- Category: G1.Sniper 3 Review
- Published on Monday, 01 July 2013 19:42
- Written by Super User
- G1.Sniper 3 Review
- Introduction, Box, Accessories, and Board
- Layout and Design
- Circuit Analysis
- BIOS Walkthrough/Gallery
- Air Overclocking(LLC Testing Included)
- Liquid Nitrogen Testing Results
- Test Setup and Benchmarks
- PLX8747 vs. NF200 vs. Native Z77 and Z68
- Creative Core3D (CA0132) Audio Testing
- USB 3.0 Tests and Software
- All Pages
Air Overclocking(LLC Testing Included)
A note on LLC and Software monitoring accuracy: So last time I did my reviews I tested out the LLC and compared software(easytune, cpuz, hwmon) to hardware readings(voltage read points and multimeter). It showed some interesting discrepancies which are good to note as you don’t want to over-volt your CPU or use more LLC than you should. Software voltage monitoring isn’t really accurate enough for precision voltage monitoring or testing. I also had requests from some to use more practical voltages than 1.45v to test LLC, as no user is going to use 1.45v 24/7. I used 1.45v to show how strain has an effect on LLC, but the results at 1.4v are the same as the algorithms are set very strong as the PWM and VRM can handle it. Also my CPU frequency is increased to increase the current used by the CPU and the frequency will have a big effect on LLC as well. That means that at stock, LLC extreme will raise your VCore more than it will at 4 GHz and 4 GHz more than at 4.5 GHz. That means that LLC high will droop more at 4.5 GHz than at 3.5 GHz. So just be aware of this.
Very nice results mate? Yes indeed the G1.Sniper 3 has the same LLC and voltage regulation as the UD5H, but I painted the background green. I added an extra column to show you the percent difference between idle and load, minus means the voltage drooped while positive means the voltage rose. What is interesting is that software shows much more droop than hardware monitoring, and hardware monitoring is 99% accurate while software is about 75%. So what does this mean? This means that setting LLC extreme until about 1.6v will raise your voltage under load, not keep it constant as CPUz reports.
Percent Difference = ((Load-Idle)/((Load+Idle)/2)
Percent Error is calculated = (Fake Load-Real Load)/Real Load
So let’s take a look at how well the new Sniper 3 overclocks.
Here is the maximum water/air overclock of 5.3GHz on my 3770K:
Booya! 5300MHz or 5.3GHz, is my maximum overclock using a Zalman AIO water cooler:
I think it works pretty well and stays quiet, not as good as H100 though. I am not stable at 5.3 GHz, I can do 5 GHz stable with a lot of heat, or I can settle for 4.8GHz stable:
The voltage I used for 4.8 GHz is nice. I will share the profile with you guys, so don’t worry.
Now let’s look at some of my air cooled memory OC results with a normal ULV kit from Kingston:
It is a simple 1.35v 1600mhz 9-9-9-24 T2 8GB(2x4GB) kit.
Now this memory is based upon Micron memory, and it does about 2200MHz max clocks, I didn’t spend much time on it but you can tighten those timings up a bit:
Also 1600 MHz with XMP but T1 instead of T2 works just fine:
If you are wondering why I used SuperPI32M for stability of memory, it is because it is a good memory stability tester. I used stock voltage (1.35v for the run above) and 1.55v for the 2200 MHz.