I5 750 overclocking

[MrK6]

PLL voltage is at 1.9V, didn't touch the PCH.

 

[MadScientist]

MrK6,
Thanks for the info.

I'm running my PLL voltage at 1.90 too, but had to bump up the PCH to 1.10V. I'm going to give it a few more trys at lower PLL voltages but if it needs 1.90V that's ok since Intel's recommended PLL max is 1.98V.



Asrock P55 Pro, bios V1.90, i5-750 3.80 Ghz @1.325V bios, Cooler Master Hyper 212 Plus (2 fans) with TIM Consultants T-C Grease 0098, G.Skill Ripjaws 4GB (2 x 2GB) DDR3 1600 @ 760 Mhz 9-9-9-24, 1.5V, OCZ ModXStream Pro 700, MSI NX8800GT, Win 7 Ultimate 64bit

 

[imported_mingus]

Can someone give me basic O/C settings, ASRock P55 Dlx, i5, G.Skill DDR3-2000 F3-16000CL9D-4GBTD . http://www.gskill.com/products.php?index=217

3.4ghz to 3.5ghz is fine for now.

Thx

 

[MadScientist]

Originally posted by: mingus
Can someone give me basic O/C settings, ASRock P55 Dlx, i5, G.Skill DDR3-2000 F3-16000CL9D-4GBTD . http://www.gskill.com/products.php?index=217

3.4ghz to 3.5ghz is fine for now.

Thx

You should be able to do 3.4 to 3.5 Ghz on stock voltage settings.

My Asrock P55 Pro MB bios settings I think are identical to your Deluxe.

First flash the MB to the latest bios. Link

Try these settings for a start:
Overclock Mode - Manual
Bclk Freq - 170 to 175 for 3.4 Ghz to 3.5 Ghz
Spread Spectrum - Disabled
QPI Frequency - anything over 150 - 160MHz base clock, select the lower multiplier
CPU Voltage ? Auto, select Manual if you have to raise voltage
VDrop Control ? Without Vdrop
Dram Voltage - Auto
Dram Freq - select the highest choice that will still keep your ram below its spec rating of 1000 Mhz (2000). This option on this MB really sucks. It only gives you 3 ram ratio choices.
VTT Voltage- Auto
PCH Voltage - Auto
CPU PLL Voltage - Auto

Under CPU Configuration:
Enhanced Halt State ? Disabled
Intel Virtualization - Disabled
CPU Thermal Throttling - Enable
No-Execute Memory Protection - Disabled
Active Processor Cores - All
A20M - Disabled
Intel SpeedStep - Disabled, this will also disable Turbomode.
Intel C-State - Disabled
Disable everything you are not using, floppy, firewire, unused SATA ports, etc.

If interested, Xbitlabs has a review of your MB. Link.

Good Luck.

 

[imported_mingus]

Wow, more than basic, thanks much. So basically the clock changes and the multiplier stays the same?

Will most likely make more sense once i get in there.

thanks again, will post back with results.

 

[AznAnarchy99]

I really dont know whats wrong with mine. Today when i turned on my rig i had some horrible screeching sound problems. Set bios back to normal settings and it went away. I was running 3.6ghz for the last week without a problem Linx stable...

I guess bad luck with chip or what....

 

[MadScientist]

Originally posted by: mingus
Wow, more than basic, thanks much. So basically the clock changes and the multiplier stays the same?

Will most likely make more sense once i get in there.

thanks again, will post back with results.

On the Asrock MBs the CPU clock ratio (mutiplier) for the i5-750 is locked at 20, unless you are using TurboMode or the auto OC functions in the bios.

DDR3 2000 is a bit of overkill if you are going to oc to 3.4 - 3.5 Ghz.

 

[fearrun]

I am currently using the following configuration for running at stock settings:

CPU Configuration
C1E - Disabled
SpeedStep - Disabled
TurboMode - Disabled
C-STATE - Disabled
Ai Overclock Tuner - Manual
CPU Ratio Setting - 20
BCLK Frequency - 133
DRAM Frequency (divider) - DDR3-1333
DRAM - 7-7-7-21-2N 1.5V
CPU and PCIE Spread Spectrum - Disabled
Load-Line Calibration - Disabled
All other settings are at defaults or auto, including voltages.

This has passed 24 hours stability with Prime95.The highest temps in Celsius displayed by Core Temp were 54, 53, 52, 54. According to CPU-Z, core voltage at idle is 1.176V and when all cores are at full load it drops to 1.136V.

After changing the BCLK Frequency to 166 and switching the memory divider to preserve it at DDR-1333 ratings. I was able to achieve 3.33GHz which was stable running LinX, 100 times using 3072MB of memory (Problem size 20014), finishing in 3 hours 35 minutes and 20 seconds at 48.3355 GFlops peak.

The highest temps displayed in Celsius were 63, 63, 61, 64. CPU-Z was showing the idle core voltage at 1.248V and with all cores under full load it slightly fluctuated between 1.184-1.192V.

Is the decrease in core voltage, under full load, the normal functioning of VDroop?

I have not had time yet to run another 24 hour Prime95 stress at that overclock. Would it even be necessary for that long? Also, for the temps while running LinX, do they appear to be reasonable?

Thanks for any feedback.

 

[MadScientist]

Originally posted by: fearrun
Is the decrease in core voltage, under full load, the normal functioning of VDroop?

Yes, you can enable Load-Line Calibration to correct for it, but if you are stable leave it disabled.

 

[AznAnarchy99]

I threw on a second fan and things got much more stable. Temps dropped about 5-7C. Only thing that its a bit louder but Ill get used to it. 3.6 is fully stable after an hour of Linx and temps are about 33 Idle, 65 Load. I kept 1.35 Core and 1.3 VTT. Going to try at going up til 3.7 later.

I tried 3.8 at 20x 190, and 21x 172, at those voltages but it wasnt stable for more than 10-15 minutes with Linx. Im using 20x at 180 blck right now for 3.6. Which is better lower or higher multiplier?

 

[rchiu]

Man I am probably one of the few unlucky ones with my i5 750. Only able to get it stable at 3.4ghz on ~1.25 vcore. 3 of the 4 core seems to be okay on higher speed, but this one core keep on craping out on prime95 at > 3.4ghz.

Using coolermaster 212, idle ~34C, but full load gets ~80C with only one fan, although my room temp is ~30C. If I up my vcore to >1.3, the temp will go above 85C and the pc will freeze.

Guess it's a combination of one bad core and hot temperature where I live.

 

[AznAnarchy99]

Originally posted by: rchiu
Man I am probably one of the few unlucky ones with my i5 750. Only able to get it stable at 3.4ghz on ~1.25 vcore. 3 of the 4 core seems to be okay on higher speed, but this one core keep on craping out on prime95 at > 3.4ghz.

Using coolermaster 212, idle ~34C, but full load gets ~80C with only one fan, although my room temp is ~30C. If I up my vcore to >1.3, the temp will go above 85C and the pc will freeze.

Guess it's a combination of one bad core and hot temperature where I live.

I feel ya, I cant get past 3.6 for anything. I can do 1 hour of Linx stable but when I get to 3.7 I can do 5 minutes..

 

[MadScientist]

Originally posted by: rchiu
Man I am probably one of the few unlucky ones with my i5 750. Only able to get it stable at 3.4ghz on ~1.25 vcore. 3 of the 4 core seems to be okay on higher speed, but this one core keep on craping out on prime95 at > 3.4ghz.

Using coolermaster 212, idle ~34C, but full load gets ~80C with only one fan, although my room temp is ~30C. If I up my vcore to >1.3, the temp will go above 85C and the pc will freeze.

Guess it's a combination of one bad core and hot temperature where I live.

86F room temp. Do you live on the equator? There's this thing called an air conditioner that's been around for awhile.

 

[MadScientist]

TH has posted a follow-up article to the one in which they toasted an Asrock P55 Pro MB with 1.40Vcore. http://www.tomshardware.com/reviews/p55-motherboard-overclock,2460.html

In summary:
"Our previous page shows that our processor consumed over 200W when fully overclocked at 1.45V, while ASRock’s engineer told us the P55 Pro’s voltage regulator is designed to sustain 150W. While many might suspect a hardware change might be required, ASRock already had a hardware solution designed into its motherboard. It turns out that the motherboard already had over-current protection built-in, and ASRock had simply neglected it in the BIOS, believing that average overclockers wouldn't need it. A BIOS update to version 1.80 was the sole requirement to fix the P55 Pro. We were first pleased to find that a simple BIOS update allowed us to fully test the P55 Pro’s overclocking capability without a single sign of fatigue to the board. It completed several hours of eight-thread Prime95 stress testing at 1.35V, and dutifully shut off when we tried to apply even more stress through increased voltage and clock speed.
It has been our contention that no motherboard used for overclocking should be able to push enough energy through the CPU voltage regulator to damage it, because any motherboard that includes overclock settings should also include over-current protection. We understand that cheaper boards use lower-capacity voltage regulators, but lower capacity simply adds to the need for such protection. Any motherboard too cheaply made to include over-current protection shouldn’t even have manual voltage control, but should instead be marketed towards the non-overclocking whitebox market.
ASRock proved itself by using nothing more than a BIOS update to enable the over-current protection already included in P55 Pro. While leaving the feature disabled in previous BIOS revisions was a big mistake, redemption in the form of BIOS revision 1.80 is enough for us to have some faith in a board that can be pushed to 1.35V and 4.0 GHz with our Core i7-870 processor. Anyone who wants to use higher voltage to reach even greater clock speeds shouldn’t have a big problem choosing a higher-priced product. On this board, upper-range features such as dual eSATA ports, onboard Port 80 diagnostics display, onboard power and reset buttons, a rear-panel CLR_CMOS button, and IEEE-1394 FireWire now make the board a perfect fit for enthusiasts who prefer not to push their luck at high overclock settings."

How has Asrock proved itself with a bios that enables over-current protection? It's still a cheap-ass MB with the same cheap components that can not be oc'ed over 1.35Vcore. They should change the tech specs of this MB to include, "For those who prefer not to push their luck at high overclock settings". You maybe able to reach 4.0 GHz @1.35vcore with an i7-870, but I doubt you can do it with an i5-750 with this MB.

Bios V1.80 is not available on their site. They are now up to V2.00. V2.00 moved bios settings around and gives you more vdimm settings lower than 1.559V. This bios also causes CPUZ or Everest to not report core voltages.
I gave their bios Easy OC feature another try. It set the bios to:
3.8Ghz, 190, Vcore – 1.33125, Vdimm-1.615, VTT – 1.174 PCH – 1.100, PLL – 1.903, Ram 8-8-8-21 @570 Mhz
Speedstep – Enabled
Intel C-State – Enabled
C State package limit setting – Auto
changed SATA mode from AHCI to IDE

With Speedstep enabled when you boot into Win 7 you immediately get a hardware BSOD. These settings, with Speedstep disabled, also failed Prime95 Blend after 6 hours and LinX (all memory) after 3 minutes.

Asrock also has new Win 7 64bit versions of their OC Tuner and IES (Intelligent Energy Saver) programs, but to enable IES you have to enable Speedstep in the bios. IES only useful feature is that it's telling me my CPU @1.325V is consuming 82W idle, 91W full load.

Taking a cue from N7, I ran LinX (all memory) on my current 3.8 Ghz oc,(1.325Vcore bios, VTT-1.209V, PLL -1.903V, PCH -1.100V, Vdimm-1.559V) which was 12 hours Prime95 Small FFts and Blend stable. It failed after 5 minutes. I tried increasing the vcore up to 1.35V and then the other voltages but it still failed after 5 mintes. I can't get it stable at 3.8 ghz, could be the CPU, MB, ram, or the CPU/ram interface. I tested the ram with Memtest86+ at 760 MHZ, 1.559V, 9-9-9-24 for 6 hours with no errors.

I dropped down to 3.7 Ghz and same voltages. It failed LinX on the 47th run. I upped the vdimm to 1.603V and it passed 3-50 runs at all memory(94-98% mem usage, problem size 21166, 3435MB, 53 Gflops, 65C load), and Prime95 Blend for 12 hours (60C load). Even though LinX saves me hours of testing, I'm still going to use Prime95 Blend as my final test. Old habits die hard. Now let's see if I can lower my ram timings.

LinX, like Intel Burn Test, is Intel's Linpack with an easy to use GUI. The author suggests running it for 50 -100 runs at all memory. The latest version can be downloaded here: http://www.xtremesystems.org/forums/showthread.php?t=201670

Final Answer:
I tried lowering my ram timings to 8-8-8-24, no go, failed LinX. So I'm stuck at 3.7Ghz, 1.325vcore bios, 1.601vdimm, VTT 1.203V, PCH 1.100V, PLL 1.903V, LinX full memory 50 runs and Prime95 Blend 12 hours stable.



Asrock P55 Pro, bios V2.00, i5-750 3.70 Ghz @1.325V bios, Cooler Master Hyper 212 Plus (2 fans) with TIM Consultants T-C Grease 0098, G.Skill Ripjaws 4GB (2 x 2GB) DDR3 1600 @ 740 Mhz 9-9-9-24, 1.6V, OCZ ModXStream Pro 700, MSI NX8800GT, Win 7 Ultimate 64bit

 

[imported_mingus]

Well, we went 170 and it seemed to run well at 3.4, then we ran OCCT. 15 minutes in temps were 49C which seemed fine, then at about 45 Minutes, error CPU Core 1. hmm.. Sort of locked up on us too. Rebooted, works fine.

Maybe raise CPU voltage slightly? No voltage changes were made.

You should be able to do 3.4 to 3.5 Ghz on stock voltage settings.

My Asrock P55 Pro MB bios settings I think are identical to your Deluxe.

First flash the MB to the latest bios. Link

Try these settings for a start:
Overclock Mode - Manual
Bclk Freq - 170 to 175 for 3.4 Ghz to 3.5 Ghz
Spread Spectrum - Disabled
QPI Frequency - anything over 150 - 160MHz base clock, select the lower multiplier
CPU Voltage ? Auto, select Manual if you have to raise voltage
VDrop Control ? Without Vdrop
Dram Voltage - Auto
Dram Freq - select the highest choice that will still keep your ram below its spec rating of 1000 Mhz (2000). This option on this MB really sucks. It only gives you 3 ram ratio choices.
VTT Voltage- Auto
PCH Voltage - Auto
CPU PLL Voltage - Auto

Under CPU Configuration:
Enhanced Halt State ? Disabled
Intel Virtualization - Disabled
CPU Thermal Throttling - Enable
No-Execute Memory Protection - Disabled
Active Processor Cores - All
A20M - Disabled
Intel SpeedStep - Disabled, this will also disable Turbomode.
Intel C-State - Disabled
Disable everything you are not using, floppy, firewire, unused SATA ports, etc.

If interested, Xbitlabs has a review of your MB. Link.

Good Luck.

 

[mav451]

Man I am probably one of the few unlucky ones with my i5 750. Only able to get it stable at 3.4ghz on ~1.25 vcore. 3 of the 4 core seems to be okay on higher speed, but this one core keep on craping out on prime95 at > 3.4ghz.

Using coolermaster 212, idle ~34C, but full load gets ~80C with only one fan, although my room temp is ~30C. If I up my vcore to >1.3, the temp will go above 85C and the pc will freeze.

Guess it's a combination of one bad core and hot temperature where I live.

Err 47C's jump from idle to load sounds a bit ridiculous. I was gonna question contact, but your idle temps are normal...just your load is off the charts. Your load temps are maybe what I'd expect at 4.2Ghz+ with HT on, but this is a 750 without HT we're talking about here.

 

[MadScientist]

Err 47C's jump from idle to load sounds a bit ridiculous. I was gonna question contact, but your idle temps are normal...just your load is off the charts. Your load temps are maybe what I'd expect at 4.2Ghz+ with HT on, but this is a 750 without HT we're talking about here.

That is a large difference between idle and load temps. At 30C ambient you would think he would be getting higher idle temps. I run about 11-12C higher idle than ambient, but I'm running at 1.325Vcore, he's at 1.25Vcore. We have the same CPU heat sink, but I'm using 2 fans. I'd question his contact too. The method for applying thermal paste to a HDT heat sink is slightly different than a flat bottom heat sink. You first have to fill in the channels between the pipes and the mounting base and then apply TIM to either the mounting base partitions, or as some do to the heat pipes, as illustrated here: http://benchmarkreviews.com/index.php?option=com_content&task=view&id=170&Itemid=1&limit=1&limitstart=5

 

[n7]

After updating to BIOS F3n, i decided to recheck my OC.

Well, re-attempt higher, mainly.
No luck though. It still requires a massive increase past current speeds to get anywhere.

21x198 (4158 MHz) DDR3-1980 9-9-8-24-74 1N
Vcore - 1.375
Vtt - 1.23
PCH - 1.05 (stock)
PLL - 1.8 (stock)
VDIMM - 1.6

 

[MadScientist]

Nice oc n7, and on stock PCH and PLL voltages. I'm about ready to fleabay my POS Asrock board, although I could have gotten a bad oc'ing i5-750. Luck of the draw.

Do you think 10 runs of LinX is enough? I just tried lowering my ram timings from 9-9-9-24 to 8-8-8-24 and LinX failed on the 24th run, but it will pass 6 hours of Memtest86+ at 7-7-7-24 all tests, and test #5 looped for 3 hours.

 

[n7]

Nice oc n7, and on stock PCH and PLL voltages. I'm about ready to fleabay my POS Asrock board, although I could have gotten a bad oc'ing i5-750. Luck of the draw.

Do you think 10 runs of LinX is enough? I just tried lowering my ram timings from 9-9-9-24 to 8-8-8-24 and LinX failed on the 24th run, but it will pass 6 hours of Memtest86+ at 7-7-7-24 all tests, and test #5 looped for 3 hours.

Memtest86+ isn't a stress test, it's an error checking test.

In short, it means that while the RAM may be error free, it does not mean it's stable (as you've discovered).

I've found when OCing, Memtest86+ is NOT a good stability indicator in modern systems with larger amounts of RAM.

It's a good initial basic check, & good for finding defective RAM.
It's not good for ensuring overclocked/tweaked stabilty.

Whenever you tighten timings/increase speed/decrease vDIMM, etc, you'll want to use something like HCI Memtest (as many instances as you have logical cores) & LinX @ maximum RAM setting for verfifying stability.

Also, while i do personally run LinX for 10 or less runs for quick checks, i'd recommend going longer to be certain, as i too have seen it error past the 10th pass many times.

I just ran my OC for 50 passes today while @ werk to be sure.
All is well ^_^

 

[rchiu]

That is a large difference between idle and load temps. At 30C ambient you would think he would be getting higher idle temps. I run about 11-12C higher idle than ambient, but I'm running at 1.325Vcore, he's at 1.25Vcore. We have the same CPU heat sink, but I'm using 2 fans. I'd question his contact too. The method for applying thermal paste to a HDT heat sink is slightly different than a flat bottom heat sink. You first have to fill in the channels between the pipes and the mounting base and then apply TIM to either the mounting base partitions, or as some do to the heat pipes, as illustrated here: http://benchmarkreviews.com/index.php?option=com_content&task=view&id=170&Itemid=1&limit=1&limitstart=5

I think my problem is the weakass fan not moving enough air. I am in the process of buying a better heatsink (a prolimatech Mega Shadow) and a more powerful fan that I can turn up when the cpu is at load, possibly add a second fan like you have. Will reapply the paste as well when changing the heatsink. Hopefully will get a little better result this time.

 

[MadScientist]

Thanks for the advice n7.
I agree. I have had other builds where my ram settings would pass Memtest86 and fail Prime95 Blend.

 

[wwswimming]

My main problem I foresee, is that I went with an Asrock P55 Pro, which has great features, but is cheap, and tom's hardware says they had stuff blowing up at 1.425v, with 2 separate boards, because the boards aren't designed for high CPU voltages, at least the P55 Pro isn't, I think the Extreme version has 16 phase, whereas mine will only have 8.

there's absolutely nothing wrong with Asus/Asrock's 8 phase power processing circuitry.

i remember the Anandtech review of the P5N32, when Asus first introduced their 8 phase design in 2006. it was absolutely the most glowing positive review i've ever heard.

good luck getting 50% OC !

 

[lopri]

I just ran my OC for 50 passes today while @ werk to be sure.
All is well ^_^

Those Linpack numbers are fascinating. According to AT server reviews Deneb/Shanghai is relatively strong with Linpack, yet it looks like i5-750 is still a good 20%~25% faster than my 955 BE, clock for clock.

How much does frequency affect the temperature of i5-750? Can you aim below 70C by lowering frequency, or frequency doesn't really matter? If you can, what's the realistic frequency I can expect while maintaining <70C temperature?

 

[n7]

Those Linpack numbers are fascinating. According to AT server reviews Deneb/Shanghai is relatively strong with Linpack, yet it looks like i5-750 is still a good 20%~25% faster than my 955 BE, clock for clock.

How much does frequency affect the temperature of i5-750? Can you aim below 70C by lowering frequency, or frequency doesn't really matter? If you can, what's the realistic frequency I can expect while maintaining <70C temperature?

Generally i find vcore is all that affects temps in a large way.
They'll likely go down slightly if i was running a lower clock, but it's minor.

Maintaining 70C under LinPack load?
-great cooling
-low vcore (& therefore lower clocks)

Keep in mind that for LinX/IBT @ max RAM, us 8 GB users are stressing more RAM than the 4/6 GB users, which creates a larger problem size & consequently results in higher temps.

I really have never worried too much about temps much with LinX as long as it's below thermal throttling, since there's nothing that ever gets even close to those temps under normal usage.
Even P95 doesn't even get close, nevermind something like gaming or folding, etc.