Aleutia Relia Industrial PC Review: Ivy Bridge & Q77 in a Fanless Chassis
by Ganesh T S on December 4, 2012 10:00 AM EST- Posted in
- Industrial PC
- HTPC
- Ivy Bridge
- Aleutia
Thermal Performance:
One of the drawbacks of completely passively cooled systems is that the CPUs reach their maximum temperature much faster than actively cooled systems. The CPU throttles down its performance as soon as the limits are reached. We proceeded to test out the thermal performance by loading up both the CPU and GPU with Prime 95 and Furmark. After hitting the thermal limits, the CPU speeds were throttled to 900 MHz.
One of the points to note in the above screenshot (taken with the unit at room temperature - 72 F) is that the chassis becomes hot enough for the HDDs to report a temperature of 59 C. This is even without any hard drive activity in the loading process. 55 - 60 C is considered the maximum permissible temperature for hard disks to ensure reliable operation. It doesn't seem advisable to use mechanical hard disks in the system if it is expected that the workloads for the system would end up throttling the CPU. We pinged Aleutia about this, and they indicated that the advisable ambient temperature for our test system was 30 C (for a pure SSD only solution, they upped it to 33 C). They also indicated that the i3-based system could go higher in terms of ambient temperature, but didn't mention specifics. For users with a requirement for operation at higher ambient temperatures, Aleutia indicated that they have custom solutions available too.
Our next plan of action was to determine the efficiency of the thermal solution under the above constraints. To test this out, we placed the unit in a temperature chamber and recorded the power consumption under full load. As soon as throttling hits, the power consumption of the unit starts to drop. We repeated the test with the temperatures at 0 C, 15 C and 30 C. After each run (i.e, subjecting the unit to full loading and letting the CPU cores reach the Tjmax of 94 C), we let it idle for 30 minutes. The core temperatures after the idling with the ambient at different temperatures are provided below.
| Aleutia Relia Cooling Efficiency | |
| Ambient Temperature | Core Temperature 30 min. after Tjmax |
| 0 C | 31 C |
| 15 C | 36 C |
| 22 C | 39 C |
| 30 C | 51 C |
The power consumption graphs are provided below for the unit's operation over a 20 minute duration after the start of the loading process.
The thermal solution works very well at low temperatures, with throttling kicking in after 9 minutes or so. With rising ambient temperature, the throttling kicks in within 3 minutes (at 30 C). These graphs are with absolutely no air flow in the temperature chamber, and are hence, worst case scenarios.
At 30 C, we also have detailed graphs of the temperature, load and frequency for one of the cores. These were generated by CoreTemp 1.0 RC4.
Facebook
Twitter
RSS
47 Comments
View All Comments
ganeshts - Wednesday, December 5, 2012 - link
Nope, we could clearly hear the noise from the hard drives.deltaTdawg - Wednesday, December 5, 2012 - link
This article makes it clear that the author has very little experience reviewing fanless systems.From an engineering perspective, there are 3 elements of interest in the thermal system - the CPU block, the heatpipes, and the heatsinks.
1. The author never looked at the thermal block, never took it off, never examined the thermal paste, never checked how flat or smooth the block surface was. Flatness and roughness play a HUGE role in cooling. Most importantly, thin metal mounting systems tend to flex when overtightened. This leads to a bowing of the CPU block, causing very poor contact with the CPU.
2. These heatpipes are 6mm diameter. Heatpipes have thin walls, so if they are bent too tight they can crimp and severely impact the flow of evaporated fluid which actually dissipates the heat. Look at the gallery. See the heatpipes? In the 5th image especially it is clear how many bends and crimps exist. Each crimped point severely impacts cooling performance. To examine this, the author could use an IR thermometer to look at the heatpipe temperature before and after each crimp. The flaws would be immediately evident as the heatpipe would be significantly lower temperature after each crimp. This indicates the heat is not dissipating properly.
3. Simple: Finned heatsinks are not ideal for convection cooling. Also, the mounting bracket does not apply even pressure, and no effort is made to distribute the heatpipe condenser evenly over the heatsink surface.
From an engineering standpoint, this is a disaster. My conclusions are supported by the experimental findings in this review. A cooling system that causes an ULV CPU to throttle. Hmm. I have personal experience with completely passive systems that can easily cool a i7-3770K under full CPU+GPU load - under 80C.
ganeshts - Wednesday, December 5, 2012 - link
The point of the review's thermal performance section was to present to readers whether / how much the CPU gets throttled under different ambient temperatures and not meant to go into a thesis on how Aleutia could improve the thermal performance.Anyways, responding to your points:
This is a loaner sample and meant to get back to the manufacturer in working condition. Not all systems that reach our labs can be subject to full teardowns. So, even assuming that I took out the thermal block and judged the smoothness of the block surface, those stats matter zilch to the reader once the thermal performance stats (how long it takes for the system to get throttled under load) is presented. FWIW, the loading process was repeated countless times before the system made it to the temperature chamber. In effect, taking out the thermal block and analyzing it would have probably made sense in a dedicated review of the Streacom cooling solution and not for the Relia itself.
Btw, the i7-3770T is NOT officially a ULV CPU. It is just power optimized and operates at lower base clocks (2.5 GHz) instead of the default i7-3770 (3.4 GHz) to bring down the TDP from 77 W to 45 W.
We would like to definitely hear about the components used in your passively cooled system. What is the size of the system, and how long does it operate at max performance without throttling and at what ambient temperature?
deltaTdawg - Wednesday, December 5, 2012 - link
Fair enough.I didn't mean to come off aggressively, and I apologize if I did. Certainly there is a limit to what testing can be performed. The reason the cooling system matters, in my mind, is because it is performing so poorly. SPCR reviewed a unit from the OEM chassis provider - Streacom FC5 OD - using a 65W i3-2100 processor. In that review, the system didn't throttle; rather, it hit 74C (53C above ambient) on full CPU+GPU load. This is not terrific performance - but it is much better than the Aleutia system.
So from that I can only conclude that Aleutia's particular design is what has negatively affected results - which is why I feel it is of interest for this review, not for a Streacom review. In my mind, a system whose cooling system can't handle a burn-in load is not fit to ship. Most vendors do a burn-in/stress test prior to shipping, to ensure the system components are not faulty. Aleutia obviously wasn't able to do this; so how can they ensure the components are stable? From an industrial market perspective, stability and performance in a hot environment should be the utmost concern.
My system is 13"x10"x5" (WxDxH). It is running a Gigabyte GA-Z77N-WIFI, an i5-3750K, 8Gb DDR3 1600MHz, and a 128GB Intel 520. It is using a custom chassis and custom cooling system, and has zero moving parts. Ambient is 18C. The system idles at 31C, and full CPU+GPU load stabilizes at 56C. Return to idle temps when the load is removed occurs in under 30 minutes.
The system runs overclocked at 4.1GHz. Idles at 37C, full GPU+CPU load at 78C, which stabilized after 45 minutes and stayed stable thereafter. Return to idle temps occurs in under 30 minutes.
Overclocking is limited by the motherboard's lack of voltage control. The hardware maintains stable temperatures around 80C at 4.3GHz - but the system is not stable. I will be testing next with the Asus P8Z77i motherboard, which is overclocker friendly.
One last question: Why did you unlink the gallery? And why, if you find the gallery link (http://www.anandtech.com/Gallery/Album/2429) did you remove all the inside photos? This seems to come right on the heels of my remarks on their cooling system.
ganeshts - Wednesday, December 5, 2012 - link
The Streacom FC5 OD is a uATX chassis while this one is mITX. I guess the larger surface area of the FC5 OD will definitely contribute to better cooling.Streacom is planning to introduce a modified version of this chassis in the market as the FC2 (with an open backplate, obviously). I am looking forward to someone making a better performing system with the FC2 (not that I am defending Aleutia here, but I just want to see how much of the issue is with Streacom rather than Aleutia).
Your system definitely sounds interesting and I am sure readers would appreciate some pictures of the internals to understand how to design their own passively cooled systems.
Btw, the gallery is still there in the piece and was never 'unlinked'. The URL you are referring to:
http://www.anandtech.com/Gallery/Album/2429 : This is from the previous pipeline piece where I covered the launch of the Relia.
The inside photos are in this gallery:
http://www.anandtech.com/Gallery/Album/2473
deltaTdawg - Wednesday, December 5, 2012 - link
Yeah, sorry about that. That's what happens when you sit down at the computer before you have your morning coffee :)Interestingly, the chassis dimensions are not really relevant. What actually matters for heat dissipation is the heat sink surface area. I can't account for the size of the fins, but the mounted surface area on the two systems is virtually identical, as the FC5 uses only one heatsink to dissipate, whereas this unit uses both. The FC5 heatsink spans 30.2 square inches, while the Relia spans 31 square inches.
I would love to show off my system. However, it's a near-launch unreleased product, so I can't share photos online yet. I'll email you separately, if you're interested. Suffice to say, silent cooling like I've described is challenging, if not impossible, to achieve with consumer parts. But it's a real passion for me, so I'd be happy to continue the conversation by email.
ganeshts - Wednesday, December 5, 2012 - link
Sure, feel free to send over the details to my e-mail ID (can be got from the author by-line). I can keep it under wraps till launch.zilexa - Wednesday, December 5, 2012 - link
Seriously?? Who the HELL is going to spend >600 on this device?People who do that clearly do not have a clue what to do with their money.. *sigh*
I love the fact it uses mSATA and has enough options to connect. But as HTPC, just buy a Zotac AD12 or AD06 or similar and you are done (preferrably the AMD option since they support out-of-order execution and feel faster then the Atoms, and have better graphics built in).
There is absolutely no use for a Core i7 in a HTPC. It doesn't have a dedicated GPU so it is also not meant for gaming. This thing is really to shake money out of peoples pocket.
ganeshts - Wednesday, December 5, 2012 - link
Hmm.. This is an 'enterprise' play. You pay for quality and support too.I can't find any other pre-built PC with similar configuration for a lower price (read, fanless mITX with dual GbE and i7- class processor). People looking for industrial PCs will know the value :)
Btw, usage as a HTPC is just an additional application. There are use-cases such as hotel room TVs and signage applications where reliability is key and the benchmarks I have presented in the review are helpful.
Would I recommend this for the standard living room HTPC? Probably not.. The target market, as I explained in the final section, is something different.
Sikku - Wednesday, December 5, 2012 - link
Does this system use a PicoPSU??Would like to see reviews of this type of PSU...