The TBG Ultimate Gaming Mini-ITX PC Build

[Steve N. Mavronis]

Steve - the Gallery profile of your sweet system is now live. Thanks for this great submission!

Thank you very much Ari, it looks nice! I've settled on a manual UEFI BIOS overclock for my 6700K of 4.6 Ghz and 3200 Mhz XMP profile for my TridentZ memory.

[Steve N. Mavronis]

I settled on manually overclocking my i7-6700K to 4.5Ghz at 1.32V and enabled core voltage adaptive mode. It can also do 4.6Ghz at 1.33V-1.35V but decided to choose stability and lower core temperatures over speed. At 4.6Ghz running the Asus RealBench test I monitored with CPUID HWMonitor and a couple cores hit 80c under load. At 1.32V an OC of 4.6Ghz crashed during the heavy multitasking test. At 4.5Ghz they only reached 78c so cooler at least.

A lot of this just ends up being bragging rights to see what you can achieve. But in the end we want our computers to last a long time. I think a 12% CPU speed boost from 4.0Ghz to 4.5Ghz is still good.

Frankly before I built this PC while researching my parts (https://pcpartpicker.com/b/96JV3C) wasn't really expecting to even hit 4.5Ghz so I'm very happy with it indeed, especially with my G.Skill DDR4 TridentZ memory being at the rated 3200Mhz using XMP.

[Ari Altman]

You're using a great board, and you seem to have a pretty good chip. There are a lot of examples on the internet of people running 4.7GHz at 1.4v, but frankly I am not in the least bit impressed with that. When their processors end up degraded by the end if the year, I hope they realize why. Skylake is not some miracle chip... It overclocks just like the previous three generations, which means you top out at 4.5GHz if you intend to actually be using your chip for a few years.

I personally run all my chips at 4.4GHz (3770K, 4690K, 4770K, 5820K, 6600K, and 6700K), and the next big TBG article will compare them all in some of the latest games. Stay tuned!

[Steve N. Mavronis]

Skylake is not some miracle chip... It overclocks just like the previous three generations, which means you top out at 4.5GHz if you intend to actually be using your chip for a few years. I personally run all my chips at 4.4GHz (3770K, 4690K, 4770K, 5820K, 6600K, and 6700K), and the next big TBG article will compare them all in some of the latest games. Stay tuned!

So what is the 6700K estimated lifetime at 4.4Ghz compared to 4.5Ghz?

What's that like a 10% improvement over 6700K stock 4.0Ghz?

Maybe 1.3V or 1.31V would be a starting point to test a 4.4Ghz overclock at for stability?

[Ari Altman]

Skylake is not some miracle chip... It overclocks just like the previous three generations, which means you top out at 4.5GHz if you intend to actually be using your chip for a few years. I personally run all my chips at 4.4GHz (3770K, 4690K, 4770K, 5820K, 6600K, and 6700K), and the next big TBG article will compare them all in some of the latest games. Stay tuned!

So what is the 6700K estimated lifetime at 4.4Ghz compared to 4.5Ghz?

What's that like a 10% improvement over 6700K stock 4.0Ghz?

Maybe 1.3V or 1.31V would be a starting point to test a 4.4Ghz overclock at for stability?

It's hard to say where degradation starts to occur, but I think that as long as you're under 1.35v you'll be ok.

My 6700k runs at 4.4GHz with 1.29v. Skylake definitely requires more voltage than the previous two generations, which needed about 1.2v for the same overclock.

[Steve N. Mavronis]

My 6700k runs at 4.4GHz with 1.29v. Skylake definitely requires more voltage than the previous two generations, which needed about 1.2v for the same overclock.

Right now I have my 6700K at 4.4Ghz at 1.28V and Asus RealBench passed so far I didn't try lower than 1.275V yet. I'm debating on attempting 1.26V. CPUID HWMonitor reported the highest core temp was 71c at full stress test and the package TDP stayed less than 95W in the 80's.

What do you think? Should I remain at 1.28V anyway just for a little overhead and stability?

[Ari Altman]

My 6700k runs at 4.4GHz with 1.29v. Skylake definitely requires more voltage than the previous two generations, which needed about 1.2v for the same overclock.

Right now I have my 6700K at 4.4Ghz at 1.28V and Asus RealBench passed so far I didn't try lower than 1.275V yet. I'm debating on attempting 1.26V. CPUID HWMonitor reported the highest core temp was 71c at full stress test and the package TDP stayed less than 95W in the 80's.

What do you think? Should I remain at 1.28V anyway just for a little overhead and stability?

That sounds very good to me. Low temps and safe voltage. Up to you whether you want to take the time to stability test at lower voltage.

[Steve N. Mavronis]

That sounds very good to me. Low temps and safe voltage. Up to you whether you want to take the time to stability test at lower voltage.

Skylake seems to be a very forgiving processor. Today I wondered if it could do 4.5Ghz at "less than 1.3V" and here are two runs of Asus RealBench at 1.28V and 1.29V and I think the score results will fluctuate with marginally higher or lower results running at the same settings. Not sure what to make of it vs keeping things at 4.4Ghz?

----------------------------
Mon Apr 4 2016 20:05:11
(4.5Ghz @ 1.28V 90.55W 73c Max Load)
Image Editing: 197331 Time: 25.784
Encoding: 179560 Time: 53.397
OpenCL: 100032K Samples/sec:  2626
Heavy Multitasking: 184685 Time: 52.955
System Score: 145395
----------------------------
Mon Apr 4 2016 20:31:06
(4.5Ghz @ 1.29V 92.01W 74c Max Load)
Image Editing: 197331 Time: 25.784
Encoding: 182014 Time: 52.677
OpenCL: 100032K Samples/sec:  2630
Heavy Multitasking: 183002 Time: 53.442
System Score: 145588
----------------------------

[Ari Altman]

Absolutely run it at 4.5GHz/1.28v, but be aware that you may find instability at these settings in certain applications that realbench has not picked up on. If you do, just drop the speed 100MHz and you'll likely be fine.

[Steve N. Mavronis]

Absolutely run it at 4.5GHz/1.28v, but be aware that you may find instability at these settings in certain applications that realbench has not picked up on. If you do, just drop the speed 100MHz and you'll likely be fine.

I'll try 4.5Ghz @ 1.28V again tonight. I usually just run RealBench and CineBench once as a survival test after a settings change before proceeding. How many cycles would be good to test general stability and reliability? If it fails after further testing with voltages below 1.3V then I'll fall back to 4.4Ghz.
 
I just start doing normal stuff like GIMP image editing and playing games like Quake Live to see if it crashes LOL. I plan on playing the new Doom shooter when it comes out in May.
 
Also, I tend to enable XMP before entering the core multiplier and manual core voltage so all are done at the same time. Then after testing I boot back into BIOS and set it to adaptive voltage, same value. Any advantage to doing one or the other (XMP/Core) separately or in what order, verses at the same time?

[Steve N. Mavronis]

I re-tested my 6700K overclock settings for a target of 4.5Ghz at minimum stable voltage with low CPU thermals and best benchmark performance. I measured temperature changes with CPUID HWMonitor. Based on my results below it looks like 2.80V is the most promising VCore setting for me to work with.

Both RealBench and CineBench results seem to trend with one another. My test range was between 1.30V-1.27V in 0.01 increments between tests. I didn't even bother trying 1.26V because the results started trending lower at 1.27V and didn't expect any different other than more of a chance for instability. Here are my test runs and results:


Tue Apr 5 2016 18:26:58
4.5Ghz @ 1.30V 92.81W 74c Max Load
Image Editing: 197010 Time: 25.826
Encoding: 178348 Time: 53.76
OpenCL: 100032K Samples/sec: 2628
Heavy Multitasking: 180719 Time: 54.117
System Score: 144020
----------------------------
[CineBench CPU 968cb]


Tue Apr 5 2016 19:51:31
4.5Ghz @ 1.29V 92.62W 73c Max Load
Image Editing: 197807 Time: 25.722
Encoding: 178883 Time: 53.599
OpenCL: 100032K Samples/sec: 2628
Heavy Multitasking: 184375 Time: 53.044
System Score: 145267
----------------------------
[CineBench CPU 985cb]


Tue Apr 5 2016 20:19:42
4.5Ghz @ 1.28V 90.93W 73c Max Load
Image Editing: 197094 Time: 25.815
Encoding: 182499 Time: 52.537
OpenCL: 100032K Samples/sec: 2622
Heavy Multitasking: 183438 Time: 53.315
System Score: 145759
----------------------------
[CineBench CPU 989cb]


Tue Apr 5 2016 20:50:44
4.5Ghz @ 1.27V 89.68W 71c Max Load
Image Editing: 196850 Time: 25.847
Encoding: 181529 Time: 52.818
OpenCL: 100032K Samples/sec: 2626
Heavy Multitasking: 184090 Time: 53.126
System Score: 145618
----------------------------
[CineBench CPU 986cb]

[Ari Altman]

When you say the results started trending lower, I'm not sure what you mean. As long as the system is stable through the benchmark, you're OK. It won't perform slower if unstable, it will just crash.

[Steve N. Mavronis]

When you say the results started trending lower, I'm not sure what you mean. As long as the system is stable through the benchmark, you're OK. It won't perform slower if unstable, it will just crash.

Just looking at the RealBench system score and CineBench score and encoding together. Yeah 1.28 and 1.27 are real close probably within the margin of error I suppose. I could try 1.26 maybe...

I was reading an Skylake OC guide that said after the crash, take the last successful voltage and add 0.025V to it for an added margin of stability. If I took that recommendation, if 1.26V crashes then it would put me to 1.27+0.025=1.295V. If 1.25V crashes then it would put me to 1.26+0.025=1.285V.  If 1.24V crashes then it would put me to 1.25+0.025=1.275V and so on.

Still, I could try to closer approach stock voltage. I guess I'm being pragmatic about it?

Also one observation I'm curious about. When I set a manual voltage it always differs from what the VCore line says in BIOS. That is all but 1.28V which always matches like it's special.

[Ari Altman]

When you say the results started trending lower, I'm not sure what you mean. As long as the system is stable through the benchmark, you're OK. It won't perform slower if unstable, it will just crash.

Just looking at the RealBench system score and CineBench score and encoding together. Yeah 1.28 and 1.27 are real close probably within the margin of error I suppose. I could try 1.26 maybe...

I was reading an Skylake OC guide that said after the crash, take the last successful voltage and add 0.025V to it for an added margin of stability. If I took that recommendation, if 1.26V crashes then it would put me to 1.27+0.025=1.295V. If 1.25V crashes then it would put me to 1.26+0.025=1.285V.  If 1.24V crashes then it would put me to 1.25+0.025=1.275V and so on.

Still, I could try to closer approach stock voltage. I guess I'm being pragmatic about it?

Also one observation I'm curious about. When I set a manual voltage it always differs from what the VCore line says in BIOS. That is all but 1.28V which always matches like it's special.

I typically go by .005v increments, but .025 is a more conservative approach that will lead to less tweaking, and potentially a higher voltage/temp.

The vcore in the UEFI is not an accurate representation of vcore in use, as it does not use power saving or adaptive routines.

[Steve N. Mavronis]

Okay here's a recap of RealBench and CineBench scores. I added 4.5Ghz @ 1.26V - 1.24V just one pass trying to gauge relative performance between them. Yeah the temps keep trending down with the lower voltages but score results seem to be stagnating somewhat. The W value is the package power, I'm assuming the CPU power heat dissipation as related to the default Intel 6700K spec of 95W.

How low should I go and still expect overclock stability with 24/7 use at 4.5Ghz? I'm surprised 4.5Ghz is being so accommodating. At 4.6Ghz I crashed when trying less than 1.32V for example. Maybe that has to do with the Digi+ VRM power control turning off at 4.6Ghz and above.

Tue Apr 5 2016 18:26:58
4.5Ghz @ 1.30V 92.81W 74c Max Load
Image Editing: 197010 Time: 25.826
Encoding: 178348 Time: 53.76
OpenCL: 100032 KSamples/sec:  2628
Heavy Multitasking: 180719 Time: 54.117
System Score: 144020
----------------------------
[CineBench CPU 968cb]

Tue Apr 5 2016 19:51:31
4.5Ghz @ 1.29V 92.62W 73c Max Load
Image Editing: 197807 Time: 25.722
Encoding: 178883 Time: 53.599
OpenCL: 100032 KSamples/sec:  2628
Heavy Multitasking: 184375 Time: 53.044
System Score: 145267
----------------------------
[CineBench CPU 985cb]

Tue Apr 5 2016 20:19:42
4.5Ghz @ 1.28V 90.93W 73c Max Load
Image Editing: 197094 Time: 25.815
Encoding: 182499 Time: 52.537
OpenCL: 100032 KSamples/sec:  2622
Heavy Multitasking: 183438 Time: 53.315
System Score: 145759
----------------------------
[CineBench CPU 989cb]

Tue Apr 5 2016 20:50:44
4.5Ghz @ 1.27V 89.68W 71c Max Load
Image Editing: 196850 Time: 25.847
Encoding: 181529 Time: 52.818
OpenCL: 100032 KSamples/sec:  2626
Heavy Multitasking: 184090 Time: 53.126
System Score: 145618
----------------------------
[CineBench CPU 986cb]

Wed Apr 6 2016 19:31:38
4.5Ghz @ 1.26V 89.10W 71c Max Load
Image Editing: 198781 Time: 25.596
Encoding: 180195 Time: 53.209
OpenCL: 100032 KSamples/sec:  2626
Heavy Multitasking: 183479 Time: 53.303
System Score: 145615
----------------------------
[CineBench CPU 993cb]

Wed Apr 6 2016 19:58:04
4.5Ghz @ 1.25V 87.84W 70c Max Load
Image Editing: 197692 Time: 25.737
Encoding: 178208 Time: 53.802
OpenCL: 100032 KSamples/sec:  2630
Heavy Multitasking: 184337 Time: 53.055
System Score: 145060
----------------------------
[CineBench CPU 993cb]

Wed Apr 6 2016 22:28:09
4.5Ghz @ 1.24V 87.05W 69c Max Load
Image Editing: 197815 Time: 25.721
Encoding: 178414 Time: 53.74
OpenCL: 100032 KSamples/sec:  2632
Heavy Multitasking: 184149 Time: 53.109
System Score: 145096
----------------------------
[CineBench CPU 993cb]

Note: Testing @ 1.23V crashed Windows 10 (bluescreen) during the RealBench encoding benchmark test.