Micro-ATX Cases - Shoebox Showdown
by Joshua Buss on January 2, 2007 12:40 PM EST- Posted in
- Cases/Cooling/PSUs
Benchmark Comparisons - Temperature
Ever since our socket 754 DFI motherboard has been giving us problems producing reliable temperatures for benchmarking, we've been on the lookout for a new micro-ATX motherboard for testing. We've had great luck with ASUS sensors in the past, and settled with an A8N with the popular socket 939. Using a dual-core CPU will allow us to see bigger differences in CPU heat output, and will let us better evaluate how cases handle the two extremes of cooling loads.
This new motherboard proved to be an excellent choice for our test bed, as it offered a total of eight temperature sensors - nine including the hard drive. We nailed down the location of pretty much every sensor by trial and error and some canned air sprayed upside down to provide localized cooling. Interestingly, we found the board was really showing four different temperatures for roughly the same area, which was right around the CPU. We also found a sensor in a Winbond chip in the top right which works nicely as an ambient motherboard temperature indicator, and will be shown simply as "MB" in our temperature results tables.
Our 6600 GT continues to provide both a core temperature and an ambient figure read via the NVIDIA driver, and speedfan reflected these numbers too. The only problem is that probably because we were using a separate graphics card, we couldn't find a single sensor tied directly to the north bridge, as unlikely as that may seem. For benchmarking, we took an average result after temperatures had leveled off for over 20 minutes of testing at the idle and full load tests. For loading we used two copies of the Folding@Home project (one for each core) and the popular rthdribl GPU benchmark.
Since the Qmicra is such a customizable case in terms of its thermal performance, we tested it in two completely different configurations; one with just a single Antec Tri-Cool fan on its slowest setting and the CNPS 7000 CPU heatsink, and one with four Arctic Cooling 80mm fans set at each of the designated mounting points and a CNPS 9600 heatsink. The drastic difference in overall airflow affected the temperatures of our components for sure, but we still got very interesting results.
We had to run our temperature tests over, and over, and over to be sure these numbers we were seeing were correct. Needless to say, they are not what we expected. With the Qmicra able to hold the beastly CNPS 9500, we figured for sure that combination would sweep the CPU temperature benchmarks, but instead it was beat pretty handedly by the MicroFly with the smaller CNPS 7000 cooler, both under load and at idle. We even re-mounted our 9500 several times to ensure it wasn't a fluke, but these figures here were always representative of what we saw.
The only explanation we really have is that our particular CNPS 7000 is better at making good contact against our 3800+ X2, and the Ultra's larger 120mm fan located right above the CPU area is keeping only cool air around the processor. As many have pointed out and experienced first hand, the Aria runs a lot hotter than the other offerings, getting the CPU up to a toasty 68C under full load. The good news is that we never encountered any stability problems, so really, these high temperatures don't even need to be taken as that much of a caution - just know that overclocking and the Aria might not be a smart combination.
Other interesting findings that caught our attention included the cooler-than-open-air temperatures of the hard drives and motherboard in the Qmicra once we populated all four fan mounts, suggesting that the all-aluminum design and heavy airflow is doing an excellent job of cooling those areas of the case. The huge decrease in hard drive temperature shows how targeted airflow can be more important than merely having a spacious interior. It would have been really interesting if the 9500 could've fit into the MicroFly, but even as it stands it's clear that the Chenming 118 has a good overall design for performance and the other cases designed in its style work just as nicely. Where the Qmicra really shines here though is in the flexibility. Users can come up with their own selection of fans based on their cooling needs and have custom-tailored performance.
Ever since our socket 754 DFI motherboard has been giving us problems producing reliable temperatures for benchmarking, we've been on the lookout for a new micro-ATX motherboard for testing. We've had great luck with ASUS sensors in the past, and settled with an A8N with the popular socket 939. Using a dual-core CPU will allow us to see bigger differences in CPU heat output, and will let us better evaluate how cases handle the two extremes of cooling loads.
New mATX Test Bed | |
ASUS A8N-VM nForce 430 AMD 3800+ X2 Dual Core CPU OCZ 512MB DDR2 x 2 Zalman CNPS 7000cu Heatsink or Zalman CNPS 9500 Heatsink Maxtor 80gb IDE Hard Drive Chaintech GeForce 6600GT Zalman VF-700 Cu GPU Cooler MSI DVD-CD/R/RW Combo drive Zalman ZM460-APS PSU |
This new motherboard proved to be an excellent choice for our test bed, as it offered a total of eight temperature sensors - nine including the hard drive. We nailed down the location of pretty much every sensor by trial and error and some canned air sprayed upside down to provide localized cooling. Interestingly, we found the board was really showing four different temperatures for roughly the same area, which was right around the CPU. We also found a sensor in a Winbond chip in the top right which works nicely as an ambient motherboard temperature indicator, and will be shown simply as "MB" in our temperature results tables.
Click to enlarge |
Our 6600 GT continues to provide both a core temperature and an ambient figure read via the NVIDIA driver, and speedfan reflected these numbers too. The only problem is that probably because we were using a separate graphics card, we couldn't find a single sensor tied directly to the north bridge, as unlikely as that may seem. For benchmarking, we took an average result after temperatures had leveled off for over 20 minutes of testing at the idle and full load tests. For loading we used two copies of the Folding@Home project (one for each core) and the popular rthdribl GPU benchmark.
Since the Qmicra is such a customizable case in terms of its thermal performance, we tested it in two completely different configurations; one with just a single Antec Tri-Cool fan on its slowest setting and the CNPS 7000 CPU heatsink, and one with four Arctic Cooling 80mm fans set at each of the designated mounting points and a CNPS 9600 heatsink. The drastic difference in overall airflow affected the temperatures of our components for sure, but we still got very interesting results.
We had to run our temperature tests over, and over, and over to be sure these numbers we were seeing were correct. Needless to say, they are not what we expected. With the Qmicra able to hold the beastly CNPS 9500, we figured for sure that combination would sweep the CPU temperature benchmarks, but instead it was beat pretty handedly by the MicroFly with the smaller CNPS 7000 cooler, both under load and at idle. We even re-mounted our 9500 several times to ensure it wasn't a fluke, but these figures here were always representative of what we saw.
The only explanation we really have is that our particular CNPS 7000 is better at making good contact against our 3800+ X2, and the Ultra's larger 120mm fan located right above the CPU area is keeping only cool air around the processor. As many have pointed out and experienced first hand, the Aria runs a lot hotter than the other offerings, getting the CPU up to a toasty 68C under full load. The good news is that we never encountered any stability problems, so really, these high temperatures don't even need to be taken as that much of a caution - just know that overclocking and the Aria might not be a smart combination.
Other interesting findings that caught our attention included the cooler-than-open-air temperatures of the hard drives and motherboard in the Qmicra once we populated all four fan mounts, suggesting that the all-aluminum design and heavy airflow is doing an excellent job of cooling those areas of the case. The huge decrease in hard drive temperature shows how targeted airflow can be more important than merely having a spacious interior. It would have been really interesting if the 9500 could've fit into the MicroFly, but even as it stands it's clear that the Chenming 118 has a good overall design for performance and the other cases designed in its style work just as nicely. Where the Qmicra really shines here though is in the flexibility. Users can come up with their own selection of fans based on their cooling needs and have custom-tailored performance.
37 Comments
View All Comments
jmke - Wednesday, January 3, 2007 - link
Do you have a picture of your test room? 15dB (I guess A weighed) seems awfully low :)JoshuaBuss - Wednesday, January 3, 2007 - link
It is very low.. in our old location we couldn't get any lower than 23 dB and now we're further away from the highway. All I know is that's what my new meter says if everything is silent and the dogs next door aren't barking.JoshuaBuss - Wednesday, January 3, 2007 - link
Whoa, I'm sorry.. it was supposed to be 18 dB. Updated.KayKay - Tuesday, January 2, 2007 - link
I built a machine for my brother using the Ultra microfly and it is a great case. The deciding factors for this was the ability to accept a Full-Size Power supply and the removable motherboard tray. They make this in several colours, as well as clear-sided versions. A small box with the ability to put some powerful components in it! 3800x2 with a 7600GT, this thing runs super quietCuriousMike - Tuesday, January 2, 2007 - link
I *just* built a machine three days ago using the Aria (NSK1300) as the basis; I just wanted something smaller than a tower case. I didn't see the review mention the case as being the NSK1300--- the Aria (IIRC) was all black. The case you reviewed is identical to the NSK1300.My build included a Frys X2 4200 "EE" combo deal with ECS C51GM motherboard.
The retail box AMD fan must have been running full tilt all the time (3000rpm sound right?)... it was the noisiest in the case. The CPU would reach 70' under Prime95 load within about 20 minutes, idleing at around 50'.
I replaced the stock HSF with a Zalman CNPS8000, which was reviewed poorer than the 7000 used in this review. The 8000, at anything other than it's lowest fan setting, is as loud as the AMD retail unit; at it's lowest fan setting, it *just* beats the cooling capacity of the stock HSF; running around 50' idle, and 69' with Prime load.
With the 8000 at full RPM, it only knocks a few degrees... 67'. The machine runs hot.
I used a evga 7600GTS and put a zalman 7000 GPU cooler on it (using the slowest fan setting); that made it go from idleing around 54' to 48', and running ATITool for 15 minutes, stock fan was 71', zalman brought it down to 65'.
With the current fan setup, the machine is tolerable noise wise. It's nowhere near silent.
The NSK1300 is cramped... almost impossible to route cables neatly.
It's pretty slick removing all the panels and the drive cage.
Myrandex - Tuesday, January 2, 2007 - link
I have the Lian Li PC-V300 and given its popularity with this crowd, I would have loved to see it included in there. I look forward to finishing the article as it is great so far.JoshuaBuss - Tuesday, January 2, 2007 - link
There are still a few more mATX cases on deck to be looked at soon.. hopefully the V300 will be one of them too.tayhimself - Tuesday, January 2, 2007 - link
I can't for the life of me find a good uATX motherboard with decent overclocking features for a core2 chip. I would like to build a uATX system but I havent gone with one because of the motherboards.Staples - Tuesday, January 2, 2007 - link
Unfortunately, good motherboards are not made in this size or even micro ATX for that matter. There are some decent P965 boards made in micro ATX sizes but the best ones are all in full sized ATX.JarredWalton - Tuesday, January 2, 2007 - link
µATX, uATX, mATX, and micro ATX are all the same thing. (u is the abbreviation used for mu a lot of the time - m already being taken by "milli" and most people not want to bother with the special symbol µ.)