HP Z2 G1i, Core Ultra 9 285K, 32GB, 1TB SSD, RTX A1000, Black

The HP Z2 G1i SFF packs a 98th percentile Core Ultra 9 285K CPU into a compact 4kg chassis, making it a standout for CPU-heavy professional work. The RTX A1000 GPU is only average and the 8GB VRAM limits AI and heavy rendering, which is the machine's biggest weakness. At $3,914 to $3,999 USD, you're paying a premium for the small form factor and port selection. Buy it for CAD and engineering work, skip it for AI or GPU-heavy tasks.

The Core Ultra 9 285K is the headline act and it delivers. With 24 cores and a 3.7GHz base clock, this chip chews through multi-threaded workloads like rendering, simulation, and code compilation. Our database puts it at the 98th percentile for CPU performance in this category, which means it's one of the best CPUs you can get in a workstation right now. If your workflow involves waiting on CPU-bound tasks, this machine will make those waits noticeably shorter. The 32GB of DDR5 at 5600 MHz is also well above average, landing at the 82nd percentile. That's enough headroom for most professional workloads, though heavy simulation or massive datasets might push you toward 64GB eventually.

The RTX A1000 is where things get more nuanced. It's a solid professional GPU, sitting at the 67th percentile in our database, which is about average for this category. For CAD work, 3D modeling, and light rendering, it's perfectly capable and the certified drivers matter more than raw speed. But with only 8GB of VRAM, it's not going to handle large AI model training or heavy GPU rendering workloads well. Our AI and LLM score for this machine is 51.2 out of 100, which is the weakest area by a wide margin. If you're planning to run local language models or do serious GPU compute, this card will hold you back. The 1TB NVMe drive is also middle of the pack at the 69th percentile, which is fine for a boot drive and active projects, but you'll likely want external or secondary storage for larger asset libraries.

The Apple Mac Studio M3 Ultra is the most obvious alternative if you're not tied to Windows. Apple's machine will absolutely crush this HP in GPU-bound tasks and AI workloads, and the M3 Ultra's unified memory architecture means you can get far more than 8GB of VRAM for GPU compute. But the Mac Studio runs macOS, so if your workflow depends on Windows-only software or NVIDIA CUDA, it's a non-starter. The HP also has a much better port selection for legacy peripherals, with all those USB-A ports that Apple has largely abandoned.

The Lenovo Legion Tower 5a Gen 10 and MSI Aegis ZS2 are interesting comparisons because they're gaming-oriented towers that often come with much stronger GPUs for less money. If your work involves real-time rendering, GPU-based simulation, or you just want to game after hours, those machines will deliver better graphics performance per dollar. But they lack the professional GPU drivers and ISV certifications that the RTX A1000 brings, which matters if you're running SolidWorks or similar software that demands certified hardware. The Dell ECT1250 and CLX SET are also worth a look if you want more storage and GPU headroom in a traditional tower, but you'll give up the compact footprint and some of that port flexibility.

Value is a tricky conversation with this machine. At $3,914 to $3,999 USD, you're paying a premium for the small form factor and that monster CPU. The Core Ultra 9 285K is genuinely excellent, but the rest of the build doesn't quite match the price tag. The RTX A1000 is a capable professional card, but it's not a high-end workstation GPU, and 8GB of VRAM feels tight for a machine in this price bracket. The 1TB SSD is also modest for a workstation, especially when you consider that many competitors in this range offer 2TB or more.

That said, the SFF form factor carries real value for certain buyers. If desk space is at a premium or you need to move the machine between sites, the compact chassis is worth something. And the port selection is genuinely better than most competitors, which saves you from buying docks and hubs. But if you don't need the small size, you can get more GPU and storage for the same money in a traditional tower. The CAD listing at $4,108 suggests Canadian buyers are paying a similar premium. Overall, this is a machine where you're paying for the engineering that went into shrinking it, not for raw component value.

HP's Z2 G1i SFF is one of those machines that makes you do a double take when you see the spec sheet. A 24-core Intel Core Ultra 9 285K crammed into a small form factor box that weighs 4kg? That's the kind of hardware density that used to require a mid-tower and a prayer. HP is clearly aiming this at people who need serious CPU muscle for 3D design, rendering, or engineering work but don't want a full-size tower eating their desk space. And on paper, the CPU is the star here, sitting at the 98th percentile in our database, which means it's basically top of the charts for this category.

The rest of the build is a bit more grounded. You get 32GB of DDR5 running at 5600 MHz, a 1TB PCIe 4.0 NVMe drive, and an NVIDIA RTX A1000 with 8GB of GDDR6. That GPU is a professional card, not a gaming card, so think certified drivers for SolidWorks, AutoCAD, and similar tools rather than high frame rates in Cyberpunk. The port selection is genuinely impressive for a small box, with 9 USB-A ports, 2 USB-C, 4 Mini DisplayPort outputs, and Ethernet. HP clearly understands that workstation users plug in a lot of stuff.

Who is this for? Mostly professionals running CPU-heavy workloads who need a compact, quiet-ish machine that won't dominate their workspace. Architects, mechanical engineers, data analysts, and anyone doing heavy multitasking with professional software will find a lot to like. Gamers and AI enthusiasts should probably look elsewhere, and we'll get into why later. The price is also worth noting. We're seeing it listed between $3,914 and $3,999 USD, which puts it firmly in premium workstation territory.

If your work involves AI model training, large-scale GPU rendering, or real-time ray tracing, this machine will frustrate you. The RTX A1000's 8GB of VRAM is simply not enough for those workloads, and our AI score of 51.2 out of 100 confirms it. You'd be much better off with a tower that has a high-end consumer GPU like an RTX 4080 or 4090, or a workstation with a more powerful professional card like an RTX 4000 Ada or better. The Lenovo Legion Tower 5a Gen 10 or MSI Aegis ZS2 are worth a look if you want more GPU power for less money.

Gamers should also skip this. The RTX A1000 is not designed for gaming, and while it can run games, you'll get far better performance per dollar from a gaming tower. And if you're a general home user who just wants a fast PC for browsing, office work, and media, this is massive overkill. A mid-range desktop with a solid CPU and integrated graphics would handle all of that for a fraction of the price. This is a specialist tool for CPU-bound professionals, and if that's not you, your money is better spent elsewhere.

If you're a professional who needs top-tier CPU performance in a small, desk-friendly package, the HP Z2 G1i SFF is a compelling option. The Core Ultra 9 285K is genuinely one of the best workstation CPUs available, and the port selection means you won't be hunting for adapters. For CAD work, 3D modeling, engineering simulations, and heavy multitasking, this machine will feel fast and responsive, and the professional GPU drivers mean your software will actually be supported the way it should be.

But if your work leans on GPU compute, AI, or large-scale rendering, this is not the right tool. The RTX A1000's 8GB of VRAM is a real limitation, and our AI score of 51.2 out of 100 tells the story. You'd be better served by a tower with a stronger GPU, even if it means giving up the small form factor. And if you're just looking for a powerful general-purpose desktop, the price is hard to justify when gaming towers with better GPUs and more storage cost less. This is a specialist machine for CPU-bound professionals who value compactness, and for that specific audience, it's a strong choice.