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

[Cibic]

Yes but isn't the idea that because of vdroop, with LCC Enabled you can have it set lower than with it disabled?

Enabled: CPU Volts set to 1.4V in Bios, 1.36V in CPU-Z
Disabled: CPU Volts set to 1.4V in Bios, could dip beneath 1.36V under load and therefore unstable?

That was my understanding of it.Cibic thanks for the awesome reply, hopefully tomorrow evening I'll get some spare time, read over it properly, take some notes and start some intensive testing ^^

Asus Striker II Extreme with LLC Disabled has a vdroop of 0.05v so if you want your cpu at 1.36 on full load you really want it at 1.41.
By the way, ram at high speeds are most likely to perform at design specs with dual cores. On quad cores you should relax timings and increase voltage to sustain operation, or relax timings just a bit, increase voltage a notch or two and downclock it. Either way, it should be about the same performance. Nvidia chipsets get the most performance with 1:2(or 1:1 with ddr2) ratio, 1900FSB linked and synced with my ram, so I can't decrease ram frequency.

Wish I had someone to help me when I got this MB, lets hope I save you some frustration.

[powellandy]

Has anyone tried using linpack to tune NB GTL and skews. I tried the following and wonder if it is a valid method and whether the results are repeatable.

Stable at 450Mhz FSB on Q9300 (specs below)
CPU GTL tuned as here - http://67.90.82.13/forums/showpost.p...&postcount=425

CPU skew / NB skew / NB GTL - 3 linpack Gflops (vista x64 max memory - about 3070MB)

A/A/+40 - 46.4 46.9 46.8
A/A/+20 - 47.3 47.5 47.4
A/A/A - 47.2 47.5 47.4
A/A/-20 - 46.4 46.9 46.7
A/A/-40 - 46.8 46.9 46.8

So opted for NB GTL + 20

-100/N/+20 - 47.2 47.5 47.4
N/-100/+20 - 47.3 47.4 47.3
-200/-100/+20 - 47.2 47.4 47.3
-400/-300/+20 - 47.4 47.5 47.4

So opted for -400/-300.
Just finished a 12h OCCT RAM (Large) run and it is stable.

Any thoughts? Are there any other combos of skews you would try?
Andy

[mikeyakame]

Cryptik,

Bought myself a DFI LT-X48 T2R as a spare board while my Rampage is getting unbroken and thought what the hell I'd have a play with it.
I'm in love here Taken me 3 days of a bit of playing around and finally got 400mhz FSB stable heh. GTL adjustment is simply awesome though. Printed myself out a copy of the VTT/GTL value map tables DFI put together for reference because god knows I could ever remember which value means what at which Vtt.

Fine tuning is pretty damn amazing though. 0.008v GTL fine adjustment values mapped all the way from 1.06v Vtt to 1.60 or 1.70v Vtt. The tables themselves cover roughly 0.533x to 0.780x if I recall, don't have the desire to rip out a calculator to confirm that heh.

http://dfics.dfi.com.tw/dfi_cs/LTX48/X38X48_vtt.xls

If you're interesting in dropping your jaw take a look at that excel spreadsheet. It's definitely jaw crashing to the floor worthy

Also explains what the hell all the numbers mean I always wondered myself, now it's perfectly clear to me. And the adjustment is flawless in all aspects. CPU 0/3 and 1/2 are paired which is fine, and NB is on its own on the LT X48 I got.

[CryptiK]

Nice, I think I'm getting a DK-T2RSB Plus to have a play with. Been tossing the idea around for a couple of months, just have to make the final decision.

[mikeyakame]

DK lacks the adjustability of the LT, I compared both and went for the LT. The price difference was only $30aud or so, it wasn't very much. The LT I found to be the much better buy, as it also has a 8 phase digital volterra VR and a pretty cool northbridge cooler made by thermalright. You can attach an 80mm fan to the back of it vertically. The UT model had a silly heatpipe joined cooling assembly and probably not worth the extra cost.

[CryptiK]

Yeah the LT does have 8 phase, but the DK has a better clockgen, it uses the same one as one the Rampage Extreme (ICS 9LPRS918JKLF) which is the reason the DK can hit higher FSB than either the UT or the LT (DFI actually said that). The DK still has 6 phase digital PWM with 18 fets so its not too much worse in that department.

I do wish they just modified the LT to have the same clockgen as the DK. It would be good if you had a nice E0 wolfdale to test the LT with.

EDIT - the DK (when 'advanced' BIOS mode is selected) has a large amount of options, at least all the ones Ive seen you mention the LT has. Perhaps the LT has more though, I haven't seen bios SS's from it, nor do I own one.

[Falkentyne]

People, having Loadline Calibration Enabled (or any other vdroop reducer or eliminator) reduces CPU life, CPU needs more voltage and strains system, in my case causes instability in overclocking. I was thinking of enabling it but since its a bad thing I will not.
Full report here http://www.anandtech.com/cpuchipsets...spx?i=3184&p=6



Having a higher voltage on non cpu use will not raise your temperatures, on load though lets say 1.42 LLC Disabled does less heat from 1.42 LLC Enabled.

The hysteria over loadline is nothing but that..hysteria. It IS true that vdroop is there for a reason, but this is supposed to be for cpus running at *STOCK* voltages and speeds. Yes, Intel knows we overclock; heck they're even overclocking now (and as everyone knows they have taken a more overclock friendly approach than the hardline approach from years ago; witness the extreme editions--who would want an unlocked multiplier EXCEPT overclockers? It's completely absolutely USELESS for a normal consumer).

While it is true that droop is supposed to prevent voltage spikes on load/idle changes, there is still absolutely zero proof anywhere that cpus have been damaged by it. The only warnings have occured from loadline causing -overvolting- at load instead of matching the BIOS settings. There was a huge recent thread about this, which turned out to be nothing but a bunch of hot air and flaming. Oh and keep in mind that vdroop only existed starting with the pentium 4 line. There was no noticable vdroop (at least nothing of this magnitude) on the pentium 3 chips; and those chips used -higher- vcores (2.0 for old P3; 1.75v for coppermines) than what we have now.

I will also tell you this--from personal experience. High voltage does cause degradation, far more so than heat, although each chip responds differently. And it is MORE dangerous to set a higher idle voltage without loadline or vdroop mod, than it is to set a lower voltage and using loadline/vdroop mod, as long as your load voltage doesn't go higher than idle. Voltage is the #1 cause for degradation followed by temps. I had degradation *just* from high idle voltages on a P4 northwood by putting 1.7v idle into it and leaving the computer on idle constantly. Load voltage was 1.62v, a lot lower. And it wasn't the load voltage that caused GNDS here, it was the idle voltage. If there had been a non soldering vdroop mod back then (at least that I had known about and been able to use), my chip would have lasted alot longer without reducing in clocks so sharply.

My X6800 also reduced in clocks (tho not as much as the P4's) by needing 1.55v for 3.7 ghz-which required 1.625v set in BIOS for idle. It was the high idle voltage that hurt the chip; it lost about 100 mhz; needed about 0.05v more. 1.55v idle would have been a LOT easier on the chip. I would have much rather taken the chance with random spikes, than having this happen! And even 1.55v idle was too high; would have been best to have the vcore at 1.5v max (idle) on air.
Subzero completely changes the game.

BTW I didn't know anything about vdroop mods when I had the X6800 either, but my P5W is now vdroop modded and my QX chip hasn't been damaged anywhere near what the X6800 without mod was; there has been very light degradation (something along the lines of 0.01v more needed) but that's it, and I don't keep it at 4 ghz unless I need it anyway.

[BababooeyHTJ]

The hysteria over loadline is nothing but that..hysteria. It IS true that vdroop is there for a reason, but this is supposed to be for cpus running at *STOCK* voltages and speeds. Yes, Intel knows we overclock; heck they're even overclocking now (and as everyone knows they have taken a more overclock friendly approach than the hardline approach from years ago; witness the extreme editions--who would want an unlocked multiplier EXCEPT overclockers? It's completely absolutely USELESS for a normal consumer).

While it is true that droop is supposed to prevent voltage spikes on load/idle changes, there is still absolutely zero proof anywhere that cpus have been damaged by it. The only warnings have occured from loadline causing -overvolting- at load instead of matching the BIOS settings. There was a huge recent thread about this, which turned out to be nothing but a bunch of hot air and flaming. Oh and keep in mind that vdroop only existed starting with the pentium 4 line. There was no noticable vdroop (at least nothing of this magnitude) on the pentium 3 chips; and those chips used -higher- vcores (2.0 for old P3; 1.75v for coppermines) than what we have now.

I will also tell you this--from personal experience. High voltage does cause degradation, far more so than heat, although each chip responds differently. And it is MORE dangerous to set a higher idle voltage without loadline or vdroop mod, than it is to set a lower voltage and using loadline/vdroop mod, as long as your load voltage doesn't go higher than idle. Voltage is the #1 cause for degradation followed by temps. I had degradation *just* from high idle voltages on a P4 northwood by putting 1.7v idle into it and leaving the computer on idle constantly. Load voltage was 1.62v, a lot lower. And it wasn't the load voltage that caused GNDS here, it was the idle voltage. If there had been a non soldering vdroop mod back then (at least that I had known about and been able to use), my chip would have lasted alot longer without reducing in clocks so sharply.

My X6800 also reduced in clocks (tho not as much as the P4's) by needing 1.55v for 3.7 ghz-which required 1.625v set in BIOS for idle. It was the high idle voltage that hurt the chip; it lost about 100 mhz; needed about 0.05v more. 1.55v idle would have been a LOT easier on the chip. I would have much rather taken the chance with random spikes, than having this happen! And even 1.55v idle was too high; would have been best to have the vcore at 1.5v max (idle) on air.
Subzero completely changes the game.

BTW I didn't know anything about vdroop mods when I had the X6800 either, but my P5W is now vdroop modded and my QX chip hasn't been damaged anywhere near what the X6800 without mod was; there has been very light degradation (something along the lines of 0.01v more needed) but that's it, and I don't keep it at 4 ghz unless I need it anyway.

This is something that I have been wondering about for a while. Wouldn't voltage spikes become MORE dangerous while overclocking, in theory? Im just wondering weither a higer voltage is more dangerous at idle while the chip is drawing very few amps or if its more dangerous to have a slightly higher voltage while the chip is under load while drawing much more amperage. Then I am hearing reports of loadline overvolting while on full load when tested with a DMM on Asus 16 phase boards. If you combine that with voltage spikes from overshoot (Im not sure if Im using the appropriate term here) it sounds a little more dangerous to me, but vdroop is also huge on my board 0.05 vcc from idle to load with cpu-z.

[Cibic]

The hysteria over loadline is nothing but that..hysteria. It IS true that vdroop is there for a reason, but this is supposed to be for cpus running at *STOCK* voltages and speeds. Yes, Intel knows we overclock; heck they're even overclocking now (and as everyone knows they have taken a more overclock friendly approach than the hardline approach from years ago; witness the extreme editions--who would want an unlocked multiplier EXCEPT overclockers? It's completely absolutely USELESS for a normal consumer).

While it is true that droop is supposed to prevent voltage spikes on load/idle changes, there is still absolutely zero proof anywhere that cpus have been damaged by it. The only warnings have occured from loadline causing -overvolting- at load instead of matching the BIOS settings. There was a huge recent thread about this, which turned out to be nothing but a bunch of hot air and flaming. Oh and keep in mind that vdroop only existed starting with the pentium 4 line. There was no noticable vdroop (at least nothing of this magnitude) on the pentium 3 chips; and those chips used -higher- vcores (2.0 for old P3; 1.75v for coppermines) than what we have now.

I will also tell you this--from personal experience. High voltage does cause degradation, far more so than heat, although each chip responds differently. And it is MORE dangerous to set a higher idle voltage without loadline or vdroop mod, than it is to set a lower voltage and using loadline/vdroop mod, as long as your load voltage doesn't go higher than idle. Voltage is the #1 cause for degradation followed by temps. I had degradation *just* from high idle voltages on a P4 northwood by putting 1.7v idle into it and leaving the computer on idle constantly. Load voltage was 1.62v, a lot lower. And it wasn't the load voltage that caused GNDS here, it was the idle voltage. If there had been a non soldering vdroop mod back then (at least that I had known about and been able to use), my chip would have lasted alot longer without reducing in clocks so sharply.

My X6800 also reduced in clocks (tho not as much as the P4's) by needing 1.55v for 3.7 ghz-which required 1.625v set in BIOS for idle. It was the high idle voltage that hurt the chip; it lost about 100 mhz; needed about 0.05v more. 1.55v idle would have been a LOT easier on the chip. I would have much rather taken the chance with random spikes, than having this happen! And even 1.55v idle was too high; would have been best to have the vcore at 1.5v max (idle) on air.
Subzero completely changes the game.

BTW I didn't know anything about vdroop mods when I had the X6800 either, but my P5W is now vdroop modded and my QX chip hasn't been damaged anywhere near what the X6800 without mod was; there has been very light degradation (something along the lines of 0.01v more needed) but that's it, and I don't keep it at 4 ghz unless I need it anyway.

I kind of agree, but I do this.

If without LLC my voltage is at 1.5 (so I'll get 1.45v) I leave it as is. If I do need more, then I enable LLC.

[Amurtigress]

Hello,

after having wondered about if I can manage to run my Q9550 E0 and the P45 NB at the stock GTL values after all after such a long time (Bought my CPU in October), I tried out what powellandy did, also after having read something along the lines from mikeyakame.

And, I was successful. I just felt that a narrow range of 0.66-0.69x for stability on all GTLs was not looking like the system felt overly well. Here are my results, old first:

Mainboard: ASUS P5Q3 Deluxe Wifi@n
FSB Clock: 450 MHz (x8.5=3825)
CPU VCore: 1.275V
DDR3 RAM Clock: 1350MHz (4x1 GB Corsair 1600 modules)
NB Volt: 1.4V-1.44V
FSB Term Volt: 1.36V
CPU GTL Ref (0/2): 0.66x-0.69x, using 0.68x
CPU GTL Ref (1/3): 0.66x-0.69x, using 0.68x
NB GTL Ref.: 0.66x-0.69x, using 0.68x
LLC: Enabled
CPU Clock Skew: AUTO
NB clock Sew: AUTO

New:

Mainboard: ASUS P5Q3 Deluxe Wifi@n
FSB Clock: 450 MHz (x8.5=3825)
DDR3 RAM Clock: 1350MHz (4x1 GB Corsair 1600 modules)
CPU VCore: 1.275V
NB Volt: 1.4V-1.44V
FSB Term Volt: 1.36V
CPU GTL Ref (0/2): 0.63x-0.70x, using 0.63x or 0.65x
CPU GTL Ref (1/3): 0.63x-0.70x, using 0.67x or 0.68x
NB GTL Ref.: 0.63x-0.70x, using 0.63x or 0.65x
LLC: Enabled
CPU Clock Skew: 500 ps delay
NB clock Sew: 400 ps delay

400/300 works, too, less than that tends to instability. Once again it shows that automatic settings aren't really good...

I am currently using 0.65/0.68 to not be on the razor's edge. These settings also take into account that ASUS is trying to keep the GTL references 0/2 and 1/3 apart...on top of this, the stable way higher GTL reference range indicates that this is a lot more of a stable setting than previously. However the previous settings without skews were tested 11 hours prime stable, too.

The best way of stabilizing a 45 nm quadcore above FSB430 appears to be to try and compensate with the clock skews first and leave GTL Ref. lines at defaults (not AUTO!), and then test the upper and lower limit of stable GTL settings. Then use a value that's in the middle, to be as far away from the limits as possible.

Translated into more analogue terms, this means the smaller the stable range of GTL references is, the closer the signal overshoot and undershoot are getting to the reference line. With it, the probability of one of them crossing the reference line increases drastically,and with it certainly the chance of having data corruption on the processor bus.

If you've got any questions or need my full Ai Tweaker settings, let me know.

[Cibic]

This thread is a joke, I can overclock more with all gtl refs left on auto :P

[Falkentyne]

This is something that I have been wondering about for a while. Wouldn't voltage spikes become MORE dangerous while overclocking, in theory? Im just wondering weither a higer voltage is more dangerous at idle while the chip is drawing very few amps or if its more dangerous to have a slightly higher voltage while the chip is under load while drawing much more amperage. Then I am hearing reports of loadline overvolting while on full load when tested with a DMM on Asus 16 phase boards. If you combine that with voltage spikes from overshoot (Im not sure if Im using the appropriate term here) it sounds a little more dangerous to me, but vdroop is also huge on my board 0.05 vcc from idle to load with cpu-z.

That's a very good question but I'm not sure. The loadline overvolting thing seems to be on Asus' newest boards with new bioses; apparently they didnt get overvolting on older bios (this on a core i7 board; I don't know about the p45 or x58). But if Pencil/hardware mods work properly, then people may do that instead if Asus broke loadline.

Anyway if loadline/vdroop mod is working as it should, it's much better to use it than to not use it, since otherwise you have to raise the idle vcore significantly to compensate for vdroop, and high idle voltage *can* (not saying will, but *can*) degrade a CPU. It's far better to have temporary spikes to what that idle voltage would be, without a vdroop mod, when using a mod/proper loadline, and lower idle voltage, since the CPU is exposed to less constant higher voltage. example: 1.35v idle/1.35v load with spikes to 1.41 is a lot safer than 1.41v idle/1.35v load, since the cpu is idle a LOT more than load (unless you fold 24/7 or are one of those rare people that always have a game running). Someone on ocforums posted a very clear graph explanation of this not that long ago (you can find it in a search there).

[BababooeyHTJ]

That's a very good question but I'm not sure. The loadline overvolting thing seems to be on Asus' newest boards with new bioses; apparently they didnt get overvolting on older bios (this on a core i7 board; I don't know about the p45 or x58). But if Pencil/hardware mods work properly, then people may do that instead if Asus broke loadline.

Anyway if loadline/vdroop mod is working as it should, it's much better to use it than to not use it, since otherwise you have to raise the idle vcore significantly to compensate for vdroop, and high idle voltage *can* (not saying will, but *can*) degrade a CPU. It's far better to have temporary spikes to what that idle voltage would be, without a vdroop mod, when using a mod/proper loadline, and lower idle voltage, since the CPU is exposed to less constant higher voltage. example: 1.35v idle/1.35v load with spikes to 1.41 is a lot safer than 1.41v idle/1.35v load, since the cpu is idle a LOT more than load (unless you fold 24/7 or are one of those rare people that always have a game running). Someone on ocforums posted a very clear graph explanation of this not that long ago (you can find it in a search there).

I ended up using loadline on my board because my vdroop is HUGE with my Yorkfield. It's something like 0.08 (prime load) and 0.09 (IBT load). Oddly it was half that with my wolfdale. At that point I was much better of with it enabled. I figure the 16 phase power distro for the cpu on my board should help minimize any spikes.

[ci2kla]

IMO I didn't found any clue that enabling LLC would degrade a CPU. I was playing with Asus Rampage Extreme these 3 months and I always have LLC enabled to prevent significant vdroop.
Yesterday I was accidentally have LLC disabled (E8500 @1.2v, 500x8)
Then when I tried to IBT it the vCore droop from 1.177~1.194 (By MB sensor, never trust CPU-Z) to around 1.134, of course wierd behaviour appeared / unstable / random reboot due to insufficient vCore....

[mikeyakame]

IMO I didn't found any clue that enabling LLC would degrade a CPU. I was playing with Asus Rampage Extreme these 3 months and I always have LLC enabled to prevent significant vdroop.
Yesterday I was accidentally have LLC disabled (E8500 @1.2v, 500x8)
Then when I tried to IBT it the vCore droop from 1.177~1.194 (By MB sensor, never trust CPU-Z) to around 1.134, of course wierd behaviour appeared / unstable / random reboot due to insufficient vCore....

It shouldn't damage CPU any. Could just be you never realized how much droop there is under load without LLC. I know I completely forgot until I disabled Loadline calibration. It's amazing how much of a difference it makes.

[Alien Grey]

It shouldn't damage CPU any. Could just be you never realized how much droop there is under load without LLC. I know I completely forgot until I disabled Loadline calibration. It's amazing how much of a difference it makes.

I do much better with LLC disabled. I need higher voltage under full load with LLC enabled to have the CPU stable. I don't mind that the CPU voltage is a bit higher with LLC disabled when my CPU idles this won't kill it.

[mikeyakame]

I do much better with LLC disabled. I need higher voltage under full load with LLC enabled to have the CPU stable. I don't mind that the CPU voltage is a bit higher with LLC disabled when my CPU idles this won't kill it.

Well if anything LLC reduces idle voltage. You need much higher idle voltage without regulating Vcc against the load line. But regulating against the load line strains the VR a hell of a lot more and produces a lot more heat across the voltage regs surface area.

[Alien Grey]

Well if anything LLC reduces idle voltage. You need much higher idle voltage without regulating Vcc against the load line. But regulating against the load line strains the VR a hell of a lot more and produces a lot more heat across the voltage regs surface area.

I see no benefit running it with LLC enabled. My CPU is always on full load and with LLC disabled I can run with lower voltage and less heat. Uhmmm... less heat if I was using a CPU cooler on air.

[keithallenlaw]

I read this thread all the way through and still cant see what determines a positive or negative GTL. I'm not an Algebraic person so that's probably my loss.
My current vTT is 1.378. I have all the references set to +50 only because I seen someones exact set-up as mine. My vNB is 1.56 and set to +40GTL.
I have Asus RE & q9650 on H2O. Thanks for help and yes I'm new here and yes this is my first build. Thanks! -keith

I'm an enabler.

[mikeyakame]

I read this thread all the way through and still cant see what determines a positive or negative GTL. I'm not an Algebraic person so that's probably my loss.
My current vTT is 1.378. I have all the references set to +50 only because I seen someones exact set-up as mine. My vNB is 1.56 and set to +40GTL.
I have Asus RE & q9650 on H2O. Thanks for help and yes I'm new here and yes this is my first build. Thanks! -keith

I'm an enabler.

FSB frequency and clock strength/timing/skewing determine positive or negative GTL. You need varying values depending on a lot of things. Sometimes you need higher, some times lower. Only way to calculate a base is to try values and see how they go. It just takes a fair whack of time to get it right. You'll get a feel after a while how the system and board responds to changes and different values, too high or too low or even just right.

Use windows boot up as a gauge.
I honestly use Windows Sidebar/Gadgets to take rough guess' on GTL references and clock skews

If they pop up with scroll bars something is screwed. If they take a while to load up something isnt right. Etc.

If system is running smoothly all my 5 gadgets pop up nearly same time The ones that require internet connectivy, ie ISP traffic stats and weather might take a touch longer because of ethernet adapter initializing.

Also I use Everest load up time to gauge as well. I run Everest with task scheduler at user login. If it takes aw hile to scan pci devices and load drivers something isn't right I can tell immediately. It should load up very fast with minimal delay

Another little hit I find is the timing of the Windows Startup Sound in relation to the login screen, if one happens before the other depending on how far apart they are tells me skewing or GTLs aren't right. It's up to me to figure out which one is causing it. Thats the fun of it. Nothing can tell you exactly the problem and how to fix it, only hinting at where it lies and it's up to you to use those hints to guide your adjustments

I'll give you a good piece of advice. The key to success lies in observation

[keithallenlaw]

Thank you for your insight. I was able to get this far, actually farther but I decided on my below settings for a 24/7. Your right, OC'ing takes time
and for me taking notes was very helpful. Still lost on the +/- thing. I will research. Thanks again. -keith

[radaja]

mikeyakame,i do something very similar to what you describe.after i make adjustemts to my OC i open prime95,then start blend,then i quickly click the test button and if the popup window is slow to appear or lags in any way,im pretty sure that somethings wrong and ill just stop the test and reboot and make my adjustments.but now that you mention the windows startup logon screen in relationship with the sound,i think you may have just saved me an extra step.thanks

[mikeyakame]

Thank you for your insight. I was able to get this far, actually farther but I decided on my below settings for a 24/7. Your right, OC'ing takes time
and for me taking notes was very helpful. Still lost on the +/- thing. I will research. Thanks again. -keith

THink of this way.

GTL References are calculated off Vtt. GTL Reference is a ratio or multiplier of Vtt.

So with offset voltages, you have a base GTL Reference multiplier for each GTLREF adjustment. NB is usually 0.63x, CPU 0/3 are usually 0.63x, CPU 1/2 are usually 0.67x.

What ever that multiplier * Vtt gives you in a voltage value is your set GTL Reference voltage for each.

All -/+ offset voltages are is that reference you calculated from the base + or - n[mV].

So say +50mV, with 1.30v Vtt, and NB base of 0.63x.

0.63 * 1.30 = 0.819V or 819mV

819mV + 50mV = 869mV

0.869 / 1.30 = 0.668

Therefore +50mV with 1.30Vtt and 0.63x base, changes the GTL Reference multiplier to 0.668x.

Simple as that

[Cibic]

Thank you for your insight. I was able to get this far, actually farther but I decided on my below settings for a 24/7. Your right, OC'ing takes time
and for me taking notes was very helpful. Still lost on the +/- thing. I will research. Thanks again. -keith

No problem man.

[keithallenlaw]

Thanks gentlemen. Very insightful. I'm actually thinking "leave well enough alone" where I'm at, which is basically my sig settings below.

vCore =1.356
pLL = 1.590
vTT = 1.378
vNB = 1.55
vdram=1.90 (kingston 2x2 hyper X 1600MHz @1350MHz 7-7-7-21

GTL ref. +50/+50/+50/+50 with NB ref. +40

LLC=enabled

These settings pass an Intel Burn @max load for 10 rounds with the cores reaching upper 60's Celsius.

My current 3DMarkVantage:



Maybe I need to offset cores 1 and 3 from 0 and 2.

My max stable on this build was 460x9 but I wasn't comfortable with the high voltages and upper 70's Celsius core temps with IBT.
But that test is rigid and doesn't apply to gaming. I truly expected a better bench and 3D score on H2O. I gave up and settled on a
24/7 set-up.

Thanks guys! -keith