Jump to content
Emilo

Overclocking 6-Core CPU on GA-EX58-UD5

19 posts in this topic

Recommended Posts

Note: Credits and personal thanks to d00d and others which contributed to the creation of the following topic:  GA-EX58 and GA-X58A DSDT native power management modifications

 

---- For this guide to work you must properly patch your own DSDT as per d00d's instructions from here.

       To have all Speed-Step frequencies available, you must modify the Device (PX40) section so that AppleLPC.kext loads as described in point 9 of d00d's guide.

       You can check if AppleLPC.kext is loaded on your system by going to About This Mac/System Report/Extensions/AppleLPC

       You can use my DSDT and config.plist, but only with GA-EX58-UD5 mobo on F13 BIOS after modifying SMBIOS and Graphics sections in my config.plist to reflect your System Definition & GPU.

       If you own different motherboard and have trouble editing DSDT manually, use DSDT patches by motherboard from olarila.com and apply them to your own DSDT with DSDT Editor.

 

 

If you want all C/P states working, you should have the following enabled in BIOS:

1. Intel® Turbo Boost ---- Enabled
2. CPU EIST Function ----- Enabled

3. Load Line Calibration -- Enabled

4. CPU Vcore must be set to “Normal
5. Dynamic Vcore (DVID) must be set to value optimal for your overclock ( +0.24375 value allowed me to reach 190 BCLK/4.54GHz overclock on X5670 & i7-980X )

 

   

      If you are not overclocking but want all C/P states used, set Dynamic Vcore (DVID) value using following formula:

     CPU Vcore lowest value (when set to Normal+  (DVID) set value  =  1.35V (or Vcore value shown inside M.I.T window when using Optimized Defaults in BIOS)

 

     You must reboot after each DVID change to see updated lowest and highest CPU Vcore value.             

     Updated highest Vcore value will be shown in BIOS at the bottom of MB Intelligent Tweaker (M.I.T) screen.

     Updated lowest Vcore value will be shown in BIOS at M.I.T/Advanced Voltage Control/CPU Vcore.
     HWMonitor.app should show the same lowest and highest Vcore values during idle and during maximum load.

     It may take several tries to get desired Vcore values when adjusting DVID as Load Line Calibration function applies its own adjustments as well.

 

I do not use any external SSDT, letting Clover generate C/P States. I extracted my own DSDT by pressing F4 key at Clover bootscreen and applied d00d's patches from here.

I use unmodified F13 BIOS from Gigabyte website as F13 mod10 version was giving me some trouble.
With overclock settings listed below, my Xeon X5670 & i7-980X show the following frequency states2.27GHz (x12), 4.16GHz (x22), 4.35GHz (x23), 4.54GHz (x24).
CPU Vcore value shown in HW Monitor of system using Xeon X5670 CPU change in multiple steps between 1.136V and 1.408V.

CPU Vcore value shown in HW Monitor of system using i7-980X CPU change in multiple steps between 1.120V and 1.382V.

Note: Intel specifies that maximum Vcore voltage can be 1.35V(+/-)10 percent which is 1.4175V at maximum.

Both systems pass 10 hours of IntelBurn Test and have everything working properly including shutdown, restart, sleep and auto-sleep.
After switching to Noctua DH-14 Air Cooler my CPU temperature maxes at 78C during IntelBurn Test. In regular use under load it maxes at 69C. Idle temp is 32C.

 

My overclocking method is very simple:

At the beginning I set DVID to +0.24375 as this value would give me desired maximum CPU VCore of 1.408V.

Than I left all other voltages I could at Auto and just played with BCLK value until I reached stability.

Then using Easytune6 and Hardware Monitor on Windows I reviewed actual values of BIOS parameters that were set to Auto.

Since all values looked fine to me I decided to leave them at Auto.

If you copy my BIOS overclock settings start with BCLK Frequency value 172 and see how much higher you can go before system becomes unstable.

If CPU temperature gets too high, get better CPU Cooler or reduce DVID and BCLK values until temperature drops to comfortable level.

On Mac OSX you can use Geekbench 3 "Stress Test" function to check stability of your overclock.


Here are my complete Overclock BIOS settings for Xeon X5670 & i7-980X CPU:

 

RAM Notes:

Initially I used Corsair XMS3 RAM running at 1.65V  and with "Intel® Turbo Boost" enabled in BIOS I could not obtain stability beyond 174 BCLK/4.17GHz.

Corsair XMS3 RAM is falsely advertised by many internet sellers as 1.5V type. This RAM will not work properly below 1.65V.
When I switched to Low Power CORSAIR Vengeance LP RAM and left it at default 1.5V and 9-9-9-24 timings, higher stable CPU overclock became possible.

In general I do not overclock my RAM, trying to run it at the lowest voltage as increasing RAM voltage is an obstacle in getting the best CPU overclock possible.

My Xeon X5670@4.54GHz draws 108 Watts of power during Intel Burn Test, so I also had to get better CPU Cooler to keep my temperatures acceptable under full load.

I switched to Noctua DH-14 air cooler which I highly recommend.

 

 

Advanced CPU Features:
CPU Clock Ratio ........................ [22x]
Intel® Turbo Boost Tech ........ [Enabled]
CPU Cores Enabled ................... [All]
CPU Multi Threading .................. [Enabled]
CPU Enhanced Halt (C1E) .......... [Auto]   ------  (when Enabled, I get occasional system lag after period of inactivity)
C3/C6/C7 State Support ............ [Auto]   ------  (when Enabled, another single core frequency of 4.72GHz (x25) appears only with Xeon X5670 CPU, but I can't boot into Recovery Partition and get kernel panic on reboot after each macOS Sierra update)

CPU Thermal Monitor ................. [Enabled]
CPU EIST Function .................. [Enabled]
Bi-Directional PROCHOT ............ [Enabled]
Virtualization Technology ............. [Disabled]

Uncore & QPI Features:

QPI Clock Ratio ......................... [Auto]
QPI Link Speed ......................... [6.84 GHz]

Uncore Clock Ratio ................. [x17]  ---   (Set Value = 2x System Memory Multiplier +1)
Uncore Frequency ................... [3230 MHz]
Isonchronous Frequency ........... [Enabled]

Standard Clock Control:
Base Clock (BCLK) Control .............. [Enabled]
BCLK Frequency (MHz) ................ [190]  ---  may have to be tweaked (start from 172 and see how much higher you can go before system becomes unstable)
PCI Express Frequency (MHz) .......... [Auto]
C.I.A.2 ....................................... [Disabled]

Advanced Clock Control:
CPU Clock Drive ........................ [800mV]
PCI Express Clock Drive ............ [900mV]
CPU Clock Skew ........................ [0ps]
IOH Clock Skew ......................... [0ps]

Advanced DRAM Features:
Performance Enhance ...................................... [standard]
Extreme Memory Profile (X.M.P) ....................... [Profile 1]
System Memory Multiplier (SPD) ................... [x8]  ---  may have to be lowered for other then 1600MHz/1.5V RAM type

Memory Frequency (MHz) ................................ [1520 MHz]
DRAM Timing Selectable (SPD) ........................ [Auto]

Advanced Voltage Control:

CPU
Load Line Calibration ............... [Enabled]
CPU Vcore 1.1360v ................... [Normal]
Dynamic Vcore (DVID) ............ [+0.24375]  ---  may have to be tweaked ( +0.24375 value allowed me to reach 190 BCLK/4.54GHz overclock on X5670 & i7-980X )
QPI/VTT Voltage 1.175v .............. [Auto]
CPU PLL 1.800v ......................... [Auto]

MCH/ICH
PCIE 1.500v ......................................... [Auto]
QPI PLL 1.100v .................................... [Auto]
IOH Core 1.100v .................................. [Auto]
ICH I/O 1.500v ..................................... [Auto]
ICH Core 1.1v ...................................... [Auto]

DRAM
DRAM Voltage 1.500V .................. [Auto]  ---  my RAM is 1600MHz/1.5V type
DRAM Termination 0.750v ........... [Auto]
Ch-A Data VRef. 0.750v ............... [Auto]
Ch-B Data VRef. 0.750v ............... [Auto]
Ch-C Data VRef. 0.750v ............... [Auto]
Ch-A Address VRef. 0.750v .......... [Auto]
Ch-B Address VRef. 0.750v .......... [Auto]
Ch-C Address VRef. 0.750v .......... [Auto]

 

 

Anyone with X58 board should upgrade to 6-core CPU.
Presently, used 6-core Xeon X5670 could be purchased from eBay for as low as $60.
Adding SSD Drive connected to Native Sata 3 PCI-E card will dramatically improve boot time and user interface responsiveness.

If needed, native USB 3 PCI-E Card and native WiFi PCI-E Card could also be added.
My system overclocked to 4.54GHz, score 20874 points in Geekbench and it performs like brand new machine.

 

You can combine advantages of small but fast SSD and large but slower Hard Drive by moving /Users folder from SSD to Hard Drive and telling OSX to use it.

   Here is the procedure:

 - After /Users folder is copied to HD, go to System Preferences/Users & Groups/right click on Account to be used/Advanced Options/Home directory/Choose/select Account copy located on Hard Drive

 - Beware not to change Hard Drive name after this procedure as SSD-Hard Drive connection is defined in OSX by the name of the Hard Drive.

This SSD+Hard Drive combination works as well as Apple Fusion Drive and it is easier to configure.

Also, it works without any issues in case where SSD Drive is formated as APFS and Hard Drive as HFS+

 

Note:  If you plan to upgrade from macOS Sierra 10.12 to macOS High Sierra 10.13 and new APFS file system you must first do the following:

 - Update Clover EFI Bootloader to the latest version

 - Copy contents of EFI/CLOVER/kexts/10.12 folder to EFI/CLOVER/kexts/10.13 folder.
 - Copy apfs.efi driver to EFI/CLOVER/drivers64 folder
( apfs.efi is located inside  Install macOS High Sierra/right click Show Package Contents/Contents/SharedSupport/open BaseSystem.dmg/usr/standalone/i386 )
You should use APFS file system only on SSD drives. It does not work well on conventional Hard Drives.

 

 

 

For anyone willing to try my setup, I also attached all my files used with Clover bootloader:

- Attached config.plist can be used after modifying SMBIOS and Graphics sections to reflect used System Definition & GPU (I use MacPro5,1 definition).

  (Don't enter any value in Cpu/Frequency MHz field of config.plist as it will cause occasional Kernel Panic during boot on overclocked system)

- My patched DSDT.aml file comes from F13 BIOS of GA-EX58-UD5 mobo and could be used with any CPU and any System Definition.

- Attached FakeSMC.kext 6.23.1417 is patched for smc-thurley (MacPro5,1 platform) and in Plugins folder contains kexts required for HWMonitor.app to show sensors data.

  (Don't add/install ACPISensors.kext to Plugins folder. It can cause many different problems, even in systems with properly configured BIOS and perfectly patched DSDT)

- Attached AHCI_3rdParty_SATA.kext contains patch allowing to use keyboard eject button to eject/inject CD/DVD Drive.

- Build-in LAN works perfectly with attatched RealtekRTL81xx.kext

- Toleda script from here must be run to enable onboard audio.

- For macOS Sierra use Clover_v2.4k_r4003 or newer. Prior versions of Clover for Sierra break sleep and/or create problems with Wi-Fi cards on GA-EX58-UD5.

- Remember that some changes made in config.plist take two reboots to be used by OSX.

 

So far this overclock has been confirmed to work on seven different X58 systems besides two of my own. If it works for you please do not forget to confirm in post reply.

EX58-UD5_F13.zip

post-199991-0-58444000-1483912725_thumb.png

post-199991-0-76600800-1483924052_thumb.png

post-199991-0-21776300-1484063941_thumb.jpg

HWMonitor.zip

Edited by Emilo

Share this post


Link to post
Share on other sites
Advertisement

Thanks! Very interesting, but i can't start with yours dsdt and config.

I have GA-EX58-UD5 rev1.0 BIOS 13 Intel i7-920 and nvidia GT730 (web driver)

 

Is there information about video card in DSDT? or about memory? because i have different memory too 

Share this post


Link to post
Share on other sites

Thanks! Very interesting, but i can't start with yours dsdt and config.

I have GA-EX58-UD5 rev1.0 BIOS 13 Intel i7-920 and nvidia GT730 (web driver)

 

Is there information about video card in DSDT? or about memory? because i have different memory too 

 

 

Your Intel i7-920 is 4-core CPU which requires different BIOS settings, config.plist and FakeSMC.kext patches.

My DSDT does not contain any information about my Video Card or Memory.

As it is, my guide will work well only for X58 systems with 6-core CPU.

Share this post


Link to post
Share on other sites

Thank you, Emilo.

 

My system becomes unstable beyond 175 BCLK, but very happy anyway :) I've reduced DVID to 0.14375 (temps rocketed to 90ºC under stress test with your settings, now reach the mid-70s) and increased QPI/VTT to 1.335 to improve stability. Memory specs are 1066/ 7-7-7-16 @1.5V, they are running stable now at 1400/6-6-6-18 @ 1.640V

 

It seems to be a consensus about disabling virtualization, but I've always kept it enabled without problems.

 

HWMonitor shows every sensor with your patched FakeSMC (updated to latest version).

 

I already had a DSDT modified as per d00d, so I had sleep working, but it only wakes up with the power button. Does yours work with keyboard/mouse?

Share this post


Link to post
Share on other sites
-  I already had a DSDT modified as per d00d, so I had sleep working, but it only wakes up with the power button. Does yours work with keyboard/mouse?
-  It seems to be a consensus about disabling virtualization, but I've always kept it enabled without problems.
-  My system becomes unstable beyond 175 BCLK, but very happy anyway. I've reduced DVID to 0.14375 (temps rocketed to 90ºC under stress test with your settings, now reach the mid-70s) and increased QPI/VTT to 1.335 to improve stability. Memory specs are 1066/ 7-7-7-16 @1.5V, they are running stable at 1400/6-6-6-18 @ 1.640V

 

-  With my DSDT and config.plist, every function works properly including wake from keyboard and mouse.

-  I did not find Virtualization Technology setting to interfere with anything, but I do not use Parallels or VMWare so I keep it disabled anyway.

-  Initially I used Corsair XMS3 RAM running at 1.65V  and with "Intel® Turbo Boost" enabled in BIOS I could not obtain stability beyond 174 BCLK/4.17GHz.

Corsair XMS3 RAM is falsely advertised by many internet sellers as 1.5V type. This RAM will not work properly below 1.65V.

When I switched to Low Power CORSAIR Vengeance LP RAM and left it at default 1.5V and 9-9-9-24 timings, higher stable CPU overclock became possible.

In general I do not overclock my RAM, trying to run it at the lowest voltage as increasing RAM voltage is an obstacle in getting the best CPU overclock possible.

My Xeon X5670@4.54GHz draws 108 Watts of power during Intel Burn Test, so I also had to get better CPU Cooler to keep my temperatures acceptable under full load.

I switched to Noctua DH-14 air cooler which I highly recommend. It can compete with many more expensive water cooling solutions.

Share this post


Link to post
Share on other sites

- With my DSDT and config.plist, every function works properly including wake from keyboard and mouse.

 

It seems that my DSDT was not good enough, it works using yours :)

 

Thank you for the hint on the memory overclock. I'd try again when I upgrade my cooler, for now 4.1GHz feels just right. 

Share this post


Link to post
Share on other sites

Thanks again Emilo for this awesome guide. I have followed it on my X5670 / X58A-UD3R with good results. My previous OC was non-Turbo, 200 x 22 = 4.4Ghz, and with no Auto settings - I had manual settings for all voltages.

 

Here are my observations of the result of changing from that OC to using your settings, plus a couple of tweaks:

  • Using your OC settings exactly - except for RAM voltage which I left at 1.68V as I'm on Vengeance RAM - gave me the following:
    • macOS GB4: 3824 / 18425.  Within less than 1% of your scores. 100 points higher in Single-Core than my old OC, multi-core almost identical.
      • Windows is still lower than macOS, and the difference is now even greater: 3719 / 17457
      • I wonder if this is partly down to differences in GB4 between platforms rather than macOS actually running faster.  But maybe it is.
    • Better P-States:  as you said, P-States have been improved.  I now see in macOS different cores at different multipliers, where previously all were the same.  I also see a few more multipliers: specifically, I have seen x12, x21, x22, x23 and x24.  Which I suppose is just one extra compared to before (x21), as the latter two are Turbo states.
      • So in macOS I'm still not seeing the granularity I expected - I thought I'd see eight or ten possible states throughout the range of x12 to x24.  But it's still an improvement.
      • In Windows I do see more states - I've definitely seen 3610Mhz for example, which is x19.  But it's hard to monitor in Windows as HWMonitor's P-state monitoring is now partly broken for some reason (see below)
  • I then made one further tweak: turned off Auto SPD, and configured timings manually so I could select Command Rate 1 instead of 2.  I was using CR1 on my previous OC with success, and I'd read that CR1 vs CR2 can be equivalent to lower RAM timings, eg equivalent to setting 8-8-8 instead of 9-9-9.
    • This boosted my macOS GB4 multi-core, giving scores of 3821 / 18728  (an increase of about 1.5% on the multi-core figure.)
    • However when I then upgraded to 40GB RAM, it dropped quite a bit, to 3803 / 17814.
I did briefly test pushing things further - eg 195 x 24 with a couple of attempts at higher +VIDs.  It did boot into Windows and GB benchmark ran OK, but I got some hard freezes and invalid results in IntelBurnTest, and the temperatures were way too high; I stopped my last test when one core hit 99C.  I don't have the cooling to even attempt this I think, nor is it really worth the risk and hassle to get another 5%.

 

Beside the different RAM voltage and now the manual SPD, I also have one other setting different to you: Virtualisation enabled.  I had this enabled on my previous OC and it was fine, and I plan to use VMWare Fusion, so this will potentially be useful.

 

What I find interesting is how this new OC has some quite different behaviour to my previous one, in several ways:

  • I now see the true, expected VID values, eg maxing at 1.440V.  Before, my VID never went above the high 1.3x's, and I could never understand why - several times when tweaking I added more to Dynamic VID and saw exactly the same maximum values, which seemed weird but I never investigate further.
  • Windows HWMonitor now shows that my max power is 108W, where before it never exceeded 95W.   I didn't realise until now that that was unusual - I thought the chip simply couldn't go over 95W.  But with the new settings, it can and does.  This is obviously related to the first point, the VID not going as high as expected.  Something in my old settings was capping it.
  • For some reason P-State monitoring is now broken in Windows HWMonitor:  after launching it, the processor multipliers don't appear at all for at least a minute, and after that they don't update properly.  They work fine in macOS HWMonitor, and Windows CPU-Z also seemed to show it OK, but it only displays one figure for the whole CPU.
  • My Northbridge temperatures are much lower with the new OC.  With the old one, I was going above 75C quite often, and I actually put a Noctua 40mm fan on the heatsink to make sure this didn't go too much higher.  That took me to an average of 68C.  With the new OC, and that fan still in place, I'm seeing 55C max.  So I can probably now remove that fan.
That VID capping also means that I was wrong when I previously said I was at +0.28V Dynamic.   That's what I set, but clearly I never reached it.   I was never quite sure on that, because I wasn't certain what the base voltage was meant to be - it seemed like every time I went into the MIB settings, it showed a different base voltage for "Normal", and whatever it showed never seemed to match the figures I saw quoted  So I couldn't be sure whether the VID I saw in HWMonitor was applying the full +Dynamic or not.  I just knew that I sometimes had to boost Dynamic two or three steps before I saw my actual VID go up.   Now I see that actually something was capping it to the max 95W TDP of the chip.

 

One possible explanation for the capping is that I was running Load-Line Calibration level 2, which I suppose does more smoothing and could create a max power/VID cap?  If not that I suppose it was due to my never using Auto, manually setting every voltage; one or more of those settings may have needed Auto in order to go beyond 95W.

 

I still find it weird that we're getting higher Multi-Core scores in GB4 with a lower non-Turbo speed.  If it truly runs all cores at once, surely 6 x 4.4Ghz is better than 6 x 4.18Ghz?  Perhaps it doesn't truly stress all CPUs at once, such that the Turbo sometimes has a chance to kick in.  I suppose I could test that further with some non-synthetic benchmarks - some long running task, like compiling GCC or Firefox with -j12 - that could be used to verify on my old and new OC to compare against GB4's synthetic benchmarking.  Then again that sounds time consuming and I am happy with what I have so it'd probably only be for academic interest.  I think I will definitely stick with the new OC even if some benchmark did prove the old one to be 2% or whatever faster in some specific case.

 

Unfortunately since I upgraded from 24GB to 40GB RAM (5 x 8GB Corsair Vengeance DDR3-1600; one slot empty) I have lost all the GeekBench 4 Multi-Core gains I made. Even after I tried some better timings (8-8-8-21), my current score is 3845 / 17815. That's the best ever single-core result I've had, but I've lost over 1000 points in Multi-Core. I don't know if this is because of having one DIMM slot empty, or if it's just that more RAM is slower in these benchmarks. I am now tempted to buy a sixth and final 8GB DIMM to see if that improves things ;) But I will at least wait to see if I can get a good price, as I've spent rather too much already!

Share this post


Link to post
Share on other sites
One odd thing:  my voltage is peaking at 1.44V, where yours is 1.41V.   I just double checked and I definitely have the same +VID, 0.24375.   I wonder what's causing that?  Perhaps our slightly different motherboards, or maybe it can vary by individual CPU?    Or perhaps because I have overclocked RAM - maybe that causes it to boost VID.

 

Also, you said your CPU runs at 150W in IntelBurnTest?  Is that as seen in Windows HWMonitor?  Or measured externally?  Because that's ~40% higher than my 108.3W (according to Windows HWMonitor) and sounds huge!

 

Final question: for FakeSMC and HWMonitor, do you know what is different in the patched version you attach versus the 'standard' kazlek version 6 branch?

 

I got my FakeSMC*.kext files from RehabMan's repo of kazlek's version 6 branch; current version 6.26.  This includes FakeSMC.kext, FakeSMC_ACPISensors, FakeSMC_CPUSensors, FakeSMC_GPUSensors and FakeSMC_LPCSensors.

 

These seems to be working fine for me, except I noticed I don't have a GPU VCore figure that you do have in your screenshots.   I'm just wondering if there's any reason to try the patched version you attached versus the (later) ones I have now.

 

I also just tried Slice's Version 3, but I got a KP on boot.  I afterwards realised I probably missed a file, as I didn't copy a kext for the SuperIO chip (eg it says "ITEIT87x.kext for ITE LPC chip, usually for Gigabyte motherboards".)   

 

I am not yet sure if there's any benefit to the HWSensor 3 branch versus 6, but I figured I'd try it out and find out.  I will have another go tomorrow.

Share this post


Link to post
Share on other sites

 

- One odd thing: my voltage is peaking at 1.44V, where yours is 1.41V. I just double checked and I definitely have the same +VID, 0.24375. I wonder what's causing that?
- Also, you said your CPU runs at 150W in IntelBurnTest?  Is that as seen in Windows HWMonitor?  Or measured externally?  Because that's ~40% higher than my 108.3W
- Final question: for FakeSMC and HWMonitor, do you know what is different in the patched version you attach versus the 'standard' kazlek version 6 branch?
  I got my FakeSMC*.kext files from RehabMan's repo of kazlek's version 6 branch; current version 6.26.
- I noticed I don't have a GPU VCore figure that you do have in your screenshots.

 

- In general Max CPU VCore = Min CPU VCore + DVID value, but on the top of that it is also adjusted by "Load Line Calibration" setting in BIOS.

  Check what is the actual minimum VCore value of your CPU. Lowest Vcore value will be shown in BIOS at M.I.T/Advanced Voltage Control/CPU Vcore.

  Also on my MOBO I can only turn "Load Line Calibration" ON or OFF. On your MOBO this setting has choice of several "Levels". Try playing with different levels.

- Here are results of new test that I performed today using 64bit HWMonitor 1.34 and IntelBurnTest set to Maximum Stress Level:

   At Powers/CPU/Max it shows 195.35 Watts.

   At Intel Xeon X5670/Powers/Package/Max it shows 108.55 Watts.

- My HWMonitor, FakeSMC and Sensors kexts were extracted from HWSensors.6.16.1372.pkg using Pacifist.

  FakeSMC.kex was patched for smc-thurley (MacPro5,1 platform) which is described in point 13 of d00d guide.

  Sensors kexts were placed inside FakeSMC/Show Package Contents/Contens/Plugins folder. You can find it all in attatched EX58-UD5:F13.zip.

  Don't add ACPISensors.kext to Plugins folder. It can cause many different problems, even in systems with properly configured BIOS and perfectly patched DSDT.

- Ability to access and display GPU VCore sensor value depends on video card model, HWMonitor and Sensors kexts. It works with my evga GTX 780 and attatched kexts

 

 

EX58-UD5:F13.zip

Share this post


Link to post
Share on other sites

Thanks Emilo.
 
Any particular reason you're still on HWSensors/FakeSMC 6.16?  No reason to upgrade to latest I suppose?
 
I have edited the Info.PList of my FakeSMC 6.26 to apply the d00d patches.  It was working OK before and still is now, but can't hurt to apply any recommended patch.  I also removed ACPISensors as per your recommendation - I thought I needed this for HWMonitor sensors, but clearly not.
 
One difference I note with the latest 6.26 FakeSMC is it no longer uses Contents/Plugins for the other kexts.  They are all at the top level - ie all installed side by side in EFI/CLOVER/kexts/Other:
 
KxLM0jo.png
 
Regarding max VID:  I am using Load Line Calibration Level 1 as I assumed that was same or closest to your "On" setting.  But maybe there is some difference between mobo.  I don't mind that I am peaking at 1.44V, I just thought it interesting that there was a difference despite seemingly identical settings.
 
 

Check what is the actual minimum VCore value of your CPU. Lowest Vcore value will be shown in BIOS at M.I.T/Advanced Voltage Control/CPU Vcore.

 
That is the interesting thing - at least on my motherboard, this number is not the absolute minimum of the chip, but seemingly the minimum of the current settings.  Also, strangely, I currently see a lower voltage in HWMonitor than in MIT.
 
Here is my current MIT/CPU Voltage value, with your OC settings: 1.21875
TdeHG2Km.jpg?2
 
However HWMonitor shows the lowest value as 1.184:
SM0eojp.png?1
 
And this is what I saw in MIT/CPU Voltage a few weeks ago, with my old OC - a different lowest VCore: 1.11875:

cEbjVy5m.jpg?1
 

 

And I am sure that when I last played with OC, I remember seeing other figures as well - that's what I meant when I said in an earlier post that I was never sure what my base VID was, because it kept changing and the figure I saw never matched figures I saw others quote.  Clearly you must have had yet another different figure, because you said you chose +0.24375 so as to achieve a maximum of 1.408, so I guess you saw a Normal of 1.16425?

 

So at least on my mobo, I suppose it would be most accurate to say the figure on MIT/CPU Vcore is "the lowest achievable VCore with current MIT settings"?   Though that still doesn't explain why HWMonitor shows a lower value in actual usage.

 

I thought maybe this CPU Vcore Normal would be affected by Line Load Calibration, so I tested all three settings for that - Standard, Level 1 and Level 2.  But, at least with my current OC settings, each of these three LLC values always gave me the same VCore Normal of 1.21875.

 
In terms of maximum VID, here the BIOS MIT and HWMonitor agree: 1.440V.   However this is not what would be expected from 1.21875 + DVID of 0.24375, which would be 1.4625.  But maybe this figure is affected by Line Load.
 
To be clear, I am sure none of this really matters.  All that does matter is a stable OC with decent performance.  I just find it interesting to investigate all the different options and what they mean - something I never really did with OC before; I just followed a guide, got it stable, and got on with it.  Now I am more curious to see what everything does and what everything means  :)
 

Share this post


Link to post
Share on other sites

Any particular reason you're still on HWSensors/FakeSMC 6.16?  No reason to upgrade to latest I suppose?

 

- Just as you mentioned, I find no reason to upgrade.

- When it comes to Max VCore:

  I would keep lowering DVID and testing stability to determine what is minimum DVID value (or VCore) required to maintain this overclock on your MOBO.

Share this post


Link to post
Share on other sites

Well I picked up a second hand X5670 as suggested after reading some very encouraging experiences all over the internet. So fingers crossed. Only thing I don't understand. Why do some sites suggest my MB X58A-UD3R Rev 1 doesn't support this CPU, while other sites suggest it will work? Seems like people who have tried it with the Rev 2 at least said it worked, so I'm guessing Rev 1 will also suit?

 

A bit more on topic... it seems that playing around with DSDT comes after the installation of MacOS?

Share this post


Link to post
Share on other sites

Well I picked up a second hand X5670 as suggested after reading some very encouraging experiences all over the internet. So fingers crossed. Only thing I don't understand. Why do some sites suggest my MB X58A-UD3R Rev 1 doesn't support this CPU, while other sites suggest it will work? Seems like people who have tried it with the Rev 2 at least said it worked, so I'm guessing Rev 1 will also suit?

 

A bit more on topic... it seems that playing around with DSDT comes after the installation of MacOS?

 

Yes the X5670 is a very worthwhile upgrade over the i7 930 you had before.  I did the exact same upgrade, i7 930 to X5670, last year and couldn't be happier - my GeekBench scores went up by about 60% (so more than just having 50% more cores.)   In my experience the X5670 overclocks just as easily, or even more so, than the 930.  You'll need a decent CPU cooler for that of course.

 

Officially no Xeon CPUs are supported on consumer X58 motherboards.  This confused me as well for some time - I had an i7 930 from August 2010 until last year, and it took me that long to finally realise that actually I could use any of the X58/LGA1366 CPUs on this motherboard.  When I finally googled it properly, I wished I had done so years before!

 

However I have a Rev 2.0.   I assume the Rev 1 works just the same.  But I haven't heard specific confirmation of that.  If I were in your position I'd certainly buy the chip and try it.   I think the Rev 1 boards are rarer, most people I've seen post here have Rev 2 like me.

 

I'm sure it'll be fine.   Another thing that isn't officially supported but works fine: the RAM limit is 48GB not 24GB.  I have no idea why the motherboard manufacturers never mentioned this, seems like a big sellling point to me.  You can also use ECC RAM if you want, at least when you have a Xeon processor.

 

Yes, the DSDT is only likely to affect you post-install.  However it's good to get it in place before the upgrade.  And the affect of having a bad DSDT in High Sierra with latest clover can be as severe as your BIOS resetting to defaults every time you shutdown macOS.  That's what happened to me after I upgraded.  My system worked nearly perfectly without a DSDT at all in Yosemite and El Capitan, but I had to have a properly edited DSDT to get High Sierra running without major problems like BIOS reset and minor problems like no sleep/wake.

 

Just stick the DSDT I sent you in EFI/CLOVER/ACPI/patched and you should be able to forget about it after that.  I doubt the DSDT changes between Rev 1 and Rev 2 of the mobo;  yesterday I linked you to the general Gigabyte GA-X58 ACPI thread, and the DSDT patching instructions are the same for all variants, including our -UD3R, and the -UD5 and -UD7.   The DSDT I uploaded for you already has all those patches applied and is confirmed to work on UD3R Rev 2, and I'm sure will be fine for you as well.

Share this post


Link to post
Share on other sites

What are your core temps like? Mine are a bit all over the place. I have tried applying thermal paste twice. The first time, it wasn't ideal, some bits of the CPU were dry, but the second time, I'm pretty sure I got all of the CPU covered. However, one of my cores is significantly hotter than the others. I'm talking 10 degrees different. I got a similar result both times I tried applying thermal paste.
 

EDIT: Problems with over clocking: I reduced multipliers for my CPU, memory and QPI because I wanted to start playing around with the bclock. It wouldn't post. I had all the power savings modes turned off, so after a CMOS reset, I thought I'll do the same except keep all the power saving modes on as per default. Still no beuno. Still not posting. My uncore is x2 of DRAM i.e 1600 vs 800, so that's all good... hmm...

I'm just following the same guides as for i7 930. Is there something I'm missing?

 

EDIT: Thanks TheBloke for the write up, some good tidbits in there. I think ECC is a bit overkill after reading about what it's for.

post-2318321-0-92802500-1520039720_thumb.jpg

Share this post


Link to post
Share on other sites

 

EDIT: Problems with over clocking: I reduced multipliers for my CPU, memory and QPI because I wanted to start playing around with the bclock. It wouldn't post. I had all the power savings modes turned off, so after a CMOS reset, I thought I'll do the same except keep all the power saving modes on as per default. Still no beuno. Still not posting. My uncore is x2 of DRAM i.e 1600 vs 800, so that's all good... hmm...

 

I'm just following the same guides as for i7 930. Is there something I'm missing?

 

Why don't you follow the guide in the OP of this thread? 

Share this post


Link to post
Share on other sites

I didn't want to follow the guide because @ 3.84Ghz, my hottest core was already at 70C. An extra Ghz right off the bat would I feel bust me through the 81C limit of these chips. I don't want to risk it.

Bit of an update. I have tried everything and all things point to a MB that is acting up. I was lucky to find any overclock at all. It's stable which is a good thing, but it's the only overclock that I was able to find. Seems I must have damaged the MB when I was overclocking the i7 930. No matter what I tried with the X5670, it wouldn't post. Oh well, better than nothing so I'm not complaining. It's still 2ce as fast as my old setup, so good enough.

Share this post


Link to post
Share on other sites

Personally I am ignoring the 81C limit, at least for burn test purposes.   During extended IntelBurnTest, I will peak around 96C.  I consider that OK because:

  • It's only in IntelBurnTest - in normal day-to-day usage, the highest I've so far seen in macOS is 70C for a few minutes
  • The temperature limit of most CPUs I've used before has been stated as 100C, and I prefer to think the Xeon will likely hold up to that level as well, at least in short bursts
  • Worst case, I only paid £35 for this Xeon.  If I fry it, I will buy another - and not push the temperature so high on the second.  And prices will only fall over time.  But I've had this one six months, have taken it above 90C several times, and so far it's fine.

I don't know what to say about your issues and thinking you've damaged the motherboard.  It does sound pretty unlikely to me - unless you were previously running your Northbridge at 100C a lot?  I suppose it's possible

I said a few times that you should follow the guide because the guide already covers multiple speeds.  Emilo says specifically to start at a BCLK of 170, which is only 4.08Ghz at Turbo, barely faster than the 3.84 you already had.  And of course you could go even lower; BCLK 160 would be the same 3.84 you have now.

The important point of following the guide is that all his other settings are well tested.  In one of your posts that got deleted in the forum migration you mentioned some settings that sounded to me quite different to the guide - like changing Uncore and stuff like that.

Ultimately if you can find a working, stable OC then that's all that matters.  But I would certainly have started from the point of the guide - with whatever BLCK you are comfortable with - and change from there, rather than trying random settings.

What RAM do you have?  From what I've seen and learnt recently, RAM plays a huge part in the stability of an OC.  Emilo with his Corsair Vengeance LP is able to leave RAM at Auto voltage and achieve 190x24.  With my Corsair Vengeance I can also achieve 190x24 (or 200x22), but only by setting RAM at the much higher 1.66V.  Other RAM may require other voltages, or may not be stable at all.

So personally I would suspect the RAM before I suspected the whole motherboard.  What DIMMs do you have and what voltages did you try in your tests where you couldn't even POST?

If you're happy with your stable OC then no need to push further.  But I would be surprised if your mobo was somehow damaged.    Of course, it could be that your Rev 1.0 mobo is less OC'able than my 2.0 and Emilo's later UD5.  That's certainly possible I guess.

Share this post


Link to post
Share on other sites
Posted (edited)

I'm definitely not 100% sure it's the motherboard. Just suspecting it's the case. When I was overclocking my i7 930, I got it stable at 4Ghz @ 200 bclock then started playing around with it (I tried to go down to 3.8Ghz to reduce temps a bit and something happened (not magic smoke) but basically I tested the 4Ghz overclock for an hour on both OCCT and Prime95 and they were both stable for an hour but after playing around and trying to go for 3.8Ghz with different bclock, the computer became totally unstable even at low overclocks and 4Ghz broke, I literally could only run the comp at stock speeds only.

As for my memory, it was working fine at 1600 i.e. 1600/200 so multi of x8 and 1.54v and slightly tighter timings than factory i.e. factory was 9 9 9 ?24?, I got it down to 8 9 8 24. That was 1 hour mem test stable. That's the settings I was using as the upper limit for my memory because I knew my memory was stable at those levels. I have GSkill mem, not ripjaws which I think are the expensive ones, I just have the relatively average ones x6 sticks 12Gb.

I was asking for advice on like 4-5 forums about overclocking this chip. None of the advice helped. Nothing that I tried resulted in anything other than no post. I think I got lucky in the sense that I stumbled onto an overclock that worked literally the first stab in the dark that I took and unlucky in the sense that I agree, these chips do seem very easy to overclock, but some part of my system isn't playing nice, can't do much about it unfortunately. I tried everything but nothing worked. 3.84Ghz is good enough for what I need for now. The next step is to get a new computer. I was only extending the life of this current machine by a few years. I'm pretty happy with it to be honest. It does what I need it so I'm just going to stick with it.

Edited by Atomic_Sheep

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

  • Recently Browsing   0 members

    No registered users viewing this page.

×