Quote:
Originally Posted by mikeyakame
:p:
Printable View
Quote:
Originally Posted by mikeyakame
:p:
i was wondering what multiplier/percentage asus are using for the nb gtl ref on the rampage extreme too (0.67?). anyone know? or know how to find out please?Quote:
Originally Posted by ci2kla
Hi Mikeyakame, and many thanks :up: for taking the time for the concise reply although I don't understand it 100% :confused2 it's good to know that it is in theory possible and that I haven't got things horribly wrong :)Quote:
Originally Posted by mikeyakame
guys, is there a method for calculate the NB GTL?
Yeah mate it's exactly the same as for the cpu.
GTL Ref x Vtt = NB Vref
If you have +/- mv settings use them to alter the ratio.
For example my NB GTL Ref = 0.64x
So 0.64 x 1.2v = 0.768v
Anyone with Rampage Formula X48 and 45nm Wolfdale chip? I really need a little help now with lowering my voltages...my current settings below:
CPU Ratio: 8.5x
FSB Freq: 530
PCIE Freq: 100
Clock Skew: Auto
DRAM Freq: 1062Mhz
Timings: 5-5-5-15-5-65 rest on Auto
DRAM Static Read: Auto (enable or disable = crash)
Ai Clock Twister: Lighter
Performance Level: 9
CPU Core: 1.50v (1.489v real)
CPU PLL: 1.50v (1.60v real)
NB v: 1.47v (1,49v real)
DRAM: 2.02v (2.04v real)
FSB Term: 1.48v (1.40v real)
LoadLine Calibration: Enabled (disable crash)
SB v: 1.10v (1.16v real)
SB1.5: 1.50v (1.56v real)
CPU GTL: x0.63 (x.062 became unstable suddenly. 65 & 67 crash)
NB GTL: x0.63 (anything lower no boot; higher crash)
Rest on Auto, all Spectrum disabled. C1E, TM Function, SpeedStep disabled.
Please any experts, some inputs are appreciated! I can't lower any of my voltages at all...
subbed for use later :)
thanks mate :)Quote:
Originally Posted by CryptiK
So, how much should be NB Vref? Cause i can get stability with 1.49vNB and 0.64 - not sure if this is safe.Quote:
Originally Posted by CryptiK
HELP!
Sorry to be that guy, but i just need to know what to set mine too...
Currently I have it set to +30mv on GTLVREF 1 only. all others are auto or 00mv.
I have a 780i Board, and IT ONLY LETS YOU GO +MV there are NO -MV. It goes form +05mv to +100mv.
I also DO NOT SEE AN OPTION FOR VTT so what would I use for VTT?
I am looking to stabilize my QX9650 at 4.0-4.1 ghz. I would also like to lower the Vcore if possible (currently 1.45 set in bios)
What GTL VREF setting do I want to use for
GTL VREF 0
GTL VREF 1
GTL VREF 2
GTL VREF 3
???
What Vtt (vFSB) are you using? I'm not sure what the upper limit of Vref for the NB is but you'd have to be pushing very high Vtt to go beyond the safe limit.Quote:
Originally Posted by spoof
Do you have a voltage called vFSB? Vtt is another name for vFSB. if so what is it set too? It also depends what type of CPU you have, and how it specifically behaves. GTL's arent something where you can ask "what do I set it to" and have it be the same answer for everyone. You must experiment yourself to find the best settings. If you tell me your vFSB, I can tell you where to start from.Quote:
Originally Posted by Xenozx
EDIT - actually I'm not sure if nvidia chipsets call Vtt vFSB, if anyone knows what they call it can you chime in here pls.
Back to theory for a moment with a question...
Since the GTLs change the thresholds, I assume that they also change the timing of the clock signals by making the clock high/low at different points on the signals's edge slope?
So, (assuming that that is correct ;)) -- if I also change the timing directly using slew, do I need to recheck GTLs?
EDIT: do I need to retune GTLs?
As far as i know vNB and it's GTL are only MB related. Currently i'm using 1.36 VTT with 0.67x GTLREF, which is normal(and is < 1). But it turns out, that VTT and GTLREF doesn't fix the lockups. If you remember, i have lockup problem after 455Mhz. It didn't boot after 470Mhz, but with 0.63 GTLREF on the NB and 1.53 vNB it did it with 473, even was able to log into Vista(than it froze). So i think, that the value, which will help me fix the lockups is vNB+NB GTLREF. So, i'm not sure what are the safe borders. When it's vNB+NB GTLREF<=1, i'll not be able to run in with more than 465, but when it's more, may be i'll able to do it (after all, i'm under water, high voltage isn't the problem).Quote:
Originally Posted by CryptiK
It's not vNB + GTL = less than 1.0v, for the cpu anyway, it is Vtt x GTL Ref = less than 1.0v
Just keep playing with it you will find the sweet spot. However your board is not renowned for 450 - 480 FSB stability with a quad, I would be very surprised if it could do it.
These are the voltage settings in my bios.
http://techreport.com/r.x/nvidia-nfo...os-voltage.jpg
my exact settings for each one in order is
vcore 1.45625v
CPU FSB 1.4v
ram 2.125v
SPP 1.55v
CMP 1.55v
SPP - MCP 1.40v
Yes, i know that Vtt x GTLREF = less than 1.0v. I made mistake, i wanted to type vNB x GTLREF = something. So, if i run 1.53(vNB)x0.67(GTLREF for vNB) = 1,0251 there will be no problem?Quote:
Originally Posted by CryptiK
And, if i can do 470 FSB with P5E X48@Rampage Formula(0408), which board should not have problem with quad and high FSB?
Not sure on the first part tho it seems kinda high.Quote:
Originally Posted by spoof
For the question as to what board does decent FSB for a quad, my Asus P5Q Deluxe is doing well with my Q9400 at 486 fsb daily. Had it up to 490 fsb but it was getting a bit flakey and couldn't get er stable, could be the board/ my chip or just an oddball setting.
Larry
Is this true for the P45 also. From experimentation (system specs below). I was able to get stable at FSB 425 with vNB 1.28 and then drop one step to 1.27 by changing NB GTL to +30 from auto (aiming for about 0.67) and at 450 to drop again from 1.29 to 1.27 by going to +40 from +30. Each time I was stable (8h OCCT RAM) at the higher vNB and not at one step down, which became stable by altering the GTL up - whereas most posts here seem to suggest 0.58-0.6 is ideal GTL target for NB - which as I understand it would require a negative GTL offset as the default on this board is 0.64.Quote:
i agree with that guy too! nb gtlref calculation is not exactly a science, well even less so than the cpu counterpart. i have a few simple guidelines i follow when setting the nb gtlref value...
1. you will almost never need to use a higher value than cpu gtlref unless you are attempting to use fsb frequencies which don't want to be used! if thats the case change the fsb frequency it's an actual solution to your problem! its so much easier than jacking voltages all over the place to find stability when it's in fact hiding from you because it lies about its good looks and fun times to feel a sense of importance!
CPU is on auto/-40/auto/-40 aiming for 0.63 on all.
Confused!! :confused::confused:
Andy
What are the default multiplier settings for your board for the 4 GTL Ref's?Quote:
Originally Posted by Xenozx
It should say in your manual or in the bios when you highlight each one. For example on my board they are
GTL Ref 0 = 0.635x
GTL Ref 1 = 0.667x
GTL Ref 2 = 0.635x
GTL Ref 3 = 0.667x
Per the manual.
�� GTLVREF Lane 0
This function defines the voltage level for GTLVREF Lane 0. Use the Page
Up and Page Down keys to select a voltage or select [Auto]to automatically
set the voltage.
�� GTLVREF Lane 1
This function defines the voltage level for GTLVREF Lane 1. Use the Page
Up and Page Down keys to select a voltage or select [Auto]to automatically
set the voltage.
�� GTLVREF Lane 2
This function defines the voltage level for GTLVREF Lane 2. Use the Page
Up and Page Down keys to select a voltage or select [Auto]to automatically
set the voltage.
�� GTLVREF Lane 3
This function defines the voltage level for GTLVREF Lane 3. Use the Page
Up and Page Down keys to select a voltage or select [Auto]to automatically
set the voltage.
what does it mean when you can game, and do everything on your PC perfectly fine at say 4.2 ghz. can run 100 back to back cinebenchR10 quad core. but as soon as I launch prime 95, and start the torture test my computer BSOD's and reboots?
like, thats not the CPU failing is it? Is that a sign that more voltage is needed, the FSB is too high, bad memory, or what?
I think all boards do follow that general rule, as they all use clockgens which have preferred frequency multiples they produce from the base frequency of 33.33MHz. However if this thread has shown one thing, it's that GTL Ref requirements follow a general trend for each chip type (including the NB chips) but each individual chip differs, meaning you need to experiment and find exactly what your chip likes.Quote:
Originally Posted by powellandy
My NB on my M2F does not like anything over or under 0.64x, I lose stability if I change it from that. It appears yours behaves differently, and perhaps the fact you are running a quad alters its behaviour.
Not specific enough (not your fault its the manuals lack of info). You're going to have to do this blind. Try setting +50mv for all four of them. However when testing CPU GTL Ref settings you must be absolutely certain your ram & NB are stable, otherwise you will have no idea where errors are coming from if they occur.Quote:
Originally Posted by Xenozx
Fail small FFT's = usually CPU (but can be ram/board if theyre very unstable)Quote:
Originally Posted by Xenozx
Fail Large FFT's = usually ram/MCH unstable
Can you pass linpack? That is quite cpu oriented also. If you fail small FFT's, increase vcore or vFSB, tweak your GTL Ref's, or lower your cpu clock. if you fail large FFT's, increase vdimm or vNB, alter your timings, or lower your ram clock.
It's bizarre isn't it. I have now managed (with those GTL settings) - at 450MHz FSB - to get vFSB down to 1.27 and vNB down to 1.20 (the latter is much lower than that I needed at 425 - although the vFSB is higher). Just finished 13h OCCT RAM - the last 1h of which had FurMark running as well!Quote:
Originally Posted by CryptiK
Temps BTW where NB=61, SB=56, power=54 - so, again, I seem to be lucky at stable at higher temps than most.
Cheers,
Andy
Nice work, those are pretty high temps for only 1.2 vNB, I'm really surprised you're still stable. What are your ambients?
Did you find that BIOS 1501 helped gain stability with your quad compared to 1307?
They are aren't they - I am not sure why. I did the washer trick on the back screws (but that made at best 1-2oC difference anyway). I now have a 60mm fan blowing on the NB from a zalman bracket. Resting temps are case=38, NB=47, SB=50, CPU=42 (tjmax=100). Room temp is about 22.Quote:
Originally Posted by CryptiK
With OCCT Large (3.0.0b08) - NB=56,SB=54,CPU=66,case=41. When I add in FurMark with OCCT still going NB=60,SB=56,CPU=68,case=44.
I think the problem is I have the SupremeFX card, a modem and then the GPU (8800GTS) with stock cooling that runs 67 idle and 79 stressed in the first 3 slots so the NB is croweded by the cards (and CPU cooler) so that even with the 60mm fan the air is still hot there.
When I was stable at 425 I ended up with vFSB=1.20600 and vNB=1.27225 and still temps of NB=60 (before the 60mm fan) with OCCT. I am now at 450 with vFSB=1.27225 and vNB=1.20600 - I can only assume therefore that vNB and vFSB both contribute about equally to NB heat (especially with the heatpipe system) - I must say I never tried raising vFSB and lowering vNB at 425MHz.
I suppose at some point I should take the whole heatpipe assembly off and put some better TIM underneath and ensure it is all fitting OK as the temps are a little high. Although given I am stable at temps where most others seem to struggle I also ? the sensor - do you know where the vNB sensor is (ie. in chip, under heatsink etc.). I will try and get an external reading from the heatsink itself with a probe and see what it is.
I upgraded to 1501 when I was at 425 - with the settings I posted in the other thread - I was stable with these with both 1307 and 1501. I then went to 450 with 1501 - haven't been back to 1307. I tried 1602 but that crashed OCCT in 2 min with my higher voltage 450 settings - let alone these new ones. My SATAs are on 1,3,5 so I don't suffer from that problem with 1501 and I actually quite like it :).
Cheers,
Andy