Ok I kept on reading that link till page 8, then my head got overloaded with info. I have taken notes on what I should roughly do, so I will try it out after work and I'll let you guys know how I get on. Thanks for the help SoulsCollective :up:
Printable View
Ok I kept on reading that link till page 8, then my head got overloaded with info. I have taken notes on what I should roughly do, so I will try it out after work and I'll let you guys know how I get on. Thanks for the help SoulsCollective :up:
It's easy - for a default of 1.1 vTT, GTL is 0.63.
1.1 * 0.63 = 0.693, our stock GTL value. Accordingly, 0.693 / vTT = GTL multiplier. For a vTT of 1.2 that gives us 0.693 / 1.2 = 0.5775, and for a vTT of 1.26 that gives us 0.693 / 1.26 = 0.55. This gives us the default GTL values every time, but some users (such as me) are finding that a slightly higher or lower GTL value gives better stability. Work out what the default value should be for your current vTT, then experiment with raising or lowering the GTL multiplier slightly. Using Linpack (I recommend IntelBurnTest by AgentGOD of these forums), the results should be immediate within minutes - from 0.57 GTL to 0.58 I saw results go from two successful runs to ten.
Thanks for that Soulscollective, well I had some success yesterday, this is what I did:
I started by setting both CPU GTL v Ref (0/2) & (1/3) to 0.63, that didn't work, boot failed, so I tried 0.64 for both and it froze within one pass of Dos memtest, so I raised it one more so that both (0/2) & (1/3) were at 0.65 (1.30v x 0.65 = 0.845 so within range). Dos memtest passed 10 runs of Test 5 and one run of test 7. So I booted into windows, I ran windows Memtest on full ram for one hour, roughly (430% coverage) and it was fine, no errors found. So then I stopped memtest and launched Linx now within 3 min I got the dreaded blue screen and the dumping of physical memory message, and the computer restarted. So I decided NOT to change any settings, and just run Linx again from a fresh reboot to see if it'll run the test longer (as it had already passed all the others) so I started Linx for 100 runs on full ram, and after 3hrs and 17min it had passed with no errors, (kind of confused why it crashed at the beginning of the test earlier and now it passed the whole test with no problem), so anyway I restarted the computer to run Prime95 over night. Woke up this morning to find that it had frozen 2hrs 56min into Prime, but no errors on any of the cores up to that point.
Here's a screen shot of Linx finished:
http://img.photobucket.com/albums/v2...atos/LinxP.jpg
So now I'd like suggestions as to what I can do to make my system more stable, it already passes memtest (windows and dos), LinX, but froze on Prime 95? I will also run OCCZ and I'll download Intel burn test and see if they pass as well. I just got two simple questions:
1- Why did Linx crash in the beginning after windows memtest but passed with flying colors after I rebooted the computer?
2- What could have caused Prime95 to freeze nearly 3hrs into priming? (no errors were found up to that point) my Bios settings are below:
CPU Ratio: 8
FSB Strap: 333
FSB Frequency: 400
PCIE Frquency: 100
DRAM Frequency: 1601
DRAM Command Rate: 2N
Timings: 7-7-7-24
DRAM CLK Skew A: Norm
DRAM CLK Skew B: Norm
DRAM Static Read & Write: Disabled
CPU v: 1.29375
CPU GTL v Ref (0/2): 0.65
CPU GTL v Ref (1/3): 0.65
CPU PLL v: 1.54
FSB Termination: 1.30
DRAM v: 1.90
NB v: 1.49
NB GTL v Reference: 0.67
LLC: Norm
CPU SS: Disbaled
PCIE SS: Disabled
Anything I haven't listed above is on Auto
Downloaded Intel Burn test, and it passed the 5 runs and 10 runs:rolleyes::
http://img.photobucket.com/albums/v2...IntelBurns.jpg
BTW: Is Linpack the same as LinX? or is Linpack Intel Burn Test?
and am I to assume that my system is 24/7 stable or not really? I've still got to run OCCT, Prime 95 overnight (again) and Memtest overnight, or am I just being a bit anal? :shrug:
I would say you might need to up your CPU voltage a little bit more and then run Prime95 overnight. I had the same problem you did, I passed Memtest86, HCI Memtest, LinX 100 runs but failed in prime in the 1-2 hour range. I upped my Vcore one notch and the system has been 100% stable ever since. My system is running extremely fast right now and I am very happy.
When I raise voltage for my CPU would I have to do all the tests agaiin? Or could I rest assured that it will not change the results & only Run prime?
Well to say you are 24/7 Prime95 stable you would need to run Prime95 again. I ran LinX 100runs and passed then ran Prime95 overnight after upping the Vcore one notch and everything passed error free. I also looped Vantage 3 times and that ran great.Quote:
Originally Posted by Defratos
Ok well, I increased my vcore to 1.30v last night, but I also increased the VTT to 1.34v (was @ 1.30v) because I've read VTT should be a few notches more than Vcore. Since I upped the VTT I also had to play around with the GTLs and I lowered it from 0.65 to 0.64. Anyway Prime ran error free for 3h 18m before it froze. So now I'm going to try Vcore @ 1.30v, VTT @1.30v and GTLs both @ 0.65.
I'm kind of confused why I would need 1.30v for vcore when I'm only running 3.2Ghz or is this normal? I know I'm soo close to being 24/7 stable, there just seems to be something holding my system back after 3hrs of stress testing on any program :shrug:
Try to lower your vtt instead? or have you done that? im at 1.22 and more just gets it more unstable.. tweaking the GTL is like night and day when you get it right..
I use the method, 0.80/VTT=GTL value in multiplier, as a basic instep, it usually wants like 0.8v REF or little more, im at 1.22 with 0.69 thats 0.84v ref. but every chip is diffrent.
but as a start this is good.
Getting right GTL is important, when setting the wrong for me it was error from 3 mins, to rigt GTL over an hour.
I haven't tried to lower VTT, the lowest I've set it is 1.30 just because I read somewhere that I should "always" keep VTT a few steps above vCore which I'm not sure if its right or not. I'll try lowering it and see if that gives me any help. Increasing Vcore is not helping its most stable around 1.29375 so I'll try lowering VTT and see how that goes, and yeah I agree once I got GTLs correct I could get my computer to pass every stability test, but anything over 3.5hrs seemed to freeze the computer.
Well, I've got it that you should never keep VTT in Line with vcore, doesnt matter if its lower or higher, but not in line with it, it will interfere with the current and make "white noise" or how you say.. :) even if I matched up the GTL when raising Vtt, it did only get worse. maybe if you goal is to get 550fsb stable then you need that much.
ok I see, I just lowered VTT to 1.28 and had Vcore at 1.29375, computer froze while priming around 30min into the test. so I'll try a lower VTT and see how that goes
Good :) and dont forget the GTL either.. remember my formula :) it works actually ..
Be prepared that Vdimm and Latencys can affect things more than you think to, its not all about the voltages, one notch wrong on vdimm will make an hour in stabilty sometimes. Man i have been in trial and error SO MANY times finding these things out, im in one now where Large FFT's crashes in 7 mins, now i have to go trial and error for hours :)
Ok, now I need help too :)
i Cant for the world not get my NB/Ram settings stable.... CPU are stable as i have tested that to make sure, but LARGE ffts crashes, my changes just increase or decrease a little in my tests, and now are we talking about max i get is 7 mins, lowest about 2 mins, and if i change more than this, it doesnt get better, so i dont get it actually, why?
CL settings 7-6-5-16 60 , vNB 1.49~ VTT 1.20 , nb GTL 0.61, NB skew auto, CPU skew Auto, right now im trying advance 100 and advance 200 on CTL skew..
but it just wont work, 4mins now as counting... higher or lower on the voltages dont help, found 1.98-2.02 for vdimm nothing else more stable, lower than 1.49 unstable on vnb, higher even so... so must be some skew tweaking that has to be done?
Thanks for that Infa, yeah I recalculate my GTL ref voltages every time I change VTT, I keep it in the 0.80-1.0 range, usually 0.8 something. As for Vdimm and Latencys I have kept this at standard recommended settings from OCZ. The main issue with me is the ram does not like the motherboard I have, so they need more tweaking, I can get the system stable at lower than standard speed i.e. anything lower than 1600Mhz will work, but I want it to run at it's advertised and "stock" speed which is 1600.
Bloody hell you're running 4.5Ghz, I'd be satisfied with 3.6Ghz ;), still trying to get 3.2Ghz completely stable. Hope you find a solution to your problem, I know I won't be of much help, and yeah I know all about trial and error, I've already been through over 60 tests and bios settings to get this far (and I've got the proof) I'll post it up once I've gotten things stable.
My aim today is to drop VTT down to 1.2 and work my way up and see what gets better stability.
Heh, No problems man :) i like to share my knowledge! well i got a dual and you a quad they are quite diffrent from each other though...
now the test ran 13 mins... ahh...more testing, driving me crazy :)
Usually quads need more VTT, but not allways as every chip is different, why not aiming higher? and try to get it stable from there instead? 3.6? 9450 is multi 8 right? how high can you get your FSB?
Yeah it is multi 8, I haven't really tried to push it above 400fsb (noob here) but I would be over the moon if I could get it to 450fsb. I just want to try and get a stable system sooner rather than later and seeing how I've spent so much time on 3.2ghz I'd rather get that 100% stable and then work on 3.6Ghz.
I forgot to ask something. When setting CPU GTL v Ref I place the same figure for both (0/2) & (1/3), now at stock I believe one is at .63 and the other is at .67, should I keep these two separated by the same amount? because so far I've put the exact same GTL settings for both.:shrug:
OMG now i have trail and errored for 2 days, NOTHING works damnit.... it has been like this before on this board, i find a stable setting, then suddenly it's not anymore. and i have to retweak or take out and in the RAM, change place or something, and it got stable again, but now nothing works, tried more, less volts, diffrent cpu and nb skew, IT DONT work, it allways fail in OCCT after 3-6 mins. help me? i usally get this right on my own, but now im lost, perhaps the RAM has gone bad? been keeping it at 2v all the time and that shouldnt kill them all of a sudden... will keep on trying some stuff, but suggenstions would be nice :)
regards.
EDIT: maybe a little bit unclear, but Im talking about RAM stability :)
Ok, now i dont get whats up, memtest now passed 270%, and no error :) OK! maybe its OCCT 3.0 thats up? a bug? couse this makes no sense :)
will test Prime LARGE instead and see....bbl.. :)
EDIT: Hmm, Dunno whats up again :) but prime fails too, but ONLY at FFT size 1792, its fine until it gets there, then it BSOD,or reboot.. is it to large to handle? what sizes are usual to test? Prime's own settings for "large" fft's are 128-1024, and i take it as if THAT passes then its fine?
but BLEND mode uses upto 4096, and i can't do anything that will help my p.o.s to get past 1800fft ?
gaah, bad memory? i can put dual memtest when i sleep to see if it errors, but it didnt error after 1h10m atleast.
I just got done trying to beat my 505FSB wall. I was able to bench 1M and 32M @ 510FSB which is my highest yet. I beat both my 1M and 32M personal best. The 32M beats my HWbot 32M Q9450 1st place record.:eek:
I also lapped my Q9450 and True120 Extreme Black. They were both extremely concave. Temps are much better now;)
Nice :) Any chance you can update your linked to profiles for 0605?
Sure, what profile are you looking for? 510FSB Profile? I ran this in my chiller. Idle temps 19C / Load max 47C
I would not advise running this profile with air or plain water cooling.
I was hoping for all of them :p: Seriously though, I was referring to those linked to in sig. Would be good to have a look at what worked for someone else with a quad.
Well you can download all the ones I have uploaded by clicking on the link in my sig. I will upload the 2 I have been playing with for the past 2-3 weeks.Quote:
Originally Posted by SoulsCollective
Here's a picture of my Q9450 before I lapped it. Both my Q9450 and True120 Extreme are nice and shiny with a mirror finish now!
I know the feeling Infa, and yeah it sucks. I'm still trying to get my 3.2 stable, I've gotten a bit further but I still need a good week of testing I'm hoping before its 100% stable.
What is lapping? and why do people lap their cpus? how do you lap a CPU?:shrug: