Overclock.net › Forums › Specialty Builds › HTPC › War of the Celerons! Power Consumption - G1850 vs J1900
New Posts  All Forums:Forum Nav:

War of the Celerons! Power Consumption - G1850 vs J1900

post #1 of 2
Thread Starter 
I wanted to share some information to hopefully shed some light on a few things I have noticed relating to low power consumption PCs. I have been using a Haswell Celeron G1850 2.9GHZ CPU for several months which has been very good to me. I don't ask too much of my computer and it has done well while reducing power usage compared to my older Core2Duo machine. At the end of December I was planning on moving the Haswell over to replace the C2Duo as my HTPC. The power supply happened to crap out on me that day, which motivated me to test out something I had been wondering for a while.

I have the M3-ATX power supply from mini-box that I bought several years back along with a 12Volt 8.5Amp power brick. I remembered roughly what my power consumption findings were with the old 200 watt standard 120 volt PSU and after switching to the M3 I saw an average drop of 10 watts. This may not be huge for someone with a modern gaming rig, however when going from under 40 watts idle to under 30 watts idle we are talking a significant percentage! I knew there was a big difference between the power usage reported by reviews by websites and customer reviews when looking at boards such as the J1900 Celeron boards. The answer is in the power supply.

When I first got the G81M-H board put together, I immediately went into BIOS or UEFI or whatever people call it these days, and checked the voltages. comparing to older Gigabyte boards I noticed there wasn't as much room up or down. Limited to a minimum of .9V, I said what the heck, save and reboot, and it worked flawlessly. I was able to set a number of CPU voltages to their minimum and benchmark the PC to see power usage that I was very pleased with. With renewed interest in my power consumption, I decided to head into BIOS to give things a better look. The first thing I noticed was that the DDR voltage of the RAM (1.35 low voltage stick) showed the proper number reported, but also showed it at 1.35v shaded in grey to the right. The RAM voltage is adjusted on a different page and does not go lower than 1.5v, and I could have sworn every time I checked in the past that it read 1.35v but showed 1.5v next to it as the applied voltage. I don't have any other way to test the RAM voltage but I'm guessing it is using the RAM at the lower voltage setting.

The other thing I found that I didn't remember seeing before was a setting for a dummy load. It said something to the effect that certain power supplies will not operate with too low of a load. In all of the reviews I have seen for low power Celeron and AMD boards, I do not recall ever hearing about this option. I'm sure I have seen a few dozen reviews from different sites and with the people looking into these kinds of computers being interested in more than just synthetic benchmarks (say power consumption, noise, etc), I was a little surprised by this. Although it is possible that some boards do this automatically and do not have a BIOS option.

In any event, I moved the Haswell with a slapped together case inside and mounted it to the VESA mount on my TV. Losing a fan by getting rid of the now defunct 200w PSU made things a little quieter. Then it happened - I found an ASRock Q1900M motherboard on sale with rebate for a stupidly cheap price. I moved the M3-ATX over to the new board, installed Linux Mint, and have been testing it out for about 1 day. Absolutely no sound from the board. The cheap HDD I bought is mostly silent unless I put my ear near it, and that is the only moving part. I think the power brick has a small fan in it, but hasn't kicked in since using it on the Haswell or this J1900 and is barely warmer than ambient temperature. So far using Linux Mint 17.1 Cinnamon I have been getting around 17 watts while web surfing, 22 watts watching videos, and about 26 watts max playing 0 A.D., Super TuxKart, and a couple steam games.

I am going to bench this J1900 Bay Trail and the G1850 Haswell under windows 7 using the same HDD, RAM, and M3-ATX PSU to try and get a more accurate picture of power consumption. You can find performance stats all over the web, so I will only give general impressions of how the PCs perform under my usage. Apparently there is an issue with Cinnamon where there are random hangs, and it has displayed itself differently between the 3 computers I have set up as my HTPC. So, I will test out Mint XFCE and LXLE (finally got this to download properly!) and maybe some others and give a general impression of how each PC performs for me as an HTPC.

I realize this is a little late in the game, with new chips coming soon to raise the bar on performance per watt, but I hope this will serve as a benchmark. I know a bunch of people interested in low power PCs and intend to do the same tests when I can get my hands on newer parts. Stay tuned for a full run down of the power specs this weekend!
post #2 of 2
Thread Starter 
I finally got the testing done. I didn't do any extensive graphics testing, as the game I tested with (0 A.D.) was at the limit of what the J1900 could handle. I only ran the tests with a single 4GB stick of RAM and stock graphics speeds and voltages. People looking for a basic HTPC out of a Celeron shouldn't be expecting a top of the line gaming rig. I'm sure the performance would increase if more RAM is used in dual channel mode.

For the testing I used a single stick of Crucial 4GB low voltage (1.35V) RAM, M3-ATX PSU, old 2.5" 60GB Seagate SATA HDD with Windows 7, wired mini keyboard, wired mouse, and LAN connected. I disabled the serial/parallel ports in BIOS on both boards and left all other ports turned on. The CPU voltages on the G1850 were dropped as far as they would go. The wattages were recorded watching the output from a Kill-A-Watt meter. If I get my hands on a better multimeter, I will try and compare power consumption off of my small solar battery bank to these findings. Also, I see they have a Pico PSU 80 watts and 60 watt brick package for cheap. I am going to probably pick up 2 and compare how well those do with power consumption. The smaller brick with the lower wattage boards may be more efficient.

I ran each test 3 times, going through each test one by one, then powering down and starting over. The highest wattage during each test was recorded. During most of the tests the wattage fluctuated less than 1 watt. Results were pretty consistent and the averages are as follows:

G1850

Idle - 22.8
Super Pi - 34.2
Prime 95 - 46.8
0 A.D. - 52.4

J1900

Idle - 14.4
Super Pi - 16.9
Prime 95 - 20.3
0 A.D. - 22.3

I did do some other tests but the results were nearly equal to what I have listed here. The G1850 did have the fan speed up during the tests, and the J1900 doesn't have a fan, but barely got warm. It was laying in the open. If I were cramming it in a tiny box and shoving it in an entertainment center I would probably rig up a better heatsink than the one that comes with the Q1900M. It is sufficient for my needs as-is for now. Any case or PSU fan should have no problem providing enough airflow. The silence of no fans is pretty nice. An SSD would be even better, as there would be no moving parts to fail. The motherboard and the M3-ATX both have all solid caps, so it should last a very long time, and outlive its usefulness.

Video playback with Hulu and Netflix has been problem free for both machines in Linux. The real kick in the pants for me has been the Cinnamon desktop hanging issue. Certain computers get it worse than others and unfortunately this Q1900M seems to get it bad. It plays videos fine, but some simple tasks like making a bootable USB drive, moving files, even using the Steam browser to look at games all can cause long hangups. One of the things that caused short hangups on the G1850 was a few steam games, which seem to be fine on the J1900 board. I've tried a few distros and now I'm about to install LXLE. That should work better and I can get more time in as an HTPC with this machine.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: HTPC
Overclock.net › Forums › Specialty Builds › HTPC › War of the Celerons! Power Consumption - G1850 vs J1900