Here's my best so far but it got hot as hell, up to 95 degrees on core 0. That was running Prime95 that was stable for 2 hours. I tried to find a lower Vcore / QPI combination but all I've got is blue and black screens.
Attachment 89387
Printable View
Here's my best so far but it got hot as hell, up to 95 degrees on core 0. That was running Prime95 that was stable for 2 hours. I tried to find a lower Vcore / QPI combination but all I've got is blue and black screens.
Attachment 89387
wtf.. you're using a true, 3.8ghz and hitting 95ºC?
isn't that too much?
it's only 1.4v
The base is probably not making good contact, my base is convext, but im lucky that atleast the middle makes good contact, the sides are left out, temps arent bad, so i wont lap it yet, since i am not at it. I have spoked to several people who sell the TRUE, all say the base problem is on most of them, especially the IFX 14.
and some pi action with stock i7 cooler..
http://www.imageshack.gr/files/p2q17huyiwpv030cs5aw.jpg
How much of a performance hit would I take used an old DDR 3 1600 2 x 2Gb kit vs triple channel?
anyone got a prime95 run yet?
whats max tempie for this babies?
That's exactly what I've seen on the graphs from the RAM companies like Corsair, and review sites. It's almost like a 1-2-3 linear increase. The bandwidth numbers are shocking with this platform. I about flipped out of my chair when I saw one of Fugger's runs. This platform has "knock your head off" bandwidth! :D
My TRUE only makes contact in the middle as well, but my temps aren't nearly as bad as those.
My question was more like... "Is the i7 hotter than my Kentsfield?"
cause we're both using the same volts (1.408) and the temps here are pretty high (28ºC)... I get 35 idle and 66 full... My temps only go over 70 when my room is above 30ºC
sorry for running out of topic, I'm just really curious as to how hot is the i7
Water cooling mine with dtek fuzion (old model) and a very ugly 30 minute homemade 1366 bracket (awaiting real ones), temps max 68C under linpack, 66C prime smallffts. 4ghz looks like will be pretty easy to get stable, but just started testing few hours ago. Way too many numbers to play with, going to take a while to figure out what I am doing.
TJMax is 100C.
When overclocking you'll probably lose long term Prime stability near 90C.
I finally got RealTemp sorted out today so it should work correctly with Core i7 now. Version 2.86 is available here:
http://www.fileden.com/files/2008/3/...alTempBeta.zip
VID has changed in Core i7 so I need to do some more reading and testing to get that working correctly.
There are retention brackets for the 1366 at coolermaster.com for the V8. Contact Coolermaster support and buy some if thats the reason your not making good contact. until then put your tower on it's side and put something heavy on the V8 and downclock your CPU until your shipment for those brackets come in.
Idk but i do know these things can handle a bit more heat than the core2 series. i've seen people at 90C and even 100C and it not fry there chips.. but i'm sure they were'nt running there system 24/7 like that.. Intel likely will not release the TJ Max values until these things are about to go off the market :(
As long as it's stable and not above 100C I wouldn't worry about it.. Well I'm not worrying about it. :p:
I made some changes and it's now running cooler.
BLCK: 190 (was same)
DRAM: DDR3-1523 (was same)
CPU Volt: 1.4V (was 1.4125V)
DRAM Volt: 1.64V (was 1.66V)
QPI/DRAM V: 1.375V (was 1.38125V)
CPU PLL: Auto (1.80V) (was Auto 1.88V)
Memory Timing: Auto (8-9-9-24-1T) (was set @ 9-9-9-24-1T)
HT: Disabled (was Enabled)
Attachment 89422
Here's 12 hours of Prime95. The max core temp has been 79 degrees.
Attachment 89423
interesting, is HT really usefull for non benchers ?
for the 965 users....and especially those who already run their 965 with SS,dice,cascade etc etc....
what do you think of the potential of this on cold?
http://www.imageshack.gr/files/q6udfdqkmiwteza0m3ow.jpg
I've noticed that aswell, on the old corsair 2x2gb 1600mhz 9-9-9-24 1.8v kit where it was topping out @ 11k~ read, the new 3x2gb 1600mhz 9-9-9-24 1.65v kit is going to 17k~.
I've read that it's way too fast even at stock for desktop systems to take advantage of it, and that it'll only really be utilized in enterprise models for things like socket-to-socket communication.
...if some NHM owners can post somes screens of this beta (Read only version): CPU-Tweaker-beta1.zip
I need it before continuing writing...
Note that it's a new version wich support only NHM and K10 CPU (with integrated Memory Controller).
The interface is the same for both CPU, and I use WinRing0 driver now (unstead tvicport in memset).
I keep memset for chipsets with integrated memory controller (P35, P45...) and athlon64.
Thank for the report... :)
I hope this helps!
Attachment 89425
Update:
If the memory timings are correct then CPU-Z is wrong. CPU-Z shows CAS# Latency @ 8 while CPU-Tweaker shows 10. They were set to Auto in the BIOS. I also noticed the 2x2048 when it should be 3x2048.
http://www.isarapix.org/pix68/1227361031.png
First thing I see is that it says 2x2GB even though it's 3x.
4ghz (4.2 turbo) for 24/7 prime stable, linpack on max stable 60. Need 1366 WB to try higher 24/7.
Both pics same settings.
Turbo enabled, loadline off
Vcore bios 1.39, cpuz idle 1.36, cpu load 1.31
Vtt 1.30 with uncore 18x
QPI 36x
Vdimm 1.62, mem 8, 8, 8 24 (cpuz 1.48.5 bug says CAS 7, it is 8)
need to get the newest superpi my friend
SuperPi mod 1.5
TJMax is stored within each Core i7 CPU. RealTemp and others have no problem reading that value from MSR 0x1A2.Quote:
Jasonlmore: Intel likely will not release the TJ Max values until these things are about to go off the market.
Bowman---where can you download CPU-Tweaker 1.0 Beta1??--sorry found it
=> eccu: See post #328
Cas# and memory size, I note it for the finale version...
2 cores enabled.....stock i7 cooler....crappy mem....cpu-z wrong....CAS10....
must change mem....stop....this is crap....stop....
I flashed my P6T Deluxe BIOS to 0903 and it's not working out. Have any of you tried it?
The flashing process wasn't smooth as some of the options weren't available on the first few boots but eventually that got squared away. Now when I try to oc I haven't been able to get SuperPI to complete its run. I've changed the voltages but I haven't found any that will work.
I had 0805 up to 3.8 GHz Prime95 stable for over 12 hours. I'm about ready to flash back to 0805.
Question... what cooling are you guys reaching 4+ using? Are you air or water? And of those what brand?
I'd love to know what kind of cooling you needed to break 4.0ghz....
...a new beta: CPU-Tweaker-beta2.zip
-Read Cas# should be better.
-Fix Reading Memory size.
-Fix a XMP detection.
Water,PA120.3, 3 yates 1300rpms, ambient 26C, homemade ugly bracket on DTek v1 (pic)
http://www.xtremesystems.org/forums/...5&postcount=19
4ghz prime temps are 70C w/ load vcore 1.31 after 9 hrs small ffts. At 4.2 and higher, temps get toasty on load...just starting to bench now, will await new waterblock for higher mhz prime.
Interestingly, the turbo mode does not affect stability, despite being at 4.2 ghz...and this is likely from less vdroop with only 1 core loaded.
When more than one core loaded, no turbo, and 8 instances prime 4ghz is 9hrs small ffts stable w/ 1.31 load vcore (post 331).
When only one core loaded turbo kicks in to 4.2, but vdroop is to only 1.344 vcore (decreased load of one core), which is enough for that core at 4.2 (4.186 mhz turbo core 1 23x182) to be prime/orthos small ffts stable, tested by using taskbar to run prime/orthos small ffts w/affinity set to core 0,1 of 8.
I can leave turbo enabled, and get extra boost to 4.2 on single core tasks.
I would bet that is the case for most, if not all. Mine is definitely that way, going from 4 to turbo 4.2 or 4.1 to turbo 4.3. The vdroop is so much less with one core + running one core is easier to overclock. It might not hold true if overclocking all cores to 4.4 or so, ie if turbo reaches past limit of chip, also if put loadline on may alter things, have not tried that yet.
As far as I know everything is correct on the main form. Should the spd button work?
Attachment 89525
Attachment 89526
... :D could you send a cpu-z register dump at memset@hotmail.fr please.
This is what I'd like to know too...right now with Rampage II Extreme (0602 BIOS) and a Core i7 940, stock Intel cooler I cannot get over 3.4-3.5GHz to be Prime95 stable on 4 cores/8 threads. Still waiting on my TRUE, but if I cannot get 4GHz with it then I will likely want to go water instead...
What's the maximum reasonable speed expectation for something like TRUE or Noctua?
I'm starting to get the feeling 4GHz is not going to happen on air...not unless TRUE shaves off 20C when compared to stock cooler....
Also, what is a good water setup for this rig? Which rad/pump/block(s)? I have an Antec 900 btw...never done water b4.
sub9 spi, on water with ambients 26C. Best mine will do with 1.55 vcore bios.
spd fine here....
I found that my stock cooler was not flat. It was a PITA to lap too. I had to take a special machined block I have that was just the right size and "reverse lap it by hand. If you do this make sure that what you use to sand with is very true...it pretty much has to be machined, and the paper must be affixed to it completely flat with some spray adhesive ina very thinn coat to avoid 'waves". I took the block and pressed it onto the opaper on a glass top on my workbench I have, but any piece of thick glass would work to make sure it was totally tru and flat.
Care must be taken to avoid any uneven pressure. It's a laborious process.
It also could be that the IHS is not flat, but that gets into seriously voiding your warranty. If that's not a concern lapping may help. Some have reported bad Thermal Interfaces.
Saying that I am definitle going water whether I have a cool running CPU on air or not. I'm one of the "cooler is better" crowd, and even if it would do 4GHz on stock air, I know that it might do 4.5 on water. I'm not expecting that much, but if I were to get it that would be great. I already have some of the parts here for water, but will testing first with air. After studying water cooling I sriously belive it's a more efficient way of cooling. It just takes a bit more care.
Man, I am hoping that it is not the IHS. With AMD I had no problem because I'd just pry off the IHS and cool the CPU directly, but with a $700 CAD CPU I am not about to even attempt this...makes a huge temp difference though, at least on AMD's parts. In terms of the stock cooler, meh, I do not want to bother lapping it. That one's gonna go anyway. I hate crappy push-pin heatsinks. The one that comes with the Core i7 920 and 940 is even lousier than the one shipping with the Core i7 965.
In terms of water, I want to go water, but I am not sure Antec 900 will accomodate it nicely. Maybe I will try the TRUE first, and then see from there. I can always sell it if I want to do water. If I can get a 4GHz 4 core/8 thread stable OC I will be more than happy. I do agree though that water cooling is the way to go. It's just a bit $$$ for good parts.
By the way, you can tighten those timings on the Dominator sticks. I have the same ones, but my timings are 7-7-7-16 at 1603MHz, 1.65V :)
...fixed in this version: CPU-Tweaker-beta2.zip
Started to test Gigabyte EX58-Extreme. What confuses me is the CAS and dividers. With 7-7-7-20 I get 6-7-7-20 with 133MHz BCLK, with 222MHz BCLK I get 8-7-7-22.
Also, if I am raising the divider I cannot boot, not even with CAS9 or anything and low BCLK. Have any of you experienced any simulair?
Screenie with stock cooler:
http://loppmarknader.se/blossas_albu...222_stabil.JPG
cpuz doesnt read cas ok all the time, should be fixed soon.
Ok, so it is not directly related to just Gigabyte-motherboard then I guess... Strange though that same CAS-setting changes within cpu-z depending on BCLK.
Thanks for your reply. :)
The spd button now works for me.:D The CPU-Z Timing Table doesn't match with yours but is either wrong?
Attachment 89559
Hi
ihave my i7 920 @20 x 190 3800
on a gigabyte UD5
water cooling on cpu Fuzion v1 with temp bracket
the northbride seems to get quite warm
T balancer reports 55 c from termal probe
works fine but can not get in boot @ 20 x 200
lower the cpu x & ram x
volts
1.325 cpu
ram 1.65
vtt/qpi 1.28
also is better to disable the specail cpu features
such as hyper trending etc
cheers
like xtreme ice says, put uncore to 16x (uncore has to be at least 2x mem divider however), and you may also have to raise qpi/vtt to 1.32 or so which feeds uncore. Also assuming you put qpi to 36x, lowest setting.
btw....does anyone know if qpi pll helps to stabilize qpi, if overclocking qpi (from necessity at higher base clock b/c not enough qpi dividers)?
And anyone know why not more lower qpi dividers, need to have 30x and maybe 26,24x?...not to mention mem divider I need is 9:2.
Hi
yes had uncore @ 2x mem divder & qpi @36
will try higher qpi voltage
will not start pushing it until get 1366 bracket for fuzion 1 or may be new water block:rolleyes:
cheers
some quick tests at 1.4Vcore load on SS...
1.275 Vqpi,1.8Vmem...
for some reason cpu-z shows only 4 threads not 8.
I checked bios, hyperthreading activated.
using a gigabyte ud5 board and a i920.
Did you setup the new system with a clean OS and set ACPI Multiprocessor PC?
OK, you're good then. In nLite you can set that ACPI stuff in the "pre-setup" or whatever they call it. If you see 8 in the Device manager the OS is working correctly. Just was checking to be sure the OS was configured right. It is. It has to be right or it wouldn't show all 8 in the device manager.
I don't post much here, but I figured I'd throw this out there so everyone can get a better understanding of how the i7 overclocks.
Stock Intel heatsink:
http://hardware-geek.com/Pictures/i7bench12.png
Since I'm on the stock heatsink I disabled 2 cores to try and lower temps...saved me a whole 2C but doesn't allow me to OC any higher w/o causing instability.
My 920 in a 29deg celsius room temperature, just fooling around to see low vcore stable, EX58-UD5, XMS3 1520mhz, True black air cool. Vcore set 1.31v and QPI/VTT 1.24v, dram 1.6v that's all.
http://i360.photobucket.com/albums/o...4Nov241942.gif
http://i360.photobucket.com/albums/o...8/PB240317.jpg
http://i360.photobucket.com/albums/o...8/PB240314.jpg
http://i360.photobucket.com/albums/o...8/PB240315.jpg
http://i360.photobucket.com/albums/o...8/PB240316.jpg
It would be nice, if every one post their BIOS settings.
Hi guys !!
Got mine core i7 today
Corsair 3GB 1600 Tri-channel
Gigabyte UD5
1366 Enzotech waterblock retention
Core i7 920 OEM @ 3836A
http://img376.imageshack.us/img376/3765/nhelamenbp8.jpg
http://img246.imageshack.us/img246/3466/cpubatchvx8.jpg
http://img127.imageshack.us/img127/1570/sistemalv6.jpg
3800Mhz @ 1.28v
26º/31º Idle
60º/65º full
http://img246.imageshack.us/img246/6730/idleag8.jpg
http://img397.imageshack.us/img397/2...estavelmm2.jpg
16500Mb/s & 28ns
http://img80.imageshack.us/img80/8361/bandaturbowf1.jpg
Setings:
Multi: x19
BCLK:200
Performance : Turbo
Loadline : Enable
Uncore : x16
Memory : x8
QPI/VTT : 1.28v
Vdim : 1.60v
thnks to user gligy for the bios settings ! :up:
regards
Agreed, thats kinda the whole point of the thread of you think about it. With ever person that posts up bios and system specs it gives us a better overall picture of the i7. I'm still waiting for a couple parts to come in, but I'll post in with bios/pics when I can.
2nd boot
4Ghz @ 1.36v (1.33 Full)
http://img231.imageshack.us/img231/3913/4ghzui4.jpg
I only run SPi and everest , didnt tested with prime yet !
regards
@mascaras...nice results...congrats. :up: I would have been happy with 3.8GHz at 1.28v. I hope I am as fortunate. I don't get mine until tomorrow. :(
Thank you and gligly for posting the bios settings.
Hi
been testing 920 Costa rica A
3835A766
Cpu @190 x 20 3800
cpu @1.325
Qpi =1.28
Dram =1.64
Turbo mode enabled
mem is run @ 1520 (not too bad)
but have made mistake of geting corsair 1333 tri channel
should have gone for 1600
if i lower the mem x to 6 and uncore to 12 so 200 x6 =1200
computer will not boot
thanks to gligly for orginal advise on volts
cheers
Loadline ON 1.35v bios, 1.328 idle, 1.31v load is small ffts stable (12+hrs, pic below)
Loadline OFF 1.39 bios, 1.36 idle, 1.31 load is small ffts stable (9+hrs, pic previous post)
overshoot should be .08v from 1.31v in both cases (if GB followed intel recommendations and overshoot =bios setting loadline off - load v)....so for me LOADLINE ON FTW.
Bios F3, 182x22, turbo on, (turbo tested prime stable by setting affinity to cores 0/1).
QPIx36
uncore 18x (18x182 or 3276mhz) w/1.28v qpi/vtt (for 20x w/182blck need 1.32v)
pcie 100
performance turbo
8 mem divider (for now, OC last, need 9 mem divider)
DRAM 8,8,8,24 at 1.60v (rated 1600 1.65v)
rest stock/auto, all that matters on mine is above settings.
Hi again
Settings:
Multi: x19
BCLK:200
Performance : Turbo
RAM: DDR3-1600Mhz @ 9-9-9-24
Loadline : Enable
Uncore : x16
Memory : x8
QPI/VTT : 1.26v
Vdim : 1.60v
Vcore : 1.31v Bios (1.264 Full)
CPU settings :BIT & Multi Core ON , rest All OFF
3800@ 1.264v
http://img407.imageshack.us/img407/5625/126ow3.jpg
i will try lower the vcore and see if runs prime95
regards
Is that stock cooling mascaras?
just checking, mine is at 3.2GHz w/ 35C idle on stock cooler, if you got to 3.8GHz with 35C i would have been insaaaaanely jealous
hehehehe :p:
btw: if anyone interested , i posted a few tests with Core I7 920 & Gigabyte GX58-UD5 here >> http://www.xtremesystems.org/forums/...d.php?t=208811
:up:
? bug in F4e GB extreme bios where 23 multiplier (turbo) always active, (would not think that even possible?) When priming 8 cores, multiplier should be 22 (turbo only active with one core) but multi stays 23 per cpuz with all 8 cores loaded only on F4e bios, works correctly F3 bios. Also, F3 prime stable 12 hrs, but same settings in F4e crash immediately almost as if really running at 4.186...(though could be memory channel subtimings were different as well in F4E). Anyone else run into this issue? to late this pm to try and figure it out...
I'm working on 4.0 with my Gigabyte EX58-UD5 and i7 920 It's a little work but I think I can get there Vcore is a little stronger than I would like, but the board has blown me away how graceful and perfect it Overclocks.. I haven't ever really owned a Gigabyte because they always seem to leave out some feautures that I love (esata) but I am absolutely astonished at how much quality this board brings to the table for the price... I suggest this board without reservation for anyone considering an X58 build I will post my settings in the morning...
Hi
hmmm it seems F3 bios is better ( im using F3)
yep , im also enjoying the UD5 , really great mobo , the only problem i found is that i need to use the White Ram Slots , if i use the Blue Slots the system wont boot ,(2 friends also had same problem) do you have same problem??
best regards
as per manual for 3 triple channel start with white slots http://i4memory.com/f90/gigabyte-ex5...os-info-11500/
http://nx.i4memory.com/x58manuals/ex...figuration.png
Hello, I am trying to overclock my MSI X58 Platinum and i7 920. Can some tell me a software application that will tell me what the current QPI voltage is? The MSI does allow control of that voltage but it does not ever show what it is in BIOS. The auto setting for CPU voltage does not work correctly when over clocking so I do not trust the QPI voltage to the AUTO setting. Thank you.
I didn't notice this before but CPU-Tweaker 1.0 beta2 shows the Bus @ 195.6MHz along with some other changes.
BIOS Settings:
CPU Ratio Setting: 20
BCLK Frequency: 190
DRAM Frequency: DDR3-1573
DRAM Timings: All Auto
Attachment 89639
I then checked the stock setting and now CPU-Tweaker shows the Bus @ 137.6MHz.
BIOS Settings:
CPU Ratio Setting: 20
BCLK Frequency: 133
DRAM Frequency: DDR3-1066
DRAM Timings: All Auto
Attachment 89640
I looks like that CPU-Tweaker is adding 4-5MHz to the BCLK BIOS setting. Now am I the only one that is seeing this?