I recently got my hands on an Asus Maximus V Extreme and a 3770k. Lately I was benching on the RIVE, so I kinda forgot how to tweak the Z77 platform... I have decided that I will start tweaking/re-learning and checking the results with ROG RealBench! :cool:
It proved to be a very exciting experience, so I'm willing to share it... 🙂
The objective was to get as high score as possible in RB, but keep the tweaking as simple as possible - I was curious primarily about this platform's and CPU's performance... more serious tweaking will come later...
Let's start with the hardware: Mobo:Asus Maximus V Extreme - stock cooling CPU:Intel Core i7-3770k - it is a used, de-lidded & re-lidded specimen (TIM has been changed in it), watercooled (chiller) RAM:GSkill TridentX, 2x4GB, 2600MHz, 10-12-12-31 2T, 1.65V - stock cooling SSD:Samsung 830, 128GB - for OS and applications HDD: some small laptop platter - for data OS:Windows 7 Ultimate - with all the updates installed VGA: GeForce Titan - stock cooling (has no importance in this experiment)
ROG RealBench is the newest version to date: v1.1
As it can be seen, the only watercooled part is the CPU... it's connected to my chiller and the coolant temp was set to 65F/18.33C, while the room temp was 23C...
The OS is running in Diagnostic Mode, Realbench.exe's priority is set to Realtime and explorer.exe is turned off... after each set of runs, restarted the system...
1. After setting up the system and installing all the updates, I run my first RealBench test... just to set a baseline... At this point everything was on BIOS defaults - got a system score of 570
2. The first tweak was to choose the XMP settings form Ai Tuner: 2400MHz/10-12-12-31 2T - score went up to 645
3. just to have the sensation that I'm controlling the tweaking, I set the memories at their XMP spec manually... and set the CPU to 3.9GHz (39x100)... I started to run two tests one after another - got a max score of 646
4. Then I switched the memories from 2T to 1T - just a little bit of increase in scores: 648
5. Then I have started to push the CPU, basically by only increasing the multi... but before, I disabled the CPU Spread Spectrum, Execute Disabled Bit and all C-States... CPU at 4GHz (40x100) - score 662
6. Just continued increasing the multi... CPU at 4.2GHz - score 690
7. CPU at 4.4GHz - score 719
8. CPU at 4.6GHz - score 746
9. And CPU at 4.8GHz - score 770
Please, note that I haven't changed anything else in BIOS beside of those mentioned above... However, the next step, moving to 5GHz started to present some trouble:
The system booted up fine, even the AIDA64 stress-testing was going on well... - only the RealBench was not happy...
10. So, I took some extra measures... enabled the Internal PLL Overvoltage, set LLC to Ultra HIgh and disabled VRM Sread Spectrum... also, started to push the Vcore manually.. . at 1.520V finally got a result with CPU at 5.0GHz - score 796
11. Got very excited and applied an another tweak: reduced the screen resolution to 1024x768... had to push the Vcore a little bit more (to 1.530) - and finally passed the 800 mark with a 4-core CPU! 5.0GHz - score 809
12. Then followed a long period of trial and - mostly errors - while I was pushing the Vcore further for 5.1GHz... when I passed 1.6V, I actually set the coolant temp to 60F/15.5C, just to stay on the safe side... also, enabled the Xtreme Tweaking (not sure if this one helped) and set the LLC to Extreme... I was running AIDA64 in background to monitor the temps... when I got to 1.66V, finally passed with CPU at 5.1GHz - score 805
Temps reached 80C during the second test!
13. And finally I reset the screen to the lower resolution... CPU at 5.1GHz - score 821
Here is the summary of my journey:
And a chart with the evolution of the scores:
I'm showing the each score from both of the runs, so it can be seen that it is a big difference running RealBench for the second time! Also, check how much improvement brings if you set the screen resolution lower!
One more thing... you also can see how the Vcore got increased... - please, monitor your temps while tweaking!!! The role of this thread is to show some basic tweaks (maybe, letar on some more detailed ones too) and not suggesting that your CPU can do the same clocks at the same voltage! Your CPU can be actually better, needing less voltage... or can be worse... - don't forget that I used a chiller for cooling...
So, that's it for now... hope you will enjoy it...
Sorry, Z. Didn't mean to be inconsiderate. It was just a little proof of concept test. I may tinker with it a little more and re-post a better description of the idea to a more appropriate thread when I get some spare time.
Good suff to know. I have my 3 360 rads up and running now. I am able to pass the 2nd part of the RB encoding @ 4997 but temps are high and shes throttling down during 100% load giving me a low encoding score. Nik gave me some tips on my memory witch allowed me to push my trident X memory up past 2700 MHz (thx Nik) so I may rollback my OC and push memory further.
If you're using that icy water tank, you may consider to use it alone, only for the CPU for short runs like RB... if you have that icy tank inline with your rads, the rads actually will warm up the water... - the icy water has it's temps below the ambient (pretty close to 4C), and your ambient is surely higher - the rads will just inverse their role since the air going through them is warmer than the water, so they will start heating up your coolant to room temp...
Now, if you use only the icy water, then the risk of condensation is very high! That's what I meant earlier to pay attention and insulate your mobo...
Also, keeping a distance of 30-50 cm behind the rads (can not appreciate how much is behind them now) will improve the airflow through them and subsequently improve their cooling performance...