Update: 12. March 2015
This post becomes old, however, even when it may have been stupid to call prime95 a bad choice to test overclocks it somehow may have saved peoples wallet. In the beginning when x99 was thrown onto the market, things were very new and everyone had to get closer to the new design step by step, carefully. As you have to admit, there were many articles about mainboards that were catching fire, go up in smoke, or break in any other possible way. Things are getting more stable and people are making their way through the jungle of architechture. Today i still would recommend people to be careful with prime95 but i would no longer discourage people from using it.
Update: 06. October 2014
Well i can not completely revise this.. when i started to think that prime95 doesn’t go well with Haswell-E i did run future tests with other tools like cinebench, sisoftware sandra, aida64 and had no problems even with lower voltages (1.28 V). Rendering videos, playing games and whatever you do with your computer also did work quite well. Today i tested with OCCT and it goes the same direction (1.35 V stable). With prime95 i ended at 1.37 V or even 1.39 V . So prime95 still does require lot more voltage to run stable than for example OCCT does, so im not sure yet where this journey is gonna end. I can not recommend using prime95.
The long journey comes to an end now. [Due to Update on 06. October i have to over-think this] I finally or accidently came to the “conclusion” that prime95 isn’t the right choice to test the Haswell-E processor series, like the i7-5930K to run any sort of stresstests.
In the end of the day this processor, the i7-5930K will be running at 4.5 Ghz with everything else overclocked at reasonable temperatures and stable. Well, not prime95-stable but Aida64-Cinebench-and-SiSoftware-Sandra-stable. Prime is just no longer an option or relevant for 2011 and x99.
So this is going to be
a big round-up confusing
The truth is, that i stumbled over a forum thread, while i was searching for wprime95, saying that prime95 doesn’t go well with Haswell-E. So, while it’s still saturday night/sunday in the very morning and i was about to leave the system for today as it is, i decided to repeat some steps on the Asus x99 Deluxe, i did in the past weeks. In the beginning of the OC-adventure, i started with prime95 only and later took SiSoftware Sandra into the boat.
Quick Overview of the System (or have a quick look at previous posts):
- Asus X99 Deluxe (white/black) Motherboard
- Intel i7-5930K at 3.5 GHz (will use MX-4 paste)
- Corsair Vengeance LPX DDR4, 16 GB, 2800MHz 4-Modules-Memory-Kit
- Corsair RM1000 Power Supply
- Kraken X60 Watercooling
- 4x Corsair 140mm Fans (purple/violet)
- Corsair Obsidian 450D Case
Now prime95 has been replaced by Aida64 and Cinebench, while SiSoftware Sandra stays in the club. Benchmarks and graphs are always nice, but i intentionally use the software to make sure components are working fine together in harmony.
It’s already a month now since x99 and friends have been thrown on the market. A lot of you guys were sure smarter than i am/was and already knew that prime95 is a bad idea – when it comes to Haswell-E. While the hole world already OC’ed their CPUs with success i will just stick to some pictures. After i decided to no longer use prime95, i revoked / undid most of the changes i made there, to start from zero.
|CPU i7-5930K in GHz:||Corsair LPX DDR4 in MHz:||Asus x99 Deluxe Baseclock in MHz:|
|3,7 @ base values, unknown voltage (release BIOS 0505)||2166 @ base values, unknown voltage (release BIOS 0505)||100 @ base (BIOS 0505)|
|–||2800 @ 1.5 V or 1.35 V (XMP didn’t work and Memory was very unstable at higher frequencies – or let’s say, with prime95 it was)||–|
|4,1 @ 1,23 V (release BIOS to BIOS 0801)||2800 @ 1.35 V (release BIOS + BIOS 0701, XMP started to work with 0701)||100 (BIOS 0505 – 0701)|
|4,2 @ 1,25 V (BIOS 0701 + BIOS 0801)||2800 @ 1.2 V (BIOS 0701 + 0801)||127,5 (BIOS 0801)|
|4,25 @ 1,25 V (BIOS 0801)||3000 @ 1.35 V (BIOS 0801)||125 (BIOS 0801)|
|4,375 @ 1,265 V (BIOS 0801)||3000 @ 1.35 V (BIOS 0801) + (BIOS 0904)||125 (BIOS 0801)|
|4,427 @ 1,31 V (BIOS 0904)||3034 @ 1.35 V / 3416 MHz Cache (BIOS 0904)||126,5 (BIOS 0904)|
|4,462 @ 1,276 V (BIOS 0904)||3000 @ 1.35 V (BIOS 0904)||127,5 (BIOS 0904)|
|4,5 @ 1,276 V (BIOS 0904)||3088 @ 1.35 V (BIOS 0904)||128,7 (BIOS 0904)|
|4,5 @ 1,35 V||OCCT stable||128,7 (BIOS 0904)|
|4,5 @ 1,37 – 1,39 V||prime95 stable, requires extreme² cooling||128,7 (BIOS 0904)|
Some values were tested on different BIOS versions, some were stable some were not, some or many of them don’t even show up here. The table doesn’t show all combinations. In all cases you can say, that the newer setting also did run at least 1 BIOS version behind. For example 4.5 GHz was also tested on 0801 with Vcore from 1.27 to 1.390 and a lot of changes in voltages for other parts like Input Voltage or Agent Voltage or Cache Voltage or Changes in the DIGI Tab. But i do not recommend this especially because this happend in the prime95 period. In a previous post i said that XMP with BIOS 0701 was unstable. Thats not correct. This was also affected by prime95.
To the edge 4.6 GHz but prime95
As i mentioned already in a previous post, i could easily clock that Haswell-E jewel to 4.6 GHz. Problem had been, that it just didn’t pass the prime95 tests and that i found myself increasing and increasing over and over again the Vcore. At this moment i was noticing something weird with prime95 the first time, but i didn’t know about that bad relationship between prime95 and Haswell before, so i kept using it. What in the end led to a super hot CPU and with a crying Kraken X60 with stock fans. I made a short clip to demonstrate the Kraken stock fans at 1000% load (yes, that extra zero was intended). [Video quick link].
|CPU – i7-5930K||DRAM – Corsair LPX 2800 DDR4||Cache||Baseclock|
|4.5 GHz||3088 MHz||3860 MHz||128.7 MHz|
|1.35 V||1.2 V||–|
On 30. Sept. 2014 (2 days after publishing this post) - got a bluescreen 0x3B occasionally during SiSoftware Sandra Multicore Efficiency Test (this test did already run a few times with no error or bluescreen); will increase Vcore slightly from 1.276 to 1.29. This update doesn't show up in the BIOS screenshots below!!!
Now starting from zero with default settings and new crew
So starting from zero with some basic settings (CPU, DRAM, Cache and Baseclock will be set), and some of them are the result of the past few weeks. Initially will reset all voltage values and any performance related values to “Auto” again and boot the system.
Not that surprising but defaults deliver
And find myself happy with it. Now where the Vcore is back to normal and due to the absence of prime95 the temperatures during stress tests are lovely ( the funny thing here is, they never were bad, just prime95 made appear everything worse ). And more important the Kraken X60 stock fans do not have to go crazy anymore. Win-Win situation for me and the computer parts. Maybe i can get back to my Corsairs purple fans? When i run prime95 now with these settings, the system instantly goes down, but on the other hand i can normally use it. Can run any (non-prime95)benchmark, can render videos etc, can play any game … etc. somethign weird is burrowed in prime95.
The Aida64 Report has been uploaded to let you guys and girls see some graphs. If that’s not enough, i would like you to visit the SiSoftware Sandra Result / Ranking Page. No graphs but benchmaaaarks. SiSoftware Link 2
Temperature peak during tests 69 °C
the temperatures are very good in my opinion and the Kraken fans aren’t even close to the speed they were running when i used prime95 in previous overclocking attempts and tests.
Since all changes i made in the BIOS during the overclocking-journey are no longer necessary, it remains pretty much basic. Just for you, the…
Asus x99 Deluxe BIOS settings, i currently, 28. September 2014, run my system with.
.. meanwhile will run another SiSoftware Sandra Overall Bench.