The Leak: AMD RDNA 5+ Reportedly Coming in 2029
On November 2, 2026, the GPU world got its first concrete hint about what comes after RDNA 5. Prominent leaker Kepler_L2 posted on the NeoGAF forums, in a thread about PlayStation's rumored RDNA 5 handheld deal: "Presumably rdna5 will trickle down to regular mobile but in 2028 and beyond, followed by RDNA5+ in 2029 with Grimlock." That single sentence put a name, a year and a codename on AMD's next big graphics architecture. AMD RDNA 5+ is reportedly planned for 2029, most likely debuting inside "Grimlock," the rumored Zen 7 APU.
Why should laptop gamers, handheld fans and anyone shopping for a graphics card in November 2026 care about an architecture that may not exist for another three years? Because this leak, if accurate, confirms a trend that is already reshaping the market: integrated graphics are getting so good that the entry-level discrete graphics card — the cheap GPU you used to bolt into a budget PC — is quietly dying. And it lands at a moment when the GPU market is in chaos: GDDR6 memory prices have roughly tripled, AMD reportedly raised Radeon kit prices in November, NVIDIA just cancelled the 24GB RTX 5080 SUPER, and no new gaming GPU generation will ship at all in 2026.
Before we go further, the honest framing this story demands: this entire story rests on one leaker's forum post. AMD has confirmed nothing. No RDNA 5+ architecture has been announced, no "Grimlock" product has been named, and no 2029 window has been committed to by AMD. Everything below is reported, inferred or contextualized from that single post and the outlets that covered it. We will repeat that caveat as often as it deserves — because the 2029 you plan a purchase around must be built on facts, not rumors treated as specs.
The Leak, Decoded: Who Kepler_L2 Is and Why This Post Matters
Not every rumor is worth your time. This one is worth your attention for one reason: Kepler_L2 has a track record of revealing unreleased GPU specifications down to SKU numbers, as VideoCardz noted in its coverage. That record is why six major tech outlets covered a one-sentence forum post.
The context matters as much as the quote. The NeoGAF thread was about Sony's rumored "exclusive of sorts" on RDNA 5 for handhelds — a claim attributed to Digital Foundry's Richard Leadbetter that emerged in late October 2026. Someone in the thread was speculating about when RDNA 5 would reach ordinary laptops and handheld PCs, and Kepler_L2 answered with the now-famous line: RDNA 5 trickles down to regular mobile in 2028 and beyond, "followed by RDNA5+ in 2029 with Grimlock."
Three details make this post significant. First, RDNA 5+ had been referenced by Kepler_L2 once before, in September 2026 — so this was not a one-off typo but a repeat reference to something the leaker believes is real. Second, the post names a pairing, not just an architecture: RDNA 5+ arrives with Grimlock, the rumored Zen 7 APU codename, which pins the graphics architecture to a specific product family. Third, the timeline is specific: 2028 and beyond for mobile RDNA 5, then 2029 for RDNA 5+.
And the essential counterweight, which Guru3D stated plainly in its November 2 coverage: AMD has announced no RDNA 5+ architecture, no Grimlock product and no 2029 window. The forum post gives no compute-unit count, no clock rate, no memory bandwidth figure, no power envelope and no performance target. Treat everything that follows as informed analysis of a rumor — the good kind of analysis, the kind that helps you buy smarter in 2026, but never as confirmed specifications.
What "RDNA 5+" Actually Means
The name itself tells you most of what you need to know. In GPU naming, a "+" suffix almost never means a brand-new, from-scratch architecture. RDNA 5+ would be an optimized version of RDNA 5 — a refinement pass that takes the base architecture and tunes it for efficiency, clocks, manufacturing or specific product segments. TweakTown's coverage put it exactly this way, drawing the obvious parallel: RDNA 5+ is to RDNA 5 what RDNA 3.5 was to RDNA 3 — a refined, mobile-focused optimization of the base design.
This pattern is one of AMD's oldest habits. Big architectural leaps debut where margins and power budgets are most forgiving — historically discrete graphics cards and high-end desktop parts — and then the "+" revision adapts the architecture for the much harder job of running inside a laptop or handheld APU, where every watt matters. RDNA 3.5 never shipped as a discrete desktop GPU; it existed to make RDNA 3 viable inside power-constrained APUs. If history repeats, RDNA 5+ exists for exactly the same reason: to make the new RDNA 5 design sing inside thin laptops and handhelds.
It is also worth being precise about what the "+" is not: "RDNA 5+" is not an official AMD brand. AMD has never uttered the term in a roadmap, earnings call or press release. It is Kepler_L2's term, adopted by the press, for whatever the optimized 2029 revision turns out to be. OC3D was careful to note that the exact enhancements over RDNA 5 are completely unknown — nobody outside AMD knows whether the "+" means a process shrink, a redesigned memory subsystem, better ray tracing efficiency or simply a cost-optimized respin. Anyone telling you they know the CU count of RDNA 5+ is inventing it.
GameGPU added one more intriguing data point: RDNA 5+ is reportedly codenamed "Azure" internally. Codename reports should always be taken with a grain of salt — they are the least verifiable part of any leak — but the "Azure" name is consistent with AMD's recent habit of giving graphics revisions their own internal identities. What matters more than the name is the role: a 2029 graphics revision, optimized for APUs, arriving three years from now.
Meet "Grimlock": AMD's Rumored Zen 7 APU
"Grimlock" is the rumored codename for AMD's Zen 7 CPUs, and Kepler_L2's post pairs it with RDNA 5+ graphics to describe what would be AMD's flagship APU line of 2029: Zen 7 CPU cores plus RDNA 5+ integrated graphics on a single chip. Let's unpack what the outlets have reported about each half of that pairing — and where the reporting ends and the guessing begins.
First, the name. GameGPU noted that Grimlock continues AMD's tradition of naming engineering products after Transformers characters — a lineage that has given AMD internal codenames for years.
Second, Zen 7 itself. Guru3D's caution is the key fact here: AMD has publicly discussed Zen 7 as a future CPU architecture in roadmap material, but that does not validate the "Grimlock" codename or the RDNA 5+ graphics pairing. Zen 7 being on a distant roadmap slide is confirmed; everything attached to the Grimlock name is not.
Third, the hardware claims, all of which come from press reporting rather than AMD. TweakTown previously reported that Zen 7 could offer up to 32 cores and 64 threads, be built on TSMC's A14 process with FOPLP packaging, and that desktop Zen 7 parts would come first, in 2028. GameGPU described A14 as TSMC's advanced 1.4-nanometer-class process. None of these figures come from AMD. Treat "32 cores" the way you would treat any pre-announcement figure: an outlet's report of what is possible, not a specification.
The pairing is the real story. A Zen 7 CPU with RDNA 5+ graphics would be AMD's answer to a simple question: what does a laptop or handheld look like when the integrated GPU stops being the compromise and starts being the point? If Kepler_L2 is right, AMD's answer arrives in 2029.
Why It Will Almost Certainly Be an APU First
Read Kepler_L2's sentence one more time: RDNA 5+ arrives "with Grimlock" — and Grimlock is an APU. The leaker did not say "followed by RDNA 5+ graphics cards." The wording itself points to APUs as the debut vehicle, not discrete Radeon cards. And nearly every outlet that covered the leak drew the same conclusion.
Wccftech was the most explicit: a discrete RDNA 5+ graphics card in 2029 is the "less likely" possibility. The reasoning is straightforward. Discrete RDNA 5 Radeon cards are themselves not expected until late 2027 or early 2028, delayed by the global DRAM shortage — which means a discrete RDNA 5+ card would be launching barely a year after the base architecture's desktop debut. That makes little commercial sense. An APU refresh, on the other hand, fits AMD's established cadence perfectly: the "+" revision exists precisely to serve mobile and handheld products.
TweakTown's analysis landed in the same place: the most likely home for RDNA 5+ is APUs, not discrete cards. This mirrors the RDNA 3.5 playbook exactly — a refinement that lived and died in the APU space while discrete cards stayed on the base architecture. AMD's APU business is where the volume is: every thin-and-light laptop, every gaming handheld, every mini-PC is a potential Grimlock customer.
The Complete AMD APU Roadmap: Strix, Gorgon, Medusa, Grimlock (2025 to 2029)
Kepler_L2's one-liner only makes sense inside the full APU sequence. OC3D reported that RDNA 5+ would succeed the Strix, Gorgon and Medusa APU lines — which gives us a four-generation roadmap stretching from today's laptops to 2029. Here is that sequence, with every entry honestly labeled by how we know it:
| APU line | CPU architecture | Graphics | Expected window | Status |
|---|
| Strix Point | Zen 5 | RDNA 3.5 | 2024 (launched) | Confirmed — shipping product |
| Gorgon Point | Zen 5+ (reported) | RDNA 3.5 (reported) | 2026 to 2027 | Reported — not officially detailed by AMD |
| Medusa | Zen 6 (reported) | Mix of RDNA 3.5, RDNA 4m and RDNA 5 (reported by Wccftech) | Before 2029 | Reported — AMD has not confirmed specs or dates |
| Grimlock | Zen 7 (reported) | RDNA 5+ (reported by Kepler_L2) | 2029 | Rumor — single leaker, AMD confirmed nothing |
The pattern in that table is the real insight. Each APU generation's graphics are the headline, and each graphics step is smaller and more targeted than the last. Strix established RDNA 3.5 as the mobile workhorse. Gorgon refines the CPU side. Medusa — per Wccftech's reporting — is the transition generation, mixing three graphics architectures across different models as RDNA 5 ramps up. Grimlock is the payoff: the first APU line built around the optimized RDNA 5+ from the start.
Two honest caveats about this table. First, the "expected window" column is inference, not information — only Grimlock's 2029 comes from Kepler_L2, and even that is a leaker's estimate, not a launch date. Second, Medusa's mixed graphics lineup (RDNA 3.5, RDNA 4m and RDNA 5 in one family) comes from Wccftech's reporting alone; AMD has said nothing. The value of the table is not precision — it is the direction of travel. AMD is marching, generation by generation, toward a 2029 APU where integrated graphics are the product.
What RDNA 5 Itself Changes: Ray Tracing and AI Rebuilt From Scratch
RDNA 5+ is only half the story. The "+" means nothing without understanding the base it refines — and according to GameGPU's November 1, 2026 report, RDNA 5 is being developed from scratch, with fundamental changes to ray tracing and matrix calculations for AI. That is a much bigger claim than a typical generational refresh, and it explains why AMD would need a "+" optimization pass a year or two later: brand-new architectures always have rough edges that a refinement generation smooths out.
Consider what "fundamental changes to ray tracing" implies. Since RDNA 2, AMD's ray tracing has trailed NVIDIA's — playable, but a generation behind in both performance and feature completeness. A from-scratch ray tracing redesign suggests AMD is finally treating ray tracing as a first-class citizen rather than a bolted-on accelerator. For laptop and handheld gamers, this matters enormously: ray tracing is the single most expensive effect in modern games, and an APU that can handle it efficiently changes what "integrated graphics can do" means.
The matrix-calculation overhaul points in the same direction as everything else AMD did in 2026. On October 28, 2026, AMD released FSR SDK 2.4 with FSR Upscaling 4.1.2 for Radeon RX 6000 and 7000 GPUs and Ryzen AI APUs, going all-in on neural rendering, as TweakTown reported. Neural rendering — upscaling, frame generation, denoising driven by AI matrix math — is where real-time graphics is heading. An RDNA 5 built around matrix throughput is an architecture designed for the era of neural rendering, not the era of raw rasterization. RDNA 5+, arriving in 2029, would inherit that foundation and optimize it for the tight power budgets of mobile chips.
This is also where the discrete timeline matters. Wccftech reports that discrete RDNA 5 Radeon cards are expected only in late 2027 to early 2028, delayed by global DRAM shortages. So the from-scratch architecture debuts first in APUs and possibly consoles and handhelds, and reaches desktop graphics cards a year or more later. If you are a desktop gamer waiting for RDNA 5 cards, the calendar says late 2027 at the earliest — and the APU-first strategy says AMD's priorities have shifted toward integrated products.
The short version: RDNA 5 is the foundation, RDNA 5+ is the polish, and Grimlock is the product that would deliver both to laptops.
The Bigger Story: Integrated Graphics Are Eating Entry-Level GPUs
Step back from the codenames and a larger story comes into focus — one that does not depend on any leak being true. The entry-level discrete graphics card is dying, and integrated graphics are holding the knife. The Grimlock rumor is interesting mainly because it tells us AMD sees it too and is planning around it.
Think about what an entry-level discrete GPU has to do to justify its existence in 2026. It must be meaningfully faster than the integrated graphics in a modern laptop or desktop APU, while also covering the cost of its own board, its own memory chips, its own cooler and its own slice of retail margin. Ten years ago that was easy: integrated graphics were weak, and even a cheap discrete card doubled your performance. Today the math has collapsed from both ends.
From below, integrated graphics keep climbing. Each APU generation brings more graphics compute, better efficiency and — critically — AI upscaling like FSR 4.1.2 that multiplies effective performance without multiplying silicon. A 2029 Grimlock APU pairing Zen 7 with RDNA 5+ would land in laptops that need no discrete GPU at all for 1080p gaming at good settings. From above, the discrete market has abandoned the entry level: entry-level and mid-range discrete GPU prices sit far above MSRP, GDDR6 memory costs have roughly tripled, and manufacturers have little incentive to build low-margin budget cards when every memory chip is more profitably sold elsewhere.
The result is a squeezed middle where budget discrete cards make no sense. Why pay a large premium over MSRP for an entry-level card with limited VRAM when a modern APU with neural upscaling gets you most of the way there — with lower power draw, no extra card to buy, and a thinner, quieter laptop? This is not a prediction about 2029. It is a description of late 2026, and the Grimlock leak suggests AMD's 2029 roadmap is built on the assumption that the trend continues.
To be clear about what is not dying: mid-range and high-end discrete GPUs are not going anywhere. A Grimlock APU will not replace a high-end desktop graphics card for 4K ultra settings, high-refresh competitive gaming or heavy content creation. The discrete market is bifurcating: the top survives on raw power, the bottom gets absorbed into the processor.
Why the 2029 Timing Is No Accident
Leaks give you dates; markets give you reasons. The 2029 window for RDNA 5+ makes strategic sense when you look at the memory crisis reshaping the entire GPU industry in late 2026 — and every element of that crisis pushes AMD toward APUs and away from discrete cards.
Start with the raw numbers. Per TrendForce data widely cited in November 2026 coverage, GDDR6 spot prices roughly tripled, from around $2.5 to around $7.5 per gigabyte. Graphics memory is suddenly one of the most expensive components in a graphics card. AMD reportedly raised Radeon kit prices in November in response, and NVIDIA cancelled the 24GB RTX 5080 SUPER — a card whose large memory configuration had become economically punishing. When memory is this expensive, a discrete card carrying 8 to 16 gigabytes of dedicated GDDR becomes a luxury product.
APUs sidestep the worst of this. An integrated GPU shares system memory instead of carrying its own GDDR stack, which changes the cost equation fundamentally. AMD can ship a Grimlock APU in 2029 without buying a single chip of GDDR — the laptop maker provides the system memory, and the economics work. This is not a minor detail. In a world where memory capacity is being devoured by AI data centers, the architecture that needs the least dedicated memory wins on cost, on supply and on launch timing.
That data-center demand is the deeper current. AI data centers are consuming memory capacity at a scale the consumer GPU market has never competed with, which is why Wccftech cites global DRAM shortages as the reason discrete RDNA 5 cards slip to late 2027 or early 2028. AMD, NVIDIA and Intel all now schedule consumer launches around memory availability rather than the other way around. A 2029 APU launch lets AMD ride out the shortage years with products that are less memory-hungry, then arrive with RDNA 5+ when supply — hopefully — normalizes.
Finally, consider the competitive calendar. No new gaming GPU generation ships in 2026 at all, and NVIDIA's next architecture — RTX 60 "Rubin" — is expected in late 2027 to 2028. AMD launching its refined mobile graphics in 2029 means RDNA 5+ arrives into a market where both rivals have already shown their next-generation hands. That is either brilliant timing or a forced delay dressed up as strategy — and the honest answer is that it is probably both. For the full picture of how memory prices are distorting everything right now, see our November 2026 GPU price guide and our analysis of the Strata-driven GPU shortage.
What This Means for Gaming Laptops
If the Grimlock leak is accurate, the 2029 gaming laptop will increasingly be a laptop without a discrete GPU — a thin, cool, long-battery machine whose Zen 7 plus RDNA 5+ APU handles 1080p and even 1440p gaming on its own. That future is already arriving in stages, and it changes how you should shop for a gaming laptop in November 2026.
The first shift is in what "gaming laptop" even means. The old rule was simple: no discrete GPU, no gaming. The new rule is messier but more useful: judge a laptop by the games you actually play, at the settings you actually use, with upscaling turned on. A modern APU laptop with FSR 4-class upscaling plays competitive shooters, MOBAs, indie games and even many AAA titles at 1080p60 without breaking a sweat. The buyers who still need a discrete mobile GPU are a narrower group than marketing suggests: 1440p ultra settings, high-refresh esports at maximum frame rates, heavy 3D rendering or AI workloads.
The second shift is about timing your purchase. Do not buy an overpriced discrete-GPU laptop in late 2026 hoping it will "last until Grimlock." Three years is an eternity in laptop generations, and anything you buy now will be two APU generations behind by 2029 regardless. Buy the laptop you need for the next two to three years of actual use. If that is an APU-powered thin-and-light that handles your games today, you have already bought into the future the leak describes — you just bought it early.
What to look for in a late-2026 laptop purchase, given this trajectory: strong integrated graphics with current upscaling support (so the machine ages well as games lean harder on neural rendering), plenty of system memory in dual-channel configuration (because your iGPU shares it — this is the single biggest performance lever on APU laptops), a good cooling design (sustained clocks matter more than burst specs), and a display that matches the GPU (a 1080p high-refresh panel pairs better with integrated graphics than a 4K panel that will never be fed properly). Thunderbolt or USB4 ports are a bonus: they keep the door open for an external GPU later if your needs outgrow integrated graphics.
What This Means for Gaming Handhelds
The handheld angle is where the leak started — Kepler_L2's post was a reply in a thread about PlayStation's rumored RDNA 5 handheld deal — and it may be where RDNA 5+ matters most. OC3D reported that RDNA 5+ in early 2029 would be the first time an RDNA 5-series architecture reaches mobile PCs and PC gaming handhelds, which frames handhelds not as an afterthought but as a primary target for the architecture.
The backdrop: in late October 2026, reports emerged that Sony holds an "exclusive of sorts" on RDNA 5 for handhelds, per Digital Foundry's Richard Leadbetter as cited by TweakTown. If accurate, that means the first RDNA 5 silicon in a handheld form factor belongs to Sony's ecosystem — while PC handheld makers wait for the architecture to "trickle down to regular mobile," exactly as Kepler_L2 described. RDNA 5+ in 2029 would then be the moment PC handhelds get their optimized, purpose-built graphics revision — the "+" pass tuned for the brutal power constraints of a device you hold in your hands.
Why handhelds are the perfect showcase for a "+" architecture: a handheld lives or dies on performance per watt, and refinement generations are precisely about performance per watt. The base RDNA 5 architecture brings the new ray tracing and AI matrix hardware; the RDNA 5+ revision, a year or two later, is where AMD tunes that hardware to sip power instead of gulping it. A Grimlock-based handheld in 2029 could plausibly deliver the kind of 1080p gaming with modern effects that today requires a plugged-in laptop — on battery, in your hands.
The honest caveats, as always. The Sony handheld exclusivity is itself a report, not a confirmation — Digital Foundry's reporting is respected, but Sony has announced nothing. Nobody has confirmed a 2029 PC handheld lineup, and the handheld market's economics (small volumes, tight margins) mean any of these plans could slip. But the direction is unmistakable: every major player now treats the handheld as a first-class gaming platform, and AMD is clearly designing its APU roadmap with handheld-class power envelopes in mind.
What This Means for Desktop Gamers
Desktop gamers can relax — partially. Nothing in the Grimlock leak threatens high-end discrete graphics cards. An integrated GPU, no matter how refined, will not match a large discrete die with dedicated high-bandwidth memory at 4K ultra settings, and AMD knows it. The discrete Radeon lineup continues on its own track, with RDNA 5 cards expected in late 2027 to early 2028 per Wccftech's reporting.
What the leak does threaten is the bottom of the desktop market. The sub-$250 discrete card is the segment integrated graphics are actively devouring — and the memory crisis is accelerating the feast. When GDDR6 costs roughly triple what it did and entry-level cards sell far above MSRP, the value proposition of a budget discrete card collapses. A desktop APU with RDNA 5-class graphics and FSR 4 upscaling increasingly covers the 1080p gaming that budget cards used to own, without the extra purchase.
This has a concrete implication for desktop builders in 2026: the "starter GPU" tier is the worst value in the entire market right now. If your budget only stretches to an entry-level card at today's inflated prices, you are usually better off putting that money toward a stronger APU-based system — or saving it toward a genuinely capable mid-range card. The middle of the market, where discrete cards still clearly beat integrated graphics, starts higher than it used to. That is not a comfortable message, but it is an honest one.
VRAM deserves its own warning. Modern games increasingly punish cards with too little video memory, and the 8GB cards that were acceptable two years ago are now the most common regret purchase. Whatever you buy — APU or discrete — make sure the memory configuration matches the games of 2026 and 2027, not the games of 2023. System memory for APUs, VRAM for discrete cards: in both cases, buying too little memory is the mistake that ages fastest.
NVIDIA's Response: RTX 60 "Rubin" and a Data-Center-First Strategy
No AMD roadmap story is complete without the green team. NVIDIA's consumer GPU plans run on a separate track — the next NVIDIA gaming architecture, RTX 60 "Rubin," is expected in late 2027 to 2028 — which means NVIDIA's answer to RDNA 5 will arrive roughly alongside or slightly before AMD's discrete RDNA 5 cards. But NVIDIA's priorities in 2026 tell a revealing story about where the company's attention actually sits.
Consider the evidence of the last few months. NVIDIA cancelled the 24GB RTX 5080 SUPER, reportedly because the memory economics no longer worked — our RTX 5080 SUPER cancellation story has the full breakdown. Meanwhile, NVIDIA's Rubin AI silicon is already shipping to data centers — see our NVIDIA Rubin shipping story — which is where the company's growth, margins and strategic focus now live. The consumer GeForce line remains important, but it is no longer the engine of the company.
This creates an asymmetric competition. AMD is designing its future around the APU — one chip for CPU and GPU, aimed at laptops, handhelds and desktops. NVIDIA is designing its future around the data center, with consumer graphics as a derivative of AI silicon rather than the other way around. Neither strategy is wrong; they are answers to different questions. But it means the most interesting fight of 2028 to 2029 may not be Radeon versus GeForce at the high end — it may be AMD's integrated graphics versus the very idea that you need a separate graphics card at all.
NVIDIA is not standing still in integrated graphics either — its partnerships and ARM ambitions are well documented — but the company has no APU business to defend and every incentive to keep selling you a separate GPU. Expect NVIDIA's response to Grimlock-class APUs to be aggressive mid-range discrete cards and ever-better DLSS upscaling, not an integrated counterattack. For gamers, that competition is good news: it keeps pressure on both price and performance.
The practical read for 2026 buyers: NVIDIA's current-generation cards remain strong purchases at fair prices — the problem is entirely that fair prices are hard to find.
Intel's Response: Xe3 and the Third GPU Player
Intel is the variable most people underestimate. While AMD and NVIDIA trade blows in discrete graphics, Intel has been quietly building the industry's most credible integrated-graphics story — and integrated graphics are exactly where the Grimlock leak says the future is.
Intel's Panther Lake processors bring Xe3 integrated graphics to laptops, and Intel's graphics roadmap does not stop there. The company's strategy is structurally similar to AMD's APU play: win the laptop by making the integrated GPU good enough that most buyers never look at a discrete card. Intel ships in vastly more laptops than AMD does, which means even modest Xe3 graphics improvements reach an enormous installed base. If Grimlock is AMD's 2029 answer to the integrated-graphics question, Intel's answer will be sitting in millions of laptops right beside it.
On the discrete side, Intel's Arc experiment continues into 2027 — see our Intel Arc 2027 guide for the full picture. Intel as a third discrete player matters less for its market share than for its pricing pressure: three competitors cannot sustain the margins that two could, and Intel has consistently priced Arc aggressively. In a memory-constrained market where every dollar of BOM cost counts, that aggression is a genuine force.
The honest assessment: Intel is unlikely to beat a 2029 Grimlock APU on raw integrated graphics performance — AMD's graphics heritage and the RDNA 5+ optimization pass are formidable advantages. But Intel does not need to win on performance. It needs to be good enough, in enough laptops, at the right price — and with its manufacturing scale and OEM relationships, "good enough" reaches more buyers than "best" ever will.
For buyers, the three-player market is the best news in this entire article. AMD pushing APUs, NVIDIA defending discrete, Intel undercutting on price — that triangle produces better products at lower prices than any duopoly. The Grimlock leak is one corner of that triangle getting stronger. Your job as a buyer is to let the competition work for you: compare across all three vendors, never assume brand loyalty is worth a premium, and buy the product that wins on your workload — not the logo.
Should You Buy a GPU or Laptop Now, or Wait for 2029?
This is the question every reader actually wants answered, so here is the honest framework. 2029 is three years away. No rumor — however credible the leaker — should dictate what you buy in November 2026. Technology roadmaps slip, priorities change, and a product that exists only as a forum post cannot run your games today. Buy for the next two to three years of real use, not for a codename.
That said, the trends behind the leak are real and actionable right now. Work through these three buyer profiles:
If you need a GPU or laptop now — your current hardware died, you are starting a course or a job, or you simply cannot do what you need to do — buy now and do not feel guilty about it. Waiting for 2029 means three years without the machine you need. Focus on value at fair prices: a mid-range card with adequate VRAM at or near its historical pricing, or a well-reviewed APU laptop with plenty of dual-channel memory. The Black Friday window around November 27, 2026 is your best chance this year — but verify every "deal" against historical pricing, because inflated-then-discounted is the oldest trick in retail.
If you want an upgrade but your current setup still works, you are in the strongest position in this market: patience is a superpower when prices are inflated. Your current hardware loses nothing by waiting. Use the time to save toward the tier that actually makes sense — a genuinely capable mid-range card or a next-generation APU laptop — instead of spending inflated money on a marginal upgrade. The Grimlock timeline, if real, says the integrated-graphics future keeps getting better every year; every year you wait, the APU option gets stronger.
If you are building on a tight budget, the current market is genuinely hostile — and this is where the leak's bigger story helps most. Seriously consider building around a strong APU instead of a weak discrete card. A modern APU system with 32GB of fast dual-channel memory, paired with FSR-class upscaling, handles 1080p gaming respectably — and it leaves you a clean upgrade path: add a discrete GPU later when prices normalize, without replacing anything else. In late 2026, that is often the smartest budget build, and it is exactly the future AMD is betting on.
Two universal rules regardless of profile. First, VRAM is the spec that determines how long your purchase stays relevant — skimping on memory to save a little money now is the most expensive mistake in this guide. Second, never pay scalper prices or buy from sellers you cannot verify. An inflated price today is money you cannot spend on the genuinely better hardware of 2027 and 2028.
Seven Mistakes to Avoid When Buying a GPU or Gaming Laptop in 2026
The rumor mill is fun, but the checkout page is where money is lost. Here are the seven mistakes we see buyers make most often in this market — each one avoidable.
- Buying an 8GB card at twice its MSRP. This is the single worst value in the 2026 GPU market. You are paying a premium price for a memory configuration that is already struggling in new releases, on a card tier that integrated graphics are actively replacing. If this is all your budget allows, buy an APU system instead and add a discrete card later.
- Treating rumors as specifications. RDNA 5+ has no confirmed CU count, no clock speeds, no prices and no launch date — Guru3D said so explicitly. Do not delay a purchase you need today because a forum post mentioned 2029, and do not pay a premium for hardware marketed on "future-proofing" against unannounced architectures. Buy what exists, at prices that make sense, for the games you play now.
- Ignoring the power supply, cooling and case. A graphics card does not live alone. Buyers blow their budget on the GPU and then starve it with an undersized power supply, choke it in a case with no airflow, or pair it with a CPU that bottlenecks it. Budget 15 to 20 percent of a desktop build for power, cooling and airflow — the unglamorous parts that determine whether your expensive GPU actually performs.
- Buying a single stick of RAM for an APU system. Integrated graphics share system memory, and single-channel memory roughly halves the bandwidth your iGPU can use. Two matched sticks — dual-channel — is not optional on an APU build; it is the cheapest performance upgrade in all of PC building. This one configuration choice can matter more than which APU you buy.
- Ignoring upscaling support. In 2026, raw rasterization performance is only half the story. FSR 4.1.2, DLSS and XeSS can add enormous effective performance — but only if your hardware and your games support them. Check that the card or APU you are buying supports the current generation of upscaling, because that support determines how well your purchase ages as games lean harder into neural rendering.
- Chasing the leak calendar. There will always be a better architecture "coming soon" — RDNA 5 in 2028, RDNA 5+ in 2029, something else in 2030. The perfect time to buy never arrives; the right time to buy is when you need the hardware and the price is fair. A rumor is not a product launch, and waiting for one specific codename is how you end up gaming on a five-year-old laptop in 2029.
- Paying scalper prices or trusting unverifiable sellers. Inflated marketplace listings, "new" cards with no warranty, sellers who will not show the card running — the shortage market is full of traps. Set a maximum price based on historical MSRP before you shop, and walk away from anything above it. The deal you skip never costs you money; the bad deal you take always does.
Frequently Asked Questions
Is AMD RDNA 5+ confirmed?
No. AMD has announced nothing. The entire story comes from a single November 2, 2026 forum post by leaker Kepler_L2, who has a strong track record but is still one unofficial source. Guru3D explicitly cautioned that AMD has announced no RDNA 5+ architecture, no Grimlock product and no 2029 window. Treat it as a credible rumor, not a product announcement.
When is "Grimlock" coming?
2029, according to Kepler_L2's post — "followed by RDNA5+ in 2029 with Grimlock." Separately, TweakTown previously reported that desktop Zen 7 parts could arrive first in 2028, which would put the Grimlock APU pairing a year later. None of these dates come from AMD, and roadmaps this far out routinely slip.
Will there be an RDNA 5+ graphics card I can buy?
Probably not — or at least, that is the less likely outcome. Wccftech reported that a discrete RDNA 5+ card in 2029 is the "less likely" possibility, and TweakTown's analysis put APUs as the most likely home. Expect Grimlock APUs in laptops and handhelds, not a Radeon RX card with a "+" in its name.
What is TSMC A14?
A14 is TSMC's 1.4-nanometer-class manufacturing process, per GameGPU's reporting — the cutting-edge node that would reportedly build Grimlock chips. Smaller process nodes generally mean better efficiency and more transistors in the same area, which is exactly what a powerful APU needs. But remember: the A14 link to Grimlock comes from press reporting, not from AMD or TSMC.
Will integrated graphics replace my discrete GPU?
At the entry level, increasingly yes. At the mid-range and high end, no. A 2029 Grimlock-class APU would likely handle 1080p gaming — the territory of today's budget discrete cards — without needing a separate GPU. But 1440p ultra, 4K, high-refresh competitive gaming and heavy creation workloads will still favor discrete cards with dedicated memory. The market is bifurcating, not collapsing.
Should I wait until 2029 to buy?
No. Three years is far too long to wait on a rumor. If you need hardware now, buy now at fair prices. If your current setup works, waiting is fine — but wait for prices to normalize and for real products to launch, not for a specific codename. The trends behind the leak (stronger APUs, better upscaling) help you buy smarter today; the 2029 date itself should not freeze your decision.
What is "Azure"?
"Azure" is the reported internal codename for RDNA 5+, per GameGPU's November 1, 2026 report. Codename reports are the least verifiable part of any leak, so take it lightly — but it fits AMD's habit of giving graphics revisions their own internal names. It has no official standing whatsoever.
What does the Medusa APU have?
Per Wccftech's reporting, Medusa — the Zen 6 APU line expected before Grimlock — will mix RDNA 3.5, RDNA 4m and RDNA 5 graphics across different models. That makes Medusa the transition generation: older graphics on affordable models, new RDNA 5 on premium ones, as the architecture ramps up. AMD has confirmed none of this.
How reliable is the Kepler_L2 leak?
More reliable than most, but still a single source. VideoCardz notes Kepler_L2's history of revealing unreleased GPU specs down to SKU numbers. But one accurate leaker is still one person: AMD could change plans, cancel the project or rename everything. Reliability of the source is not confirmation of the product.
Where can I read more about RDNA 5 and the GPU market?
Start with our RDNA 5 deep dive for the base architecture, our November 2026 GPU price guide for the memory-crisis pricing context, and our NVIDIA Rubin shipping story for the competitive picture. The Sources section below links every outlet whose reporting this article is built on.
Sources and Related Reading
Every fact in this article comes from the reporting below. AMD has confirmed none of the 2029 details — all of them trace back to Kepler_L2's November 2, 2026 forum post and the outlets that covered it.
- TweakTown — AMD RDNA 5 Plus reportedly planned for 2029, could debut inside Zen 7 Grimlock APUs (November 2, 2026 — the original report: Kepler_L2's quote, Zen 7 up to 32 cores/64 threads, TSMC A14, FOPLP packaging, desktop Zen 7 in 2028, RDNA 3.5 analogy, Sony handheld exclusivity via Digital Foundry)
- VideoCardz — AMD RDNA 5 reportedly coming in 2029 with Zen 7 Grimlock APUs (November 2, 2026 — Kepler_L2's track record down to SKU numbers, prior September 2026 RDNA 5+ reference)
- Wccftech — AMD RDNA 5 GPU architecture 2029, Zen 7 Grimlock APUs (November 2026 — discrete RDNA 5 cards late 2027 to early 2028 delayed by DRAM shortages, Medusa's RDNA 3.5 / RDNA 4m / RDNA 5 mix, discrete RDNA 5+ card "less likely")
- OC3D — AMD plans RDNA 5+ GPU architecture for early 2029 (November 2, 2026 — early 2029 window, first RDNA 5-series graphics on mobile PCs and handhelds, succeeds Strix/Gorgon/Medusa, mass production late 2028 suggesting CES 2029, enhancements unknown)
- Guru3D — AMD RDNA 5+ and Grimlock APU rumor points to a 2029 Zen 7 pairing (November 2, 2026 — the caution: no AMD announcement, no specs given, Zen 7 on public roadmaps does not validate the codename or pairing)
- GameGPU — AMD Grimlock hybrid processors: Zen 7 with RDNA 5+ graphics in 2029 (November 1, 2026 — RDNA 5 from scratch with ray tracing and AI matrix overhaul, "Azure" codename, Transformers naming tradition, TSMC A14 1.4nm, RDNA 3.5 continuing on affordable mobile until 2029)
- TweakTown — FSR SDK 2.4 released with FSR Upscaling 4.1.2 (October 28, 2026 — FSR Upscaling 4.1.2 for Radeon RX 6000/7000 GPUs and Ryzen AI APUs, AMD's neural rendering push)
Related reading from this blog:
The honesty disclaimer, one last time: AMD has confirmed nothing in this article beyond what it has already said publicly about Zen 7 as a future CPU architecture. Every 2029 detail — the RDNA 5+ name, the Grimlock codename, the APU pairing, the dates, the process node — comes from a single leaker's forum post and the press coverage around it. Plans change, codenames get cancelled, and timelines slip. Use this article for what it is: informed analysis of a credible rumor, grounded in the real market forces of late 2026 — not a product announcement.
Comments
Post a Comment