Results and Conclusion
To test the HEX 2.0 cooler I decided to use my small form-factor setup, considering the intended use of a compact 95 mm cooler. In larger systems it will be far more cost-effective (and likely better performing) to use a larger all-in-one liquid cooler, but not every case will support even a 120 mm solution. My current mini-ITX setup is no slouch, with an Intel Core i5-6600K processor on an EVGA Z170 motherboard with solid overclocking support, and that last part is important as this Core i5 doesn't present much of a thermal challenge at stock speeds.
I tested the HEX 2.0 along with a couple of other options on hand, with both stock and overclocked CPU loads. For a comparison with a standard air solution of the same size I chose the Noctua NH-U9S, which occupies the same 95 mm footprint and also uses a 92 mm fan. My trusty Corsair H75 120 mm liquid cooler is also included for reference, but, again, the HEX is more intended for situations where even a 120 mm liquid option is not possible, or liquid is simply not desired.
| Test Platform | |
|---|---|
| Processor | Intel Core i5-6600K |
| Motherboard | EVGA Z170 Stinger (mITX Intel Z170) |
| Memory | Crucial Ballistix Sport 8 GB 2400 MHz DDR4 |
| Graphics Card | XFX AMD Radeon 5450 (Fanless) |
| Storage | OCZ Vertex 460 120GB SSD |
| Power Supply | Corsair TX 650W PSU |
| OS | Windows 8.1 64-bit |
Temperatures and Noise Levels
Stock results are a little surprising, as the Noctua cooler is more effective with these loads. But the i5-6600k behaves much more like a 65W part than a 91W part at stock workloads (indeed, the non-K Core i5-6600 variant is a 65W part), so to truly test the capabilities of the HEX 2.0 I needed to overclock the CPU. My goal with OC stability is always Prime95, as I use this benchmark to stress the CPU in all of my cooler testing, and I use the highest power draw torture test available in Prime95. I ended up with a rock solid 4.7 GHz overclock from my i5-6600K, but to speed the process up I set a target voltage of 1.34v. All cores were stable at 4.70 GHz throughout all testing, but the voltage was automatically adjusted to a whopping 1.4v (1.399v, to be precise) by my EVGA Z170 Stinger board during those punishing Prime95 runs.
The performance of the HEX 2.0 in this more extreme example is decidedly more impressive, as we see it leapfrog the Noctua cooler to finish behind the Corsair H75 liquid solution. I didn't expect the Hex 2.0 to be able to beat the Corsair H75, but I was pleasantly surprised by how well it fared compared to the dual-fan 120 mm liquid cooler with the overclocked CPU.
As to noise levels, the HEX 2.0 performed better than I was expecting for a design using just a single 92 mm fan, with my SPL meter registering idle noise of just 32.2 dBA (fan speed ~1160 RPM), with load noise of 34.3 dBA (fan at ~1690 RPM), and a high of 40.9 dBA (~2300 RPM) under stress. (All SPL measurements made with a 31.3 dBA noise floor.)
I also measured power consumption from the wall during testing, and the HEX 2.0 certainly will add to your total system power requirements when it is dealing with an aggressive thermal load using the 'insane' preset. I observed power levels of ~39W idle and ~97W load with the 'standard' setting, which is about what this i5-6600K system without a powerful GPU will generally produce under load, but 'insane' mode (which had an identical ~39W idle) produced a high of 138W. Those extra ~40W were not constant, as the software controls the TEC function, using it only as needed to control temps.
Conclusion
We have seen TEC/Peltier products in the past, and even reviewed them in years gone by; but implementing this kind of cooling technology in a consumer CPU cooler has obviously not resulted in a shift in the cooler market to this point. Closed-loop liquid CPU coolers have become commonplace, standard air cooling is better (and increasingly less noisy) than ever, and in general the need for TEC coolers just has not been felt for consumer CPU cooling. I think what Phononic has done with the HEX 2.0 is show the advantage of a TEC in a demanding small form-factor implementation, where liquid coolers will not always fit. A great example of this is a case I have in for review currently: the Lian Li PC-Q17. The only way to install a liquid solution like the Corsair H75 (which I used as a liquid cooler comparison in benchmarks) in the PC-Q17 is to mount it outside of the enclosure, just as with the In Win 901 mini-ITX enclosure I reviewed (way back in 2014). In enclosures like these a small solution is required for a streamlined build, and the HEX 2.0 is a great liquid alternative here.
The HEX 2.0 is overkill for a stock Intel CPU, and didn't fare better than the less expensive Noctua NH-U9S in those tests. Where the HEX 2.0 really showed its potential was in overclocked testing, as I was feeding my Core i5-6600K 1.34+ volts to get Prime95 stable on all cores. The results were far above the Noctua cooler with the voltage that high, showing just how much overhead one has if using the cooler in its most aggressive setting (appropriately called "Insane" mode). I would love to re-test the HEX 2.0 with an overclocked AMD Ryzen CPU to see just how far a little 95 mm cooler can take a processor like the R7 1700, but that will have to wait. For now I hope the 4.70 GHz OC from my i5-6600K is sufficient to provide an idea of the thermal management this HEX 2.0 is capable of.
Strengths
- Premium build quality
- Very high performance for its size
- Low noise output from the 92 mm fan
- Easy to use software
Weaknesses
- Cost
- Usefulness depends on specific usage scenarios
For the best thermal performance I have seen from a cooler of its size, the Phononic HEX 2.0 is a great option for builds where a liquid cooling loop can't be managed, but aggressive CPU performance (and additional overclocking headroom) is desired. It is expensive, but you are buying a unique product for a specific purpose. If it fits your particular needs, I can think of no better option for demanding thermal loads in this specific form factor.
How does it deal with
How does it deal with condensation on the TEC?
The design objective and
The design objective and proven in operation is that you will not see any condensation with the HEX 2.0 unlike previous TEC based coolers. That is the benefit you get out of our Intelligent Control of the Active/Passive operation.
Don’t worry about it, this
Don’t worry about it, this thing can’t even cool the proc below ambient at idle.
How would it lower temps
How would it lower temps below ambient if the TEC only turns on at high thermal loads?
That would be the worst time
That would be the worst time for it to run. At high thermal loads, a TEC would only add more heat to get rid of. Anyone who designed such a system clearly knows nothing about thermal transfer.
So, I’d avoid anything they designed like the plague it is.
“I would love to re-test the
“I would love to re-test the HEX 2.0 with an overclocked AMD Ryzen CPU to see just how far a little 95 mm cooler can take a processor like the R7 1700, but that will have to wait”
No it needs to be done ASAP for those that are looking at the 1700’s best savings for the overclocker’s money! Maybe it can be done when some of the tweaked Motherboards arrive with faster DDR4 memory support. Also test the RX 580 Polaris refresh with any builds for an affordable build option with that 1700’s overclock potential and affordable sweetspot of a price/performance metric.
If I could have purchased a
If I could have purchased a motherboard I probably would have already. Now that I've waited this long I might just go with an R5 1600/1600X instead. And the MSI X370 Gaming Pro Carbon I've had my eye on is still out of stock as I type this…
Well yes try the R5s along
Well yes try the R5s along with the R7 1700 with this cooler SKU for some overclocking benchmarks.
And what do you make of any 16 core/32 thread Zen/workstation SKUs using these rumored(1) chipsets? I’d love to see any 16 core Zen/Worksttation SKU benchmarks on the single socket ASUS X390 motherboard(?), and remember that for Zen/Naples the Infinity Fabric was supposed to offer an NVlink like ability to connect up more in a direct attatched GPU fashion to any Vega radeon Pro WX/instinct SKUs that will also make of of the Infnity Fabric IP. So there will probably be a 16 core Zen/Workstation variant on an MCM module with some of that extra Server/workstation IP on the MCM module. I do not think there will be any 16 core Ryzen/consumer branded SKUs.
(1)
“Rumor: AMD X390 and X399 chipsets diagrams leaked?”
https://videocardz.com/67594/rumor-amd-x390-and-x399-chipsets-diagrams-leaked
I saw thermoelectric in the
I saw thermoelectric in the title and immediately decided I had traveled back in time. Who on earth uses Peltier’s to transfer large amounts of heat at or above idle temps.
*ambient, not idle. Excuse
*ambient, not idle. Excuse me
No one. They’re way too
No one. They’re way too inefficient to move any reasonable quantity of heat. For every Watt they remove, they generate several more. Clearly that’s not going to scale beyond a dozen Watts or so. So, yeah, it may cool down your idle part below ambient, but what’s the benefit of that? Worse yet, they will act as insulators at load.
Peltier coolers will always
Peltier coolers will always (unless someone has a really incredible breakthrough) have a COP (Coefficient of performance) below 1.0, meaning they will always consume more energy than the heat they remove.
The only possible way of actively cooling (by active I mean not just pumping a secondary fluid to reject the heat) computer components is to use refrigeration cycles. The problem is they are far too complex to make it worth it so far. And it’s hard nowadays finding a refrigerant that is reasonably ecological and safe to use.
Absorption and adsorption cycles can also work, but require a large setup. And if noise is a concern then maybe thermosiphons will work, I’ve heard of a company using a thermosiphon to cool a pc with no moving parts required, but it was expensive and relatively large since it requires large heatsinks to condense the fluid.
@Sebastian.
I’d love to see
@Sebastian.
I’d love to see temp graphs combined with dB information…Not sure how that would work without causing info overload.
It could be a great way to get a complete picture of how each cooler performs without having to look at temp and dB graphs to figure out which cooler suits you best.
I’ve toyed with combined
I've toyed with combined graphs in the past with mixed results. That's an interesting idea, however – have not attempted combining noise/temp data like that before. If it made sense visually and was on a fixed scale it could work.
I’ll stick to a Noctua DH-15
I’ll stick to a Noctua DH-15 thanks, This might do a decent job, but it’s inefficient and you need to run it insane mode to achieve decent cooling when OC’d. No thanks.
I cannot see using this thing
I cannot see using this thing in a PC because existing methods are good, and because I am scared of what happens if power too the Peltier element were to fail under load.
However to cool very high power LED systems used for horticulture or large area lighting it could be extremely useful.
While idea is certainly
While idea is certainly interesting it has potential to be a catastrophic. When powers go down, CPU will be fried in no time at all.
If you really want tower cooler standard constructions are much more safe and proven.
And make no mistake. I’m all for innovating things, but practically comes first. Couple fans on tower cooler work perfectly well. Thank you.
I’ll definitely be trying
I’ll definitely be trying this out in my next build instead of a AIO, thanks.
Was the NOCTUA fan at maximum
Was the NOCTUA fan at maximum RPM?
This was the main confusing thing I noticed. If the Noctua fan had more RPM available then perhaps it could close the gap (not that it needs to for THIS exact OC as it’s well within temp).
Don’t understand this design,
Don’t understand this design, if its supposed to turn on when its needed then it means that the heat dissipation of the heatsink cant keep up with the processor and in that scenario the TEC just adds more heat.
TEC isnt viable anymore due to its inherent inefficiencies. A TEC element is a set of XxY smaller heatpumps where each specific pump is limited in how much it can handle. To manage the loads generated by modern CPU’s the size of the element needs to be very large and it’ll be a beast in power/heat.