Posted on by Jean-Luc Aufranc (CNXSoft) - 9 Comments on Beelink EQi 304 review – Part 3: Ubuntu 26.04 on a Wildcat Lake mini PC

Beelink EQi 304 review – Part 3: Ubuntu 26.04 on a Wildcat Lake mini PC

After checking out the hardware with an unboxing and a teardown of the Beelink EQi Wildcat Lake mini PC, we reviewed the Intel Core 3 304 computer with Windows 11 in detail. We’ve now had time to test the Beelink EQi mini PC running Ubuntu 26.04 with Linux 7.0/7.1 kernels.

We’ll report our experience with feature testing, benchmark results, 2.5GbE, 10GbE, and WiFi 6 networking evaluation, storage and USB/Thunderbolt tests, and more, including power consumption measurements.

Beelink EQi Wildcat Lake mini PC Ubuntu 26.04 review

Ubuntu 26.04 installation and system information

I’d usually resize the Windows partition to install Ubuntu alongside it in dual boot configuration. However, since the Beelink EQi mini PC also comes with two PCIe Gen4 x1/x2 sockets, I installed Ubuntu 26.04 on a 256GB NVMe (PCIe Gen3 x4) SSD fitted into the PCIe x2 socket and left the Windows 11 partition on the UFS storage alone. Despite the Intel Core 3 304 chip being a brand-new CPU, I had no issues during installation.

Going to the Settings -> About window shows we have an AZW EQi computer powered by an Intel Core 3 304 penta-core processor with Intel Graphics (WCL), 16 GB of memory, and a 895.9 GB total disk capacity (built-in UFS + 2x NVMe SSDs) with Ubuntu 26.04 LTS 64-bit and Linux 7.0.0-27 kernel. For reference, Beelink is a brand owned by Shenzhen AZW Technology Co., Ltd.

AZW EQi Intel Core 3 304 Ubuntu 26.04 LTS

We can get more information with the inxi utility:


The utility detects a penta-core Intel Core 3 304 CPU clocked from 400 MHz to 4500 MHz (P Cores) and 3300 MHz (E cores), 16 GB of RAM, the three storage devices, including the YMUSAB4TF3D1C1 UFS flash, Realtek RTL8127 10GbE controller, MediaTek MT7922 802.11ax WiFi 6 (and Bluetooth) network, and an Intel I226-V Ethernet controller. The 4500 MHz is a little odd, since the maximum Turbo frequency for the Intel Core 3 304 is 4300 MHz on Intel Ark, but more on that a little later.

Let’s start the benchmarks with the usual sbc-bench.sh script:


First off, sbc-bench.sh is a neat little utility since it detected the wrong reported CPU frequency (4500 MHz), and instead measured 4285 MHz on the P core, much closer to the advertised 4.3 GHz Turbo frequency.

The 7-zip multi-core benchmark shows a decreasing performance starting with 17,533 MIPS in the first run, and dropped to 15,880 and 15, 666 MIPS in the second and third runs for an average of 16,360 MIPS. This usually points to thermal throttling, but if we check out the full log, we’ll notice it’s not necessarily the case, and power limits look to be the main culprits:


This is especially evident for the cpuminer benchmark, since at 64-65°C, we are far from the T-junction temperature (100°C).

The single-core 7-Zip benchmark started on the P Core and later moved to an E Core. It’s unclear whether it’s normal behavior from the script to test both types of core, or an issue, since the results only show one single-threaded score (5180 MIPS)


Let’s check the power limits:


PL1 is set to 15W, PL2 to 35W, and PL4 to 80W, all of which match the values reported in Windows 11. As a side note, PL1 and PL2 can be adjusted in the BIOS if needed.

We’ll now check the single-core and multi-core performance of the Intel Core 3 304 processor with Geekbench 6.7.1.

Intel Core 3 304 Wildcat Lake CPU Geekbench 6 benchmark

The AZW EQi mini PC scored 2490 points in the single-core test and 6657 points in the multi-core benchmarks

Since the Intel Core 3 304 also features an Intel AI Boost NPU, I tried to run Geekbench AI 1.7.0. However, only the CPU backend was available. No GPU, no NPU:


At this point, I also noticed Firefox would not load. The kernel reports the following:


I’m not sure it’s directly related to the issue I had in Firefox, but I upgraded the system to Linux 7.1 since the Wildcat Lake got a few more commits.


I can confirm we are now using Linux 7.1:


But still no luck with GeekbenchAI. So I installed the latest Intel Linux NPU driver (and firmware):


We’ll notice the firmware file used has been updated. The first four lines are before the update (NPU was already detected), and the last three lines after the upgrade (with the newer firmware version).


Let’s reboot to make sure everything is updated. I also installed OpenVINO and confirmed the NPU was detected:


But still no luck in GeekBenchAI… I suppose the benchmark needs to be updated.

Let’s now test the GPU performance with Unigine Heaven Benchmark 4.0. I tested it first with Linux 7.0, and then with Linux 7.1, and there was not a big difference. 16.6 FPS and 419 points in Linux 7.0, and 15.8 FPS and 399 points in Linux 7.1 at the standard 1920×1080 resolution.

Intel Core 3 304 Widlcat Lake Unigine Heaven Benchmark 4.0

Next, I played YouTube videos at 4K and 8K resolutions in Firefox and Chrome.

Beelink EQi YouTube 4K 8K Firefox Ubuntu 26.04
YouTube 4K/K video playback in Firefox

I was really impressed by YouTube video playback on Windows 11 since I could play anything smoothly up to 8K 60 FPS. It’s not quite the case on Ubuntu 26.04 on Firefox, 4K 30 FPS was just fine, 8K 30 FPS was mostly fine with one or two micro freezes, 4K 60 FPS had a few more loading issues (buffer health was fine, so not related to networking), and 8K 60 FPS was unwatchable, worse than a slideshow.

Beelink EQi YouTube 4K 8K Chrome Ubuntu 26.04
Youtube 4K/8K video playback in Google Chrome

Google Chrome was a bit better with 4K 30 FPS and 60 FPS playing smoothly, 8K 30 FPS mostly fine (a few micro freezes), and 8K 60 FPS was not quite a slideshow, but still unwatchable due to the large number of dropped frames. You’ll also note the mix of VP9  and AV1 video codec during playback, while it was all AV1 in Windows, unless I look for a specific VP9-only video.

Let’s now test web browsing with Speedometer 2.0 in Firefox. The system managed 329 runs per minute, quite impressive.

Intel Core 3 304 speedometer 2.0 benchmark

Since Speedometer 2.0 is deprecated, I repeated the test with Speedometer 3.1 for future reviews. Score: 23.2 points

Intel Core 3 304 speedometer 3.1 benchmark

Now that we have gathered Ubuntu 26.04 benchmark results, it’s time to compare the Beelink EQi Wildcat Lake Core 3 304 mini PC against the previous generation Intel Alder Lake-N platforms, such as the GEEKOM Mini Air12 (Intel N100) mini PC and ODROID-H5 (Intel Core i3-N300, actively cooled) SBC, as well as a mid-range ($500+) system: GEEKOM A5 Pro 2026 Edition.

Here are the high-level specs of the four systems.

Beelink EQi Wildcat LakeGEEKOM Mini Air12ODROID-H5GEEKOM A5 2026 Edition
SoCIntel Core 3 304Intel Processor N100Intel Core i3-N300AMD Ryzen 5 7530U
CPU5-core (1P+4E) Widlcat Lake processor up to 4.3 GHz4-core Alder Lake-N processor up to 3.4 GHz8-core Alder Lane-N processor up to 3.80 GHz6-core/12-thread up to 4.5GHz
GPU1-core Intel Xe3 Graphics @ 2.3 GHz24 EU Intel UHD Graphics up to 750 MHz32 EU Intel HD Graphics @ 1.25 GHz7-core AMD Radeon Vega Graphics @ 2.0 GHz
Memory16GB LPDDR5-640016GB DDR5-480016GB DDR5-4800 SO-DIMM
(user installed)		16GB DDR4-4800
Storage512GB UFS 3.1512GB M.2 NVMe SSD512GB M.2 NVMe SSD
(user installed)		1TB M.2 SSD
Default OSWindows 11 ProWindows 11 ProN/AWindows 11 Pro
Linux OS Ubuntu 26.04Ubuntu 24.04Ubuntu 26.04Ubuntu 25.10
Price (July 19, 2026)$509 (pre-order), $609 MSRP$475 on GEEKOM website (with default coupon)About $500+ with case depending on configuration$499 on Amazon

And now for the benchmark results…

Beelink EQi Wildcate LakeGEEKOM Mini Air12 LiteODROID-H5GEEKOM A5 2026 Edition
Linux distributionUbuntu 26.04Ubuntu 24.04Ubuntu 26.04Ubuntu 25.10
sbc-bench.sh
- memcpy21,001.1 MB/s9,577.9 MB/s12,034.0 MB/s16,443.7 MB/s
- memset48,200.1 MB/s8,403 MB/s21,967.5 MB/s15,861.3 MB/s
- 7-zip (average)16,36012,75023,82036,370
- 7-zip (top result)17,53313,56623,87238,503
- OpenSSL AES-256 16K1,523,187.71k1,231,295.83k1,376,447.15k1,224,955.22k
Geekbench 6 Single2,4901,2281,3891,980
Geekbench 6 Multi6,657
3,1225,1787,024
Unigine Heaven score419276475686
Speedometer (Firefox)329157174352

The Intel Core 3 304 stands out with its high memory bandwidth and single-core performance similar to high-end Intel CPUs against all three competitors. It’s much faster than the quad-core Intel N100 CPU in all aspects, and only gets beaten by the ODROID-H5 (octa-core) Alder Lake-N platform in some multi-core benchmarks, as well as in the Unigine Heaven 4.0 Benchmark for 3D graphics acceleration. Those are also the main weaknesses against the AMD Ryzen 5 7530U mini PC, which is significantly faster for 3D graphics and multi-core workloads, as we’d already seen on Windows 11.

Storage and USB performance

Since we’ve installed Ubuntu 26.04 on the M.2 (PCIe Gen3 x2) SSD with EXT-4 file system, I’ll start storage testing with that one:


About 1600 MB/s read and write sequential speeds as expected. Let’s run the same test on the M.2 (PCIe Gen3 x1) SSD with EXT-4 file systems:


Again, as expected, with 863 MB/s reads and 832 MB/s writes. Note that all ports are Gen4, so you could expect doubling those values with a proper PCIe Gen4  SSD.

I also tested the UFS (Windows) partition:


We have 3.7 GB/s sequential read speed and 649 MB/s sequential write speed. The read speed number is probably wrong because of caching, as the NTFS file system does not support the direct I/O option in iozone. For reference, we got 2,013.58 MB/s sequential read speed with CrystalDiskMark on Windows 11.

We also tested the USB4 and USB 3.2 ports on the mini PC with an ORICO M234C3-U4 M.2 NVMe SSD enclosure and the two USB 2.0 Type-A ports with a Seagate USB 3.0 HDD, using botctl, lsusb, and iozone utilities as appropriate.

For reference, here are the results for the left USB 3.2 Type-A port on the front panel…


… the left USB4 port on the rear panel…


… and one of the USB 2.0 ports:


Here’s a summary for all six USB ports from left to right:

  • Front panel
    • USB-A – USB 3.2 – 10 Gbps – Read speed: 996 MB/s; write speed: 985 MB/s
    • USB-C – USB 3.2 – 10 Gbps – Read speed: 1,001 MB/s; write speed: 988 MB/s
  • Rear panel
    • USB-C #1 – Thunderbolt 3 –  40Gbps – Read speed: 2,942 MB/s; write speed: 2,617 MB/s
    • USB-C #2  – Thunderbolt 3 –  40Gbps – Read speed: 2,942 MB/s; write speed: 2,623 MB/s
    • USB-A #1 – USB 2.0 – 480 Mbps – Read speed: 41 MB/s; write speed: 42 MB/s
    • USB-A #2 – USB 2.0 – 480 Mbps – Read speed: 41 MB/s; write speed: 42 MB/s

All ports work as expected, and I didn’t come across the same reliability problems as in Windows, pointing to driver issues on the Microsoft operating system.

I also decided to test the Khadas Mind Graphics 2 dock with an NVIDIA GeForce RTX 5060 Ti graphics card. The dock was detected, but the not the NVIDIA graphics card.


We can see the audio part (speaker), an extra 2.5GbE USB port, and a 7.4GB SD card from the Khadas device, but no new NVIDIA graphics device.  The Intel Tapex Creek peripheral is also detected. I could not find any NVIDIA reference in lspcie, dmesg, etc… I’m not quite sure how to make that one work, and it’s the same problem as in Windows.

Networking (2.5GbE, 10GbE, and WiFi 6) performance

Let’s now test networking on the Beelink EQi Wildcat Lake Core 3 304 mini PC, which offers 2.5Gbps Ethernet, 10Gbps Ethernet, and Wi-Fi 6 (802.11ax) connectivity.

We’ll start with the 2.5GbE RJ45 port using the iperf3 utility and AAEON’s UP Xtreme i11 Edge mini PC on the other side.

  • Download
  • Upload
  • Full duplex

All good. The 10GbE port was tested through the iKOOLCORE R2 max mini PC running QWRK, a fork of OpenWrt.

  • Download
  • Upload
  • Full-duplex

Again, perfect! Even better than in Windows for the bi-directional test.

I also tested 5 GHz WiFi 6 using a Xiaomi Mi AX6000 router.

  • Download
  • Upload

840 Mbps download and 728 MB/s upload is not too bad, but for some reason, it’s much lower than the 1.77 Gbps download speed and 1.65 Gbps upload speed we got on Windows 11 Pro with the exact test environment. Some tweaks may be required to extract more performance.

I also quickly tested Bluetooth connectivity by pairing the Beelink EQi to an OPPO A98 5G Android smartphone, using the mini PC as a de facto Bluetooth speaker and transferring a file. No issues at all here. The file was even transferred while listening to music over Bluetooth.

Beelink EQi 304 Bluetooth File Transfert

Thermal performance of the Beelink EQi 304 mini PC on Ubuntu 26.04

To evaluate the thermal performance/cooling efficiency of the Beelink EQi Wildcat Lake mini PC on Ubuntu 26.04, we ran a stress test on the penta-core Intel Core 3 304 CPU while monitoring CPU temperature and frequency with Psensor temperature monitor and sbc-bench.sh script.

Beelink EQi 304 stress test Ubuntu 26.04

We can see a sharp CPU temperature increase from 38°C to 80°C at the beginning of the stress test before it stabilizes at around 65°C over the long test.  That probably means we could play around with the PL1 (long duration) power limit to slightly increase the performance. The CPU frequency stabilizes at 2900 MHz (P Core) and 2600 MHz (E Core), and probably reached the maximum 4.3 GHz/3.3 GHz during the initial spile, which was short enough that sbc-bench.sh didn’t pick it up in monitor mode.


In any case, that means the cooling solution is perfectly adequate.

Fan noise

As noted in the Windows review, the Beelink EQi Wildcat Lake Core 3 304 mini PC is probably the quietest actively cooled mini PC we’ve ever tested, as it stays very quiet even under heavy loads. We measured the fan noise with a sound level meter placed around 5 centimeters from the top of the device:

  • Idle – 38.9 – 39.1 dBA
  • Stress test on all 5 cores – 40.1 – 40.2 dBA

For reference, the meter measures around 36.4 – 36.8 dBA in a quiet room.

Beelink EQi Wildcat Lake power consumption on Ubuntu 26.04

I finally measured the power consumption with a wall power meter:

  • Power off – 0.9 Watts
  • Suspend – 1.6-1.7 Watts
  • Idle –
    • WiFi 6 only – 4.0 – 4.2 Watts
    • WiFi 6 + 2.5 GbE – 4.7 – 4.9 Watts
    • WiFi 6 + 10 GbE – 5.6 – 5.7 Watts
  • Video playback – 13.6 – 18.6 Watts (YouTube 4K 60FPS in Firefox)
  • CPU stress test (stress -c 5)
    • First few seconds – 33.9 – 34.1 Watts
    • Longer runs – 18.8 – 18.9 Watts

During testing, the mini PC was connected to Wi-Fi 6, a USB RF dongle for a wireless keyboard and mouse combo, and a CrowView Note laptop shell via HDMI.

Conclusion

The Beelink EQi Wildcat Lake Core 3 304 mini PC works relatively well on Ubuntu 26.04, considering it’s based on an SoC from a brand new processor family. As noted in the Windows review, it delivers excellent single-core performance, close to high-end Intel processors, which should be great for workloads such as web browsing, outstanding 2.5 GbE and 10GbE networking performance, and support for up to three 4K displays. The system is also the quietest actively-cooled mini PC we’ve reviewed, and power consumption is fairly low for an Intel computer.

There are some differences between Windows and Linux. WiFi 6 throughput was acceptable at 800 Mbps in Linux, but it’s still half of what I achieved under Windows in the same testbed.  While I was truly impressed by 4K and 8K YouTube video playback on Windows, the system struggles more under Linux, using either Firefox or Chrome, since 8K 60 FPS videos are completely unwatchable, and 4K 60 FPS and 8K 30 FPS experience a few micro freezes from time to time, and only 4K 30 was smooth at all times. One plus on Ubuntu 26.04 was that I found the Thunderbolt 4/USB4 to be more reliable, as I had no issues with an ORICO NVMe enclosure. However, I was still unable to use the Khadas Mind Graphics 2 dock with an NVIDIA graphics card. Another small issue I had on Linux was that the default Firefox (snap) would not start, whether Linux 7.0 or 7.1 was used, so I had to remove it and install Firefox from apt instead. Finally, while the NPU on the Wildcat Lake processor was detected on Linux, GeekbenchAI benchmark could not find it.

Compared to the Alder Lake-N and Twin Lake families, the new Wildcat Lake family delivers higher performance in every way, except for some multi-core benchmarks (Geekbench) and Unigine Heaven 3D graphics, where the Intel Core i3-N300 octa-core processor found in the ODROID-H5 was slightly faster. The Beelink EQi 304 is also slower than a similarly priced AMD Ryzen 5 system for multi-core and 3D graphics workloads, but delivers much higher memory bandwidth and single-core performance.

From my early testing of the Beelink EQi 304 mini PC, I’d say Wildcat Lake mini PCs look promising, but I would not rush to pre-order a system, since there’s more driver and software work to be done to improve interoperability of the Thunderbolt ports (Windows especially, but also Linux), and various other compatibility/performance enhancements are needed o Ubuntu Linux.

I’d like to thank Beelink for sending the Wildcat Lake Core 3 304 mini PC for review. The model reviewed here, with 16GB LPDDR5 and 512GB UFS 3.1 storage, is available for pre-order for $509, and the MSRP is shown to be $609 after the pre-order period. Shipping is still scheduled to started in approximately 35 days…

Share this:
FacebookTwitterHacker NewsSlashdotRedditLinkedInPinterestFlipboardMeWeLineEmailShare

Support CNX Software! Donate via cryptocurrencies, become a Patron on Patreon, or purchase goods on Amazon or Aliexpress. We also use affiliate links in articles to earn commissions if you make a purchase after clicking on those links.

Radxa Dragon Q8B Edge AI SBC with Snapdragon 8cx Gen3 SoC

9 Replies to “Beelink EQi 304 review – Part 3: Ubuntu 26.04 on a Wildcat Lake mini PC”

  1. Meh. Waiting for the Gorgon Halo PCs with 192GB RAM to show up (preferably in 16″ or larger notebook format). Please wake me up when they do.

    Reply

    1. Strix/Gorgon Halo is obviously a wholly different segment with a different purpose. I assume you’ll be paying at least $5,000 for your 192 GB monster.

      It’s a cheap mini PC that would look a lot better if it was $200-300.

      Reply

  2. but is this wildcat lake the succesor of alder lake N? is it confirmed oficially by intel? if so, it is a decent upgrade using the latest node.. sadly we wont get anything better for the next decade

    Reply

    1. I’ve thought that Wildcat Lake should be considered the successor to Alder Lake-N. It makes sense to give 1-2 P-cores to the segment. However, it has less PCIe lanes and generally uses more power. Some vendors are using it in similar ways to Alder Lake-N.

      There should be a Wildcat Lake Refresh coming (2027?) that has up to 4 P-cores instead of 2, with no other changes as far as we know.

      In May, DigiTimes reported on “Moon Lake” coming in 2028. That may be a true E-core-only successor to the Alder Lake-N and Twin Lake family that has the correct feature set for embedded applications, and maybe lower pricing.

      Taking a significant step back in single-thread performance is not good for users of cheap PCs, but Wildcat Lake isn’t really cheap when it’s around $500 for the worst SKU. AMD might do something interesting and compete in this muck with a “Bumblebee” design (2x Zen 6 + 2x Zen 6C + 2x Zen 6 LP). It would probably have similar pricing intended for MacBook Neo clones, but perhaps superior CPU/GPU performance.

      Reply

      1. as far as i understand, intel is ditching the BigLittle architecture for a unified one once they get 14A working.

        Reply

        1. Titan Lake supposedly. But I don’t think it means they’ll move to only using one type of core. LPE-cores will probably coexist in most products to provide a low power mode, and “E-cores” as a subdued P-core or beefed up LPE-core could persist.

          Since Wildcat Lake technically has LPE-cores, not E-cores, we’ll have to see if the description of E-core-only Moon Lake is wrong (if it materializes).

          Reply

  3. > The single-core 7-Zip benchmark started on the P Core and later moved to an E Core

    As designed, please see the full log. There we see the P core scoring 5180 (at 4300 MHz) vs. 3770 for the E core (at 3300 MHz). Which is just a 6% difference in performance relative to clockspeed.

    Would be interesting to do a ‘MaxKHz=3300000 sbc-bench -G’ run since if supported by the kernel this would test both P and E cores at the same clockspeed with GB6 (which is not a great benchmark but may show at least some more differences between the two core types)

    Reply

Leave a Reply

Your email address will not be published. Required fields are marked *

Boardcon MINI1126B-P AI vision system-on-module wit Rockchip RV1126B-P SoC
Boardcon MINI1126B-P AI vision system-on-module wit Rockchip RV1126B-P SoC