Sunday, November 30th 2025
AMD Lists Ryzen 7 9850X3D CPU with 5.6 GHz Boost Frequency
AMD has launched an official website placeholder for the Ryzen 7 9850X3D processor, which was leaked in October. As a part of the "Granite Ridge" Ryzen 9000 X3D series, AMD is reportedly developing two additional updates to the lineup. One is a flagship model, the Ryzen 9 9950X3D2, and the other is a refined eight-core, 16-thread X3D design. Today, AMD's website confirms the Ryzen 7 9850X3D, featuring 8 cores and 16 threads, 96 MB of L3 cache, and a boost clock up to 5.6 GHz, with a 120 W power envelope. This represents a 400 MHz increase over the Ryzen 7 9800X3D model, which has the same cache and core configuration. AMD is likely using higher-binned dies it has saved over the past few months to offer a better alternative to Intel's upcoming "Arrow Lake Refresh," scheduled for early 2026, and "Nova Lake," scheduled for late 2026.
However, details about the new flagship model, the Ryzen 9 9950X3D2, featuring 16 cores and 32 threads, remain uncertain. This chip reportedly includes 3D V-Cache on both chiplets, totaling approximately 192 MB of L3 cache. It is said to operate at a base frequency of 4.30 GHz, with a boost up to around 5.6 GHz, and has a TDP of about 200 W. The 9950X3D2, with its new naming convention, suggests a tradeoff of slightly lower peak boost speeds (-100 MHz) compared to the single-stack X3D Ryzen 9 9950X3D model, but offers significantly increased cache capacity. Today only the higher-clocked Ryzen 7 9850X3D is confirmed by AMD, so we might have to wait on the dual-chiplet X3D2 SKU a bit more.
Sources:
AMD, via VIdeoCardz
However, details about the new flagship model, the Ryzen 9 9950X3D2, featuring 16 cores and 32 threads, remain uncertain. This chip reportedly includes 3D V-Cache on both chiplets, totaling approximately 192 MB of L3 cache. It is said to operate at a base frequency of 4.30 GHz, with a boost up to around 5.6 GHz, and has a TDP of about 200 W. The 9950X3D2, with its new naming convention, suggests a tradeoff of slightly lower peak boost speeds (-100 MHz) compared to the single-stack X3D Ryzen 9 9950X3D model, but offers significantly increased cache capacity. Today only the higher-clocked Ryzen 7 9850X3D is confirmed by AMD, so we might have to wait on the dual-chiplet X3D2 SKU a bit more.
148 Comments on AMD Lists Ryzen 7 9850X3D CPU with 5.6 GHz Boost Frequency
But enthusiasts keep voting with their wallets telling manufacturers they don't care about power consumption so we have CPU's running past the point of efficiency, same with graphics cards that have melting prone power connectors drawing enough power to heat a room.
A fitting automotive analogy is how lots of car manufacturers these days will put a "sport tuned" suspension on cars with stiff uncomfortable over bolstered seats when 99% of people are driving their car to work or the grocery store, not taking corners on a track at 80MPH. And yeah that braking test is an ironic complaint when 150-0 is a pointless test when all that matters is 60-0 and how safely can a vehicle brake in wet or adverse conditions.
And unless you are purposefully going out, finding BIONIC workloads that use AVX2, and running 0 offset, you are not touching peak power. Most bionic workloads do not use AVX2, those that dont wont hit peak power limits.
And if it is a problem? Lower the peak power limit. You dont need Lisa Su to release a specific CPU for you to do that. Again, mountains and molehills. Just set the power budget to match a 7000 series then go make some tea.How many are there? please, elaborate on how many 5080 buyers use 9900ks? And what that has to do with the data provided by 720p testing?
Guarantee you there are people out there doing that.This is a strawman argument. The existence of poorly optimized software does not negate the results of testing.Learn to plug your GPU in correctly, and if it is that much of a concern, lower the power target. That's what consumers who want lower power parts do, instead of whining they need specific SKUs to meet their needs.
Or, you know, they just buy the T series chips. Nobody is stopping you from buying 35w CPUs today. You're not forced into buying the highest end parts.This is a great example of what I am talking about. There is no shortage of cars with soft suspension and seats with little or no bolstering. Yet people will whine about "sports tuned" cars.
Stop buying the sports packages dummy! Dont buy a Mustang then whine its not a minivan or econobox. Or better analogy, dont buy a Golf R then complain its mileage sucks and the seats have too much bolstering when the normal Golf exists.And you have fallen for the great filter. It wasnt pointless, because the brake systems that could handle 150-0 could ALSO handle 60-0 far better then systems designed just for 60-0. As the British found out the hard way as their terrible cars failed one after another and the Germans at their lunch. All the companies that only designed for legal highway speed and no higher went out of business and saw their assets bought by other companies in other countries, their plants left to decay and their engineers unemployed on the streets.
But thank you for proving my point. Some people just cant see the forest through the trees.
Also perhaps manufacturers should dial down on the CPU running on the edge of its limits and allow users to have some OC room if they care about another 2-3FPS in synthetic benchmarks.The point is people buying a 5080 or 5090 aren't going to be CPU bottlenecked.It isn't and game optimization is a much more serious issue than focusing on a game running at 1000fps at 720P.Ah yes, the "plug in your gpu correctly" cop out to defend the company pushing a defective connector on everyone. Tell that to everyone with a melted or burnt connector who plugged it in correctly.
Only the enthusiasts lower power targets, the average gamer isn't tinkering in the UEFI or undervolting.
Also could you show me where the low power version of a X3D chip is? Oh right one doesn't exist and you're missing the point.If you've looked at cars within the last decade you'd see there aren't many choices out there, everything is a boring crossover with stiff suspension or an overpriced pickup truck, few choices exist for any reasonably priced sports car.British car manufacturers didn't go bankrupt because of shitty brakes, they went out of business for all kinds of quality and reliability issues compared to every other car maker.
The car analogies don't really apply here, you're simply using them to support your opinion.
It's pretty simple I think, at least from a primarily gaming perspective. The a faster CPU is, the higher the FPS ceiling will be in any given game. That could be above what I need in the game I'm playing right now, but so might the 5800X3D I came from. Games are often also smoother and load / deal with traversals better. I want to be as GPU limited as possible as often as possible, so I'll tend to pick the fastest gaming CPU available at the time I upgrade my while system, the relative performance metrics are such an easy way to discern that.
I upgraded to AM5 when the 9800X3D was just coming out, it didn't make much sense to me to settle for the slpwer 7800X3D for not a lot less money, especially when my reusable cooler and comfortably deal with it. Never seen it use 155w either, more like 130w completely unrestrained, but typically 50-80w (usually more toward 50w) when gaming. From what I gather, that's maybe 15% more power and less than 10w on average more than a 7800X3D... Not losing any sleep (or sweating more) there.
I could see people building an AM5 rig when the 9850X3D is out buying it, because it will be the fastest, and I'd give them the nod. Not a useful upgrade from a 9800X3D or probably even a 7800X3D though.
280 AiO + 1000rpm fan + ycruncher vt3 :D
Temperature =! heat.So why cant you UC and sacrifice 2-3 FPS to hit your power limit? Why should manufacturers cater to your whim instead?How is that related to 720p game testing, exactly?Game optimization is a seperate issue. But you;ve already demonstrated you cant understand why the 720p test exists, so et tu, Hecate91.Tens of millions of nvidia GPUs run fien without melting.
Cope and seethe. Post a source for the rate of failure if you are so confident.You're SOOOOO close to getting it.And you lost it. Just undervolt an x3d and stop crying.Honda civic
Hond accord
VW jetta
VW golf
Kia K4
Kia K5
Hyundai elantra
Toyota camrey
Toyota corolla
Mazda 3
The list goes on, and this doesnt include all the simple road cars discontinued because nobody was buying them. STFU about not having choice.
And most crossovers are not "stiff". They are floaty garbage that couldnt handle their way out of a paper bag. Go drive a GLI/GTI or a BMW M series, THOSE are stiff.Of which poor braking and handling were a part. Again, forest::trees.A metaphor is a figure of speech that, for rhetorical effect, refers to one thing by mentioning another. It may provide clarity or identify hidden similarities between two different ideas. Metaphors are usually meant to create a likeness or an analogy.
The fact you get hung up on the car part and cant understand the underlying implications speaks wonders. See the forest through the trees, young apprentice.Some generations are duds. Zen 4 and 5 are very similar, yes there was not much movement. So if you want more efficient....buy zen 4.
In AVX 512, zen 5 is much more efficient.And it really isnt an issue. Sure, it sounds like it, but there are tens of millions of laptops that regularly hit tjmax doing basic tasks that run on and on for years without failure. Hell look at Macbooks and how hot they regularly get.See? There is an efficient CPU for those who want efficiency. Mountains, molehills.TPU has 137k members. Based on Steam's monthly active users, we represent 0.001% of the gaming world.
I'll let you ponder on how realistic those statistics would be, and if the existence or lack thereof in a TPU sig means such things do not exist.
The fact that you are hung up on what the 9800X3D consumes baffles me.
It is stronger than a highly tuned 5900X in R23.. that is a big deal.
I would personally avoid paying money in real life for a card with the connector as long as possible because 8 pin and 6+2 works fine and doesn't have to be babysat, avoid bending, or require use of a special GPU included cable.
I haven't said there isn't a problem, obviously there is. But lets be frank, most of us are not running 4090s or 5090s. Those have always been on the esoteric side of the spectrum.
Also probably because benchmarkers tend to never adjust / compensate for differences in power consumption, and that might make 144W PPT AMD SKUs look worse than 253W PL2 Intel ones (at least to laymen).
That's just my two cents.
One of my friend doing comp-engineering has the 'screw the TDP just gimme more performance' mentality and he doesn't mind the 575W on the 5090 at all. If anything he'll push the power limit even higher... at least, that's his words.I wonder if this problem can be solved by simply placing the ECO setting in the OS, e.g., Ryzen Master or hell even Radeon Adrenaline, so even the less technical ones get to choose the power portfolio they prefer.
Or just actually fix the damned Windows to let it have the ability to switch between power plans automatically (e.g., no boost at desktop or after I/O inactivity for x minutes). Which they'll never do because it's Microsoft.
> Spent an industrialized nation-worth of power to train GPT models.
> Pushes CoPilot and AI slop to all copies of Windows.
> Fails to even notice an 8-hour long idle and enable core parking and use powersave schedulers like Android and linux did a decade ago with plain old human-written C++.
To illustrate, if I do a 95 W TDP (128 W PPT) on my 7950X3D, that'll keep it nice and cool under my NH-U9S in all-core work, but it'll still boost to the moon and reach Tjmax when I only have 4-6 cores active. I'm that case, those 4-6 cores get all of the 128 W headroom, which is too much heat density for the little U9S.
I think AMD should adopt Intel's way of managing boost on a per-active core count-basis and allow independant timers for different PLs; it always bugged me that I have to choose a single frequency and power limit for all cores and workloads.I just thought of a probably horrible idea of specifying in PBO a TDP and EDC limit of 95W/128+A, in order to limit the power under high-voltage (e.g., high boost) scenarios but still allow all-core (low-voltage) to behave somewhat well. I pulled this straight out of my head and this might be a wildly bad idea, though.
Now they're no longer the cash-strapped AMD ten years ago maybe they'll start hiring more for marketing; and I hope to the earth with that money they do not instead go pay a bunch of executives to do vendor lock-ins and extracting more margins a.k.a. Skylake-era Intel.
Lisa Su seems to have ran AMD well in that it's not as much a corporate ar*hole as the other players (Intel/Nv/Qualcomm).
Hmmm... Let's do some comparing and see what 350-400 dollars worth of GPU gets you
Oh.. okay :)
Now... thought experiment, how many bottom-end GPUs like these do you think you can run off a 9800X3D? I reckon you might stretch to 3-4 upgrades and still have CPU oomph to spare.
So expensive... Your comment kinda shows how poorly people actually look at what they buy. Over 20 years of PC gaming I really can't say this is an expensive hobby. I've spent more on Magic cards in the last 3,5 years than I spent on PC gaming in 15 years... with Magic spending at 4.7K EUR since 2022, 2,7K invested and 2K resold. All my gaming PC expenses on hardware amount to 3,5-4K at best - over the course of many more years, as well, and trading/buying up to nearly the top of the performance stack. PC Gaming is dirt cheap.
I'm gonna spend the rest of the day trying to find a proper facepalm pic for this situation. Wow.
The only correct response here is that we should educate.
As you can see, depending on scene and content, you can get a radical FPS impact even on last gen 4090 which is today, a 5080, and next gen is a 6070 (well within range for most), which is when the gap between a 7800X3D and a 9800X3D (two top end gaming CPUs) will make itself known quite well. The Stalker 2 video already shows you a gap of 20+ FPS, and those are frames within the range of 'I kinda do want them', ergo 70-120 FPS.
The issue with people reading CPU reviews is exactly the sentiment that @Harun8 presented as well: a short sighted view on what the PC should do today. You don't buy a platform and CPU for one or two years and a single GPU, I hope - and if you do, you're doing it wrong.