Upgrade testing of the industrial mini giant KIWI310
Hello everyone, I am Dr.K. Since we introduced KIWI310 earlier, today we have obtained first-hand updated samples of KIWI310. Based on the specifications we have obtained, the main upgrade is the CPU, which has been upgraded from the original Celeron N3350 (referred to as N3350 below) to Celeron J3455 (referred to as J3455 below), and the PCB circuit board has also been partially adjusted according to J3455. Based on the technical data we obtained from Intel's official website, the main difference between the two is that N3350 is a dual-core processor while J3455 is a quad-core processor. The power consumption of the cores has also been increased from 6W for the dual-core to 10W for the quad-core. Both processors belong to the Intel Apollo Lake generation, so technically they should be "Pin to Pin compatible". However, Axiomtek has redesigned the power supply circuit for the 6W and 10W differences, indicating a very high pursuit of stability.
Next, let's take a look at the comparison table between J3455 and N3350 (data sourced from the Intel website).
Intel Celeron J3455 |
Intel Celeron N3350 |
|
Vertical Segment |
Desktop |
Mobile |
Processor Number |
J3455 |
N3350 |
Total Cores |
4 |
2 |
Total Threads |
4 |
2 |
Burst Frequency |
2.30 GHz |
2.40 GHz |
Processor Base Frequency |
1.50 GHz |
1.10 GHz |
Cache |
2 MB |
2 MB L2 Cache |
TDP |
10 W |
6 W |
Processor Graphics |
HD Graphics 500 |
HD Graphics 500 |
Graphics Base Frequency |
250 MHz |
200 MHz |
Graphics Burst Frequency |
750 MHz |
650 MHz |
From the Vertical Segment item, it can be seen that Intel positions N3350 in the mobile market, while J3455 is positioned in the desktop market. There are some differences in their target audiences, so there are also clear differences in design and specifications. J3455 has a complete Quad Core with 4 CPU cores, with a Base Frequency of 1.5GHz, higher than N3350's 1.1GHz, and a slightly lower Burst Frequency of just below N3350's 2.3GHz. The built-in HD Graphics 500 core also has an operating base frequency that is increased from N3350's 200~650MHz to 250~750MHz. The increase in core count and clock speed also results in different TDPs for the two, with J3455's TDP increasing from 6W to 10W. Perhaps it is due to these design differences that Axiomtek ultimately optimized the power supply design when upgrading to J3455 in the KIWI310.
Let's take a look at the physical appearance of the KIWI310 J3455 version. As you can see, the board itself maintains the same compact 1.8-inch mini size, but the engine core has been upgraded to a physical 4-core processor, with both the base frequency and graphics processor frequency significantly increased. We can expect an improvement in performance based on tests.
With the TDP of the CPU increased to 10W, Axiomtek also provides an optional heatsink module with a fan. With a 10W processor, this should be sufficient to handle the heat. Currently, Intel CPUs have a very good temperature protection mechanism, which will dynamically adjust the frequency according to the processing requirements and heat dissipation conditions to stabilize the system. In general, the CPU is likely the most reliable component in the entire computer.
Fig xx. Heat sink for KIWI310
the overall appearance has a texture unique to bare machine enthusiasts. If you have sharp eyes, you will notice that the corners of the heat sink are rounded, which increases the cost of CNC machining and edge repair when using handheld devices for assembly. Finally, the anodized iron gray treatment adds a level of delicacy that is not inferior to that of mobile handheld devices.
The fan itself requires an additional 5V power supply and can be directly connected using a 2-pin pitch 2.54mm Dupont Connector. This connector is currently the easiest to obtain and can be connected directly to the +5V and GND signals on the 40-pin GPIO.
According to the functional labeling of the 40-pin GPIO, Pin 4 is the 5V power supply, which connects to the red positive wire of the fan (indicated by the red arrow in the figure). The black wire is connected to Pin 5's ground (indicated by the green arrow in the figure).
Next, let's take a look at the parameters on the comput
- KIWI310 N3350 CPU-Z information
- KIWI310 J3455 CPU-Z information
As you can see, the main difference between the two is the difference in CPU specifications, while the speed and CL value of the RAM are the same.
We plan to use three testing software today, which are:
- Pass Mark Performance Test V9
This software mainly focuses on comprehensive performance testing of individual items such as CPU, 2D graphics, 3D graphics, memory, and disk, and is currently one of the most referenceable comprehensive performance testing software.
- 3D Mark v2.8.6546
This software mainly focuses on heavy 3D graphics testing of the DirectX 12 3D API and is a comprehensive 3D gaming testing indicator specifically for independent graphics cards. As the KIWI310 does not have an independent graphics card, it only uses Intel's built-in HD Graphics as reference data.
- AS SSD Benchmark 2.0.6821.41776
This software mainly focuses on comprehensive read and write performance testing of storage devices, covering the scope of system operation and user read and write, such as large file read and write, 4K random read and write, 4K-64Thrd simultaneous 64-thread 4K random read and write, Acc.time search time, and the final comprehensive score.
Testing Platform:
- OS:Windows 10 IOT
- SBC1:KIWI310 N3350 4G RAM/64G eMMC
- SBC2:KIWI310 J3455 4G RAM/64G eMMC
- Power supply:FSP-Group FSP065-A1BR3 65W PD Adapter
- Monitor:ViewSonic VX2476 24inch
- Mouse:ASUS ROG STRIX IMPACT II RGB USB
- Keyboard:ASUS ROG STRIX SCOPE NX TKL 80% RGB
Before starting the tests, there is an adjustment that needs to be made. Generally, in fanless heat dissipation systems or single-board computers, the Turbo Mode of the CPU is turned off by default. This is because these types of computers are designed for long-term operation, such as 24-7-365, and therefore the performance part is often set more conservatively.
To enter the BIOS settings screen, quickly and continuously click the Delete key on the keyboard when starting up. Then, go into the Advanced menu and scroll down to find Turbo Mode, which is disabled by default. Press Enter and use the up and down arrow keys to modify it to Enabled in order to activate it.
After opening it, you can use CPU-Z to see that under desktop mode, the CPU speed in CPU-Z will be boosted to 2.1~2.3GHz. This is not a problem with the motherboard or system, but rather the Burst Frequency refers to the maximum speed of a single core, while CPU-Z can only display the maximum speed of multiple cores. For more details, please refer to the explanation on Intel's N3350 product description page.
Next, let's take a look at the data collected from the tests.
1,Pass Mark Performance Test V9
From the test results, the PassMark Rating score varies, but the J3455 has the highest score under Turbo mode, and the default J3455 without Turbo has similar performance to the Turbo N3350.
In terms of the CPU Mark score, it mainly focuses on the computational ability of the CPU. It is clear that the budget of a quad-core CPU is greater than that of a dual-core one. The score increases as the CPU frequency after Turbo mode increases.
Regarding the Memory and 2D Graphics parts, the CPU's performance may have limited their performance to some extent. Therefore, with the help of the higher frequency and quad-core CPU, there are significant differences. However, both N3350 and J3455 can achieve the same limit after the frequency is increased, indicating that the number of cores does not make much difference in these two areas.
2,3D Mark v2.8.6546
The 3D Mark mainly tests the performance of the display core under DirectX 12 3D API, and these games are monsters that focus on graphics performance. Therefore, the graphics cores of N3350 and J3455 are indeed not competent and should only be used as a reference.
3,AS SSD Benchmark 2.0.6821.41776
Read Test
Write Test
The AS SSD Benchmark specifically tests the performance and access efficiency of storage devices. The CPU has little influence on the results of the Read or Write tests. Therefore, users who emphasize read and write performance can confidently choose any version of KIWI310 CPU.
- Heatsink temperature?
During the testing process, we also used an infrared thermometer to monitor the surface temperature of the heat sink of the KIWI310 N3350 and J3455. Under Turbo mode and in an air-conditioned room with a temperature of 26-28 degrees Celsius, the surface temperature of the N3350 heat sink when running RassMark was measured to be around 46 degrees Celsius, while the heat sink of the J3455 with a fan maintained at around 33 degrees Celsius. This is for reference only.
The surface temperature of the KIWI310 N3350 heat sink when running RassMark is around 46 degrees Celsius.
The surface temperature of the KIWI310 J3455 heat sink with a fan running RassMark is around 33 degrees Celsius.
Summarize:
The KIWI310 uses the Intel-developed System on a Chip (SoC) with the code name Apollo Lake, which provides some flexibility in the default CPU selection. However, since this type of SoC must be directly mounted on the PCBA during assembly, ordinary users cannot replace it themselves. Although SoCs in the same series are hardware-compatible, there are still some optimizations needed for power supply, heat dissipation, BIOS, etc. In addition, actual applications involve communication with software and other related hardware, and cannot be evaluated based solely on one item, as in our testing.
According to the overall test results, as long as the processor does not need to run at full speed frequently, the standard KIWI310 N3350 is sufficient for most requirements. When a large amount of CPU calculation is required in the actual usage environment, and the CPU heat dissipation is ensured, turning on Turbo Mode can provide additional CPU performance, offering users a flexible choice. If these adjustments are still insufficient, the upgraded version of the KIWI310 J3455 can provide quad-core processing capabilities and higher processor performance, allowing for more imagination in selection and evaluation. These adjustments can be completed without changing the casing or software, making it suitable as a development platform in the product development phase. It can support ARM platform interfaces such as GPIO/SPI/UART/PWM IO, as well as x86's most common interface, M.2 Key-E PCIe x1 & USB 3.0, with flexible overall costs and upgrade space. We also look forward to the KIWIboard platform launching more eye-catching new products in the future!