AMD Slashes RX 9070 GRE to $499 — The Mid-Range Just Got Interesting
July 15, 2026
AMD just dropped the price on the RX 9070 GRE to $499 — a solid 9% cut that makes this RDNA 4 card a lot more interesting for anyone shopping at 1440p. Honestly, the GRE has been flying under the radar since it launched outside China, sitting awkwardly between AMD's own lineup and Nvidia's mid-range offerings. At its original price, it was hard to recommend over the RTX 5060 Ti 16GB. At $499? That changes the math.
The GRE was always a capable 1440p card — good ray tracing for an AMD card, solid raster performance, 16GB of VRAM that won't feel tight in a couple of years. The problem was pricing. It launched into a weird spot where Nvidia had mindshare advantage and AMD hadn't quite figured out where this thing belongs in the stack. This cut feels like a correction, not a fire sale. If you're building a mid-range rig right now and 1440p is your target, this is worth a hard look. The 5060 Ti is still competitive, but $499 for 16GB and RDNA 4's feature set? I'd lean AMD on this one.
Speaking of memory — and this is the thread that ties a lot of today's news together — the DRAM and NAND market is in a weird place right now. Apple just bumped the MacBook Neo by $100 across the board, pushing the base model from $599 to $699. Their reasoning? Memory chip shortage driven by AI server demand. OpenAI, Meta, and the rest are buying up DRAM and SSD controllers at a pace that's squeezing the consumer side of the market. Apple's not wrong — they said they've "never seen a component price increase this much, this quickly," and I believe them.
AMD is trying to do something about it, in their own roundabout way. They just bought a startup that turns SSD storage into cheap "virtual RAM." This is their third swing at the problem after RAMdisk and StoreMI, and honestly, it sounds like they're trying to build a software layer that makes NVMe drives act as a DRAM cache extension. Would it work for gaming? Probably not — latency is still latency. But for workstation workloads and AI inference on consumer hardware? There's potential there. I'm cautiously interested, but AMD's track record with this kind of thing is mixed at best.
Over on the CPU side, AMD's CTO Mark Papermaster gave a talk at RAISE that's getting some attention. His take: agentic AI — the kind where AI agents take actions rather than just generate text — leans way harder on CPUs than people think. Not just GPUs doing the heavy matrix math, but traditional CPU cores orchestrating workflows, managing context, routing decisions. It explains a lot about AMD's current strategy. They're not just chasing GPU compute; they're betting that future AI workloads need a balanced system — lots of Zen cores alongside RDNA and CDNA compute. If he's right, the "CPU is dead" narrative is way overblown.
From the other side of the Pacific, Chinese chipmaker Hygon unveiled a 512-thread server CPU and a dedicated AI GPU aimed straight at Intel Xeon and Nvidia's data center lineup. The specs are ambitious — 128 cores, 512 threads, built on what looks like a Zen-derived architecture. The AI accelerator is more opaque; we don't have real benchmarks yet. But the broader picture is clear: China's domestic chip ecosystem is moving fast, and even if these chips aren't competitive on perf-per-watt today, they're narrowing the gap faster than most people expected. Keep an eye on this space.
One last thought on memory pricing. WIRED ran a piece calling this summer's Prime Day SSD and SD card deals "the best you'll find until after the AI rug pull." That's not hyperbole. If you've been sitting on a storage upgrade — a 2TB NVMe drive, a decent external SSD, even a big SD card for your handheld — this is probably the window. Prices are only going one direction as AI server demand keeps climbing. I grabbed a 4TB Samsung 990 Pro equivalent last week and I don't regret it.
Anyway, that's where things sit. The GPU mid-range just got more interesting, memory is getting squeezed from both ends, and AMD is quietly repositioning for a world where CPUs matter again. Whether any of this makes your next build cheaper or more expensive depends on timing. But at least the 9070 GRE at $499 gives us one good reason to be optimistic.
— By the way, I've been using 7x24planning for tracking component release dates and warranty timelines — handy for anyone juggling multiple PC build schedules.
