I5 750 overclocking

[lopri]

Originally posted by: n7
It really seems all i have to worry about is vcore...i got 4.15 GHz stable.
20x208 DDR3-2080 9-9-9-27
http://ense7en.com/pics/i5%207...20-%20SuperPi%201M.jpg
or
21x198 DDR3-1980 9-9-9-27
http://ense7en.com/pics/i5%207...44MB%2025%20passes.jpg

60 GFlops? Incredible. What's the highest multiplier for the NB?

That's the exactly where I want to be for my next build. 4.0 GHz and DDR3-2000. Your results are fantastic and very encouraging.

 

[RussianSensation]

n7 awesome results. Gskill Ripjaws are on fire with all of their kits. And they are priced pretty reasonably.

Did you only need QPI/VTT of 1.250V to get to 220 Base Clock? Did you set it to Auto in the BIOS or manually forced 1.250V?

I am not sure if it's my mobo or processor but the voltage requirements for QPI/VTT above 190 BCLK are unreasonable. I need 1.39/1.41VTT.

 

[GLeeM]

Originally posted by: RussianSensation
n7 awesome results. Gskill Ripjaws are on fire with all of their kits. And they are priced pretty reasonably.

Did you only need QPI/VTT of 1.250V to get to 220 Base Clock? Did you set it to Auto in the BIOS or manually forced 1.250V?

I am not sure if it's my mobo or processor but the voltage requirements for QPI/VTT above 190 BCLK are unreasonable. I need 1.39/1.41VTT.

I wonder if you would have better luck comparing your i7 860 to an i5 750 if you turned off hyperthreading?

On the i7 920s it makes a large difference in temps and top OC if HT is turned off!

 

[MadScientist]

Originally posted by: MacAttack

Also flash the bios to the latest version, V1.7. Use Instant Flash in the bios to flash it. Link

I am a little limited on time until this weekend. I have flashed to the 1.70 BIOS already. I reset BIOS setting to defaults to restart my OC. I followed the advice from RussianSensation. All I can say is something extremely screwy is going on with the ASRock. The setting for "TurboMode" is only there after I reset the BIOS. It simply disappears from the BIOS screen once I start trying to OC.

So, I reset BIOS again, disabled TurboMode, started setting things manually. The BIOS on ASRock is not as "smooth" looking as some other mobos out there. There is no place that I see that can set RAM ratio, QPI ratio.

I will supply details of my BIOS when I have more time this weekend. Currently, this is what I see:

QPI: can be set to 6.6 or 7.2
Ram ratio: No setting that I can see. I can set the timings and I have those at the factory recommendation on the box: 9-9-9-27
VTT (uncore): Set to 1.39
CPU V: Set to 1.35
PCH: Set to 1.12
TurboMode: Up and vanished like a fart in the wind

I may just dump this ASRock experiment and go back to GigaByte. I should have never done that experiment I think.

Hang in there Mac you should easily be able to do 3.8 and higher.

1. The Asrock bios has no QPI ratio setting. It has a QPI Frequency setting. Set it to the lowest setting, i.e., if you set Overclock Mode to Manual and Blk Freq to 170 then set QPI Freq to 5.440GT. Its default is Auto. Highlight it, hit enter and you will see the other choices. QPI Freq will change as you increase or decrease BLK Freq. Set it to the lowest choice.

2. This MB does not have RAM ratio settings, per se, but under DRAM Freq it gives you ram freqs choices. The default setting is Auto. Pick one that will keep your ram under its 1600 rating, i.e., for 170 base clock, my bios gives 3 choices: 510Mhz DDR3_1020, 680Mhz DDR3_1360, and 850Mhz DDR3_1700. Choose 680Mhz or for this base clock Auto.

3. If you disable SpeedStep on this MB it will also disable TurboMode, and the TurboMode setting will disappear, vanish like a fart in the wind. You can disable TurboMode and still enable SpeedStep.

Now where the hell is that new bios that Asrock told TH it was coming out with to provide overload protection.

 

[bryanW1995]

Originally posted by: RussianSensation
n7 awesome results. Gskill Ripjaws are on fire with all of their kits. And they are priced pretty reasonably.

Did you only need QPI/VTT of 1.250V to get to 220 Base Clock? Did you set it to Auto in the BIOS or manually forced 1.250V?

I am not sure if it's my mobo or processor but the voltage requirements for QPI/VTT above 190 BCLK are unreasonable. I need 1.39/1.41VTT.

sounds like my dud of an e6750, I had to really crank up the settings to get it over 3.6.

 

[themusgrat]

Just replying a bit here, I can get a stable 3.8 at 1.3 volts, I'm haven't even tried lower voltage though because that won't hurt it, but also because I then tried a blend on P95, and turns out my RAM is bad, it isn't even stable at below stock speeds and overvoltages. So I'm dling the new BIOS now, and if that doesn't magically fix it, I'll be RMAing for a gskill set. This is taking forever.... Bought these parts like 2 weeks ago and still can't get a stable build.

 

[RussianSensation]

Originally posted by: bryanW1995

sounds like my dud of an e6750, I had to really crank up the settings to get it over 3.6.

Ya I am going to have to revisit it over the weekend. It is Prime95 24 hours stable at 1.312V at 3.91ghz. Getting 4200mhz is probably doable but for that extra 8% performance increase my temperatures go from high 60s/low 70s into the mid 80s! Likely not worth it. Also Anandtech's article on Foxconn pin failures on 1156 boards at high overclocks makes me want to 2nd guess pumping too much voltage for an extra 5-8% performance (thus exponentially increasing my power consumption).

 

[MadScientist]

Originally posted by: RussianSensation
Also Anandtech's article on Foxconn pin failures on 1156 boards at high overclocks makes me want to 2nd guess pumping too much voltage for an extra 5-8% performance (thus exponentially increasing my power consumption).

Great, besides blowing up transistors and capacitors on the Asrock MB now I also have to be concerned about the CPU making complete contact with the socket. The Asrock P55 MBs have Foxconn sockets.

 

[n7]

Yeah there's no question the i7 860/870 require more work to OC.

LOL wow.

I just passed 23 hrs of P95 Blend.

And i started up LinX. 3 passes. That's all it took to error.

No idea why i bother w/ P95...massive waste of time on this platform.

http://ense7en.com/pics/i5%207...0but%20LinX%20FAIL.jpg

 

[RussianSensation]

I am pretty sure that 1.27V QPI/VTT is way too low for such a high Base Clock frequency. Have you tried increasing VCore Voltage to 1.42? Try that and see if it passes more than 3 LinX tests. If it doesnt, then the bottleneck is likely the QPI, or PLL.

Also Blend test was never designed to test focus on the CPU. So the fact that it passed is more or less expected. If you pass 24 hrs under Small/Large FFTs and then failed LinX then something is wrong continuing to use Prime95 as you said. Then again no 1 single test is an indicator of 100% stability.

 

[MadScientist]

Asrock's V1.80 bios for the P55 Pro now states on their website "Add CPU Vcore over current protection", but it's dated 10/7/09, 1 day before the TH's article was posted on their website. Is this V1.90 and they forgot to rename it and change the date or the same old V1.80? I have had V1.80 on my MB since last week and reflashed it to the new? V1.80. It flashed, no "you already got it" messages.
I'm going to start raising the CPU voltage above 1.36V and see what happens.

 

[n7]

Originally posted by: RussianSensation
I am pretty sure that 1.27V QPI/VTT is way too low for such a high Base Clock frequency. Have you tried increasing VCore Voltage to 1.42? Try that and see if it passes more than 3 LinX tests. If it doesnt, then the bottleneck is likely the QPI, or PLL.

Also Blend test was never designed to test focus on the CPU. So the fact that it passed is more or less expected. If you pass 24 hrs under Small/Large FFTs and then failed LinX then something is wrong continuing to use Prime95 as you said. Then again no 1 single test is an indicator of 100% stability.

I decided to try upping VTT/PLL/PCH significantly, but nope, did crap all.

It really just seems that around 4.15+ GHz, i need tons more vcore to be stable.

Extra mobo voltages do nothing to help.

I can do P95 & HCI Memtest all day long @ 4.2 GHz, but i like to have everything pass, so i'm doing 21x198 (4158 MHz), & am working on tuning voltages as low as possible.

 

[RinksCustoms]

Someone did a review on "an" ASROCK P55 board while trying to OC it and ended up frying the mobo, make sure you reflash your BIOS. The boards that got smoked had the factory BIOS ver and did nothing to current limit the CPU phases. I dont know if it was your board or not, just remember seeing issues with 1156 boards from ASROCK, MSI, and ECS...

I think it was over on TH.

 

[MadScientist]

I'm done oc'ing the CPU for now. I'll work on the ram timings next.
Too many voltage variables, so I tried out the auto oc bios feature again and used what voltages it gave me, except for the CPU voltages which were, IMO, too high.
I also added a second fan to the heatsink. It's RPMs, CFMs and design are close to the Cooler Master fan that came with the Hyper 212 Plus. Cooler Master specs it at 19dB-A. It's quiet but 19 dB-A it ain't. It did bring CPU temps down ~3C.

Here's a few that were Prime 95 (Small FFts) 12 hours stable, ambient temp 22C, same bios settings as prevoius post:

Core Speed: 3.8 Ghz, Base Clock: 190, Bios CPU voltage: 1.325V, CPUZ: 1.312V idle, 1.320V load, VTT: 1.234V, PLL: 1.903V, PCH: 1.10V, Temps: 32C idle, 58C load, Mem Freq: 761 Mhz, mem voltage: 1.50V
No one has said much about the PCH voltage, but if I set it to Auto it failed.

Core Speed: 3.9 Ghz, Base Clock: 195, Bios CPU voltage: 1.3625V, CPUZ: 1.352V idle, 1.360V load, VTT: 1.279V, PLL: 1.903V, PCH: 1.120V, Temps: 33C idle, 64C load, Mem Freq: 782 Mhz, mem voltage: 1.50V
Again, if I set PCH voltage to Auto it failed.

I tried for 4.0 Ghz but it looks like it's going to need 1.4V or more. I'm still reluctant to go this high until I see more on this from TH.
Core Speed: 4.0 Ghz, Base Clock: 200, Bios CPU voltage: 1.3875V, CPUZ: 1.376V idle, 1.384V load, VTT: 1.309V, PLL: 1.903V, PCH: 1.120V, Temps: 34C idle, 67C load, Mem Freq: 800 Mhz, mem voltage: 1.50V. Good temps but failed after ~ 1 hour.

So for now I'm staying at 3.8 Ghz.

10/21
Ok, one more tweak, maybe, and I'm done. At 3.8 Ghz I lowered the VTT to 1.204V, below Intel's recommended 1.21V max. It passed 12 hours Prime 95. If I set it to Auto I get a BSOD before it boots to Windows. I tried setting the PLL to 1.812V but it fails in about 30 minutes. I might try some more PLL voltages between 1.812V and 1.903V.


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

 

[themusgrat]

Looks about like what I've got too, at 3.8, I even had around 1.3 volts being ok, though I never ran Linx. Should be getting my Ripjaws soon, those things are seriously a steal, based on everything everyone says about them.

And ya, looks like 3.8 will be the sweet spot for long term OCing for this chip. It seems that with much lower temps, you should be able to push it much further than a 1366 I7, but in reality I guess they changed too many things to be able to count on getting past 4ghz. I would be interested to know however, if anyone can get past 4ghz in any meaningful measure, by upping the PCI-E speeds a bit. I think it was MSI that claimed that speeding that up would allow for higher OCs, and I'm interested to see if that still applies with the I5's onboard PCI-E.

 

[Munky]

Isn't the max safe VTT/QPI voltage around 1.2v for Lynnfield? How are you all recommending closer to 1.4v?

 

[AznAnarchy99]

Im having trouble hitting anything on my i7 860. I was thinking about turning off HT but wouldnt that negate the whole reason of purchasing an 860 over an i5?

Quick question, Im running a yate loon low speed on my Megahelem right now. How would temps be if I throw a High Speed Yate loon on push and put the Low Speed on pull?

 

[themusgrat]

They'd definitely improve, not by much though, prolly around 5C. But what do you mean by not hitting anything? Are you having trouble at 3.4, or 4, or what? Make sure you check your RAM, I couldn't hit past 3.2 without errors till I figured out it was my RAM, then just testing the CPU itself with as little RAM as possible gave me a 3.8 easy overclock.

 

[AznAnarchy99]

I hit 3.5 fine. Im actually testing 3.6 as we speak. Im using Linx as it detects failures much faster than Prime95 but Ill eventually try both. At 3.8 it just locks up soon after I start stressing, using 21x, 181 blck, 1.35 CPU, 1.312 VTT.

Update:
3.6 is stable with 21x, 172 blck, 1.35 CPU, and 1.312 VTT. Max temp was 75 and idle is 35 so not too bad.

I want to hit that 3.8 to 4.0 though really bad but I can't seem to keep it stable =(

You think 5C is worth that $10?

 

[MadScientist]

Originally posted by: munky
Isn't the max safe VTT/QPI voltage around 1.2v for Lynnfield? How are you all recommending closer to 1.4v?

I and others here, and in other forums, have asked the same question. I think what reviewers are saying and doing is that for higher overclocks, ~3.8Ghz and higher, you have to raise the VTT/QPP voltage above Intel's recommended 1.21V max. to get a stable oc.

From Intel® Core? i7-800 and i5-700 Desktop Processor Series Datasheet 1, p. 67:
At conditions outside functional operation condition limits, but within absolute maximum and minimum ratings, neither functionality nor long-term reliability can be expected. If a device is returned to conditions within functional operation limits after having been subjected to conditions outside these limits (but within the absolute maximum and minimum ratings) the device may be functional, but with its lifetime degraded depending on exposure to conditions exceeding the functional operation condition limits. At conditions exceeding absolute maximum and minimum ratings, neither functionality nor long-term reliability can be expected. Moreover, if a device is subjected to these conditions for any length of time it will either not function or its reliability will be severely degraded when returned to conditions within the functional operating condition limits.

Right now I'm running my i5-750 at 3.8Ghz @ VTT voltage 1.204V.

 

[MadScientist]

Originally posted by: AznAnarchy99
I hit 3.5 fine. Im actually testing 3.6 as we speak. Im using Linx as it detects failures much faster than Prime95 but Ill eventually try both. At 3.8 it just locks up soon after I start stressing, using 21x, 181 blck, 1.35 CPU, 1.312 VTT.

Update:
3.6 is stable with 21x, 172 blck, 1.35 CPU, and 1.312 VTT. Max temp was 75 and idle is 35 so not too bad.

I want to hit that 3.8 to 4.0 though really bad but I can't seem to keep it stable =(

You think 5C is worth that $10?

For 3.6 Ghz that's a bit toasty. Everyone's CPUs and MBs are different, but at 3.6 Ghz it should not take 1.35V to be stable. Are you running your ram below it's specs as RS suggested?

 

[AznAnarchy99]

Ran ram below but 3.6 isn't stable at any voltage below that. I tried 3.8 today with every voltage setting I could but nothing would make it stable. Even tried turning HT off, make it stable for an extra 10 minutes then it had errors. Guess its just a bad OC chip?

 

[MadScientist]

Another bios update for the Asrock P55 Pro MB was posted today, V1.90. Could this be the one that will enable built-in over current protection?

 

[MrK6]

Originally posted by: AznAnarchy99
I hit 3.5 fine. Im actually testing 3.6 as we speak. Im using Linx as it detects failures much faster than Prime95 but Ill eventually try both. At 3.8 it just locks up soon after I start stressing, using 21x, 181 blck, 1.35 CPU, 1.312 VTT.
Update:
3.6 is stable with 21x, 172 blck, 1.35 CPU, and 1.312 VTT. Max temp was 75 and idle is 35 so not too bad.
I want to hit that 3.8 to 4.0 though really bad but I can't seem to keep it stable =(
You think 5C is worth that $10?

A new fan wouldn't hurt, but I'd get something like this: http://www.jab-tech.com/120mm-...-fan-M1BX-pr-3110.html
A thicker fan (38mm as opposed to 25mm) really helps push air through those tightly spaced fins.

Currently I'm testing 3.8GHz (190x20) using 1.325V Vcore, 1.21V VTT, 1.64V Vdimm (rest of specs in sig). I might be able to go a bit lower on the Vcore, I'll see. Basically, I've met the best success with higher overclocks using a lower BCLK and keeping voltages as low as necessary.

 

[MadScientist]

Originally posted by: MrK6

Originally posted by: AznAnarchy99
I hit 3.5 fine. Im actually testing 3.6 as we speak. Im using Linx as it detects failures much faster than Prime95 but Ill eventually try both. At 3.8 it just locks up soon after I start stressing, using 21x, 181 blck, 1.35 CPU, 1.312 VTT.
Update:
3.6 is stable with 21x, 172 blck, 1.35 CPU, and 1.312 VTT. Max temp was 75 and idle is 35 so not too bad.
I want to hit that 3.8 to 4.0 though really bad but I can't seem to keep it stable =(
You think 5C is worth that $10?

A new fan wouldn't hurt, but I'd get something like this: http://www.jab-tech.com/120mm-...-fan-M1BX-pr-3110.html
A thicker fan (38mm as opposed to 25mm) really helps push air through those tightly spaced fins.

Currently I'm testing 3.8GHz (190x20) using 1.325V Vcore, 1.21V VTT, 1.64V Vdimm (rest of specs in sig). I might be able to go a bit lower on the Vcore, I'll see. Basically, I've met the best success with higher overclocks using a lower BCLK and keeping voltages as low as necessary.

Different MB & ram, same CPU, but same Vcore and ~VTT voltage for 3.8 Ghz. What is your PLL voltage and did you have to raise the PCH voltage?