Why you may belief TechRadar
We spend hours testing each services or products we evaluate, so that you could be positive you are shopping for the most effective. Find out more about how we test.
I’m not going to get to have the Acer Veriton GN100 for lengthy, so that is extra of a hands-on dialogue than an precise evaluate.
My first response, out of the field, to the Acer Veriton GN100 is that all of it appears remarkably acquainted. A chic mini-PC model case with a car-grill aesthetic, a number of USB-C ports alongside a 10GbE LAN port and the mercurial NVIDIA ConnectX-7 SmartNIC.
Whereas it’s bodily a bit of smaller, the ports on this machine are equivalent to these on the ASUS Ascent GX10, as each manufacturers adopted Nvidia’s Blackwell system plan precisely.
The one important distinction is that the place the Asus supplied entry to the only M.2 NVMe drive that was put in, the Acer Veriton GN100 is a completely sealed unit, the place no matter storage it has can’t get replaced or upgraded.
This can be a remarkably highly effective platform that has makes use of in information science, medical picture evaluation, robotics and AI mannequin improvement. To be clear, this isn’t a Home windows PC, and an understanding of Linux is required to make use of it.
Because the specification suggests, this isn’t a cheap merchandise, beginning at $2999 for the US-supplied {hardware}, however £3999.99 in Europe. For many who need a extremely compact and environment friendly improvement setting, particularly for AI, the Acer Veriton GN100 is an possibility, however it isn’t the one machine accessible utilizing the identical platform.
Acer Veriton GN100: Value and availability
- How a lot does it price? From $3000, £4000
- When is it out? Accessible now
- The place are you able to get it? Accessible from Acer and on-line retailers
To keep away from any confusion about specs, Acer determined there could be just one SKU of the Veriton GN100, with 128GB of LPDDR5 and 4TB of storage.
Inexplicably, it prices £3999.99 direct from Acer within the UK, however solely $2999.99 from Acer within the US. Why will we pay 82% extra for an equivalent half when the UK would not tariff Taiwanese items, Acer?
Additionally, this product doesn’t appear to be accessible elsewhere, so discovering it cheaper on Amazon, for instance, isn’t at the moment potential.
The options constructed across the identical platform are the ‘founders version’ Nvidia DGX Spark Private AI Supercomputer, ASUS Ascent GX10, Gigabyte AI TOP ATOM Desktop Supercomputer, and MSI EdgeXpert Desktop AI Supercomputer.
The Nvidia DGX Spark Private AI Supercomputer, because the originator modestly calls it, undercuts the Veriton GN100 within the UK and prices £ 3699.98 for a system with 128GB of RAM and 4TB of storage. However, it’s dearer for US prospects, costing $3999 on Amazon.com.
The ASUS Ascent GX10 value on Amazon.com is $3088.94 for the 1TB storage SKU (GX10-GG0015BN), and $4,149.99 for the 4TB storage mannequin (GX10-GG0016BN).
Even with the present value of M.2 modules, that may be a exceptional value hike for the additional storage capability.
For UK prospects, the 1TB ASUS Ascent GX10 mannequin value is £3713.02, however I discovered it through on-line retailer SCAN for a tempting £2799.98. SCAN additionally carries a 2TB possibility for $3199.99 and the 4TB mannequin for £3638.99.
The Gigabyte AI TOP ATOM Desktop Supercomputer 4TB mannequin sells for £3479.99 from SCAN within the UK, and could be discovered on Amazon.com for $3999.
And the ultimate mannequin with the identical spec as most is the MSI EdgeXpert Desktop AI Supercomputer, promoting for £3,598.99 from SCAN within the UK, and $3998.01 on Amazon.com for US prospects.
The conclusion is that the US value is troublesome to beat, whereas the European pricing is wildly out of what rivals are charging for this know-how.
Acer Veriton GN100: Specs
|
Merchandise |
Spec |
|---|---|
|
ARM v9.2-A CPU (GB10) (20 ARM cores, 10 Cortex-X925, 10 Corex-A725) |
|
|
GPU: |
NVIDIA Blackwell GPU (GB10, built-in) |
|
RAM: |
128 GB LPDDR5x, unified system reminiscence |
|
Storage: |
4TB M.2 NVMe PCIe 4.0 SSD storage |
|
Growth: |
N/A |
|
Ports: |
3x USB 3.2 Gen 2×2 Kind-C, 20Gbps, alternate mode (DisplayPort 2.1) 1x USB 3.2 Gen 2×2 Kind-C,with PD in(180W EPR PD3.1 SPEC) 1x HDMI 2.1 1x NVIDIA ConnectX-7 SmartNIC |
|
Networking: |
10GbE LAN, AW-EM637 Wi-Fi 7 (Gig+) , Bluetooth 5.4 |
|
OS: |
Nvidia DGX OS (Ubuntu Linux) |
|
PSU: |
48V 5A 240W |
|
Dimensions: |
150 x 150 x 50.5 mm |
|
Weight: |
1.2kg |
Acer Veriton GN100: Design
- Outsized NUC
- Join-7 scalability
- Zero inside entry
Whereas the GN100 seems to be like an outsized NUC mini PC, at 1.2kg, it is heavy, though it’s lighter than the ASUS Ascent GX10 by over 200g.
With a view to drive the monster silicon inside, Acer included a Delta-made PSU that’s rated to 240W over USB-C.
All of the ports are on the again of this method, and nothing is on the entrance, not even the ability button, apart from some visible styling and the Acer emblem.
These embody, equivalent to the ASUS Ascent GX10, 4 USB-C ports, considered one of which is required for the PSU to attach, a single 10GbE LAN port and a single HDMI 2.1 video out.
This association permits a single monitor utilizing HDMI, however extra ones utilizing the USB 3.2 Gen 2×2 ports in DP Alt mode, though one is completely wanted to energy the unit.
Why Nvidia selected USB 3.2 and never USB4 appears curious, for the reason that fashions and information processed on this unit will ultimately must make it elsewhere, and the most effective networking on provide is 10GbE, which equals roughly 900MB/s switch speeds.
And for these engaged on the {hardware}, the shortage of any USB-A ports for mice or keyboards seems to be a bit foolish.
Nevertheless, this {hardware} is meant for use ‘headless’ utilizing a distant console, so maybe that isn’t a difficulty within the better scheme of issues.
The place this design sheds any resemblance to PC {hardware} is with the inclusion of a ConnectX-7 Sensible NIC, a know-how acquired by Nvidia when it purchased Mellanox Applied sciences Ltd, an Israeli-American multinational provider of pc networking merchandise primarily based on InfiniBand and Ethernet.
On this context, ConnectX-7 is like these annoying cables that Nvidia used to make video playing cards work collectively, once they cared about video playing cards. Besides that the capability quantity of bandwidth that may journey over ConnectX-7 is considerably extra.
The port has two receptacles, with every able to 100GbE, permitting 200GbE to circulation between the GN100 and one other, doubling the variety of AI parameters from 200 billion in a single machine to 400 billion when buddied as much as one other.
Acer Veriton GN100: Options
- ARM 20-core CPU
- Grace Blackwell GB10
- AI platforms in contrast
The Nvidia GB10 Grace Blackwell Superchip marks a notable development in AI {hardware}, created by way of a partnership between Nvidia and ARM. It arises from the rising want for specialised computing platforms to maintain tempo with the speedy improvement and deployment of synthetic intelligence fashions. In contrast to a typical PC, the GB10 is designed across the ARM v9.2-A structure, incorporating 20 ARM cores (10 Cortex-X925 and 10 Cortex-A725). This displays a wider business transfer in direction of ARM-based choices, that are extra power-efficient than PC processors, and probably extra scalable for AI duties.
The capabilities of the GB10 are spectacular. It combines a sturdy Nvidia Blackwell GPU with the ARM CPU, reaching as much as a petaFLOP of AI efficiency with FP4 precision. This stage of energy is very appropriate for coaching and inference of huge language fashions and diffusion fashions, that are elementary to a lot of immediately’s generative AI. The system is additional supported by 128GB of unified LPDDR5x reminiscence, enabling it to deal with demanding AI duties effectively.
The caveat to all this energy and reminiscence is that PC architectures aren’t designed to take advantage of them successfully, and Microsoft Home windows reminiscence administration has lengthy been a difficulty.
Due to this fact, to be environment friendly and talk successfully with different nodes, the GB10 wants Ubuntu Linux, modified with NVIDIA’s DGX OS, to harness the platform’s energy and deal with multi-node communications.
As I already talked about, the GB10 delivers as much as 1 petaFLOP at FP4 precision, very best for quantised AI workloads. However that’s nonetheless lower than the multi-petaFLOP efficiency of NVIDIA’s flagship information centre chips, the Blackwell B200 or GB200.
Nevertheless, the place it goes toe-to-toe is in respect of energy effectivity, since this node solely consumes round 140W, which is way lower than the Blackwell B200, which may eat between 1000W and 1200W per GPU. The GB200 combines two B200 chips and a Grace CPU, and the ability demand can bloom to 2,700W. Though these methods would possibly provide as much as 20 Petabytes of efficiency, at round 19 occasions the ability.
The steadiness right here is that the GN100 can sit in your desk without having any particular companies or setting, whereas the datacentre {hardware} wants a specialist location and companies to make sure it doesn’t overheat or trigger the native electrical energy community to fail.
On this respect, the GN100 and its counterparts characterize the extra life like finish of the AI wedge, however how helpful they are often relies on what you are trying to do, and if this a lot energy is sufficient on your functions.
Acer Veriton GN100: AI Actuality Test
In my prior Asus GX10 protection, I talked at size about AI, and the way there are many folks having a bet that it’s the subsequent huge factor, and others who’re way more essential of the know-how and the way it’s creating.
I’m not going to rehash the plain flaws of AI, or the shortage of a path to handle all of these, however I’d strongly advocate researching earlier than beginning any AI endeavour and creating expectations that both can’t be met with present know-how, or the ability on this bodily small pc.
What I can say is that latest AI releases have considerably improved over earlier generations, however entry to those superior fashions, like ChatGPT 5.3 Codex and Claude 4.0, is ringfenced for paid subscribers utilizing the Cloud.
Clearly, the great thing about a tool just like the GN100 is you could obtain these fashions and run them by yourself {hardware}, even when getting essentially the most out of them requires them to be linked to the Web to supply info.
For these , working GPT-5.3-Codex on this {hardware} requires you to put in Tailscale and an area inference engine like Ollama, pull the codex to the GN100 utilizing the suitable instructions. Then you may open up an Open WebUI from one other system, ideally, and use the mannequin.
For anybody accustomed to Linux, none of that is particularly taxing. However to make it even simpler, the ChatGPT crew (or is that the AI?) has made a Codex App that does a lot of the legwork for you.
If you wish to attempt one thing else, some fashions are available in pre-prepared Docker Containers that may merely be put in and executed, making deployment remarkably simple.
While you first energy the system up, you might be introduced with an online interface created for the Nvidia DGX Spark, and set up VS Code, the DGX Dashboard with JupyterLab, Open WebUI with Ollama and Comfortable UI.
To be clear, this isn’t like a Home windows software set up. You might be given directions as to the instructions it’s good to execute, which set up the instruments and libraries which can be wanted. Those that don’t use Linux day by day will discover it a problem, however ultimately, even I managed to get virtually every thing working, no less than sufficient to load fashions and create some output. Picture era was particularly spectacular, though a number of the deep pondering fashions aren’t that responsive for those who ask them one thing genuinely difficult.
Some folks would possibly assume that as a result of the code for these fashions is being run domestically on the GN100, the price of a subscriber mannequin needs to be cheaper. However I’ve not seen that up to now, you simply get higher efficiency and save the creators’ electrical energy invoice. What working a mannequin on this {hardware} offers you is that you simply personal the mannequin, it could possibly’t be faraway from you, and there’s the potential for you management the mannequin, customising it in a particular and private means.
For these exploring AI in a severe means its crucial to make use of the newest fashions, and that usually has a price implication, even for those who personal the {hardware} platform.
What you definitely don’t wish to do is set up some free mannequin from a few years in the past, after which be dissatisfied with the outcomes. The steepness of the curve of improvement on fashions is extraordinary, and even variations of the most effective ones from six months in the past have been overtaken by the newest releases.
For these working on this space, utilizing fashionable AI is like making an attempt to get onto a bus once they don’t cease and even decelerate to permit passengers on or off. Being conscious of the place the mannequin of curiosity is, and when it has been overtaken, is essential to not being fully old-fashioned earlier than the mission is accomplished.
Acer Veriton GN100: Early verdict
NVIDIA determined to ship its Grace Blackwell know-how in an entry-level kind and created a blueprint for that of their NVIDIA DGX Spark Private AI Supercomputer, its companions are delivering their variations, just like the Acer Veriton GN100.
Apart from the outer case and some different small decisions, it is debatable how a lot variation we’re prone to see between these machines. It’s not just like the GPUs, the place the companions get to design variations and even tweak the founders’ editions for higher efficiency.
Perhaps if these change into massively in style, then we’d see extra variation, like combining two methods in a single field, or mixing the know-how with DAS storage. However for now, that is the place we’re.
That mentioned, the Acer model is completely serviceable, however when the specs are so shut its largely about value. For Individuals paying $2999 for the GN100, it is in all probability one of many most cost-effective choices, and for these in Europe, oddly, it is one of many dearer. Maybe Acer can repair that for Europeans, however given the rising price of RAM, it is extra seemingly the USA must pay extra.
The one weak spot of this design is the shortage of entry to the SSD, and if that’s a deal breaker for you, another machines do have that functionality.
As ever, Acer has delivered a workable answer for demanding computing duties, however what this model can’t assure is the talents wanted to take advantage of from their platform. Patrons want to understand that whereas the {hardware} provides greater than ten occasions the AI processing of a high-end PC, benefiting from what it could possibly do requires a specific ability set.
For extra compact computing, see our information to the best mini PCs you should buy
Source link
