How to set up GTL Ref Values for 45nm & 65nm

[CryptiK]

I'm certain that is what he meant. For multithreaded programs that take advantage of all 4 cores of a quad, a moderately clocked quad will be better than a dual core, but for applications that do not fully utilize all 4 cores, a faster dual core cpu would be better.

[lichyetan]

can anyone provide a guide for DFI P35 UT GTL reference guide for 63-61% reference... i found its different from other board...

currently running 61-61-61 for my e8400 can it works fine... but still curious to know wat percent of reference i am in for vtt of 1.2v as i tweak the current values with trial and errors.... which takes lotsa time.

Been Googled around for a long time, only found those 67% references tables and explaination...

[globi-wan]

for e8xxx series gtl is 50 -55% of vtt

[TheKarmakazi]

very nice OP and great responses, big thanks to mikey for layman's explanation. Ive learned alot more here, than with the ultra technical explanations on techrepository and anandtech.

[davidk21770]

Nah, I think it's not the opposite because Mike stats that when you don't use the quad to its full potential a dual will be faster. Meaning if you don't use it for encoding and stuff you would be fine with a dual core CPU. At least that's how I interpreted it.

It just didn't hit me that way... The way that you and mikeyakame are interpreting it, it makes sense.

[aka_titan]

i am currently running my X38 P5E with the maximus formula 1207 bios

i run my Q9650 at 3.6 ghz (9x400)at 1.20v(1.18v in everest). vNB 1.33v, NBGTL 67x, CPUGTL 63x, DDR2 2.1v. pretty sure vFSB is 1.2v.
i will check my other settings when i get home, maybe i can find alot more speed from this setup...

whats the max safe voltages for CPUPLL and VFSB? from what i read on this thread CPUGTL Voltage should not exceed 1.0v for a 45nm Quad (calculate and adjust CPUGTL so not to exceed this)and a vFSB should be about the same as cpu Vcore. correct?

anyone know what the max NBGTL voltage is for the X38?

[mikeyakame]

its relative to vNB. Higher GTLREF voltage = higher likelyhood of noise in signal. CPUPLL should generally be about 0.3-0.4v higher than vCC. vTT depends on what vCC and vNB are. Safe limit of vTT would probably be around 115-120% of vCC. Safe limit of CPUPLL, just make sure the differential between vCPU_PLL and vCC are not larger than 0.3-0.4v. Keep vCPU_PLL under 1.7-1.8v if you are concerned.

[aka_titan]

i am currently running my X38 P5E with the maximus formula 1207 bios

i run my Q9650 at 3.6 ghz (9x400)at 1.20v(1.18v in everest). vNB 1.33v, NBGTL 67x, CPUGTL 63x, DDR2 2.1v. pretty sure vFSB is 1.2v.
i will check my other settings when i get home, maybe i can find alot more speed from this setup...

whats the max safe voltages for CPUPLL and VFSB? from what i read on this thread CPUGTL Voltage should not exceed 1.0v for a 45nm Quad (calculate and adjust CPUGTL so not to exceed this)and a vFSB should be about the same as cpu Vcore. correct?

anyone know what the max NBGTL voltage is for the X38?

above is true, vCpu_PLL is at 1.5v the lowest available setting. thing is i cannot get this cpu to post at any higher, no matter what voltages i throw at it.. i don't understand such good overclocks with such low volts..it even passes OCCT...My Q6600 Run 3.6ghz and 450mhz FSB on this board...

[Dee]

for e8xxx series gtl is 50 -55% of vtt

Wow. There are so many conflicting opinions on this GTL stuff it's astounding. One guy says "don't go anywhere near that value", then someone else says something completely different.

Now we learn that GTL for E8xx series is 50-55%, yet an Asus board, which has two different values, sets it to 0.63% and 0.67%.

It's no wonder just about everyone is clueless, including me. I know nothing more than I did before this thread was started.

[DemonEyez]

Wow. There are so many conflicting opinions on this GTL stuff it's astounding. One guy says "don't go anywhere near that value", then someone else says something completely different.

Now we learn that GTL for E8xx series is 50-55%, yet an Asus board, which has two different values, sets it to 0.63% and 0.67%.

It's no wonder just about everyone is clueless, including me. I know nothing more than I did before this thread was started.

Why, just because one user says it's 50-55% , but hasn't elaborate on this or gave some source information?
NOFI, but the guy only posted 4 times and his join date is of april this year. So I for one am sticking with the info Mikeyakame is giving us on this issue

[Dee]

I see your point, although I don't think the amount of time a user has been registered should mean all that much.

Looking at my join date you'd assume I was a n00b here, well this isn't so. It's my second account, I've actually been here since 2005, but I kinda expect the members here to see me as a n00b to overclocking, hardware, and forums in general. I'm only new to overclocking Intel systems. AMD was my brand for the last 4 years, and now things are all different!

And it does get slightly confusing.

Anyways... back to the point. I'm just so tired of the fluctuation in opinion and never knowing who is right or wrong. I don't just mean in this thread, but all over the Internet. There is no solid info anywhere as to which values are correct, how much voltage (be it vTT, PLL, vNB, etc) is "safe", and how to set it all up correctly. Not that I don't trust Mike's posts, on the contrary, he seems a very knowledgeable guy. I'd like to follow the advice here but the things I've tried so far have gotten me nowhere, which then makes me look further into it, only to discover about a million different opinions!

I have to laugh, though. Especially as you value Mike's opinions, yet he's been here only since May 2008!

[mep916]

This thread is exactly what I've been looking for. Never really understood these GTL ref values, so maybe with this information, I can get my Q9450 stable above 3.6.

[mikeyakame]

I see your point, although I don't think the amount of time a user has been registered should mean all that much.

Looking at my join date you'd assume I was a n00b here, well this isn't so. It's my second account, I've actually been here since 2005, but I kinda expect the members here to see me as a n00b to overclocking, hardware, and forums in general. I'm only new to overclocking Intel systems. AMD was my brand for the last 4 years, and now things are all different!

And it does get slightly confusing.

Anyways... back to the point. I'm just so tired of the fluctuation in opinion and never knowing who is right or wrong. I don't just mean in this thread, but all over the Internet. There is no solid info anywhere as to which values are correct, how much voltage (be it vTT, PLL, vNB, etc) is "safe", and how to set it all up correctly. Not that I don't trust Mike's posts, on the contrary, he seems a very knowledgeable guy. I'd like to follow the advice here but the things I've tried so far have gotten me nowhere, which then makes me look further into it, only to discover about a million different opinions!

I have to laugh, though. Especially as you value Mike's opinions, yet he's been here only since May 2008!

I don't ever tell you to believe me. I'm just trying to help out. AGTL+ is AGTL+, GTL Reference multipliers are the same for all systems and all chips, why...because they all use the same design FSB. That is the reason Asus uses those values for all chips. Intel specifies between 0.65 - 0.69 multiplier, but that doesnt mean you cant deviate under or over if it works better with that particular board. I don't quote anything off forums, all that I know come from tech docs, books and real world applications. Take everything you read online with a grain of salt and form your own understanding based on common sense.

Opinion is opinion, not everything works for everybody depends on where the problems lie. Figuring out that bit is one thing nobody on the internet or even in person can automatically tell you. When I look at a system It takes me playing around to figure out where the weaknesses are, then figure out what to do from there. It's how you need to approach things. It's kinda like changing GTL Reference voltage when the problem lies in the PSU struggling to supply constant voltage, it won't fix the problem no matter how much knowledge you throw into it because the PSU is still the key weakness and without changing it all the knowledge in the world won't get you any further in your process!

I do all I can to help, I'm a pretty busy guy between work and other real life commitments I don't even find much time to work on my own system! Kinda sucks but thats life you know.

[DemonEyez]

I see your point, although I don't think the amount of time a user has been registered should mean all that much.

....

I have to laugh, though. Especially as you value Mike's opinions, yet he's been here only since May 2008!

Hmm I see your point. I admit, user time does not say much.

Nevertheless, I'm not running blind on one simple post that says it's 50- 55% which has not been backup by some fundamental info, something Mikeyakame had done numerous times.
For all we it's an assumption because mobo builders are setting vTT to 1.10v now for the 45nm processors. But another reason could be they do this so users don't set it too high like 65nm processors.

I'm following the info Mikeyakame is giving us. And so far is has helped me a lot. Before this topic I could not do 9x500 regardless of what voltage I set for vCC vTT and vNB. After reading this whole thread, writing down his posts with formula's and info about the ideal range for those voltages, I gave it another shot.
Guess what? i'm running stable now. Innitially I set GTL to 63/63/58 because a lot of people with a P45 T2RS plus use that setting for 500MHz and it also corresponds to setting GTL Ref to 55% of vTT.
Afer redoing the math following the guidelines of Mikeyakame (set GTL to 67% of vTT and close to 0.80v) I concluded that GTL had to be set around 67/65/58 or 67/66/58 for my vCC and vTT. After this change Orthos has been running for 4 hours straight, where it would fail after 10 minutes.

Btw, I don't see you as noob

[mikeyakame]

I have a theory which may be likely where vTT is partially set based on the chipset. For example the 975 express chipset uses a base vTT of 1.30v probably due to being build on 130nm? node, the p35/x38/x48 are built on 90nm node and use vTT of 1.20v, while the p45 is built on 65nm node and uses a vTT of 1.10v iirc. This may explain certain uses of different gtl reference multiplier by mobo manufactuers to accomodate different process node sized chips on a wider scale. Who knows really, you'd have to ask Asus, DFI, Gigabyte, etcs bios engineers for the exact reason cause I for one sure don't have a credible answer there.

One thing I can be certain of though is that an applicable vTT for the chip is selected via the preassigned VID bits for the processor when running in Auto voltage at default clocks. VID sets vCC and vTT through its bit mask based on its original binning criteria, if you check Intel processor spec sheets you will find this information lies there.

Those 2 statements may seem to negate each other, but perhaps they are simply parts of a selection criteria for applicable final vTT through the bios code.

[DemonEyez]

I was wondering now we're on the subject of voltages and stuff :P In my bios I also have control over the "Clockgen Voltage" which defaults to 3.45v and maxes at 3.75v.

Earlier in this topic you talked about vNB better not be the same as vCU PLL, that's it better to have a gap between the two.
In my bios it's called SB Core/CPU PLL Voltage. So is it safe to say that the Northbrige and Southbridge voltage should not be the same.

And how does one start with tuning CPU and NB skew? Is it derived from the CPU bus speed for example?

Any insights would be very much appreciated.

[truehighroller]

I think the orginal poster should just say start here with 45nm and start here with 65nm because I read all the technical stuff and just don't care any more. I just want to know where to start with my Q6600...

[Amurtigress]

truehighroller:

If you want to get somewhere, get to know what you are dealing with. There is no success without knowledge....be a real overclocker or leave it be. How do you want to draw your own conclusions if you don't want to know about what's going on on the lowest level while doing low level changes in the BIOS!?

...blind people talking about colors.

Mikeyakame:

I can tell from what you posted that you know your technology. I have had an electronics education myself and I have my devices here like DMM and a solder station of the more expensive kind, but so far I had no ways of taking a look at the in-depth specifications of Intel CPUs. My instinct led me to the following settings:

Q9300 C1/M1 on a ASUS P5Q3 Del. board (P45 based)

FSB: 460, CPU: 3450
VTT: 1.36V
VCore: 1.32V
GTL ref 0/2: 0.69x
GTL ref 1/3: 0.67x
CPU PLL: 1.58V
NB Volt: 1.28V

This puts my GTL references perfectly in the 0.8-1V range. 0.69x1.36V=0.9384V. Tho I wonder if I should try to get closer to 0.90 to have a safety margin.

During my testing phase of this new setup I had some latent instabilities that showed me how well towards stability I was when I was changing the GTL references by giving me more or less frequent calculation errors in Prime 95. This only resolved completely by putting the PCIe frequency from 100 to 101! 101 MHz gives me a way greater range of stable settings.

I've had a lot of hassle with this board, partially due to stupid memory timing defaults that ASUS seems to have taken STRAIGHT from the DDR2 versions of this board.

I seem to be hitting a FSB wall now at a bit above FSB 460.

Is there a way to circumvent it? I have had cases where I could boot the system at 480, but even more should be possible for my chip.

That's it for now.

Keep rolling folks, this is one of the most competent threads I've seen here!

[truehighroller]

truehighroller:

If you want to get somewhere, get to know what you are dealing with. There is no success without knowledge....be a real overclocker or leave it be. How do you want to draw your own conclusions if you don't want to know about what's going on on the lowest level while doing low level changes in the BIOS!?

...blind people talking about colors.

Don't talk down to me. I know alot about overclocking. I want a starting point referenced idiot.

[mikeyakame]

truehighroller:

If you want to get somewhere, get to know what you are dealing with. There is no success without knowledge....be a real overclocker or leave it be. How do you want to draw your own conclusions if you don't want to know about what's going on on the lowest level while doing low level changes in the BIOS!?

...blind people talking about colors.

Mikeyakame:

I can tell from what you posted that you know your technology. I have had an electronics education myself and I have my devices here like DMM and a solder station of the more expensive kind, but so far I had no ways of taking a look at the in-depth specifications of Intel CPUs. My instinct led me to the following settings:

Q9300 C1/M1 on a ASUS P5Q3 Del. board (P45 based)

FSB: 460, CPU: 3450
VTT: 1.36V
VCore: 1.32V
GTL ref 0/2: 0.69x
GTL ref 1/3: 0.67x
CPU PLL: 1.58V
NB Volt: 1.28V

This puts my GTL references perfectly in the 0.8-1V range. 0.69x1.36V=0.9384V. Tho I wonder if I should try to get closer to 0.90 to have a safety margin.

During my testing phase of this new setup I had some latent instabilities that showed me how well towards stability I was when I was changing the GTL references by giving me more or less frequent calculation errors in Prime 95. This only resolved completely by putting the PCIe frequency from 100 to 101! 101 MHz gives me a way greater range of stable settings.

I've had a lot of hassle with this board, partially due to stupid memory timing defaults that ASUS seems to have taken STRAIGHT from the DDR2 versions of this board.

I seem to be hitting a FSB wall now at a bit above FSB 460.

Is there a way to circumvent it? I have had cases where I could boot the system at 480, but even more should be possible for my chip.

That's it for now.

Keep rolling folks, this is one of the most competent threads I've seen here!

Thanks for the support!

Good to see there's still some people out there who want to know how things work to make things work! It's the fundamental I practice. If I don't understand something in particular I go to extended lengths to figure it out so at the end I can make my life easier and find answers to my questions, so I can pose more!
I have somewhat of an electrical background, my fathers an Electrical Engineer and I just enjoy electronics and mechanics of things! So I suppose I spend a lot of time doing all fun kinds of things from working on cars to working on computers and gizmos!

Now with your question. From my understanding the hole at 455-475ish requires a lot of GTL reference voltage to overcome the extremely small area under the curve that exists at nominal voltage. Your method of fine tuning GTL Ref voltage via Prime95 is spot on, and honestly thats exactly how I do it too. Adjust the multiplier based on calculation error frequency. The later the calculation error occurs the closer to ball park the GTL Reference voltages / multipliers is to nominal for that FSB and CPU frequency.

I've played with a DDR2 P5Q deluxe and it wasn't fun and games, it was one of the most unpredictable and poorly behaved chipsets I've encountered (p45 I mean), but it's by no means not without possibility of being tamed. Just really really irritating for the most part. I'd hate to be dealing with the DDR3 version, but its not so much Asus' fault (except for the horrible memory support) but rather Intel being stuck writing bios code to support speeds way above design specs due to user demand.

At 450MHz on the P5Q I know I was using 0/2 = 0.73x, and 1/3 = 0.715x for a 65nm chip with around 1.34v vTT, might have been 1.36v but its been a while. Perhaps try starting around there and see how that goes, I know this seems quite high but the idea is to move the crossing point where the rising and falling edges of bclk0 and bclk1 meet way above where they nominally should be, to a region where a valid logical 1 can be determined.

It would raise the value of vIH (input high), also raising vOH(output high), and since vIL(input low) and vOL(output low) can exist anywhere from 0.3v below the crossing threshold to 0v for a logical 0 there is plenty of area to determine the switch to ground for a logical 0. Raising vIH on the other hand means from my understanding that since ringback will always occur and at the point of the FSB hole you can shift the ringback margin inside the crossing threshold where it won't interfere in determining a valid logical 1. The excessive ringback occurs due to the amplitude of the voltage? required for those base clock frequencies from what I understand. I still have some gaps but I guess I try and do my best to explain it so it makes sense! Hell even I get confused explaining electrical theory sometimes! It's brain racking!

Any questions just shoot and feel free to drop ideas, theories and knowledge we can use to correlate all the data! The more gaps we fill the better we can get the understanding across to more people!

Thanks again for the kind words!

Edit: I also find PCIE freq of around 101-103mhz or 110-114mhz to work better at above 400FSB too. Not sure why exactly but its possible it's something to do with the way the frequency is divided from the base clock and multiplied from the analog PLL clock generator.

[mikeyakame]

Don't talk down to me. I know alot about overclocking. I want a starting point referenced idiot.

Whats with the poor attitude mate. You don't come in here demanding and throwing your arms around. Very poor form for such a "knowledged overclocker". Some of us started with no knowledge available and had to figure it all out on our own, and then share it with others to make it easier for them. Then some just expect it to all be available to them with little to no effort while maintaining zero gratitude or manors. It's just really poor form and I'm not talking you down, I'm simply telling you how it is.

Starting point is 0.667x multiplier. If you'd bothered to take the time to read what myself and others have been discussing you would have figured this out. Yeah forums are a mess but that's their nature.

[truehighroller]

truehighroller:

If you want to get somewhere, get to know what you are dealing with. There is no success without knowledge....be a real overclocker or leave it be. How do you want to draw your own conclusions if you don't want to know about what's going on on the lowest level while doing low level changes in the BIOS!?

...blind people talking about colors.

That is the poor attitude buddy.

[mikeyakame]

I think it's just a misunderstanding in general. It's better to both leave it at that. This being a technical thread there shouldn't be flaming on anybody's part, after all we just want to share knowledge and help each other, flaming and name calling doesn't fit the bill no matter how you look at it aye. This is one of those areas there isn't always a base value besides what the data we have tells us. From there it's all trial and error and learning. I've also discussed in many posts the ball park settings I use with my Q6600, that is why it's good to read even if you don't care cause the thing is I do care and took the time to post the settings I use even if they are annoying from your stance to find amongst all the clutter.

[truehighroller]

I just thought it would be good idea to post a good starting point. I have helped alot of people by sharing my settings that I use and just fealt this is what this thread is about and starter settings might be good idea that's all. I agree to drop it as well.

[CryptiK]

As I said in the first post, most people start with 0.667x Vtt for 65nm and 0.635x Vtt for 45nm, however intel state use 0.667x for both processes, so I'd simply advise anyone to start with 0.667x, and work from there. Some have found lowering it to work best, others need to raise it.

As Mikey said, try to aim for around 0.80v, so that means as you raise/lower Vtt, you must adjust the GTL Ref multiplier to maintain the desired target voltage. If you have a BIOS that has the fixed multiplier and the Y value, use the formula and add or subtract the appropriate Y value to give the desired output voltage.

EDIT - I also use the Orthos calculation error frequency to tune GTL Refs, it seems to be the best way of doing it.