In the highly competitive arena of high-end workstation processors, Intel is preparing a major offensive with its “Granite Rapids” Xeon lineup, directly challenging AMD’s well-established Threadripper series. Leaked specifications and pricing suggest a strategic battle is brewing, one that goes beyond simple core counts to encompass cache size, platform features, and AI performance. We sat down with Dominic Jainy, a veteran IT professional with deep expertise in high-performance computing, to dissect what these new chips mean for professionals. Our conversation explored the nuanced trade-offs between the flagship Xeon 698X and its high-core-count rival, the value proposition of Intel’s mid-range offerings, the architectural leaps driving significant AI workload acceleration, and the practical implications of the new W890 platform.
The flagship Xeon 698X is listed with 86 cores, trailing AMD’s 96-core Threadripper. How might its 336MB cache and architecture compensate for the core deficit, and for what specific professional workflows would you recommend this nearly $9,300 chip over its rival?
It’s a classic battle of brute force versus intelligent design, and while having 96 cores is impressive on a spec sheet, it’s not the whole story. The staggering 336MB cache on the Xeon 698X is a game-changer for certain workloads. Think of it as a massive, high-speed staging area right next to the cores. For professionals in scientific computing, financial modeling, or even high-end visual effects, where the processor is constantly churning through enormous datasets, that cache can dramatically reduce the time spent waiting for data from system memory. This can lead to a real-world performance gain that more than makes up for having ten fewer cores. I’d recommend this chip to anyone whose workflow is latency-sensitive and involves complex, repetitive calculations on large data pools, where keeping the pipeline fed is more critical than parallelizing across the maximum number of threads.
Let’s look at the mid-range. The 18-core Xeon 654 is priced around $1,300, undercutting AMD’s 16-core competitor. Beyond core count, what specific architectural advantages or platform features on the new W890 platform could make this the smarter buy for content creators and data scientists?
The mid-range is where the value proposition really comes into focus, and the Xeon 654 looks incredibly compelling. For about $350 less than the competition, you’re getting two extra cores, which is a fantastic start. But the real magic is in the new W890 platform it sits on. We’re told this platform brings higher memory bandwidth and improved power efficiency, which are huge quality-of-life improvements. For a content creator scrubbing through 8K video timelines or a data scientist loading massive models into memory, faster memory access feels incredibly responsive and cuts down on frustrating wait times. Furthermore, the platform’s support for up to 80 PCIe 5.0 lanes on the “Mainstream” configuration is a massive win, allowing you to connect multiple GPUs and ultra-fast NVMe storage without creating a single bottleneck.
Intel claims a two-to-three-times performance boost in mixed AI workloads over Sapphire Rapids. Can you walk us through the architectural changes that enable this leap? And how does Granite Rapids balance this power with its high 350W-380W TDPs to achieve better efficiency?
A two-to-three-times jump in a single generation is monumental, and it points to more than just an incremental update. While Intel hasn’t laid out all the blueprints, this kind of gain comes from a ground-up rethinking of the architecture. We’re talking about fundamental improvements in how the processor handles AI-specific instructions, coupled with faster interconnects to move data between cores and enhanced memory support to keep those cores from starving. The high 350W to 380W TDP might seem alarming, but you have to look at it through the lens of performance-per-watt. If you’re getting triple the performance for a similar or slightly higher power draw than the previous generation, your overall efficiency has skyrocketed. For a data center or a power-conscious workstation user, doing more work for every watt consumed is the ultimate goal, and that appears to be exactly what Granite Rapids is delivering.
The lineup introduces the W890 platform with “Expert” and “Mainstream” configurations offering 96 and 80 PCIe 5.0 lanes, respectively. Could you detail a real-world workstation build for each tier, explaining how professionals would leverage that difference in connectivity for their specific projects?
Certainly. For the “Mainstream” tier with its 80 lanes, I envision a build for a serious 3D artist or a video editor. You could comfortably run a top-of-the-line graphics card in a full x16 slot, have three or four Gen5 NVMe drives for blazing-fast project and cache storage, and still have plenty of bandwidth for a 10-gigabit networking card and other peripherals. It’s a powerful, balanced setup that eliminates common I/O bottlenecks. The “Expert” tier, with those extra 16 lanes bringing the total to 96, is for the absolute edge cases. Imagine a machine learning engineer who needs to run two or even three high-end GPUs for model training, each requiring a full x16 slot to communicate effectively. That’s where those extra lanes are crucial—they ensure every single component is running at its full, uncompromised potential. This is the build for someone running complex simulations or AI workloads where every ounce of parallel processing and data throughput is essential.
What is your forecast for the high-end workstation CPU market over the next few years, considering Intel’s renewed push with Granite Rapids and AMD’s established Threadripper dominance?
My forecast is that the market is about to get incredibly exciting for consumers. For the last few years, AMD’s Threadripper has largely owned this space by winning the core-count war. But Intel’s Granite Rapids strategy shows they are no longer playing that one-dimensional game. They are competing on architecture, with a massive cache on the flagship; on value, with aggressive pricing in the mid-range; and on innovation, with a massive claimed leap in AI performance. This forces a much more interesting choice for professionals. It’s no longer just about picking the chip with the most cores. Now, it’s a nuanced decision based on workflow: Do you need the highest parallel thread count, or will your work benefit more from a gargantuan cache and a highly optimized platform? This renewed competition will undoubtedly spur faster innovation from both sides and, hopefully, more competitive pricing across the board.