Archived

This topic is now archived and is closed to further replies.

PC Pilot Dave

Question about Overclocking?

Recommended Posts

Hello all,

 

I'm in the process of saving and planning to build a computer to run FSX. I currently run it on a laptop (which has been a bit rough), but I am hoping to build a new machine to run FSX on sometime this winter. I have been reading up online about what works and what does not work for an FSX computer. From what I understand, in order to get the most out of FSX it is necessary (or at least very recommended) to overclock your CPU. I have experience building computers, but this would be my first time overclocking a CPU. So my question right now is, would it be necessary to watercool this system if I am overclocking the CPU? I also have never watercooled a system, so this would be two new steps for me and I want to make sure I am prepared before I start investing into a system.

 

If it helps at all, here are the parts I am tentatively looking at. For some reason, PC Part Picker only allows me to select one fan, but with this current build, I'd plan to use every fan slot available in this case (I believe it is 5).

 

http://pcpartpicker.com/user/GLaDOS/saved/#edit_description_442881

 

Any thoughts or suggestions regarding the watercooling (or even the system in general) would be greatly appreciated!

 

Thank you!

Share this post


Link to post
Share on other sites
Help AVSIM continue to serve you!
Please donate today!

Quote

would it be necessary to watercool this system if I am overclocking the CPU?

No, a full blown custom water cooling setup is not required.

 

I run an i73770K Ivy Bridge. My cooler is a Noctua NH-D14 air cooler. I'm overlooked to 4.5GHz. My temps in FSX are 55 degrees. Nice and cool and very quiet.

 

Another option is a closed loop water cooler, they require zero maintenance and are very easy to install. Leaks are a possibility though.

 

 

Temps here, overclocked and under load...

 

http://www.guru3d.co..._review,13.html

 

H100i 76 degrees

Kraken X60 70 degrees [Noisy higher RPM fans]

Corsair H110 72 degrees [Very quiet]

NH-D14 72 degrees [Very quiet]

 

Noise levels here...

 

http://www.guru3d.co..._review,11.html

 

NH-D14 and H110 are the best in terms of heat sink/radiator efficiency. If they fit in your case.

Share this post


Link to post
Share on other sites

Hi Martin,

 

Thank you for the information! I think, given all the options, I'd prefer to stick with traditional fans. Something about running water through my system makes me nervous haha.

 

Thanks again! I appreciate the help!

Share this post


Link to post
Share on other sites

Glad to help.

 

To be fair, leaks from closed loop water coolers seem to be quiet rare. but of course we can't say precisely how common they are, as we don't have access to the manufacturers sales, and failure figures.

 

My opinion is similar to yours, why install a water cooler and with it the small risk of a leak, when something like my NH-D14 is guaranteed leak free, and cools just as well.

Share this post


Link to post
Share on other sites

You definitely need to buy a better cpu cooler than the one listed.  I bet the stock cooler is better than Rosewill.

Share this post


Link to post
Share on other sites

Thanks Martin! We definitely agree on that point!

 

Hi IFR700. I'm not too up to date on what the higher-quality cooling fans are. Would you be willing to make any recommendations? What about Martin's NH-D14?

Share this post


Link to post
Share on other sites

Thanks Martin! We definitely agree on that point!

 

Hi IFR700. I'm not too up to date on what the higher-quality cooling fans are. Would you be willing to make any recommendations? What about Martin's NH-D14?

 

Dave... have a read of the links I gave you above. all of the top coolers are there.

 

 NH-D14 and H110 are the best in terms of heat sink/radiator efficiency. If they fit in your case.

 

If you want to save some money, but still a achieve a reasonable overclock, the Coolermaster Hyper 212 Evo is a good choice.

 

But as I said,  reviews will tell you all you need to know.

Share this post


Link to post
Share on other sites

Dave... have a read of the links I gave you above. all of the top coolers are there.

 

 NH-D14 and H110 are the best in terms of heat sink/radiator efficiency. If they fit in your case.

 

If you want to save some money, but still a achieve a reasonable overclock, the Coolermaster Hyper 212 Evo is a good choice.

 

But as I said,  reviews will tell you all you need to know.

 

Hi Martin. Whoops - sorry, I forgot to go check those outs before I posted. I'll take a look at them today.

 

Thanks again for the great info!

Share this post


Link to post
Share on other sites

I have a 3770K OC'd @ 4.5 Ghz paired with a $30 CM Hyper 212 EVO and 2 fans in a push-pull arrangement and am very pleased with the performance. 4.5 Ghz seems to be the accepted "ceiling" of that chip without de-lidding. I've employed water cooling in my builds but went back to air for various reasons. Proper case airflow (location and placement of case fans) and a decent aftermarket air cooler are perfectly adequate for a moderate overclock. Regards

Share this post


Link to post
Share on other sites

Yep, Adam is spot on, the Hyper 212 EVO is the best bang for your buck when it comes to coolers.

 

Agree regarding 4.5 for 3770K as well. Mine will do 4.6 without any trouble, but 4.7 is a different story.

 

4.7 and the associated voltage is the point at which the Intel TIM becomes more of an insulator than conductor, and temps shoot up dramatically for me.

Share this post


Link to post
Share on other sites

The cooling solutions on the market now were designed for Sandy Bridge.

For Ivy and Haswell you have to turn to the mainstreem custom water cooling components. No delidding necessary 5GHz!

 

HLJAMES

Share this post


Link to post
Share on other sites

Martin, Adam,

 

What CPU voltages are you seeing in CPU-Z for your 3770K overclocks at 4.5 ghz?  I'm seeing 1.17V with a 80C CPU temperture, 68F room temperature, and using Aida64 FPU Stress test.  CPU temps in FSX are much lower.  Just curious what others are using in case I ever experience any instability in FSX.

 

Ted

Share this post


Link to post
Share on other sites

No way you are seeing 1.17 Ted, not under load. That sounds like the voltage at idle.

 

4.5 GHZ for me, FSX is around 55 or so, with an ambient of 20C. NHD-14 cooler.

 

Volts are 0.030 offset, which gives me 1.304 under load.

 

At idle volts are  1.168. Frequency drops to 1600.03 MHz.

 

Don't have Aida. In Prime blend test I'm at core temps of 64, 73, 69, 67. [Average 68.25] 

 

Ambient is high today in the UK, currently 24C. So that's a delta of 44.25.

Share this post


Link to post
Share on other sites

The cooling solutions on the market now were designed for Sandy Bridge.

For Ivy and Haswell you have to turn to the mainstreem custom water cooling components. No delidding necessary 5GHz!

 

HLJAMES

 

Wouldn't agree with that James. Aftermarket coolers aren't really designed for a given architecture. Rather, manufacturers compete with each other to produce the best cooling for a given price point.

 

NH-D14, Silver Arrow, H110... are all perfectly capable of handling an overclock up to 4.6 on Ivy bridge, which is plenty for most people.

 

As for the notion that we "have to turn" to a full blown custom loop for Ivy/Haswell, then no, that's not true at all. Further more, there are many out there with full blown custom loops who have still found that high overclocks are difficult. The reason for that is because it's not so much the cooling that's the issue with Ivy/Haswell, it's actually the fact that the thermal interface material under the IHS starts to act more like an insulator above a certain temp/voltage. 

Share this post


Link to post
Share on other sites

Correction to the above TED.

 

For the first time, I'm now getting BSOD's at that voltage. Error 0x00000101.

 

Have had to up the voltage. Now at 1.320 under load. I always new this CPU needed high volts, but it seems higher than I thought. Been fine for ages, but now prefers higher.

 

FSX temps at 20C ambient should now b around 59.

Share this post


Link to post
Share on other sites

Martin, When I started overclocking my 3770k I was using Prime 95 but then I came across the video below in which an Asus rep recommended using Aida64 and also stated that Prime95 could damage a 3770k. See time 3:50 - 4:30. This video was created in April 2012 so later updates to Prime95 may have corrected this.



I switched to Aida64 after seeing this. Going back to my notes I did document that at 4.5ghz the CPU required an extra 0.016V Vcore and ran 10C hotter stress testing with Prime95 (can't remember the type of test I used) than Aida64 at its default stress test settings. The Aida64 default settings stress the CPU, FPU, Cache, and Memory. After doing some more research I found out that only running the FPU test in Aida64 creates the highest CPU temperatures. Running the Aida64 FPU only stress test resulted in CPU temps 10C higher than running the default test with all 4 tests selected, so I think close to the Prime95 temperatures.

I just went back and ran a few tests again before writing this. Under load the FPU only test had a max temperature on any core of 80C. Testing with the default settings (CPU, FPU, Cache, and Memory) I had a max single core temperture of 71C in Real Temp. I ran the tests for 10 minutes each and the room temperature was 65F (18.3C). So I've got a max delta of 61.7C under FPU stress test, but as we discussed before, my air cooler isn't nearly a good as yours.

In both tests the vcore readings in CPU-Z varied from 1.168 to 1.184V. After the tests have been running for a few minutes the voltages seem to stabilize at 1.168, but they do jump up to 1.184 at times so I should quote that value instead of 1.170.

I have my bios set at a CPU offset of -0.035 volts and I am using an LLC setting of Ultra High plus a few other changes to the VRM. When I first started overclocking I used the manual method and used the Asus AI Suite2 Software because it allowed me to very quickly change CPU ratios and voltages without rebooting. However after I while I discovered that the AI Suite2 software had a number of problems and was also confilicting with my other software. After uninstalling AISuite2 I found that the overclocks were a lot more stable by setting them up in the bios.

Using the bios manual Vcore method I found that I would pass Aida64 at 1.165 volts but not at 1.160. A 1.165V bios manual settings yielded 1.168-1.176V reading in CPU-Z under stress load with an LLC of Ultra High. I found that this LLC setting allowed my CPU-Z reading to remain close to my bios voltage setting under load. I don't know if this is the correct way to use the LLC setting but it seems to work.

I then switched over to the offset Vcore mode and found that a -0.035 offset yielded the same CPU-Z vcore readings under stress test load so I left it there. I also have EIST enabled but not Cstates. I also have Hyperthreading deactivated. With my new settings my idle voltage is 0.868V at 1600mhz. With the Asus default optimized bios settings my idle vcore in CPU-Z was 0.888V. It seems to be stable in FSX so far (knock on wood) however I am running at a vcore voltage only .005V above not passing Aida64. I may find that I need to increase this in the future and that was the reason for my question in the post above.

Disclaimer: I am by know means an expert at this. Everything I did above was a result of scouring the internet for hundreds of hours over the last few months. I can't believe that the motherboard manufacturers do not provide better documentation on what their bios settings do and how they should be used. I know that I got a little long winded here but I thought my explanation might help others that are in the same bewildered position that I was in 6 months ago.

Ted

Share this post


Link to post
Share on other sites

Hi Ted.

 

Martin, When I started overclocking my 3770k I was using Prime 95 but then I came across the video below in which an Asus rep recommended using Aida64 and also stated that Prime95 could damage a 3770k.

 

 

Yep, I too have seen the video. In fact when I initially overclocked my CPU I used Aida 64 as well. I have found though, that the latest version of Prime does recognise the CPU, and seems to be absolutely fine for stress testing Ivy Bridge. My trial version of Aida has now expired. 

 

In both tests the vcore readings in CPU-Z varied from 1.168 to 1.184V. After the tests have been running for a few minutes the voltages seem to stabilize at 1.168, but they do jump up to 1.184 at times so I should quote that value instead of 1.170.

 

 

 

I'm astonished to be honest Ted. Either you have a great chip, or I'm doing something wring. Wouldn't be surprised if it was the latter.

 

 

A 1.165V bios manual settings yielded 1.168-1.176V reading in CPU-Z under stress load with an LLC of Ultra High. I found that this LLC setting allowed my CPU-Z reading to remain close to my bios voltage setting under load. I don't know if this is the correct way to use the LLC setting but it seems to work.

 

 

 

That may be the variable. I may not be using a high enough LLC.

 

I then switched over to the offset Vcore mode and found that a -0.035 offset yielded the same CPU-Z vcore readings under stress test load so I left it there.

 

 

 

Now that's interesting. You are using a minus off-set. I'm using a + off-set.

 

 

I know that I got a little long winded here but I thought my explanation might help others that are in the same bewildered position that I was in 6 months ago.

 

 

 

No, Ted, trust me, your post has been very useful. Some interesting differences in our approach there. I too am far from an expert in regard to overclocking Ivy. Anything else you could add would be appreciated.

 

I need to research the difference between an off-set minus, and plus. LLC also seems to be perhaps my issue.

 

 

 

 

 

Share this post


Link to post
Share on other sites

Done some checking. All my settings are okay. In the case of my CPU, it's not one of the best in the silicone lottery. Quite a few out there that need same volts as me to achieve 4.5.

 

De-lidding would help of course.

 

Very warm day yesterday in the UK, CPU temp higher, voltage requirement therfore higher. That's my theory anyway.

Share this post


Link to post
Share on other sites

The -0.035V Vcore offset was what it took to match the same CPU-Z Vcore reading under the Aida64 Stress test that I obtained using the manual bios setting of 1.165V. I did all my overclocking adjustments and testing using the manual Vcore method in the bios until I felt I had a stable setup.

 

WIthout using LLC I found that I had to input a much higher Vcore into the bios to overcome Vdroop under load.

 

The changes that I have made in my bios that are different from the default optimized settings are:

 

AI Tweaker Page

AI Overclock Tuner: XMP

Turbo Ratio: Manual

Ratio Synchronizing Control: Enabled (This sets all core to the same ratio as the first core)

1 core ratio: 45

Internal PLL Overvoltage: Enabled

Memory Frequency: DDR3-2133

CPU Load-Line Calibration: LLC

CPU Voltage Frequency: Manual

CPU Fixed Frequency: 350

CPU Power Phase Control: Optimized

CPU Voltage: Offset Mode

Offset Mode Sign: Minus

CPU Offset Voltage: 0.035

Dram Voltage: 1.5

CPU Spread Spectrum: Disabled

 

Advanced Page

Hyper-Threading: Disabled

CPU C3 Report: Disbaled

CPU C6 Report: Disabled

Package C State Support: Disabled

Primary Display: PCIe

Bluetooth Controller: Disabled

Wi-Fi Controller: Disabled (I'm on a LAN connection)

 

I may have some bad settings in here but they seem to work so far. If anyone sees any settings that should not be used or could be improved please let me know. The fact that there is no formal documentation from Intel or the motherboard manufacturers on the proper method to overclock amazes me. If there is, I haven't found it yet.  To me this is like buying an airplane and having the manufacturer say "Manuals? Oh, if you want to learn how to fly it or maintain it, just Google it"   :lol:

 

Hope these help,

Ted

Share this post


Link to post
Share on other sites

Very interesting Ted.

 

I have just noticed that I had accidentally set LLC to high, my previous setting was Auto. So I've just set back to auto LLC... and now, using my previous +0.030 offset all is stable again and back to low temps.

 

Go figure. I guess Auto is giving me a tad more LLC, than high. I'll have to see how it goes from here.

 

 

------------------------------------------------------------------------------------------------------------------------

 

 

These are my settings...

 

 

i73770K Overclocking to 4.5GHz

 

UEFI Advanced Mode F7

 

Ai Tweaker

AI Overclock Tuner: XMP

BCLK/PEG Frequency: 100

Asus Multi-Core Enhancement: Enabled

Turbo Ratio: Manual

Ratio Synchronizing Control: Enabled

Core Ratio's: 45

Internal PLL Overvoltage: Auto

CPU Bus Speed: Auto

Memory Frequency: 2133 [Set by XMP]

EPU Power Saving: Disabled

OC Tuner: Leave Alone

Dram Timing Control: Set by XMP

 

CPU Power Management:

Ratio Auto

Speed Step Enabled

Turbo Mode Enabled

Turbo Mode Parameters All Auto

 

DIGI Power Control

Load Line Calibration: Auto

CPU Power Phase Control: Optimised

CPU Voltage: Offset

Offset Voltage:  + 0.030 [Results in I.304 Volts Under Load]

 

 

Hyper Threading: Off

Share this post


Link to post
Share on other sites

Well I decided to stress test with OCCT:Linpack with 64bits and AVX enabled. I'm getting an error message after about 7 minutes so I guess I'm not as stable as I thought :blush:

 

This just goes to show you PC Pilot Dave, don't take what you read on the forums as gospel.

Ted

Share this post


Link to post
Share on other sites

I had to raise my vcore a bit to a manual setting of 1.19V to become stable for an hour run of OCCT:Linpack. A Vcore offset setting of -0.01V gave me the same vcore reading in CPU-Z under load. The -0.01V CPU offset voltage is my new CPU voltage setting and all of the others are still the same as above. Again don't use these settings as a template, but rather as a comparison reference. I don't know if this is the correct way to overclock, but it seems to be working for me so far.

 

OCCT:Linpack runs hotter than AIDA64. My baby Noctua was pushing 87C with these tests, so 4.5 ghz looks like my limit with this cooling system. Room temperature was 67F.

 

Ted.

Share this post


Link to post
Share on other sites

I had to raise my vcore a bit to a manual setting of 1.19V to become stable for an hour run of OCCT:Linpack.

 

I have to say, 1.19 is very low for 4.5 Ted. Take at look at the Ivy Bridge stable club volts in the link below. As you can see, your results are among the best there. Think you have a great chip there to be honest. When you get a better cooler you should be able to overclock quite impressively for a non de-lidded Ivy. I think your chip is in the top 5% actually. If it remains stable of course.

 

http://www.overclock.net/t/1247869/official-the-ivy-bridge-stable-suicide-club-guides-voltages-temps-bios-templates-inc-spreadsheet

 

I have to say though, just an hour in OCCT may not be enough to ensure total stability. The real test is if it runs your software long term with stability. That's what caught me out the other day. It wasn't a stress test, it blue screened during BF3.

 

Synthetic stress tests are fine, great tools, but we build our systems to be stable with our usual software.

 

As for your method, at the end of the day, whether high LLC, auto LLC, minus off-set, or positive off-set, it's actually about providing the CPU with enough volts. So if your method works, it works and all is well.

 

If I switched to a negative off-set, I don't think I would achieve anything better, I still need the same voltage for my less than impressive CPU.

Share this post


Link to post
Share on other sites

I agree Martin that I may not be stable with only one hour of OCCT testing. Thanks for posting that table. I am at the low end of the voltage range for my overclock. If I run into a stability problems with FSX raising the Vcore will probably be the first thing I try to resolve it. I don't feel comfortable running a 17+ hour Prime95 test with the CPU cooler I have. I cannot fit a larger air cooler in the case that I have and I don't really want to switch to water cooling. However, if I am faced with making a cooling system change or reducing my overclock in the future I'll probably be purchasing some new hardware.

 

Ted

Share this post


Link to post
Share on other sites