AMD Dominates CPU Sales While Nvidia Enters the

There's a version of this story that writes itself as a hype reel. AMD doing the impossible! Nvidia entering the CPU market! A memory chip that beats a processor upgrade! Roll the dramatic music, drop the thumbnail with the exploding GPU.

I've been watching this industry long enough to know that the hype reel and the actual story sometimes overlap — and when they do, it's worth paying attention. This might be one of those moments. Or it might be the Athlon 64 era all over again, where AMD won a cycle, Intel eventually recalibrated, and ten years later we were all pretending it never happened. I've got a take by the end. First, let me show you the terrain.

AMD owns a chart. The chart is real. The caveats are also real.

Hardware Lab's breakdown of the current CPU market opens with a number that lands differently depending on how long you've been buying processors: AMD now occupies all 15 of the top 15 spots on a major retailer's CPU bestseller chart. As the video notes, the claim comes directly from AMD's own senior director of marketing, so pin that flag where it belongs. But the structural shift underneath it is observable independent of any marketing deck.

I remember when buying a CPU meant a genuine five-year commitment. You picked a socket, you picked a platform, and you lived with it. Intel's dominance during the early-to-mid 2010s wasn't just market share — it was the default assumption baked into every build guide, every forum reply, every "what should I buy" conversation. AMD fans during the Bulldozer era had the same energy as people who still own a Zune: devoted, a little defensive, and quietly aware they'd made the harder choice.

Ryzen changed that. Not immediately, not cleanly, but durably. Zen 2 pushed further. Zen 3 refined the formula. Then came 3D V-Cache — AMD's stacked cache technology — and the X3D chips, which became a genuine landmark for gaming performance. The Ryzen 7 9800X3D is currently the standout on that bestseller list, which tracks. But the more interesting data point is that older Ryzen 5000-series chips are still selling well, years after launch. That's not momentum. That's something sturdier — value that holds up over time, which is a different kind of winning.

Some of that older-chip popularity is probably a DDR5 story in disguise. When premium memory prices stay elevated, platforms that run on cheaper memory get a second life with budget-conscious builders. It's less about AMD being special and more about AMD having built chips that age gracefully enough to catch that wave. Intel hasn't entirely vacated the field — its Core Ultra 250K has shown it can still surprise in certain workloads — but for the moment, AMD is setting the pace in consumer CPU sales.

Retail rankings fluctuate on promotions, stock levels, and timing. Check the same chart next week and Intel may have clawed back a position or two. A single snapshot doesn't rewrite market history. But the snapshot still tells you something about where the gravitational center has moved.

Nvidia is making a CPU now. I've seen this movie before.

Here's where I need to engage my pattern-recognition instincts, because Nvidia building an ARM-based CPU platform is exactly the kind of move that has a well-worn history in this industry. GPU company decides CPUs are part of its destiny. Considerable resources are committed. Demos are impressive. Then the hard part arrives: a software ecosystem that wasn't built for you, an x86 installed base with no particular reason to migrate, and the dawning realization that dominance in one silicon category doesn't automatically transfer to another. See also: Intel's discrete GPU ambitions, which took a decade of bruises to produce something competitive.

The RTX Spark is Nvidia's first serious consumer push into this territory. It's an ARM-based system-on-chip that reportedly combines a 20-core CPU with a large GPU built on Blackwell architecture — Hardware Lab cites 6,144 GPU cores, though that specific count for an integrated SoC variant isn't confirmed against official Nvidia specs, so treat it as a reported figure — paired with up to 128GB of unified memory. That last number is doing real work. This isn't "Nvidia makes a processor now." It's a fundamentally different architecture that blurs the CPU/GPU boundary in a way that makes the traditional comparison chart awkward.

Tech PowerUp got the first hands-on testing, running Alan Wake 2 and a demo of Capcom's long-gestating title Pragmata through the hardware. That second choice is interesting because Pragmata isn't a native ARM application — it ran through Microsoft's Prism emulation layer, which translates x86 code for ARM hardware. "Prism" is the name that's been circulating for Windows on ARM's x86 translation layer, though Microsoft's naming conventions in this space have shifted across versions, so I'll flag that loosely. Emulation layers have a history of introducing just enough performance tax to make demos feel a little too curated. In this case, Hardware Lab reports performance held up with no significant frame drops observed.

Here's my vendor-demo caveat, and I want to say it the way it deserves to be said: I have watched hardware demos since "multimedia PC" was a bullet point on a cardboard retail box, and I can tell you that every piece of software running in a controlled demo environment was selected because it runs well in a controlled demo environment. The Alan Wake 2 test ran with what the video describes as "DLSS 4.5" ray reconstruction and multi-frame generation enabled — and I'd flag that DLSS 4.5 versioning is unconfirmed as of current public information, with DLSS 4 being the documented current generation. Whatever the exact version number, the point is that the RTX Spark ran with every performance assist Nvidia could throw at it. That's not a disqualification; it's just the difference between a demo and a review.

What is genuinely notable: Nvidia's software stack, including its latest upscaling and frame generation features, appears to be fully enabled on the RTX Spark with no feature scaling. AMD's current RDNA 3.5 architecture may not get access to AMD's newest upscaling technology, which is a meaningful contrast in a head-to-head. On a compact integrated platform where squeezing extra frames from existing hardware matters more than on a discrete GPU tower, that software advantage could extend the RTX Spark's useful lifespan.

I'm not ready to call this ARM on Windows going mainstream. The ecosystem challenge is real, the pricing and availability questions are still open, and Nvidia has everything to prove in a market where it has no installed base. But the RTX Spark looks like a serious attempt rather than a strategic placeholder — and a serious attempt from Nvidia carries a certain weight.

The memory story has a better lede than it's been given.

AMD quietly announced support for a new ultra-low latency memory standard, and most coverage has treated it as a footnote. Let me give you the lede first, then the fine print.

AMD claims this memory standard delivers roughly 4% average gaming frame rate improvement over standard high-speed DDR5 configurations. According to Hardware Lab's analysis, that's a larger performance jump than the delta observed when moving from the Ryzen 9800X3D to the newer 9850X3D. Read that again: swapping your memory could, according to AMD's own numbers and Hardware Lab's interpretation, give you more gaming headroom than buying an entirely new CPU generation. That specific comparison is Hardware Lab's framing, not independently benchmarked, and AMD's 4% claim is self-reported — both deserve stress-testing before anyone opens their wallet. But if the numbers hold up under third-party testing, this is a quiet reconfiguration of what an "upgrade" means on Ryzen platforms.

The catch is hardware-specific: you need a memory kit built explicitly for this standard. BIOS updates are already rolling out for current 800-series boards and some 600-series boards, but the existing kit you already own almost certainly won't qualify. Advanced users may eventually be able to approximate the settings manually through subtiming adjustments, with all the stability risk that implies. For everyone else, this is a watch-this-space rather than a buy-now situation. Compatible kit prices will start high and come down, which is the standard arc for any new memory standard.

The question I keep returning to is whether this moment is structurally different from previous AMD resurgences — the Athlon 64 era, the early Ryzen wave — where AMD won a cycle, the industry recalibrated, and equilibrium returned. My honest read: the X3D architecture represents a genuine technical differentiator that doesn't just win benchmarks but solves a real problem in gaming workloads, and that's harder to dismiss than a clock-speed race. Nvidia's ARM entry adds a competitive vector to the CPU market that hasn't existed before. The memory latency work suggests AMD is compounding advantages rather than coasting on them.

That's not the same as saying Intel is finished — its Core Ultra lineup has proven it can still put numbers on the board in the right workloads — or that Nvidia will successfully port its GPU dominance into a new market category on its first serious try.

But for the first time in a while, the PC hardware market feels like it's moving rather than cycling. Whether that holds past the next product generation is exactly the question I don't know the answer to. I just know I've been wrong before about how long a lead can last.

— Mike Sullivan, Technology Correspondent, BuzzRAG