heatsink, put some mx-2 paste on it but always got temp near 70°C on it while priming (well, in my room there are 32° C). Done twice this operation, just to be sure to have done everything right, but got the same results.
So i swapped the heatsink with an TR hr05 ifx on the nb and a TR hr05 ifx sli on the sb, with a 80 mm fan on each one and now my nb temps are 47°c idle and 54° load. The sb remains at 37°C, touching 42°C while gaming.
Uh, i'm giving 1.55v on the nb.
Well, i've been playing around with this board for less than a week and i have to say to be very impressed by this board. It's very easy to oc and even on "auto" settings it reaches easily good clocks: as soon as it arrived i've pushed my q6700 at 3.6 with ram at 1152 simply setting the 360 x 10 on fsb, cpu vcore at 1.45 and ram at 2.3. Simply amazing. After spending the last eights months after buggies nvidia chipset (dfi lt680 and msi p7n diamond) trying to tune these board without a noticeable results and hardly finding stability, this mobo has opened a new world to me!
So, today i started trying to get the same 3.6 ghz with differents settings, 400x9 and 450x8. After readings anandtech's review of this mobo, it seems to me that the best settings was 400x9 with a Performance Level of 5 and ram at 1200. Well, after a couple of trying it resulted that a 450x9 with PL6 and ram at 1200 was way faster than the other setting (everest memory bench). Did i've done something wrong or can someone confirm my same results?
For now i can't get to a stable setup to keep 450x8, RAM at 1200 and PL6 - think it's my RAM that can't be stable @1200 since small fft test is stable but large fft fails after 4 minutes in prime95. And the board won't even post with different memory speed and PL6.
So i've ended to a lower settings, wich seems to be stable: 450x8 with ram at 1125 (5-5-5-15 2t) and PL7. Everest gives me these results: 9296 Mb/s read, 8073 Mb/s write, 8710 Mb/s copy and a latency of 54.8 - it is a good results?
Here is my voltage settings:
CPU PLL: 1.50 (1.552 read in bios?!?!?)
NB volt: 1.55
FSB Term: 1.30
Is there any key to get this board post (and get it full stable!) with a PL6?
PLL is phase lock loop voltage, has nothing to do with CPU temps. If you have Loadline enabled the PLL will be higher than bios settings. Mine runs 0.05-0.1v higher than I set. Ie. I set 1.68V CPU PLL and real is 1.76V or so.
Good luck with PL 6. You need a lot of vNB for that, and on 333 Strap timings are generally too tight. Try slacking AI Clock Twister down to light if you want to run PL6. I have no problems with PL7. My ram won't stay stable at 1200mhz either. It's most likely the ram. I finally managed to post at 500FSB, 7x500, using 333 strap and 5:6 divider but I get display lockup after 5-6mins of mem testing. I need to go over all my secondary/tertiary mem timings and make sure they comply with JEDEC minimum windows. My rampage refuses to post with 1:1 divider at 333 or 400 strap and 500FSB.
Those results are about the right ball park. If you want latency tighter you'll need to do a lot of trial and error. One thing to note though, FSB_VTT is designed to limit overshoot for vCPU. If you raise vCPU above VTT then the CPU's pull in noise filter doesn't do its job anymore, It's also designed to prevent crosstalk and noise jitter between clock domains from what I understand.
FSB_VTT is semi important to keep aligned reasonably well with vCPU, it limits overshoot and controls max deviation of FSB_SWING voltage between rising and falling edges of the clock signal. GTL_REF[3:0] limits the undershoot of FSB_SWING, and these both keep the voltage within V(IL_H) and V(IH_H) limits. These are respectively, Host GTL Voltage in Low and Host GTL+ Voltage In High.
http://www.intel.com/Assets/PDF/datasheet/319122.pdf
If you want an idea on what exactly you break by fixing something else, take a read of this datasheet for the X48. Should help u understand a little more deeply on how to get around some problems and what causes them.
You can also find the Processor Sequence datasheet here for the Q6xxx 65nm series. It lists all the CPU voltages in regards to chipset.
http://www.intel.com/design/processo...hts/315592.htm
Do you need all that volts on the NB? That's why your NB temps are so high.Quote:
Originally Posted by panheimer
And you could raise FSB term to 1.35 or around that.
NB voltage shouldn't need more than 1.35volts.
if you use manual GTL Ref multipliers , then you need to change these periodically as you go between higher and lower voltages across different FSB/DRAM/CPU clock frequencies. Different clock speeds will sometimes need more or less GTL_REF voltage depending on how the signal quality is at that point. I've found myself adjusting them as I've been down adjusting voltages to find stable voltages for current 8x445FSB and 3.57GHz cpu. But got a DRAM timing issue which I can't put my finger on, I've changed too many timings too loosely and seems like I've caused at least one minimum time period to be too short. tRAS or tRFC, guess I'll know which once I finish making sense of all the timings and their minimum values. DDR timings are too damn many, and confusing. Lack of collective data mostly, I've got so many data sheets and white papers I'm sorting through today on the way to making sense of all the different names used by manufacturers.
Well, on nvidia chipset i've got to give such voltages on nb to reach 400 fsb, so i've tought that even on intel's one was similar. This is my first oc trying with this board and i know there are many things to be tweaked. My intention was to give more juice than needed to find a good spot and then start lowering voltages.
Ok, i've got a new question: until now i'm using AI Clock Twister set to AUTO. This morning i've tried to set it to "stronger" and while posting it gave me a message like my usb is overvolting and the system will shut down in 15 seconds?!?!? So i cleared cmos and tried a "strong" value, post is ok, but this time it can't boot windows as it can't find a system file. So i've returned to "auto" value and everything worked fine again.
What's going on?
Sorry for so many quetion, but there is so many options and settings that i've never seen before and this is my first intel chipset mobo.
Thanks for your help!
beta bios 0408, 7/10/2008
http://www.keepandshare.com/doc/view.php?id=712947&da=y
for me it works at least the same as 0406... wonder if they fixed that 3rd party sata/raid controler support...
Updated to 0408 from 0219 - so far so good.
Will be lapping cpu in the next couple of days and remounting the heatsink using the hint from Abducted (he lowered NB temp from 60 to 40 under load) so higher clocks results will come a little later.
I need abit of help trying to get this stable . cant seem to boot into windows nor even run memtest without locking up.
using super talent ddr2-800mhz 4-4-3-8
cpu = e6300.
running at 4-3-3-8 at 900mhz for rams.
fsb 300
cpu multi 7
ram freq 900
1t/2t 2
cas 4
ras to cas 3
ras pretime 3
ras act time 8
ras to ras delay 4
ref cycle time 45
write recovery time 6
read to pretime 4
read to write delay 8
write to read delay 4
write to read delay 6
read to read delay 4
read to read delay 6
write to write delay 4
write to write delay 6
write to pre delay 14
read to pre delay 5
pre to pre delay 2
all pre to act delay 6
all pre to ref delay 6
dram static control disabled
ai clock twister lighter
ai transaction booster manual
cpu voltage 1.2625
cpu pll volt 1.6
northbridge volts 1.45
dram volt 2.3
fsb termination volts 1.5
south bridge volt 1.125
south bridge 1.5v 1.5
loadline calibration enabled
cpu gtl ref 0.67
NB gtl ref 0.67
dram ctrler volt 10
dram channel a volt 20
dram channel b volt 20
Try to lover tRP to 4, see if that helps. Also tRCD 3 for 900mhz is way too low I think. That is: see if 4-4-4-8 will work for you.Quote:
Originally Posted by Emperor
Also lovering tRD (perfomance level) will bring much bigger perfomance jump then either tRP or tRCD (3 or 4 for that metter).
Just my 2 cents :)
Anyone have a change log for 0408????
----------------------Quote:
Originally Posted by GAR
- updated M0110"676"60C, M1010"676"60C, M1010"677"705
- added M1110"67A"A07, M0110"6C1"109, M0410"6C2"208, M0810"6C2"209
add CPU code(Family+Model+Stepping)
Brama tell me if this bios is compatible with raid controllers :)
tRCD = 4Quote:
Originally Posted by Emperor
tRFC = 35
tRRD = 2
tRAS = 10
tWR = 4
tRP = 3
tRTP = 4
tRD_WR_All = 5
tWR_RD_SR = 5
tWR_RD_DR = 4
tRD_RD_SR = 4
tRD_RD_DR = 5
tWR_WR_SR = 4
tWR_WR_DR = 5
tWTP = 11
tPTP = 1
tALL_PRE_to_REF = 4
tALL_PRE_to_ACT = 3
Ai clock twister strong
VTT = 1.26V
vNB = 1.35V
vSB = AUTO
vSB_PLL = AUTO
CPU_GTL = 0.63x
NB_GTL = 0.67x
DRAM Controller = AUTO
DRAM_REF_CHA = +10mV
DRAM_REF_CHB = +10mV
DRAM_STATIC_READ = enabled
CPU_SPREAD_SPEC = disabled
PCI-E_SPREAD_SPEC = disabled
AI Trans Booster = manual, CPL = 7 or 8 to start with
vCPU = 1.29V or so
CPU_PLL = 1.54V or so
If u need adjust clock skews on CHA / CHB. ADV 50 PS CHA / ADV 100PS CHB
and go up and down either way on each from there to help find skews it likes.
Running CAS3 isnt exactly the X48's speciality though, thats more suited for tightly timed 975x chipsets.
no complains about 0408 from me either, it works at least as ok as 0406 which worked fine. No new options or anything cool like that though.
I'm also waiting confirmation that 0408 has no issue with raid controllers. After what I had with 0406 I have no intention to be a tester for asus bios. For me 0406 isn't a "beta" bios, it is a pre-alfa bios.
Interesting that the most recent BIOS on ASUS's FTP is still 403.........
Quote:
Originally Posted by OVERK|LL
That's because I get the BIOS versions directly from the Taiwan's BIOS Dept.
http://vip.asus.com/forum/view.aspx?...Language=en-us
http://vip.asus.com/forum/view.aspx?...Language=en-us
Great to know! I've responded to you on there.Quote:
Originally Posted by Kensek
All i want is my Rampage Extreme!!! Sold my FoxConn BlackOps in anticipation of this Board being somewhat superior.
Oh? Any particular features that stand out?Quote:
Originally Posted by Kensek
http://www.xtremesystems.org/forums/...hlight=rampage
It's the best board ever built in terms of OCing. Will cost $450 though.
I can wait for the New X58s and the Intel true Quad Nehalems in 3 months, but Why!
That's 3 months of fun I'll be Depriving Myself Of!!!!
I'll just buy an X58 1366socket Mobo and a Nehalem in 3 months.