Disclosure: Some links on this page are monetized by the Skimlinks, Amazon, Rakuten Advertising, and eBay, affiliate programs, and Liliputing may earn a commission if you make a purchase after clicking on those links. All prices are subject to change, and this article only reflects the prices available at time of publication.

The ACEMAGIC TANK03 is a mini PC with the kind of features you’d expect from a premium gaming laptop, including support for up to a 12th-gen Intel Core i9 H-series processor and NVIDIA GeForce RTX 3080M graphics.

But instead of putting those components in a laptop, ACEMAGIC stuffed them inside a tiny desktop computer that’s just a little larger than a Nintendo GameCube. The company also offers the little computer with up to 32GB of RAM and 1TB of storage.

On paper, those features look pretty impressive. But the ACEMAGIC TANK03 sells for $1399 and up, making it the company’s most expensive mini PC. So we wanted to test the system and find out whether it’s worth the money.

ACEMAGIC sent a review unit with an Intel Core i9-12900H processor, NVIDIA RTX 3080 graphics, 32GB of RAM and 1TB of storage to Liliputing for free, with no requirement that the computer be returned upon completion of the review. This review is not sponsored by ACEMAGIC, and the company did not modify or approve the content of this article in any way.

Design & specs

The TANK03 is a futuristically styled cube measuring 167 x 167 x 161 mm (6.6 x 6.6 x 6.3 inches). Being named “tank” I originally thought of an armoured fighting vehicle. However according to ACEMAGIC “the TANK03 draws inspiration from space starships, incorporating the starship’s large curved design with detailed color accents and lighting embellishments, evoking the image of a starship swiftly maneuvering through space and engaging in battle scenes”.

The TANK03 is entirely made from gunmetal coloured plastic and is surprisingly heavy, weighing in at around 2 kg (4.4 lbs). This is because internally there is quite a complex cooling system.

Image credit: ACEMAGIC

Sitting behind the front of the device is a fan that pulls in air and pushes it through a large radiator to cool five heat pipes from the heatsink that covers the GPU. The GPU is mounted on a Mobile PCI Express Module (MXM) board which is connected to the dual-sided motherboard. On the motherboard’s top side besides the GPU, there are externally facing I/O ports and vertically mounted slots for memory, storage and Wi-Fi.

On the bottom side of the motherboard is the CPU, which is soldered on and covered by a heat sink that has two heat pipes leading to two heat exchangers. Covering the heat sink is another fan that sucks up air through the base of the device and channels it to the two heat exchangers. The hot air is then expelled out of the case to the rear and left of the motherboard.

Image Credit LukeD

In terms of processing power, ACEMAGIC offers two 45 W mobile processor options: an Intel Core i9-12900H, as in the review unit, or an Intel Core i7-12700H. Both processors have 14 cores in total, comprising of 6 “performance” and 8 “efficient” cores, which together provide a total of 20 threads. The “i9” performance cores can boost to 5.00 GHz whereas the “i7” boost to 4.70 GHz. For the efficient cores, the “i9” can boost to 3.80 GHz and the ”i7” boosts to 3.50 GHz.

The model with an Intel Core i9 processor (M1A) is the only version to include an NVIDIA GeForce RTX 3080M (125 W, 16 GB) graphics card. Models with Intel Core i7 processors are available with either an NVIDIA GeForce RTX 3070M (120 W, 16 GB) or an NVIDIA GeForce RTX 3060M (80W, 12 GB) card.

The device’s I/O ports are located on the front and the back of the device. Looking at the device front-on, these are:

  • 1 x USB Type-A 3.2 Gen 2×1 (10 Gbit/s) port
  • 1 x USB Type-C port annotated with a Thunderbolt logo
  • 1 x full-sized SD card port
  • 1 x 3.5 mm audio jack
  • 1 x USB Type-A 3.2 Gen 2×1 (10 Gbit/s) port.

Above the front I/O ports is the on/off button.

One unusual and important feature is the dial on the front, which is surrounded by … RGB lighting. Ah, gaming!

The dial can be turned to one of three positions, each of which is annotated with a small symbol. Turning the dial clockwise to the “rocket” symbol results in the device running in “Performance” mode. If the dial is placed in the central position marked with a symbol containing lines, the device runs in “Auto” or balanced mode. The final anti-clockwise position, with a headphones symbol, changes the device to run in “Silent” mode. I will cover a full comparison of these modes together with RGB settings later in the review.

The remaining I/O ports are on the rear. These include, from left to right, top to bottom:

  • 1 x 3.5 mm audio jack (upper)
  • 1 x power jack (lower)
  • 2 x USB Type-A 3.2 Gen 1×1 (5 Gbit/s) ports (stacked)
  • 1 x USB Type-A 3.2 Gen 1×1 (5 Gbit/s) port (upper)
  • 1 x USB Type-A 3.2 Gen 2×1 (10 Gbit/s) port (lower)
  • 2 x HDMI 2.0 port (stacked)
  • 1 x 2.5 Gb Ethernet port
  • 1 x 2.5 Gb Ethernet port
  • 1 x Kensington Lock slot (upper)
  • 1 x 1.4 DisplayPort (lower)

Another interesting feature of this mini PC is access to memory and storage, which are concealed behind two lockable panels on either side of the device. Sliding the lock tab upwards at the rear of the side panel, followed by sliding the panel forward, allows it to be taken off the device.

However care should be taken when removing the left side panel as it includes a small fan which is connected to the inside of the device. Removing the left panel gives access to the two SO-DIMM slots that support DDR5 memory, and are kept cool by the fan.

In the review model they were populated with two Kingston 16 GB DDR5 5600 MHz sticks of memory configured to run at 4800 MHz, which is the highest speed supported by the Core i9-12900H processor.

Behind the right panel are two M.2 2280 PCIe Gen 4 x4 NVMe SSD slots with the left one being occupied by a BIWIN CE480V6D10H 1 TB PCIe Gen 3 drive in the review model. There is no thermal pad on the drive, and there is probably little purpose in putting one on, as the panel cover is plastic and likely wouldn’t conduct the heat away quickly enough from a Gen 4 drive.

I’m sure that if the panel cover was made from thermally conductive plastic it would have been mentioned in the specifications. It is a shame that there isn’t a fan in the bay similar to the opposite side, but maybe there is nowhere to vent the hot air to?

Behind the right NVMe slot is a M.2 2230 Wi-Fi card from Cdtech (China Dragon Technology). This Wi-Fi 6 CDW.C9852BE-00 card uses the RTL8852BE-CG chipset and also provides Bluetooth 5.2 (the LMP Firmware Version shows as LMP 11.2292). It appears in HWiNFO as a RealTek Semiconductor RTL8852BE WiFi 6 802.11ax PCIe Adapter.

Finally the box containing the mini PC also includes a large and heavy power supply from Huntkey (HKA300190A6-0A7, 19.0 V, 15.79 A, 300.0 W) and cable with a country specific plug.

How it performs

The review model came with Windows 11 Pro Version 22H2 OS build 22621.1992 which was activated once connected to the internet. After updates I was offered the upgrade to Version 23H2 which I installed so that all Windows testing was performed on OS build 22631.2715.

I then shrank the Windows partition to create a new 100 GB partition. To test Linux, I installed Wubuntu (Windows Ubuntu) 11.4.3 LTS as having read lots of reddit posts about it having multiple copyright and trademark violations, it seemed appropriate, for reasons that will become clearer later in the review.

Then to validate the benchmarks run on Wubuntu, I further shrank the 100 GB partition to create a new 50 GB partition where I installed Ubuntu 22.04.3 LTS. After updates, I also installed the Hardware Enablement (HWE) kernel on both Linux systems as I knew from experience that a 6.2 kernel supported the RTL8852BE chipset.

Prior to benchmarking, I ran some basic hardware and software checks. For some reason the DisplayPort didn’t work, but it might just be DOA and an issue only with the review unit. All of the other usual mini PC problem areas, such as Ethernet, Wi-Fi, Bluetooth, and audio output from the 3.5mm jacks, worked without issue on both Windows and Linux.

When I tried opening the pre-loaded NVIDIA GeForce Experience application I got a “Something went wrong” message and ERROR CODE: 0x0003 was displayed. After rebooting it still didn’t work, so I tried installing the latest NVIDIA driver. However this also failed with a message that the driver was incompatible. I then made the mistake of uninstalling the current NVIDIA driver and tried again. It still didn’t install, and in fact all of the official NVIDIA drivers said they are incompatible leaving me driverless.

Fortunately the “Tank03 Graphics Card Driver” is available on the ACEMAGIC website should anyone else need them.

The TANK03 uses a customised NVIDIA driver which, according to ACEMAGIC, is “essential for our structure’s compatibility that we utilize our internally adjusted and tuned drivers for our NVIDIA GPUs. They’re tailored to fit seamlessly within our design framework, ensuring optimal performance and compatibility”. Yeh, right.

The other problem is they don’t work on Linux and obviously NVIDIA’s Linux drivers also fail to install. If this is a major issue for you then just skip straight to “Key issues” below.

For the testing, on Windows the OS power mode was set to ‘High performance’ and similarly the CPU Scaling Governor was set to ‘performance’ on both Wubuntu and Ubuntu.

I started performance testing by using Crystal Dew World’s CrystalDiskMark to measure performance on Windows and a script using the “fio” or “flexible I/O tester command” to verify it on Wubuntu.

Given this is a Gen 3 drive in a Gen 4 slot, the drive’s performance was quite good with read speeds around 3450 MB/s and write speeds around 2900 MB/s.  Similar speeds were also reproducible on Wubuntu.  For comparison, I tested a Gen 4 drive and got read speeds of around 6750 MB/s and write speeds of around 4900 MB/s.

In the micro SD card reader, I tested both UHS-I and UHS-II cards. However, as the MOAI Electronics Corporation (SY-T18) MicroSD card reader is implemented internally using a USB 2.0 interface, the speeds were shockingly slow. Read speeds were around 18 MB/s and write speeds were even slower at 15 MB/s. For reference, I tested a USB 3.0 (USB 3.2 Gen 1×1) card reader and got read speeds of 275 MB/s and write speeds of 230 MB/s in comparison.

Testing the USB Type-A ports revealed that not all of them were USB 3.2 gen 2 like the specification states. The rear left pair as well as the rear right upper port are all USB 3.2 gen 1,  i.e. only 5 GB/s rather than the 10 GB/s as advertised. According to ACEMAGIC they are checking this and will update the specifications in due course.

Another port of interest is the front USB Type-C port which is annotated with a Thunderbolt logo. When I tested it I found that it was USB4 Gen 2 x 2 with dual 10 GB/s lanes allowing a maximum transfer speed of only 20 GB/s rather than the 40 GB/s offered by a true Thunderbolt 4 port.

And, whilst it appears to be a USB4 port,  I could not get my USB4 device to connect, either through plug & play or having connected it prior to booting.

I reported this finding to ACEMAGIC who replied saying “we are verifying the Thunderbolt certification and its operational capabilities to ensure compatibility with USB4 devices. We’ll provide clarification on the Thunderbolt certification and the port’s configuration as soon as we have confirmed the details”. Therefore it is worth keeping an eye on their specification webpage to see if it gets updated at some point.

Windows Storage
M.2 2280 NVMeUSB 3.2 (Front Left)USB4SD UHS-I/IIUSB 3.2 (Front Right)USB 3.2 (Rear Left Upper)USB 3.2 (Rear Left Lower)USB 3.2 (Rear Right Lower)USB 3.2 (Rear Right Lower)
Seq1M Q8T1 Read3435.901071.311061.4118.231070.23463.33463.431072.39463.32
Seq1M Q8T1 Write2924.051014.691038.3914.981016.98458.99456.181015.63461.49
Seq1M Q1T1 Read2365.72
Seq1M Q1T1 Write2865.21
RND4K Q32T1 Read846.91
RND4K Q32T1 Write588.80
RND4K Q1T1 Read83.58
RND4K Q1T1 Write260.96

The following benchmarks were all run with the front dial in “Performance” mode.

On Windows I ran:

  • PassMark Software’s PerformanceTest (general performance)
  • UL’s 3DMark (CPU and graphics) and Procyon (office productivity)
  • Maxon’s Cinebench (CPU)
  • Primate Labs’s Geekbench (CPU and graphics)
  • Unigine’s Heaven (graphics)
Performance Mode
(PL1=45 & PL2=65)
PerformanceTest 11.0
PassMark Rating9321.1
CPU Mark26902.9
2D Graphics Mark604.2
3D Graphics Mark18589.3
Memory Mark3228.8
Disk Mark22192.9
Time Spy Score10754
Graphics score10631
CPU score11510
Fire Strike Score24394
Graphics score27948
Physics score25753
Combined score12002
Office Productivity score7030
Word score6086
Excel score7485
PowerPoint score7645
Outlook score7002
CPU (Multi Core)13044
CPU (Single Core)1748
GPU (System Requirements)11101
CPU (Multi Core)748
CPU (Single Core)103
Geekbench 6.2.1
Single-Core Score2296
Multi-Core Score11518
OpenCL Score120047
Unigine Heaven 4.0

On Wubuntu and Ubuntu I only ran the following benchmarks:

  • PassMark Software’s PerformanceTest (CPU and memory)
  • Primate Labs’s Geekbench (CPU)
  • Unigine’s Heaven (graphics)
Performance Mode
(PL1=45 & PL2=65)
Performance Mode
(PL1=45 & PL2=65)
PerformanceTest 11.0
CPU Mark2699626299
Memory Mark32713199
Geekbench 6.2.1
Single-Core Score22952158
Multi-Core Score1140910850
Unigine Heaven 4.0

The Wubuntu results for the relevant benchmarks are similar to those on Windows with the exception of the Heaven benchmark. For both Wubuntu and Ubuntu, the Intel iGPU is used by default for this benchmark simply because the RTX 3080M does not have a working NVIDIA driver.

As an aside, comparing the Wubuntu benchmarks against the Ubuntu benchmarks shows they are quite similar. It is interesting that the CPU results are better in Wubuntu and the iGPU results are better in Ubuntu. But seeing how close they are, further testing would be required to rule out any margin of error before drawing this conclusion. Given the Ubuntu results, it does appear that the Wubuntu results can be trusted in the testing of the TANK03.

Networking is provided by either Ethernet or Wi-Fi. The 2.5 Gb Ethernet ports when tested on Wubuntu averaged 2.35 Gbits/sec for both upload and download.

The Wi-Fi 6 CDW.C9852BE-00 card was also tested on Wubuntu by connecting to a Wi-Fi 6 router. The upload speed for a 2.4 GHz connection was 191 Mbits/sec and the download speed was 122 Mbits/sec. For the 5 GHz band, the upload speed was 838 Mbits/sec with download at 827 Mbits/sec.

Turning next to graphics performance. First I tested how videos played in browsers and then I tested how games played in Steam. When I first played YouTube videos in Edge I noticed occasional frame dropping regardless of resolution. Interesting “Task Manager” showed that the iGPU was being used rather than the GPU.

So I manually set the graphics preference for Edge to be “High Performance” to force the GPU to be used. But even with Edge now using the GPU, the behaviour was the same and occasional frames kept being dropped.

However, if I left an application open on the screen, the video continued to play but without dropping frames.

I’ve no idea why, so I recommend using Chrome because when tested it played videos up to 4K 60FPS without issues or peculiarities like with Edge.

Finally looking at gaming performance. I first checked the connection throughput to the GPU which showed transfers averaging 12.122 GB/s. This aligns with it being connected to a PCIe 4.0 interface with a maximum link width of x8, i.e. 15.754 GB/s.

I then tried a few games which had built-in benchmarks at 1080p with default settings: GTA V (Reset, VSync off), Shadow of the Tomb Raider (High preset, no DLSS), Horizon Zero Dawn (Favor Quality), COD MWII (Recommended). The average frame rate was good in each of them, and in the same order as the game titles they were: 173, 151, 133, 92. Where reported, each showed that the game was GPU bound i.e. that the GPU was the bottleneck.

I also enlisted my son to help test some additional games and he reported that they all ran fine at highest settings. During game play, MSI’s Afterburner was run to capture the key stats as follows:

Armored Core VI Observations:

  • Short form mission (boss fight) runs well; during boss fight phase 2.5, 30 FPS was achieved on boss transition.
  • Long form mission (frame rate sat at 90 FPS for the most part).

Elden Ring Observations:

  • Capped at 60 FPS by default, runs at 60 FPS consistently.

Baldur’s Gate 3 Observations:

  •  Nothing to note, runs fine.

Hogwarts Legacy Observations:

  •  Nothing to note, runs fine.

Looking at the frame rates for Armored Core VI, Baldur’s Gate 3 and Hogwarts Legacy, the average was good in each of them. Again by order of the game titles they were: 98, 117 and 96 noting that Elden Ring was capped at 60 FPS by default.

The testing of the TANK03 mini PC shows that it is perfectly good for 1080p gaming, and also likely to be good at 1440p although some settings may need to be lowered slightly at that resolution.

Key issues

Graphics driver issues

The foremost and biggest issue with the TANK03 is that it requires customised NVIDIA graphics drivers. A lot of people do not like installing modified versions of drivers, especially where the original software is from reputable companies. We are constantly being warned about the security risks of data theft and spyware, hijacking and ransomware, data loss and malware etc.

Even after choosing to install the driver, there is then the question of ongoing support and the timely provision of modified new versions. For example, the TANK03 came with a modified NVIDIA driver version 537.42. At the time of the review the current NVIDIA version was already version 546.17. It also should be noted that some games check the driver version and refuse to play if the driver is too old or, at best, warn that the driver is out of date.

When contacted about this issue, ACEMAGIC responded:

ACEMAGIC aims to ensure optimal performance and compatibility by tailoring the NVIDIA driver. We work diligently to regularly release new versions to address security concerns, enhance performance, and provide updates. These updates occur periodically based on the evolving technology landscape and the specific needs arising from security or performance improvements.

To inform customers about the availability of new NVIDIA driver versions, ACEMAGIC employs various channels. We actively update our social media platforms and blog posts, notifying users of any new driver releases, recommended installations, and their associated benefits. Additionally, customers can regularly check our Drivers page, where we periodically update the latest driver versions and release notes.

We understand the importance of keeping our customers informed and up-to-date on driver updates and improvements.

This may force some people who need the latest driver to resort to using third-parties drivers (e.g. FrankenDriver) where, besides the same risks, they may also be charged for them.

Alternatively you can always “hack” the NVIDIA drivers yourself using tools and tutorials available on the web.

I re-ran all the benchmarks and some games using my hacked version of the latest driver. In this instance, the new driver did not provide any performance benefits, but it did lessen the number of warnings.

The primary reason why the TANK03 requires a customised driver is because the device ID for the GPU is not recognised (i.e. not defined) in the official drivers. Why this is the case is open to speculation, however it does seem to be common for some graphics cards from China.

With all the stories of graphics cards being repurposed from laptops including ex-mining laptops, it certainly is not a good look to have a GPU without a recognised device ID.

One further point to note about the NVIDIA GeForce RTX 3080M GPU is that the specifications list it as having a 125 W power consumption. However the NVIDIA driver limits the maximum graphics power to 115 W.

Limited UEFI (BIOS) options

The UEFI (BIOS) is quite restricted. It does allow the manual reconfiguration of the power limits, but most of the other settings should be left alone unless there is an absolute need to change something and you know what you are doing.

What is missing are settings related to fan controls for either of the two main fans. I tried installing several fan control applications but none of the fans could be identified.

Linux compatibility

While most of the TANK03’s hardware works with Ubuntu and Wubuntu, the fact that there’s no NVIDIA graphics driver compatible with the GPU in this system means that anyone buying this computer in hopes of using it as a Linux gaming PC is vastly overpaying for a system with discrete graphics, because you’ll only be able to use the computer’s integrated GPU.

While we are looking at Linux, here are a few observations about Wubuntu, that unofficial Ubuntu-based Windows 11 clone. Most of the operating system’s features appear to work, but it does include “nag” software reminding you to purchase a licence for PowerToys or to register your system by purchasing a licence on every boot.

Opening certain applications pinned to the taskbar takes forever the first time after a boot.

Overall, on the surface Wubuntu looks similar to Windows, but once you start using an OS specific option, for example Network and Internet, configuring Ethernet takes you straight into a separate KDE System Settings/Connections window which of course has a totally different look and feel to Windows.

The dial of power

As mentioned, the dial on the front of the TANK03 can be set in one of three modes: Silent, Auto and Performance.

When set to Silent, the CPU power appears to be limited to 35 W and the Performance Cores were seen running at 2493.9 MHz with the Efficient Cores running at 2094.9 MHz.

When running the Shadow of the Tomb Raider (SOTTR) benchmark using the High preset and DLSS set at Ultra Performance, the average FPS was 139.

Setting the dial to Auto increased the CPU power maximum to 45 W. Now the Performance Cores were seen running at 2892.9 MHz and the Efficient Cores at 2294.4 MHz.

Similarly, the SOTTR benchmark with the same settings as before, resulted in its average improving to 169 FPS.

To get the most out of the TANK03, just crank the dial to Performance mode.

The CPU power now appears to run using the Power Limit settings (i.e. a maximum of 65 W). During this particular test run, and likely due to Intel’s Running Average Power Limit (RAPL), it didn’t increase the Performance Cores which still ran at 2892.9 MHz, but it did increase the Efficient Cores to 2593.7 MHz.

The effect on the SOTTR benchmark was not as significant as seen when going from Silent to Auto, however the average FPS still improved and was now 177 FPS.

The dial also works in Linux. On Wubuntu I ran a stress test on all threads for thirty minutes at each of the dial settings, with an idle period before and after each test. In parallel I logged the CPU Utilization, Frequency and Temperature every second.

Plotting the results shows the effect of each mode on Frequency and Temperature. Silent mode’s frequency averaged 2250 MHz while running at an average temperature of 71°C.

There was a clear improvement with Auto mode whose frequency averaged 2761 MHz however the average temperature climbed to 80°C.

The best was Performance mode where the frequency averaged 3032 MHz but it was pumping out heat making the average temperature hit 90°C.

For each of the modes, the lettering of “TANK” on the dial changes colour. Silent is green, Auto is blue and Performance is red.

Besides looking at the dial to check what the current mode is, there is a preloaded application from CYX (Shenzhen CYX Industrial Co., Ltd.) which is the manufacturer of the mini PC, that shows the mode briefly when you start the program.

However the software is primarily for controlling the RGB surrounding the dial. It is very basic with a few lighting patterns to choose from and the ability to turn off RGB altogether. It does not turn off the dial’s illuminated word “TANK”.

One drawback of the software is that the current RGB setting is not remembered after a cold boot, however it does survive a warm reboot. Also there is no version of the software currently available for Linux so you’ll either have to like RGB or dual boot Windows first to set your choice.

After booting, and whilst idling, the computer is very quiet and no noise can be measured by my sound meter next to the device. But there wouldn’t be much reason to include a Silent option on the dial if the computer wasn’t capable of getting loud. So just how loud does it get?

To answer that question I took the following measurements to show how noisy it is in various games. I also included the Cinebench R23 benchmark to provide a reference point and took the measurements for each of the dial modes.

Windows Noise (dBA)
Cinebench R23 Multi Core32.242.348.1
Armored Core VI menu42.545.651.3
Armored Core VI playing41.546.151.6
Baldur’s Gate 3 menu42.545.652.3
Baldur’s Gate 3 playing42.547.152.2
Hogwarts Legacy menu42.544.552.2
Hogwarts Legacy playing43.347.152.2

The modes can be changed dynamically so you can see, or rather hear, the effect of turning from Performance to Silent immediately.

A scenario where this may be useful is if you typically use the device in Silent mode to minimise the noise, and then want to crunch some numbers in Excel, you can simply dial up the performance for the duration you need it.

Power Usage

Power consumption was measured on Windows as follows:

  • Powered off (shutdown) – 1.8 W
  • UEFI (BIOS) – 50.1 W
  • Idle – 32.2 W
  • CPU Cinebench R23 – 80.1 W (Silent Mode) 100.2 W (Auto Mode) 110.6 W (Performance Mode)


There is a lot to like about the TANK03. The form-factor is super small for a computer with such a powerful CPU and GPU. The performance is good and the temperatures, although high at times, are successfully controlled by the cooling solution.

But I’m still not convinced that putting thermal heat pads on a Gen 4 NVMe drive is going to be that effective. The fans can get noisy when playing games so headphones are most likely needed. Cleaning the fans may also pose a problem as I could not work out how to easily open the device to access them.

The front dial is super impressive. It is not a gimmick because testing showed it works and the ability to easily boost power in gaming, for example, can be very useful.

However, the TANK03 is an expensive computer by mini PC standards, and like many Chinese PC makers that sell their computers to a global audience, ACEMAGIC offers limited customer service and support. And the issue of the GPU’s device ID preventing the installation of NVIDIA drivers is arguably unacceptable for a consumer product. Not only does it mean that Windows users may be running outdated drivers, but it means Linux users cannot use the discrete GPU at all.

Coupled with some of advertised port specifications being somewhat overgenerous and bordering on misleading, anyone who buys the TANK03 without being aware of these limitations is going to be very disappointed. Hopefully ACEMAGIC will quickly address these issues.

I’d like to thank ACEMAGIC for providing the review unit. It’s available for purchase from ACEMAGIC’s website. Prices start at $1399 for a model with an Intel Core i7-12700H processor, NVIDIA GeForce RTX 3060M GPU, 16GB of RAM and 1 TB of storage.

The unit featured in this review sells for $1799 and has an Intel Core i9-12900H processor, RTX3080 graphics, 32GB of memory, and a 1TB SSD.

Support Liliputing

Liliputing's primary sources of revenue are advertising and affiliate links (if you click the "Shop" button at the top of the page and buy something on Amazon, for example, we'll get a small commission).

But there are several ways you can support the site directly even if you're using an ad blocker* and hate online shopping.

Contribute to our Patreon campaign


Contribute via PayPal

* If you are using an ad blocker like uBlock Origin and seeing a pop-up message at the bottom of the screen, we have a guide that may help you disable it.

Subscribe to Liliputing via Email

Enter your email address to subscribe to this blog and receive notifications of new posts by email.

Join 9,543 other subscribers

Join the Conversation


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

This site uses Akismet to reduce spam. Learn how your comment data is processed.

  1. This looked quite enticing, until I checked that with the 3080 it’s actually an almost 2k computer. Enticing because of the mini factor of course. Not sure what the state of the art is for MiniPCs that can run at 1440p maxed out, if there is such a thing at all, but will keep an eye out for better options.

  2. Too bad there’s no support for the GPU under Linux, I was looking into this mini PC as an option for an AI application, but that’s a show stopper.

    1. There is no official support for the NVIDIA drivers on Linux but I found a workaround on a Russian mining website that appears to work although I need to test further.

  3. Well, this is informative. Between the displayport and usb4 point not working, and the graphics card driver requiring hacks on top of hacks that only work in Windows, this is a computer best avoided.
    Things are a mess and they’ll only get worse, but at least for now we don’t have to accept not getting hardware we paid for.

  4. OOOF $1300 and you have to use a hacked driver and only windows. Nah bro, I’ll pass.

    Glad I bought the HX99G when it was on sale ($700 barebones) for Black Friday since I was eyeballing this. Ryzen 9 6900hx, RTX 6650M and the GPU works with linux.