HYSTOU P04 i3-7100U Mini PC, or the Computer I Couldn't Overheat

At the beginning of autumn, I tested and, accordingly, posted a review of my rather interesting HYSTOU FMP04 mini PC with an i5-7200U processor. Unfortunately, in that review I also showed the “Achilles’ heel” of both the model under review and many other Chinese devices: overheating.
After that, I decided to try a slightly less powerful version with an i3-7100U processor, and it really managed to surprise me. But you can read all about it in the review.


This computer is sold in four configurations:
1. The “system unit” only, without RAM or SSD. I can’t say for sure about WiFi, but based on my previous experience, it should be included — $246.17
2. Computer with 4 GB of RAM and a 128 GB SSD preinstalled — $326.09
3. Computer + 8 GB of RAM and a 256 GB SSD — $382.56
4. Computer + 16 GB of RAM and a 256 GB SSD — $448.63

The review covers option 2, but personally, I think options 1 and 4 make the most sense. Let me explain.
The processor can use up to 32 GB of RAM, but since there are only two slots and 16 GB modules are very expensive, people usually go with an 8+8 GB setup. The 4 GB version is only worth getting if you plan to make an “asymmetric” 12 GB setup, like the one in my main computer. Otherwise, the 4 GB stick will probably end up in a desk drawer.
So you can simply get the barebones unit without RAM or an SSD and buy everything separately, or get the 8 GB version and add another 8 GB later. It also comes with a 256 GB SSD, which is a better deal too.

Specifications:
Operating system: Not installed
Processor: ICORE-I3-7100U
Graphics: Intel® HD Graphics 620
Memory: max. 16 GB, SODIMM DDR3L, 2 slots, 4 GB installed initially
SATA — 2 x SATA 6 Gbps + 1 x mSATA (128 GB installed)
Wireless: 802.11b/g/n WiFi 2.4 GHz
LAN — Gigabit LAN
Display outputs: HDMI and VGA
External interfaces: 4x USB 3.0, 3x USB 2.0, 2x COM (RS232)
Audio output — 3.5 mm jack
Audio input — 3.5 mm jack
Dimensions: 234 x 100 x 50 (excluding feet, antenna, and connectors)
Weight: 1250 g

Everything is packed in a neat little white box.


The computer is packed quite well, and given its metal case, I don’t think it would be easy to damage.
The package includes:
1. HYSTOU P04 computer
2. Power supply
3. Power cable
4. SATA power cable
5. SATA cable
6. Two antennas
7. Mounting hardware
8. Manual
9. Driver CD
10. Mouse pad


The package contents are no different from the FMP04 version: the same manual, the same power supply, and even the same free mouse pad.
In the photo, the mouse pad is lying on top of mine, and you can see that mine is bigger—even though this one is already the new version, which is itself smaller than the previous one.


Anyone who read the first part about the HYSTOU FMP04 will probably say, “But it’s exactly the same thing!” And yes, you’d be 100% right. The only difference between these two models is the processor. Of course, the amount of RAM and SSD storage differs, but the FMP04 was also available with 4 GB of RAM and a 128 GB SSD, so that doesn’t count as a difference.


Because the computers are so similar, this review will be a little shorter, and today I won’t be taking everything apart “down to the last screw.” But of course, I’ll still cover all the general information.


The front has the same three USB 2.0 ports, a power button, and an opening for the LEDs.


The back has a full set of various connectors: HDMI and VGA video outputs, COM ports and USB 3.0, audio input and output, a pair of WiFi antenna connectors, and of course Ethernet and the power connector.


The case has the same impressive fins, though once again the manufacturer’s nameplate sticks out quite a bit. It’s annoying when the computer spins on it like a carousel during disassembly. I’d have preferred a nice engraving instead.


The rear panel with the antennas installed.


There are four feet on the bottom, four holes for installing an additional 2.5-inch hard drive, and four holes for mounting the computer to something. However, there’s no provision for mounting it behind a monitor, as there is with smaller models. And hanging a heavy unit behind a monitor doesn’t seem like a great idea anyway, in my opinion.
The bottom of the case is also made of aluminum and helps the main heatsink to some extent.


Inside, it’s the same old mess of wires. Of course, nobody usually sees this, but my inner perfectionist objects and wants things neat.


It would have taken so little to route the cables a bit more neatly, like I did last time. They won't get in the way when installing one hard drive, of course, but with two stacked like a sandwich, they easily could.


Let's move the cables aside for now and take a look at the motherboard and internals.


What can I say? The board is identical in every respect, right down to the revision: GSKL1-VERB


The WiFi module and mSATA drive are the same as last time. I'll complain again that a computer with a seventh-generation processor, released at the end of 2016, has no M.2 slot.
There are also two additional SATA connectors here. As before, there's only one power connector, and no Y-splitter is included; I had to rig one up myself in my computer.


Let's remove all the components, as far as possible.


1, 2. A small change: the FMP04 came with Samsung memory and a Hystou sticker, while this one has Micron memory with its original label.
3, 4. The SSD is the same, except that two flash memory chips are not populated and the capacity is 128 GB instead of 256.
5, 6. The WiFi is unchanged too: AzureWave, with a 2.4 GHz band.


Besides the main board, there are two additional boards: one with USB connectors, of which there are inexplicably three again, even though the board itself is designed for four; and a power button board with a pair of LEDs that shine everywhere except through the holes intended for them. That might actually be for the best, though.


All the internal interfaces and unpopulated connectors.


At the top you can see the pin header for connecting the front panel, COM ports, USB 2.0 connectors, and audio.


At the back is the same Realtek audio chip, which I still couldn't get working in Windows 7, although the problem turned out to be much deeper than that.
To the right are a fuse for the USB ports, the isolation transformer for the gigabit Ethernet port, and a DisplayPort-to-VGA chip.


The motherboard markings. By the way, this board is fairly common, not just in Hystou computers.


The power subsystem differs visually only in the color of a few capacitors.



Since this computer is hardware-identical to the HYSTOU FMP04 I reviewed earlier, apart from the processor, I didn't take it apart any further; there would be no point. But I did remove the rear panel and check the thermal interface between the processor and the case. It's exactly the same: an aluminum plate with thermal paste.

As before, we'll start by taking a look at the BIOS, especially since the computer is technically sold without an OS, even though the manufacturer has installed a "demo version".


Unfortunately, the option to power on automatically when power is applied is missing here too :(. I went through most, if not all, of the menus and still couldn't find that feature. Too bad.
Otherwise, everything is more or less the same.


The computer initially booted with a low screen resolution and English as the only language. I changed the resolution right away, but to change the language I had to connect to the internet to download a language pack. Nothing complicated, really; I described the process not too long ago.


After a few reboots, the computer was fully localized into Russian using Windows' standard method, so now we can see what it's like.
An unlicensed, activated copy of Windows 10 is preinstalled.


Device Manager.


The SSD is split into three partitions: the system partition, the main partition (drive C), and an additional partition (drive D).


CPU-Z showed roughly the same things as on the FMP04, with one difference: this board revision is from 2017, while the FMP04's is from 2018, which is a little odd.
The processors differ in that the i5-7200U runs at 2.5 GHz, with Turbo Boost up to 3.1 GHz. The processor in the unit under review has no Turbo mode, and its base clock is 2.4 GHz. Otherwise, it’s the same process technology, the same number of cores, and so on.
In fact, the only key difference is that this one doesn’t have a turbo mode. Let’s see how that affects overall performance.


As before, there’s no information about total operating time or the amount of data read from or written to the SSD, even though by the time I took the screenshot, the computer had already been running various tests for about a day.


SSD-Z didn’t turn up anything new either.


I tested the speed using a 45 GB file on drive D and a 30 GB file on drive C; the results were the same: sequential write speed was about 150 MB/s, and read speed about 520 MB/s. Write speed is a little low, but acceptable; no complaints about read speed.


CrystalDiskMark confirmed the previous test. I also took the opportunity to check USB 3.0 speed with an external SSD.


Now for some more useful information: performance.
First up is Cinebench. As you’d expect, the computer under review lagged behind in the math test, but was a little faster at graphics processing.
Graphics: 24.48/33.25 versus 23.30/31.55 for the FMP04
Math: 2.84/256 versus 3.64/324


But things got more interesting after that: Performance Test 8 showed a difference of just 10%. Keep in mind, though, that this is an average performance score across different tests.



Performance Test version 9 showed a similar 10% difference in favor of the FMP04.


In 3DMark, the computer under review was 10% faster than the FMP04, despite its more powerful processor.


The newer version of 3DMark also showed a 10% difference in favor of the model under review; only the Ice Storm test came out almost the same as on the FMP04.


Geekbench 4
Once again, it lagged behind in math and pulled ahead in graphics. The FMP04 scored 3499/6610 and 14558.


And yet, a test that seems essentially mathematical—working with the WinRAR archiver—showed the FMP04 trailing by about 7%. Quite unusual.


Well, I’ve gradually made it to the FMP04’s sticking point: the heat test. If you remember, I was surprised by the high temperature, given the fairly large heatsink and what should have been a low-power processor variant.

I run LinX for 20 passes, monitoring the temperature, and get just 53 degrees...


My first thought: this can’t be right. All right, I run the test a second time: 55 degrees. Mysterious. The FMP04 hit 92 even on the first run, let alone the second.


Maybe 20 passes isn’t enough. Let’s set it to 50 and run the test for half an hour. Sure enough, the temperature went up—to 59 degrees. Progress...
By the way, the model under review scored 43.94 GFLOPS in this test, while the FMP04 managed just 39.54. Yes, the difference in favor of the model under review is small, but its processor is weaker.


Case temperature just before the tests ended: the maximum I got was 45 degrees.


Well, LinX didn’t work out, so I’ll run OCCT.
After 55 minutes, the maximum temperature was 53 degrees; the FMP04 reached 86. I should probably mention that this is the second run shown in the screenshot. I usually take a screenshot just before the test finishes, but this time I simply missed it and had to run it again “so I could send a photo to the magazine.”


And now the most “heat-generating” test. Here I managed to get the processor all the way up to 61 degrees, compared with 99 on the FMP04.
Honestly, I expected the i3-7100U to run a little cooler, but not this much. Let me explain: the absolute difference was 38 degrees, 61 versus 99, but the relative difference, taking the ambient temperature into account, was twofold.


Even after three hours of testing, the case was at 49 degrees and the power supply at 48. Seriously, the case felt just a little warm; after the load was removed, the temperature returned to room temperature fairly quickly, i.e. The processor is barely warm at all in normal operation.


A gigabit wired network card and a 2.4 GHz WiFi module handle connectivity with the outside world.


Since the WiFi module is the same as in the FMP04, the results are comparable too: it found 77 access points nearby.


Connection speeds within a large room ranged from 15 to 60 Mbps, depending on distance and conditions.


With a wired connection, I got a steady 800–900 Mbps.


Last time, while testing performance in video encoding mode, I made a major and frustrating mistake that I only discovered much later. I took a project file and tried encoding it on the new computer. The result was about one and a half times faster, which of course made me happy.
In fact, I mistakenly tested encoding to a different format, assuming that this setting would also change when I copied the project file.
So this time I ran the test correctly and got 26 min 33 sec, compared with 31 min 43 sec on my old i3-5010U. Not a huge difference, but still.


But a problem came up during testing: both the i3-7100U and i5-7200U turned out to be incompatible with the Vegas version I usually use. I tried various fixes. Updating the video driver helped a little: I was able to transcode a five-minute clip, but when I tried encoding to the previous format (as in the last review), I still got an error. At first, the FMP04 worked correctly; I encoded a clip for the review several times and only noticed the problem some time later.
The problem is actually fairly easy to work around: just disable GPU acceleration. But performance drops noticeably.
The clip took 20 minutes to encode in the i5-7200U test. The i3-7100U should have been about the same, but with software-only processing it took almost half an hour. I don't really want to switch from Vegas, so I don't know what to do yet. A search online showed that this is a very common problem. It looks like this:
Start encoding the project.
After half a minute, two minutes, three minutes—it doesn't matter—the software freezes. The counter keeps ticking, but no frames are rendered.
There's no normal way to exit this state, and adding more RAM doesn't help.


The reason for the lower temperature is lower power consumption.
1, 2. In the off state and at the desktop, both systems are the same: 2–6 W.
3. But under maximum OCCT load, it was only 25 W, not 37 W like the previous one.
4. Peak power was about the same 37 W; the FMP04 reached 39 W. But I could only record this power by checking it in the wattmeter's Hi mode.


The final results are a little odd: in some tests the i5-7200U is faster, in others the i3-7100U. The LinX test results for the i3-5010U and i5-7200U are for a 4096 MB test file, while the i3-7100U was tested with 2448 MB—the program wouldn't allow more because it has less RAM. But since the score also increased with the test size, we can say that the i3-7100U still comes out ahead in this test. I also compared them under the same conditions, and the result was 39/43 in favor of the model under review.


To sum up, the i3-7100U was slower on average, sometimes noticeably, sometimes only slightly. But it also won some tests, especially graphics work, LinX, and WinRAR.
The computers are identical in design, but the difference in heat output is huge. Where the i5-7200U reached 96–99 degrees, the i3-7100U was only at 59–60. Of course, that's a drop in the ocean in terms of electricity savings, but it matters for reliability: there's less heat from the processor itself, which means less impact on other components such as the SSD, and less thermal stress on the board, which should make the system last longer.
Overall, I find this model more interesting, and it really managed to pleasantly surprise me.

The only problem both computers have is some incompatibility with Vegas. Though perhaps it's more accurate to say that this is a Vegas problem, not a computer problem, since it also occurs with other configurations. Maybe someone can suggest a fix.

The store provided a discount with the GBP04 coupon, bringing the price to $299.99.

That's all for now. I hope you found the review useful. If you have ideas for additional tests, let me know and I'll try them out.