Cincoze DS-1402 review: Part 3: Ubuntu 24.04 tested on an Intel Core i9-12900E embedded system

In part one of the Cincoze DS-1402 review, I checked out the hardware of the modular rugged computer before testing the Intel Core i9-12900E embedded computer with Windows 11 Pro in the second part, and I’ve now had time to test the Cincoze DS-1402 embedded system with Ubuntu 24.04.

I’ve tested most features in Linux, ran some benchmarks, evaluated the gigabit Ethernet performance of some of the ten Ethernet ports, checked CPU temperature under a stress test with and without the fan, measured power consumption, and more.

Cincoze DS-1402 Ubuntu 24.04 review

Ubuntu 24.04 system information

I installed Ubuntu 24.04 with the official x86_64 ISO in dual boot configuration with Windows 11 Pro. The installation process was smooth, and I have nothing specific to report here. Going to the Settings->About window confirms we have a CINCOZE-DS-1400 system (DS-1402 is part of the DS-1400 family) with a 24-thread 12th Gen Intel Core i9-12900E processor and 64GB RAM running Ubuntu 24.04 LTS. Somehow the SSD was not detected.

Cincoze DS-1402 Ubuntu 24.04 AboutWe can get some extra information in the command line:


The SSD has a capacity of 512GB but I resized the Windows partition to have 300GB for Ubuntu 24.04 since I needed at least 200GB free to regenerated the Llama 3 model when I reviewed the Radxa Fogwise Airbox.

inxi utility provides further details:


The Intel Core i9-12900E CPU P-cores are designed to work between 800 MHz and 5,000 MHz, and the E-Cores between 800 MHz and 3,800 MHz. Both the iGPU (Intel AlderLake-S GT1) and graphics card (NVIDIA TU117 – GeForce GTX 1630) are detected, as well as all 10 gigabit Ethernet ports (1x I219-LM and 9x I210) and the 512GB Cervoz CIE M8 T435 industrial-grade SSD. The idle CPU temperature is reported to be 33°C at idle and the NVIDIA GPU temperature is 39°C.

Cincoze DS-1402 benchmarks on Ubuntu 24.04

We’ll start benchmarking the Cincose DS-1402 (Core i9-12900E) with Thomas Kaiser’s sbc-bench.sh:


We’ll do a comparison later, but we can already see the power limits in action with the 7-zip test starting at 50,586 MIPS for the first run, before dropping to 42,634 and 42,664 MIPS in the second and third runs respectively. I can also notice that the I219-LM Ethernet interface is missing from the list of PCIe devices.

Let’s check the PL1 and PL2 power limits as requested:


PL1 is set to 65W and PL2 to 202W in Ubuntu 24.04, or exactly the same values as in Windows 11.

Geekbench 6.3.0 will be used to test the single-core and multi-core CPU performance of the Core i9-12900E processor.

Cincoze DS-1402 GeekBench 6.3.0 Intel Core i9-12900E

The Cincoze DS-1402 achieved 2,607 points in the single-core test, and 15,593 points in the multi-core test.

Let’s now test the GPUs with the Unigine Heaven Benchmark v4.0. The first time, I used the HDMI port on the front panel, and the system with an Intel Core i9-12900E processor and an Unknown GPU rendered the scene at 25.1 FPS on average with a score of 633 points at the standard 1920×1080 resolution.

Unigine Heaven Benchmark 4.0 Intel Core i9-12900E iGPU Ubuntu 24.04

That’s quite a low score and it’s likely the internal 32EU Intel UHD Graphics 770 GPU was used for this test. So I changed the settings in NVIDIA Settings->PRIME Profiles from “NVIDIA On-Demand” to “NVIDIA (Performance Mode)” and rebooted the system as requested.

Ubuntu 24.04 NVIDIA Settings

I also moved the HDMI cable to the HDMI port on the NVIDIA GTX 1630 graphics card, but that part did not matter since switching back to the front panel’s HDMI port and repeating the test yielded about the same results. In any case, now that the NVIDIA GTX 1630 entry-level graphics card was used the Unigine Heaven Benchmark 4.0 could be rendered at 42.8 FPS on average with a score of 1,078 points.

Unigine Heaven Benchmark 4.0 NVIDIA GTX 1630 Ubuntu 24.04

For reference, the computer averaged 48.3 FPS and achieved 1,218 points at 1920×1080 resolution in Windows 11 using Direct3D11 instead of OpenGL.

Next up was YouTube 4K and 8K video playback using Firefox while connected to the HDMI port on the front panel.

Cincoze DS-1402 YouTube 8K 60 FPS Ubuntu Firefox

I went straight to 8K 60FPS and it was a complete disaster with a massive amount of dropped frames (1,930 frames dropped out of 2,550) and I only had a still image after a few seconds.

Cincoze DS-1402 YouTube 4K 60 FPS Ubuntu Firefox

But after changing the video resolution to 2160p60 (4K 60 FPS), it played just fine with only a few frames dropped over a 7-minute period.

Cincoze DS-1402 YouTube 8K 30 FPS Ubuntu Firefox

I tried another 8K video at 30 FPS, and it played fine for 7 minutes with 31 frames dropped out of 12.663 points. I installed Google Chrome browser to see what would happen at 8K 60 FPS.

Cincoze DS-1402 YouTube 8K 60 FPS Ubuntu Chrome

It almost started fine for the first few seconds, but it quickly ended up being more or less the same as in Firefox with a massive number of dropped frames and the loading icon showing from time to time despite the buffer health doing OK. So I went to NVIDIA Settings again, and VDPAU does not list AV1 as a supported codec.

NVIDIA GTX 1630 GPU Supported Codecs

At that point, I thought maybe that must be the reason why. But in a plot twist, I turned off the computer before going to bed, then turned it on again in the morning and started playing the same 8K 60 FPS video in Firefox, and that time around the video played close to perfectly with only 104 frames dropped out of 37,102.

CINCOZE DS-1402 YouTube 8K 60 FPS Ubuntu Firefox Morning

I also checked CPU usage with the htop utility and the load was only 6 on a 24-thread system. It’s still using the AV1 codec, so I’m not exactly sure what happened before.

Finally, I tested web browser performance with Speedometer 2.0 with the latest version of Firefox.

Cincoze DS-1402 Speedometer 2.0 Ubuntu Firefox

The result was 236 runs per minute. It’s a bit lower than some on other Intel Core i9 or AMD Ryzen 9 I tested recently, so I repeated the test a few times, and each time it was 236 or 237. Maybe that’s because the Core i9-12900E is a 12th Gen CPU.

Since Speedometer 2.0 is deprecated, I also ran the newer Speedometer 3.0 to get data for future reviews, and the Cincoze DS-1402 achieved 16.3 points in the new benchmark.

Cincoze DS-1402 Speedometer 3.0 Ubuntu Firefox

Comparison of Cincoze DS-1402 benchmarks against mini PCs running Ubuntu 24.04/22.04

Time to compare the Ubuntu 24.04 benchmarks results for the Cincoze DS-1402 (Core i9-12900E) embedded computer against high-end consumer-grade mini PCs namely the GEEKOM XT12 Pro (Intel Core i9-12900H) and GEEKOM Mini IT13 (Intel Core i9-13900H), and GEEKOM A8 (AMD Ryzen 9 8945HS) running Ubuntu 22.04 or 24.04.

The table below shows the basic specifications of each system

Cincoze DS-1402GEEKOM XT12 ProGEEKOM IT13GEEKOM A8
SoCIntel Core i9-12900EIntel Core i9-12900HIntel Core i9-13900HAMD Ryzen 9 8945HS
CPU16-core/24-thread (8P+8E) Alder Lake processor up to 5.00 GHz (P-cores), up to 3.80 GHz (E-Cores)14-core/20-thread (6P+8E) Alder Lake processor up to 5.00 GHz (P-cores), up to 3.80 GHz (E-Cores)14-core/20-thread (6P+8E) Raptor Lake processor up to 5.40 GHz (P-cores), up to 4.10 GHz (E-Cores)8-core/16-thread processor up to 5.2 GHz
GPU32EU Intel UHD Graphics 770
NVIDIA GTX 1630 graphics card
96EU Intel Iris Xe Graphics96 EU Intel Iris Xe GraphicsAMD Radeon 780M Graphics
Memory64GB DDR5-480032GB DDR4-320032GB DDR4-320032GB DDR5-5600
Storage512GB NVMe SSD1TB NVMe SSD2TB NVMe SSD2TB NVMe SSD
Default OSWindows 11 ProWindows 11 ProWindows 11 ProWindows 11 Pro

Here is the Ubuntu/Linux benchmarks comparison table.

Cincoze DS-1402GEEKOM XT12 ProGEEKOM Mini IT13GEEKOM A8
sbc-bench.sh
- memcpy28,908.2 (P-Cores)22,375.8MB/s24,014.4 MB/s (P-core)20,318.5
- memset38,884.4 (P-cores)27,398.0MB/s26,647.9 MB/s (P-Core)62,156.7
- 7-zip (average)45,29040,19056,54068,790
- 7-zip (top result)50,58643,78360,98169,297
- OpenSSL AES-256 16K1,771,530.92k (P-Core)1,661,583.36k (P-Core)1,844,401.49k (P-Core)1,422,136.66k
Geekbench 6 Single2,6072,575
2,7452,661
Geekbench 6 Multi15,59310,44711,97413,275
Unigine Heaven score633 (iGPU)
1,079 (NVIDIA GTX 1630)
1,2931,3331,972
Speedometer 2.0 (Firefox)236298273298

Unsurprisingly the 24-thread Core i9-12900E processor will perform better than the 20-thread Core i9-12900H in multi-core benchmarks such as 7-zip and GeekBench multi-core. There’s more variability in single-core workloads, as both the DS-1402 and XT12 Pro have a similar GeekBench 6.3 single-core score, but the former is faster for AES-256, and the latter is faster in Speedometer 2.0. The iGPU of the Core i9-12900E is clearly not its strong point, and even the entry-level GTX 1630 graphics card installed in the computer is slower than iGPU of the other Intel and AMD processors tested, with AMD usually performing better.

Storage performance and USB ports

The 512GB NVMe SSD installed in the system was tested with iozone3 utility:


I found the lower write speed with 16KB IO requests to be rather odd. So I repeated the test two more times:


Some caching must make 1024 bytes IO requests faster during writes. The fastest read speed was around 1,783 MB/s, and I’m not sure if any of the numbers for the write operations can be used here… For reference, CrystalDiskMark reports a 2,088 MB/s sequential read speed, and a 1715 MB/s sequential write speed in Windows 11.  In any case, the Cervoz T435 is an entry-level SSD in terms of performance, but it’s also an industrial-grade SSD module designed to work in the -40°C ~ 85°C temperature range and an MTBF of over 3,000,000 hours.

An EXT-4 partition from a drive in an ORICO M234C3-U4 “USB4” M.2 NVMe SSD enclosure was used to check the speed of each USB 3.0 port along with lsusb and iozone3 command line utilities.

Here’s the output from the front left USB port:


and the top left USB 3.0 port on the rear panel:


I switched to a USB hard drive for the USB 2.0 ports. See the results for the top left USB 2.0 port on the rear panel:


Here’s a summary of the results of all eight ports from left to right:

  • Front panel
    • USB-A #1 – USB 3.2 – 10 Gbps – Read speed: 874 MB/s; write speed: 972 MB/s
    • USB-A #2 – USB 3.2 – 10 Gbps – Read speed: 871 MB/s; write speed: 983 MB/s
  • Rear panel
    • USB-A #1 (top) – USB 2.0 – 480 Mbps – Read speed: 37 MB/s; write speed: 40 MB/s
    • USB-A #2 (top) – USB 3.2 – 5 Gbps – Read speed: 437 MB/s; write speed: 447 MB/s
    • USB-A #3 (top) – USB 3.2 – 5 Gbps – Read speed: 439 MB/s; write speed: 447 MB/s
    • USB-A #4 (bottom) – USB 2.0 – 480 Mbps – Read speed: 38 MB/s; write speed: 41 MB/s
    • USB-A #5 (bottom) – USB 3.2 – 5 Gbps – Read speed: 437 MB/s; write speed: 447 MB/s
    • USB-A #6 (bottom) – USB 3.2 – 5 Gbps – Read speed: 438 MB/s; write speed: 447 MB/s

All results are as advertised in the specifications.

Gigabit Ethernet ports performance

The Cincoze DS-1402 has ten Gigabit Ethernet ports, two default ports, and eight implemented through two CMI expansion cards with four GbE RJ45 ports each. I won’t test them all and only report the results for the built-in ports, and the left RJ45 jack of each CMI module. I felt confident so I went straight to full-duplex (aka bidirectional) testing.

LAN1 (Intel i219):


LAN2 (Intel i210):


CMI module 1 (left):


CMI module 2 (right):


It’s all good. There was no issue like in Windows where we got 938 Mbps in one direction and 622 Mbps in the other while testing bidirectional transfers with iperf3. It could either be Windows or an iperf3 in Windows issue.

Cincoze DS-1402 Ubuntu stress test with and without fan

In order to evaluate the efficiency of the cooling, I ran a stress test on all 24 threads of the Core i9-12900E processor while monitoring CPU temperature and frequency with Psensor and sbc-bench.sh

Intel Core i9-12900E Stress Test Ubuntu 24.04

For the first few seconds the temperature jumped up to 77°C with the P-cores clocked at 3800 MHz and the E-cores clocked at 3600 MHz:


But the P-cores CPU frequency then stabilized at a lowly 800MHz and the E-cores ran at 2600 to 2700 MHz with the CPU temperature hovering at around 54°C after 12 minutes. That’s very far from the junction temperature (100°C), so I decided to disconnect the fan (around the 17:16:52 mark in sbc-bench.sh) and continue the test.

Cincoze DS-1402 fanless test

The CPU temperature slowly went up and after around 10 minutes, it reached 61-62°C. I went out and left the air conditioner at its current temperature (28°C).

Cincoze DS-1402 max temperature fanless stress test

I came back around 1h30 after disconnecting the fan, and the CPU temperature had stabilized around 70 to 71°C. That would imply the fan is optional, but we have to keep in mind the Cincoze DS-1402 is rated to run in the -40 to 50°C temperature range with the 65W Intel Core i9-12900E, and I did not stress other parts of the system which may further raise the package temperature. It still looks like it would be fine to run the Cincoze DS-1402 fanless even with a 65W CPU in some conditions. Note the company already offers the system in fanless configuration with 35W processors and a -40 to +70°C temperature range.

Cincoze DS-1402 power consumption with Ubuntu 24.04

I reconnected the fan and measured the power consumption with a wall power meter:

  • Power off – 6.3 – 6.4 Watts
  • Idle – 55.6 – 57.2 Watts
  • Video playback – 113.7 – 123.1 Watts (Youtube 8K 60fps (AV1) in Firefox)
  • CPU stress test (stress -c 24)
    • First 10 seconds – Quick ramp up to 219.7 – 219.9 Watts
    • Longer run – 95.5 – 95.9 Watts

During the measurements, the mini PC was connected to the LAN2 Ethernet port, two RF dongles (one for a keyboard, the other for a mouse), and a CrowView laptop monitor connected to HDMI and its own power adapter.  What’s baffling is the idle power consumption on Ubuntu 24.04 (around 56 Watts) is much lower than on Windows 11 (around 95 Watts). I can see ASPM is disabled in Linux, so it’s not the reason. There must be something that’s always on on Windows 11 that’s turned off in Linux.

Conclusion

The takeaway for Linux will be similar to the one in Windows 11, as the Cincoze DS-1402 modular embedded computer works well in Ubuntu 24.04 in the configuration under test with an Intel Core i9-12900E 16-core/24-thread Alder Lake processor, 64GB DDR5, and a 512GB SSD, and I did not come across any major issues. The industrial-grade NVMe SSD provided with the test sample is not the fastest around, but the users can always customize the computer to use a faster SSD at the time of order. Initially, I had trouble playing a YouTube video at 8K 60 FPS, but I had no such issue in a subsequent attempt the next day with the same video.

I haven’t tested the fan in detail in Linux, but I can confirm it’s still very noisy. However, as our tests showed, the Cincoze DS-1402 can still work as a fanless embedded computer under some conditions with a 65W processor (as tested) and no fan is needed when configured with a 35W processor.  Power consumption is quite high, but somehow the idle power consumption is much lower in Linux (56W) than in Windows (95W) for reason(s) that would need to be investigated. In any case, fan noise and relatively high power consumption would not typically be issues for the target applications such as oil and gas fields, factory floors, and mass transportation vehicles. What’s important is that the fan is placed outside on top of a metal case, so dust won’t get into the system easily compared to systems fitted with an internal fan.

I did not find any obvious issue, although as noted in the Windows review, the lack of faster Ethernet interfaces and a DIO header could be noted. However, it should not be a problem since the Cincoze DS-1402 is a modular embedded system and the company provides CMI expansion modules such as the CMI-10GLAN03-R10 with two Intel X550 10GbE RJ45 interfaces or the CMI-DIO02CMI module with 16x optically isolated DIO (8 in/8 out) among others.

The Cincoze DS-1402 is a solid and reliable system with impressive versatility thanks to a range of expansion modules and PCIe slots with a patented fixation system. It’s also adequate for harsh environments with shock and vibration certifications and a wide operating temperature range. The main consideration will likely be the price. You can build your own configuration on distributors such as Steatite or EG Electronics among others to find out the cost of a system that meets your requirements.

Share this:

Support CNX Software! Donate via cryptocurrencies, become a Patron on Patreon, or purchase goods on Amazon or Aliexpress

ROCK 5 ITX RK3588 mini-ITX motherboard
Subscribe
Notify of
guest
The comment form collects your name, email and content to allow us keep track of the comments placed on the website. Please read and accept our website Terms and Privacy Policy to post a comment.
5 Comments
oldest
newest
urostor
urostor
7 days ago

To use the dedicated GPU under Linux with NVidia on-demand, you can use the switcherooctl launch command.

Vall
Vall
7 days ago

> What’s baffling is the idle power consumption on Ubuntu 24.04 (around 56 Watts) is much lower than on Windows 11 (around 95 Watts) I respectfully disagree: what’s baffling is a computer with less performance than a Geekom A8 in many benchmarks, consume over 10x as much power in idle. It’s so ridiculous that the Cincoze turned OFF, consumes more power than the Geekom in idle. This IMO can’t be explained just by the usual inefficiency of its Intel CPU; something else is terribly wrong with this machine. This is even worse when you consider the Cincoze is positioned as… Read more »

Vall
Vall
7 days ago

Thanks for the additional data @Jean-Luc, really appreciated, specially removing the external GPU. So it alone was eating 18W, interesting. But my point is, the Cincoze still uses too much energy (and therefore produces too much heat) for too little computing power… certainly for my use case, and arguably (for the reasons I pointed) for the embedded market in general. Case in point: the discrete GPU you tested gives a score of just 1079 in Unihaven; the iGPU in the Geekom A8 gives us almost twice at 1972 — and eats a very small fraction of the energy that this… Read more »

Khadas VIM4 SBC