Maxsun Intel Arc Pro B60 Dual 48G Turbo (2025) is one of the more unusual workstation graphics cards to surface in the Intel Arc Pro lineup, and it arrives at the perfect moment. With local AI, on-device inference, and “AI agent” workflows becoming everyday tools for creators and developers, GPU memory capacity is quickly becoming just as important as raw compute. Intel’s push into pro graphics, paired with its latest Xeon 600 workstation platform, also makes the idea of an all-Intel content creation and AI system more realistic than ever.
At its core, Intel’s Arc Pro story started with the Alchemist A-series, then expanded with the Battlemage B-series products. Those earlier releases were largely about value and strong performance-per-dollar. Now, Arc Pro B60, B65, and B70 signal a move upmarket—especially for users who need more VRAM for professional apps and larger AI models.
What makes Maxsun’s take especially interesting is that it doesn’t stop at a standard single-GPU design. Instead, it doubles down.
A dual-GPU Arc Pro card built for AI and scaling
The Maxsun Intel Arc Pro B60 Dual 48G Turbo integrates two Arc Pro B60 GPUs on one card, effectively doubling the memory to a massive 48GB VRAM total (24GB per GPU). It keeps a dual-slot form factor, which matters if you’re building dense workstation configurations where airflow and expansion space are at a premium. This design is also positioned as a strong fit for scalable multi-GPU setups, including configurations that can run up to four cards in a system (where supported), depending on platform and chassis.
Intel Arc Pro B60 specs: Battlemage BMG-G21 with pro-focused memory
A standard Intel Arc Pro B60 uses the full Battlemage BMG-G21 GPU die, featuring 20 second-gen Xe cores, 160 XMX engines, and up to 197 TOPS (INT8) peak compute. Total board power can scale from 120W up to 200W, and the card uses a PCIe Gen5 x8 interface. Compared with the more gaming-oriented Arc B580, Intel’s own figures place B60 about 15.5% lower in peak TOPS, which aligns with B60’s more professional positioning and different power tuning.
Memory is one of the biggest highlights. Arc Pro B60 ships with 24GB GDDR6 on a 192-bit bus, delivering 456 GB/s bandwidth. It uses 19 Gbps GDDR6, and notably doubles capacity relative to the Arc B580—one reason it’s more appealing for AI workloads and scenes that simply do not fit inside 16GB cards.
Intel positions Arc Pro B60 performance against GPUs like the RTX 5060 Ti 16GB and RTX 2000 Ada 16GB. In Intel’s comparisons, Arc Pro B60 can reach up to 2.7x performance in various AI models, and the advantage grows when larger LLMs enter the picture—because 16GB VRAM often becomes the limiting factor.
Maxsun’s Dual 48G Turbo: two GPUs, 400W, and independent resources
In Maxsun’s dual-chip implementation, each GPU carries its own set of resources rather than acting like a single unified GPU. Each Battlemage GPU includes 20 Xe2 cores clocked at 2400 MHz and 24GB GDDR6 memory across a 192-bit interface. Each GPU also runs over its own PCIe 5.0 x8 link.
The full card is rated at a 400W TDP, reflecting the reality that you’re effectively powering two workstation GPUs in a single board.
Project Battlematrix: scaling to 8 GPUs and 192GB VRAM for large LLM inference
The Maxsun Arc Pro B60 Dual 48G Turbo also makes more sense when viewed through Intel’s bigger plan: Project Battlematrix. This is an inference workstation platform concept built to solve a common problem in AI—scaling memory and compute for larger models without jumping straight to extremely expensive enterprise accelerators.
With Project Battlematrix, Intel is targeting support for up to 8 Intel Arc Pro AI GPUs in one system. Partners can offer dual-GPU Arc Pro B60 designs to increase density. Importantly, the two GPUs on these cards are not tied together by a PLX chip; they show up as separate GPUs, each with its own 24GB VRAM. That means a single card totals 48GB VRAM, and an 8-GPU setup can reach 192GB VRAM overall.
In the full configuration, Intel describes a platform capable of 8 BMG-G21 GPUs, 192GB VRAM, and 1280 XMX engines—aimed at running model sizes above 70B parameters. The platform is optimized around PCIe Gen5 on the Xeon ecosystem and is expected to ship with an LLM-optimized Linux software stack plus full-stack validation, positioning it as a value-focused AI inference solution for professionals.
Maxsun Arc Pro B60 Dual 48G Turbo design and what’s in the box
Maxsun packages the card in a straightforward cardboard box with prominent “B60” branding, along with a 3-year limited warranty label on the back. Inside, there’s a thank-you letter and QR codes for support resources. The GPU itself is protected by anti-static wrapping plus foam supports to prevent damage in transit.
Accessory-wise, the included item of note is a 16-pin power adapter rated for 450W, converting to three 8-pin connectors.
Physically, this is a tall card at 300 x 110 x 40mm. The shroud is plastic and features the phrase “Touch The AI Future,” leaning heavily into its AI-first identity. On the rear, there’s a metal backplate with multiple screws securing the cooling assembly, and the backplate also covers memory modules mounted on the backside of the PCB.
Branding details include an “Intel Arc Pro” logo on one side and a Maxsun x Intel Arc sticker, while the opposite side keeps a clean, matte-black look. Power is delivered through a single 16-pin connector, and the card includes two mounting holes designed for server mounts.
For display connectivity, the rear bracket includes four outputs: two DisplayPort and two HDMI. A key detail here is that each GPU has dedicated outputs, reinforcing that the two GPUs can operate independently.
Cooling: blower-style “Turbo” design with vapor chamber plates
Maxsun uses a blower-fan “Turbo” cooling approach, designed to push a high volume of air through the heatsink and exhaust it out the rear—often preferred in workstation and multi-GPU builds where dumping heat inside the chassis can quickly become a problem.
Inside, the cooler uses two nickel-plated copper baseplates with vapor chamber cooling. One GPU uses a larger coldplate, while the other uses a smaller die baseplate. Thermal pads cover VRAM and VRM components, and the fan connects via a single 4-pin PWM header. The heatsink itself uses multiple aluminum fin arrays spaced for consistent airflow.
A look at the PCB shows two Battlemage BMG-G21 GPUs and the full 48GB VRAM total (24GB per GPU). The front side includes 12 GDDR6 modules and dual 6-phase VRM configurations for each GPU. The backside carries the remaining 12 GDDR6 memory chips (1GB dies), with thermal pads on the backplate to help manage temperatures.
Performance notes: graphics, pro apps, and AI benchmarks
In synthetic graphics testing like the 3DMark suite, the benchmark reportedly detected only a single GPU and delivered performance roughly comparable to, or slightly above, an Arc B580.
Additional testing referenced includes Geekbench 6 GPU (Vulkan and OpenCL), Blender rendering benchmarks across three scenes, and a V-Ray GPU score of 1890 vpoints. Professional visualization performance was also measured using SPECviewperf 15.1.
On the AI side, Geekbench AI results place performance in the neighborhood of GPUs like an RTX 5060 Ti to RTX 5070 range (depending on the specific test path and configuration), and UL Procyon AI benchmarking was also included as part of the test stack.
Pricing and who this card is for
Maxsun lists the Arc Pro B60 Dual 48G Turbo with a $1200 MSRP, though the provided pricing notes that it was selling for $2000+ as of 4/2/2026.
This is not a typical gaming graphics card, and it’s not trying to be. The main appeal is high VRAM capacity, workstation-friendly cooling, and the ability to scale into dense multi-GPU AI inference builds—especially for users working with larger LLMs where 16GB-class GPUs quickly hit memory walls. If your workloads involve AI inference, content creation, rendering, and pro visualization—and you want a path toward a high-density “Battlematrix” style setup—this dual-GPU 48GB design stands out as a distinctive option in the Intel Arc Pro ecosystem.Intel’s Arc Pro B60 is quickly turning into one of the most interesting value picks for local AI workloads, and the Maxsun Arc Pro B60 Dual 48GB Turbo pushes that idea even further. By pairing two GPUs and doubling the memory pool, this workstation-focused graphics card is built to handle larger generative AI models that often choke on lower-VRAM alternatives, while still being surprisingly capable outside of pure AI tasks.
A big part of the experience comes down to software support. Intel’s AI Playground suite bundles multiple AI tools into a single, responsive application, including a local AI agent and generative AI integrations. With the Arc Pro B60 in the system, text output and GenAI-style responses felt noticeably quick, making it genuinely fun to experiment with different features without the sluggishness that often comes with running AI locally.
For a more hands-on approach to large language models, LM Studio was also tested with a range of LLMs. The Arc Pro B60 Dual 48GB configuration stands out as a strong fit for 40B-parameter class models thanks to its huge memory capacity. In comparison, a single 24GB Arc Pro B60 is better suited for around 20B-parameter models, and certain 40B+ models can fail outright when VRAM becomes the limiting factor.
Getting Windows to recognize and use the full 48GB VRAM configuration wasn’t completely plug-and-play, though. It took some trial and error across different CPU and motherboard combinations, and ultimately required a BIOS update plus adjusting PCIe lane allocation to x8/x8 on the main slot so both GPUs had full access. Once that was set correctly, the complete 48GB buffer showed up properly, allowing direct performance comparisons between the single 24GB setup and the dual 48GB setup. The results favored the 48GB configuration more strongly as model size increased, where the extra memory translated into better token generation speeds and improved stability on heavier models.
What makes this card even more appealing is that it isn’t locked into “AI-only” usefulness. Gaming performance, even using the Arc Pro drivers, was better than expected. In testing, the Arc Pro B60 Dual 48GB delivered smooth gameplay and came in a bit faster than the Arc Pro B580. Cyberpunk 2077 was tested at High settings with XeSS set to Balanced at both 1080p and 2160p, with additional checks using XeSS Frame Generation turned off and on. The overall takeaway: yes, it can game, and it can do it at a level that makes it viable for modern AAA titles—especially when you consider you still have 24GB of usable VRAM on a single GPU.
From a broader value perspective, the Arc Pro B60 positions itself as a competitive AI graphics card largely because of VRAM capacity per dollar. Even top-tier consumer cards can come with significantly less memory than this 48GB dual-GPU solution, and the Intel-based option is described as landing at roughly half the price in the comparison being made. At the same time, it’s important to understand what you’re buying: in raw GPU compute and GPU-intensive workloads, you’re still dealing with B580-tier GPUs, roughly compared to RTX 5060-level performance. In other words, the real strength here is the massive memory buffer and what that enables for AI inference and large-model experimentation.
Memory-heavy professional alternatives can get expensive fast. Higher-VRAM workstation-class options can start around the multi-thousand-dollar range, and even modified high-memory enthusiast cards can sit near that same pricing territory. That’s where the Arc Pro B60 Dual can carve out a niche: it offers a large VRAM pool that’s hard to match at its price point, especially for users focused on running big local models rather than chasing maximum gaming FPS.
Design choices also support workstation scaling. The dual-slot form factor makes it practical to install multiple cards in a system that supports them, and the setup can scale up to four cards for a total of 192GB VRAM—an attention-grabbing number for anyone building an on-prem AI workstation for local LLMs, generation pipelines, or other VRAM-hungry tasks.
There are a couple of realistic trade-offs. Under load, temperatures can run high, with memory reported above 80°C, and the blower-style cooler can get loud. That’s not unusual for this kind of cooler design, but it’s still worth factoring in if you’re building a quieter workstation environment.
Overall, the Maxsun Intel Arc Pro B60 Dual 48GB Turbo stands out as a compelling AI-first graphics card that can comfortably run large LLMs thanks to its massive VRAM, while also delivering solid content creation performance and unexpectedly playable gaming results. For anyone prioritizing local AI workloads, large language model experimentation, and scalable VRAM capacity in a workstation-friendly form factor, it’s an option that’s difficult to ignore.
