Notebooks Powered by Ryzen 9000HX and 9000HX3D "Fire Range" Available From March-April

[btarunr]

One of AMD's key announcements at its 2025 International CES keynote was the Ryzen 9000HX line of processors, codenamed "Fire Range." A successor to the 7000HX "Dragon Range," this is essentially a BGA package of the "Granite Ridge" MCM, which combines one or two "Zen 5" CCDs with a client I/O die. The processor lacks an NPU and has a basic iGPU, but is meant for enthusiast segment gaming notebooks and portable workstations, as it's meant to be paired with discrete GPUs, taking advantage of the package's lavish 28 PCIe Gen 5 lanes. The Ryzen 9 9955HX3D stands out in the series as the only model with 3D V-Cache, making it possibly the fastest mobile processor for gaming notebooks—faster than even the "Arrow Lake-HX," given how the desktop 9800X3D compares to the Core Ultra 9 285K at gaming.

Notebook OEM Dream Machines put out a press release which specifies that the first notebooks powered by the 9955HX3D will ship either toward the end of March, or early-April 2025. "Fire Range" is known to have scored design wins from several popular notebook OEMs, which means the chips could have a good run at the markets this generation. Notebooks powered by the 9955HX3D and discrete GPU options that include the GeForce RTX 5090 Laptop GPU and the RTX 5070 Ti Laptop GPU, are expected to be priced around €2,530 and €3,860, respectively.



View at TechPowerUp Main Site

 

[rhqq]

I'm getting increasingly confused with the naming convention of the AMD processors, especially mobile ones. How does it stand against "ai" chips, like 395 Pro+ ? and the rest of CES announced lineup?

 

[3valatzy]

I'm getting increasingly confused with the naming convention of the AMD processors, especially mobile ones. How does it stand against "ai" chips, like 395 Pro+ ? and the rest of CES announced lineup?

Compare the specs. The 395 is 5.1 GHz max.

as it's meant to be paired with discrete GPUs, taking advantage of the package's lavish 28 PCIe Gen 5 lanes

Yeah, but 0% market share for Radeon. This is reserved for Nvidia 100%.

 

[Daven]

I'm getting increasingly confused with the naming convention of the AMD processors, especially mobile ones. How does it stand against "ai" chips, like 395 Pro+ ? and the rest of CES announced lineup?

The three digit models have NPUs and/or fast integrated graphics. The four digit models are just power limited equivalents of the desktop Ryzen chips without an NPU and meant to be paired with discrete mobile graphics.

Here is my favorite reference of all Ryzens in one place:

List of AMD Ryzen processors - Wikipedia

en.m.wikipedia.org

 

[rhqq]

Compare the specs. The 395 is 5.1 GHz max.

yeah, but which zen generation it is, the performance is a dice roll really.

The three digit models have NPUs and/or fast integrated graphics

7840 also has npu. there's little to no pattern at this point.

39x and 38x have better GPUS, but not the 36x and 37x... so it's not valid either.

 

[SL2]

The three digit models have NPUs and/or fast integrated graphics. The four digit models are just power limited equivalents of the desktop Ryzen chips without an NPU and meant to be paired with discrete mobile graphics.

Or, APU's and CPU's, respectively.

7840 also has npu. there's little to no pattern at this point.

It is, when comparing everything launched since Strix point last summer. I'm not saying it's not a mess, tho..

but not the 36x and 37x... so it's not valid either.

Not better GPU's than what?

 

[Daven]

7840 also has npu. there's little to no pattern at this point.

39x and 38x have better GPUS, but not the 36x and 37x... so it's not valid either.

I'm referring to the current generation. And products within a product family will have different performance levels. That has always been the case. Just because the Ryzen 9600X is slower than the 9950X doesn't mean it's confusing. If you use the wikipedia link above, you will see the following:

340 4 CUs
350 8 CUs
365 12 CUs
370 16 CUs
375 16 CUs
MAX 380 16 CUs
MAX 385 32 CUs
MAX+ 395 40 CUs

Numbers go up and so does the CUs for the most part. Also look for differentiators in the model name like 'MAX' and 'MAX+'.

 

[3valatzy]

yeah, but which zen generation it is, the performance is a dice roll really.

I don't think so. The Zen generation unless original Zen vs. Zen 5 makes little difference.
Look at the caches sizes, then.

 

[SL2]

AMD came up with a new system for mobile APU model numbers "because everyone hated it" (at 11:25). This culminated with the 7000 series, but even before that there was Zen2 and Zen3 mixed in the 5000 mobile APU series.

Now it's easier to follow, maybe, but people hate all the AI, PRO, MAX and + instead lol

 

[rhqq]

ok, so it seems they have 9xxx numbering for zen5 with rdna2 igpu, while 3xx for zen5 with rdna3.5. and then they use x3d cache only for the 9xxx, which is a questionable decission. i'm surprised they didn't put it into the max+ 395 pro whatever, which is the top-tier gaming chip from their lineup - and 3d cache would make perfect sense in the arrangement and target group they have there...

 

[SL2]

ok, so it seems they have 9xxx numbering for CPU's, while 3xx for APU's.

Fixed

i'm surprised they didn't put it into the max+ 395 pro whatever, which is the top-tier gaming chip from their lineup

How do you know it would be needed in an IGP and 2.66 times faster CPU (or unified really) RAM? Edit: I mean bandwidth.

9950HX3D:
Chiplets-IOD-----------------------------------GPU

395:
Chiplets-IOD&GPU

I've seen this over and over again: Lets throw V-cache at everything!!!

 

[rhqq]

Fixed

How do you know it would be needed in an IGP and 2.66 times faster CPU (or unified really) RAM? Edit: I mean bandwidth.

9950HX3D:
Chiplets-IOD-----------------------------------GPU

395:
Chiplets-IOD&GPU

I've seen this over and over again: Lets throw V-cache at everything!!!

technically they're all APUs, as all of them have igpu. I know, there''s a blurry line between, but I wouldn't call the 360 an APU by any stretch, given how little GPU power it has. only the 39x series are decent (roughly around rx7600 ?xt? ).
And I sustain my point on 3D cache on the top tier ones. They're clearly targetted at gaming, and historically chiplets for the 3d cache ones are higher bines with lower voltages making them always more efficient. Given these are mobile chips, they're prebinned for better efficiency, so all that's left is slapping the cache on top (well, bottom actually). side note - i know it is not "simple", but yeah, half way there regardless.
Now based on all the recent interviews I expect that's what going to happen pretty soon. Just few years ago we heard the 3D cache will be only for desktops, now they're having a SKU for mobile, and I feel like that trend will extend further down.

Now the efficiency - given iGPUs are usually hurt most by the performance of the non-v ram, having 3D cache would free up more transfers for the GPU - and that could yeld visible gains for the overall package.

 

[Cheeseball]

9955HX3D will probably go into an Alienware or ASUS ROG SCAR again, and probably won't pair it with their own discrete graphics, which seems to only be the RX 7800M and not the RX 7900M which was in one failed Alienware desktop replacement model that had possible hardware issues, at least thats the main reason why I returned mine.

 

[Random_User]

I'm referring to the current generation. And products within a product family will have different performance levels. That has always been the case. Just because the Ryzen 9600X is slower than the 9950X doesn't mean it's confusing. If you use the wikipedia link above, you will see the following:

340 4 CUs
350 8 CUs
365 12 CUs
370 16 CUs
375 16 CUs
MAX 380 16 CUs
MAX 385 32 CUs
MAX+ 395 40 CUs

Numbers go up and so does the CUs for the most part. Also look for differentiators in the model name like 'MAX' and 'MAX+'.

Sorry for pointing this out. There's MAX 390 with 32 CUs either (Radeon 8050S).

 

[SL2]

but I wouldn't call the 360 an APU by any stretch, given how little GPU power it has. only the 39x series are decent

The definition for an APU isn't if it's decent. Crappy APU's are still APU's, and crappy APU's has been around for a long time. It's not really just about GPU performance either, an APU usually uses less power, especially in idle, and have fewer PCIE lanes.

Also, decent? You draw the line at 39x, as in including the 12 core (TWO chiplets), but excluding the 8 core? Um, yeah no, that's an odd choice for a gaming laptop. The point isn't that it will run slower, but it will cost more, and use more power (40% more in desktop counterparts, who knows what difference we'll see here), while still having the same IGP.

The 16 core have the obvious advantage of having a better IGP.

And I sustain my point on 3D cache on the top tier ones.

You still don't know if it's worth it with an IGP, given the radically different layout I've already explained.

They're clearly targetted at gaming,

Not exclusively tho, they're equally targeted at work, if not more so. Just beacuse forum members here and there dreams about owning a SHalo for gaming doesn't mean it dictates the target market. Macbook pro is their biggest competitor, and it's not really the most common gaming laptop. SHalo will probaly be pretty good in games, but that's not where the money is.

The <128 GB unified RAM makes them useful for a lot of things that needs more than 16 GB VRAM in a laptop.

 

[rhqq]

The definition for an APU isn't if it's decent. Crappy APU's are still APU's, and crappy APU's has been around for a long time. It's not really just about GPU performance either, an APU usually uses less power, especially in idle, and have fewer PCIE lanes.

Also, decent? You draw the line at 39x, as in including the 12 core (TWO chiplets), but excluding the 8 core? Um, yeah no, that's an odd choice for a gaming laptop. The point isn't that it will run slower, but it will cost more, and use more power (40% more in desktop counterparts, who knows what difference we'll see here), while still having the same IGP.

The 16 core have the obvious advantage of having a better IGP.

You still don't know if it's worth it with an IGP, given the radically different layout I've already explained.

Not exclusively tho, they're equally targeted at work, if not more so. Just beacuse forum members here and there dreams about owning a SHalo for gaming doesn't mean it dictates the target market. Macbook pro is their biggest competitor, and it's not really the most common gaming laptop. SHalo will probaly be pretty good in games, but that's not where the money is.

The <128 GB unified RAM makes them useful for a lot of things that needs more than 16 GB VRAM in a laptop.

yes but no but yes.

I don't disagree per se, however iGPUs in APUs were supposed to provide decent value. by "real" APUs I only meant the 80x0s models. Otherwise we need to call all of the latest AMD processors "APUs" as pretty much all of them have iGPUs, and for now it seems like that's going to stay - yet everyone call's them CPUs. So by the nature of the language evolution - it's what we commonly consider the meaning of the word is the definition of the word (well, acronym in this case).

MacBook example is not a good one, at least in the gaming context, as the biggest issue is the metal api, that apple is insisting on - and lack of official vulkan support is what really hurts gaming on the idevices. The biggest selling point of the 395+ chips is the ability to assign 96gb out of 128gb of RAM to the GPU for LLMs and whatnot. Sure, that's nice. But other than testing/developing stuff, nobody is going to use that in production. Training of the models is going to stay in the servers area. Same with rendering - unless you need a mobile/portable station to do so at the customers location - sure, otherwise everything stays in the rendering farms for bigger companies and workstations with dedicated GPUs in smaller companies. All that together is relatively small group - especially - considering the devices that have been announced. Only the HP zbook ultra is looking decent. Otherwise we're seeing a super thin asus (which will rock in gaming), a TABLET and a series of tiny PCs. So it looks like the target group is smaller than potential gamers - or feels like (opinion!)

Anyhow - based on my gut feeling supported by almost two decades of system optimization worth of experience (read: an opinion) - I think 3D cache would improve gaming performance in this massively integrated APU and would make this chip way more attractive to more gaming laptops manufacturers.

 

[SL2]

I don't disagree per se, however iGPUs in APUs were supposed to provide decent value.

There's absolutely nothing "decent value" with Strix halo if pricing of Strix point is anything to go by. Well, yeah, besides the unique amount of (V)RAM.

I've seen a video where they claim SHalo ZBook starts at $1400 which is ridiculous given what SPoint Elitebooks cost ($2800 - 4300). No one in their right mind can believe that.

by "real" APUs I only meant the 80x0s models.

That's a bit like calling the RTX 4090/5090 the only real GPU.

At first you seem to think that 2x6 cores is a much better choice than 1x8 cores in an AMD laptop for games, then you changed it to 8000S IGP's which includes an 8 core. I don't know why I keep on replying..

Otherwise we need to call all of the latest AMD processors "APUs" as pretty much all of them have iGPUs

This doesn't make sense. Need? This keeps you up at night, right?

The fact that most AM5 CPU's have an IGP seems to have messed with your brains. Those are still not APU's, and I've still never seen a Dragon/Fire range laptop without discrete graphics.

CPU's have chiplets originally developed for EPYC, and APU's are primarily made for laptops, though not exclusively. Strix halo does have chiplets, but those are different from Granite ridge.

MacBook example is not a good one, at least in the gaming context, as the biggest issue is the metal api, that apple is insisting on - and lack of official vulkan support is what really hurts gaming on the idevices.

Wow, yeah I've never said it was good for games, I was pointing out the opposite.

I think 3D cache would improve gaming performance in this massively integrated APU and would make this chip way more attractive to more gaming laptops manufacturers.

Sure, for attractiveness, I bet it would sell, as in "Lets throw V-cache at everything", but that's a different point.

But other than testing/developing stuff, nobody is going to use that in production.

Work != production. You're the one bringing up production lol.

You have an idealistic approach to how this kind of work is done, it's all servers. Meanwhile, people are doing this with their own HW, including multiple Mac minis, or multiple RTX 4090's in the same system.
Just google it. I won't try to guess how large that market is, but it's evident that people are prepared to spend money here.

Also, Strix halo is both laptop and desktop according to AMD. HP have already shown one model.