• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Request for Help 4.6 on FX8350

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
Thanks guys, I really appreciate all the help.

Attached screenie is 2 hours into P95 Blend. I finally got it stable at 4.5ghz on the new board by setting the vcore to 1.375 (although it goes straight to 1.392v and stays there for the duration of P95 test). I am using bios 803 and Chipset 901. Other settings are:

CPU/NB voltage: 1.25
CPU VDDA: 2.5
NB voltage: 1.113 (Auto)
NB HT voltage: 1.214 (Auto)
NB 1.8v: 1.8 (Auto)
SB voltage: 1.1 (Auto)
VDD PCIE: 1.1 (Auto)
VDDR: 1.2 (Auto)

CPU Clock Multi: 22.5
Bus: 200
PCIE Bus: Auto
Memfreq: 1600
CPU/NB freq: 2400
HT freq: 2400
CPU Spread Spectrum: Disabled
PciE Spread Spectrum: Disabled
EU Powersave: Disabled

CPU LLC: Ultra High
CPU/NB LLC: High
CPU Current Capability: 130%
CPU/NB Current: 130%
CPU Power Phase: Extreme (do I really need this?)
CPU Volt Freq: Auto
CPU Power Duty: T.Probe
CPU Power Response Control: Auto
CPU/NB Power Response Control: Auto
CPU Thermal Power: 130
DRAM Current Capability: 100%
DRAM Voltage Frequency: 300
DRAM Power Phase Control: Auto

Everything on the advanced CPU tab (core6, apm etc) are disabled...

The little fan on the socket is doing great.

Only thing I did notice was CPU occasionally drop from 4.515 to 4.488 for a split second before going back up again - this was due to bus speed dropping to 199ish.

Does this look OK?
 

Attachments

  • sabo4.5hwmon.png
    sabo4.5hwmon.png
    25.3 KB · Views: 122
  • sabo4.5cpuid.png
    sabo4.5cpuid.png
    20.9 KB · Views: 121
Last edited:
One thing I noticed comparing this to the others is you have the NB speed at 2400 insted of 2200 and your running the volts on auto where in the other screens you were running 1.25v to the CPU_NB. This may have something to do with the inconsistency in your voltage settings.
 
Thanks Johan, Good idea but I already tried equating NB ant HT freq at 2000 at the lower vcore yesterday and it failed. Btw, this board has LOADS of settings! lol and I am a bit confused! I can see 3 voltages which mention the northbridge:

'CPU/NB voltage' - I have set to 1.25v manually.

'NB Voltage' - which is set to 'auto' but is showing 1.113 next to it - should I set to 1.25 as well?)

Then there is 'NB HT Voltage' - which is set to 'auto' but shows 1.214 next to it - should this be 1.25 again?)

Thanks again.
 
'CPU/NB voltage' - I have set to 1.25v manually. Good

'NB Voltage' - which is set to 'auto' but is showing 1.113 next to it - should I set to 1.25 as well?)
No leave it on auto

Then there is 'NB HT Voltage' - which is set to 'auto' but shows 1.214 next to it - should this be 1.25 again?) See response #2
 
OK, I think I am getting somewhere. Looks like Johan was right. I think the LLC on this board is more punchy than the Evo. eg for a given overclock speed, 'High LLC' on the Sabretooth is giving me the same idle vcore readings in CPUID as 'Ultra High' was giving me on the Evo.

Just a few .... more .... hours .... of ...........Lucas ......... aghhhhhh!
 
I think the LLC on this board is more punchy than the Evo. eg for a given overclock speed, 'High LLC' on the Sabretooth is giving me the same idle vcore readings in CPUID as 'Ultra High' was giving me on the Evo.

I expect that is exactly right. Heck, after all it is a better board with closer tolerances on the VRMs. I would not expect the same from one board to the next. Often I used to get duplicates of the same model of board when I worked for a mobo maker and seldom did they hold spot on from board to board. There was always that few thousands difference in voltage or so it seemed.
RGone...
 
Well, I seem to be getting stable at 4.6 on 1.4125 vcore which peaks out at 1.43v and idles on 1.416v
.
... but ... :shock: check out the socket temp on HWMon :shock:

I was sort of half paying attention whilst listening to the radio and then noticed it and took a double take. This has to be a glitch? Anyone else had this? The current value for the socket was near the core temp at around 51-51 when I noticed. The screenie was taken after hitting the STOP! button. Can't have been at that value for long if at all. All working fine and no reboot. All fans still spinning and no evidence of anything...

Oh well, thats it for today! I'll see if I've screwed it tomorrow...
 

Attachments

  • hotsocket.png
    hotsocket.png
    23 KB · Views: 113
  • sockethot2.png
    sockethot2.png
    21.3 KB · Views: 113
  • sockethot3.png
    sockethot3.png
    29.9 KB · Views: 112
Yeah, looks like HWMon is glitchy. That scared me a bit! This time around it showed exactly double the running vcore for 'vcore max', which is too much of a coincidence.

Probably something to do with the fact that I swapped out the board with no re-install apart from mobo drivers. Anyway, no harm done. Fully 2hr blend stable at 4.6 again.

Bios vcore 1.4125, LLC ultra, idles at 1.416 (which seems a bit heavy), peak draw is 1.428 and it sits there throughout the test. Now going to up the northbridge back to 12* and start tweaking it a bit.

Thanks for all the help.
 

Attachments

  • sabo4.6hw.png
    sabo4.6hw.png
    25.1 KB · Views: 104
  • sabo4.6p95.png
    sabo4.6p95.png
    27.1 KB · Views: 104
  • sabo4.6cpu.png
    sabo4.6cpu.png
    22.7 KB · Views: 103
Last edited:
"twocents" let me describe CPU_LLC for you.

1. Set Cpu voltage in bios.

2. Make a choice of LLC setting so that the overshoot voltage is NOT too high when load comes on. So that under load the cpu is hopefully having the voltage you set for Vcore in bios menu.

3. So when you say the idle voltage seems a little heavy it is not LLC, make that a well working LLC's fault.

4. Now many of us have gone ahead and followed my setup offset Vcore method and on my CVH, I haul booty at 4.7Ghz and idle at 1600Mhz with 4.7Ghz voltage at 1.43V and Idling doing net surfing (no real load) at 0.986V. Doing so now on air instead of my heavy duty water and I was even able to drop from Ultra High CPU_LLC to only High. But it takes some tweaking with trial and error to get it spot on as I finally did do.
RGone...
 
Thanks for explaining RGone.

I was trying to get my head around this LLC business. I got a bit lost with all the technical definitions and was just viewing it as a kind of vcore 'slack' range to provide a little extra kick when needed. I assumed it was upside range only.

So is my LLC setting of ultra high just masking the fact that my bios vcore is really a tad low for that clock:confused:

I am reading up about offset and will give it a go soon.

Thanks again.
 
"twocents" let me describe CPU_LLC for you.

1. Set Cpu voltage in bios.

2. Make a choice of LLC setting so that the overshoot voltage is NOT too high when load comes on. So that under load the cpu is hopefully having the voltage you set for Vcore in bios menu.

3. So when you say the idle voltage seems a little heavy it is not LLC, make that a well working LLC's fault.

4. Now many of us have gone ahead and followed my setup offset Vcore method and on my CVH, I haul booty at 4.7Ghz and idle at 1600Mhz with 4.7Ghz voltage at 1.43V and Idling doing net surfing (no real load) at 0.986V. Doing so now on air instead of my heavy duty water and I was even able to drop from Ultra High CPU_LLC to only High. But it takes some tweaking with trial and error to get it spot on as I finally did do.
RGone...
I use your method & it is doing real good that way. My LLC is set on high also.
Thanks RGone
 
I use your method & it is doing real good that way. My LLC is set on high also.
Thanks RGone

Yes it takes a little tweaking to get "just" right which is what I need when the jump is from 1.6Ghz to 4.7Ghz and a "load". But once done...I just about forget about manual setting anymore. I just look for what is neeed to get offset correct. It is the use of offset that allows the voltage to drop so far when doing mostly nothing at low cpu speed. Cool n Quiet could do similar but it jumps all over the landscape to me and just not so pretty to watch.
RGone...
 
Yes it takes a little tweaking to get "just" right which is what I need when the jump is from 1.6Ghz to 4.7Ghz and a "load". But once done...I just about forget about manual setting anymore. I just look for what is neeed to get offset correct. It is the use of offset that allows the voltage to drop so far when doing mostly nothing at low cpu speed. Cool n Quiet could do similar but it jumps all over the landscape to me and just not so pretty to watch.
RGone...
Well I'm not at your level of expertise so I use Cool N Quiet. It works pretty good & you are right it does jump around some.
 
OK, so I have been reading about LLC, fiddling around with my 4.5ghz profile and trying to follow RGone's advice:
1. Set Cpu voltage in bios.

2. Make a choice of LLC setting so that the overshoot voltage is NOT too high when load comes on. So that under load the cpu is hopefully having the voltage you set for Vcore in bios menu.

3. So when you say the idle voltage seems a little heavy it is not LLC, make that a well working LLC's fault.

Most of the stuff I found online made for pretty dry reading but this gives a not too technical overview for anybody interested:

http://linustechtips.com/main/topic/24019-load-line-calibration-why-overclockers-should-care/

(ps this forum gets a mention in 'The Book of Overclocking - Tweak Your PC to Unleash It's Power'!)

So I tried to find the BIOS vcore that would support a similar overclock without relying so heavily on LLC. Figured if I am going to go higher (and hopefully eventually the offset route) then I don't want any overshoots to accidentally kill the chip!

Took max vcore shown in HWMon which I know works for stable 2 hr blend (1.392). Put into BIOS manually and lowered CPULLC to 'High'. No good. Rounded up BIOS vcore to 1.4 for good measure and lowered CPULLC to Zero. No good. Left BIOS vcore at 1.4 and set CPULLC to 'High'. Bingo. Stable 2 hour blend.

So now I have 2 stable profiles for 4.5ghz compared in table.

So the High LLC profile gives me less overshoot from what I specified in BIOS but the absolute values are higher at both idle and max. On 'average' the voltage appears lower under constant load. I am guessing that the second profile is a better foundation for going down the 'offset route'?

4. Now many of us have gone ahead and followed my setup offset Vcore method and on my CVH, I haul booty at 4.7Ghz and idle at 1600Mhz with 4.7Ghz voltage at 1.43V and Idling doing net surfing (no real load) at 0.986V. Doing so now on air instead of my heavy duty water and I was even able to drop from Ultra High CPU_LLC to only High. But it takes some tweaking with trial and error to get it spot on as I finally did do.
RGone...

Thanks, I have been playing around with offset with no real progress so far. Tried subtracting VID shown in Coretemp (which never changes) from Manual known stable vcore in BIOS = no joy. Tried subtracting idle vcore in BIOS from max shown in HWMon for known stable config = no joy. The only other thing I did see were a bunch of VID codes in the date recorded by HWMon but I don't think they will help. They are:

22.5x 1.325v
21x 1.425v
20.5x 1.4v
17x 1.225v
10.5x 1v
7x 0.85v

I would be grateful if somebody could point me in the direction of link to the method described. I have searched around and can't seem to find it.

Thanks again.
 

Attachments

  • 4.5comparison.png
    4.5comparison.png
    10.1 KB · Views: 77
Yes the Sabertooth LLC on high your Vcore will drop .012 and on Ultra it overshoots .012
I take the .012 droop over the overshoot and on Extreme it really overshoots
 
the fan does not need to move much air, just enough to get the air moving back there.
I run mine as exaust and you can't even feel the flow but it keeps the heat from building up in that small, enclosed space.
 
Back