AMD Focused on Delivering RDNA 4 to Desktop, Mobile a Secondary Concern

[T0@st]

AMD's small portfolio of current-gen (RDNA 3) mobile-oriented Radeon dedicated GPUs pales in comparison to a wide selection of related desktop offerings—a certain demographic of PC gamers have wondered whether the incoming RDNA 4 generation will produce more options for portable platforms. An extensive Notebookcheck article indicates that Team Red is not (immediately) interested in catering to mobile dGPU enthusiasts—the online publication conducted an interview with Ben Conrad, director of product management (client-side). The "Navi Mobile" Radeon RX 7000M range is an uncommon sight on gaming laptops—relative to NVIDIA's wide rollout of dedicated GeForce RTX 4000 Mobile GPUs—normally, higher-end models are present on ultra-expensive specification sheets (paired with "Dragon Range" Ryzen HX CPUs). Industry experts believe that lower-end options are more likely to turn up inside external enclosures.

One of Notebookcheck's questions focused in on this topic—they believe that: "the number of AMD dGPU-based laptop SKUs have been pretty anemic." Their interviewee was ambushed with a query regarding his company's outlook for mobile RDNA 4 options. In response, Conrad stated: "our current graphics strategy is focused on the desktop market with RDNA 4. So, I think you'll see those types of products first in the future. Certainly, RDNA 4 and future graphics technologies will make it into mobile, whether they be on APUs or future products." VideoCardz has read "between-the-lines" and posits that Team Red could skip a generation—UDNA is possibly a better fit for a new wave of laptop dGPUs. A sort-of stopgap has appeared on the horizon—in the shape of AMD's forthcoming "Strix Halo" RDNA 3.5-based integrated solution. The flagship chip's Radeon 8060S iGPU looks promising when compared to a current-gen dGPU, but it will likely struggle when pitted against Team Green's "Blackwell" dedicated mobile platform. Upcoming competition in the APU field will arrive in the form of Intel's "Panther Lake" processors—slated for launch later this year. Its next-gen iGPU is said to utilize the Xe3 "Celestial" architecture.



View at TechPowerUp Main Site | Source

 

[TheinsanegamerN]

This is probably a good thing, given how little traction AMD gets in mobile dGPUs, but at the same time withdrawing from yet another market while not giving us anything concrete and playing coy with actual numbers is not inspiring confidence either.

 

[Wasteland]

If this is what "focused" looks like, then good grief.

 

[Daven]

Basically AMD is covering the laptop midrange and budget just like on desktops with Strix Point and Strix Halo. Discrete laptop GPUs are not required.

 

[CosmicWanderer]

Mobile "dedicated" I assume. I'm guessing mobile integrated RDNA4 will make an appearance.

Makes sense if that's the case, given how powerful AMD's integrated graphics has become.

 

[Bomby569]

they are doing their best to become irrelevant

 

[HOkay]

If this is what "focused" looks like, then good grief.

This made me chuckle! You're so right though, the last word I'd use for the 9000 series launch so far is "focused".

 

[hsew]

Unless AMD wants to offer a 7800M laptop for $1000, they ought to just bow out of the laptop dGPU market and focus on desktop and Strix Halo. nVidia is easily at 98-99% laptop dGPU market share at this point.

 

[qlum]

Appart from how entrenched Nvidia has been, on mobile efficiency is king, and AMD's dedicated gpu's have not been winning any prices here, so it makes sense that they don't really try.

 

[trsttte]

I think that when we get benchmark numbers for Strix Halo everyone will rethink dGPU on laptops, it just won't make any sense.

 

[Visible Noise]

Unless AMD wants to offer a 7800M laptop for $1000, they ought to just bow out of the laptop dGPU market and focus on desktop and Strix Halo. nVidia is easily at 98-99% laptop dGPU market share at this point.

Then they just eliminated themselves from 90% of the corporate market.

I think that when we get benchmark numbers for Strix Halo everyone will rethink dGPU on laptops, it just won't make any sense.

It’s going to be very very expensive. It’s a niche product, not an everyone product.

 

[trsttte]

It’s going to be very very expensive. It’s a niche product, not an everyone product.

The Asus ROG Flow tablet (it doesn't get more niche than that) is going to cost 2200$, a laptop could start bellow 2000$ which is the starting price of any decent sized workstation. It won't compete with a 4070 equiped regular laptop but it will be very interesting for thinner machines.

 

[hsew]

Then they just eliminated themselves from 90% of the corporate market.



It’s going to be very very expensive. It’s a niche product, not an everyone product.

Corporate market? Care to explain? Because I worked for several corporate enterprise and state and federal government IT departments for nearly a decade and never saw a single AMD laptop dGPU in the thousands of laptops I encountered over the years… of the rare workstation-class laptops I encountered, they were always equipped with an nVidia GPU.

AMD’s own website lists nine laptop offerings for their RX7000S/M(XT) lineup. Nine. Their dGPUs just aren’t competitive in the laptop sector and haven’t been for over a decade if we’re being honest. The performance and VRAM advantage AMD enjoys in the desktop dGPU market does not exist in the laptop space unfortunately. Hence why I said they’d have to offer a 12GB laptop dGPU to compete with the 8GB 4070M soon to be 8GB 5070M.



Strix Halo can’t be too expensive. I’ve heard some speculate it to start around $2200 for 32GB RAM, but if AMD were serious it’d have to start around $1600-1700. They can’t keep overvaluing their brand. No way is anyone going to spend $2200 on an entry-build 120W Strix Halo laptop with 32GB RAM to trade blows with a 5060 at best, when they can instead get a 5080 laptop for the same price, or even a nicely equipped 5070Ti laptop for slightly cheaper. No way José.

 

[Visible Noise]

Corporate market? Care to explain? Because I worked for several corporate enterprise and state and federal government IT departments for nearly a decade and never saw a single AMD laptop dGPU in the thousands of laptops I encountered over the years… of the rare workstation-class laptops I encountered, they were always equipped with an nVidia GPU.

It appears you are correct. My prior HP laptop had a AMD GPU in it, but it seems they have been discontinued.

 

[watzupken]

In my opinion, you can’t just focus on desktop GPUs and leave the mobile segment. It’s a dumb strategy simply because the desktop market have been shrinking and mobile segment growing. So by leaving their iGPU on RDNA 3, they are essentially ceding this segment to Intel and Nvidia. Even Intel is gaining on them with their Arc Battlemage. AMD is just fortunate Intel needs more time to further refine their drivers. For once, I think Intel’s ARC looks like a good alternative to very overpriced Ryzen AI 9 that comes with the best iGPU.

 

[TheinsanegamerN]

In my opinion, you can’t just focus on desktop GPUs and leave the mobile segment. It’s a dumb strategy simply because the desktop market have been shrinking and mobile segment growing. So by leaving their iGPU on RDNA 3, they are essentially ceding this segment to Intel and Nvidia. Even Intel is gaining on them with their Arc Battlemage. AMD is just fortunate Intel needs more time to further refine their drivers. For once, I think Intel’s ARC looks like a good alternative to very overpriced Ryzen AI 9 that comes with the best iGPU.

Their leaving the mobile dGPU market, not the whole mobile market. FFS We used to teach reading comprehension at an elementary level.

Their iGPUs are NOT rDNA3 anyway. They are rDNA3.5. These iGPUs have 0 competition. Intel's Arc based iGPUs are extremely rare and in games struggle against the 680m, the 780m leaves it in the dust and the newer models have upgraded to 16CU from 12, only further widening the gap. We havent even gotten to strix halo yet. Nvidia currently has NOTHING in this space, that may change with the rumored ARM laptop chip, but nothing is official yet.

In APUs, AMD is competitive. In dGPUs, they are nowhere to be seen, and the 7000m series was so lackluster they may as well not have bothered. Most likely, they are skipping rDNA4 for iGPUs because they're all about RT, and the iGPU is too restrictive to make use of the tech. They will likely skip to uDNA once that is ready.

 

[hsew]

In my opinion, you can’t just focus on desktop GPUs and leave the mobile segment. It’s a dumb strategy simply because the desktop market have been shrinking and mobile segment growing. So by leaving their iGPU on RDNA 3, they are essentially ceding this segment to Intel and Nvidia. Even Intel is gaining on them with their Arc Battlemage. AMD is just fortunate Intel needs more time to further refine their drivers. For once, I think Intel’s ARC looks like a good alternative to very overpriced Ryzen AI 9 that comes with the best iGPU.

I’d agree that AMD messed up by introducing RDNA4 and FSR4, but then eschewing those advancements in Strix Halo when Intel is nipping at their heels with Battlemage. It’s crazy to think that Intel are presenting a serious threat to AMD in the GPU space; nobody would have seen this coming even 5 years ago.

But I’d disagree with your premise: by building bigger iGPUs, Intel and AMD are not actually ceding the high-performance laptop graphics market to nVidia. They are simply changing their approach. Remember, nVidia already owns basically 99% the high-performance notebook graphics sector, so Intel/AMD aren’t ceding any ground there because there really isn’t any ground left for them to cede.

Strix/Lunar Halo are the new/old approach of tech consolidation: cramming more performance and functionality into a single chip. From the Intel/AMD perspective, it allows them to advertise those SOCs in higher-end markets. From the OEM perspective, it presents the attractive proposition of reduced complexity vs having to design a notebook with a dGPU. So maybe they don’t quite achieve 4070M performance. But who cares? It turns out, x70 and higher dGPUs represent an increasingly small size of the market anyway! It’s the x60 segment and below where most customers are, and that’s exactly what needs to be targeted for any real market share to be recovered from nVidia.


And, bigger picture, even with AI, it is highly likely that the mobile dGPU market as a whole will begin to experience a decline in the coming years if it hasn’t already. Besides the fact that technology has a well-established tendency to consolidate, which we have already seen for years with APU/iGPU, the majority of the most actively played games are either several years old or not graphically demanding to begin with. That leaves professional/creative work, which comes with the general preference of those markets for sleeker, more professional looking devices over edgy and RGB. They, for the most part, are happy accelerating their workloads with iGPU, especially when they’re away from their desks.


With all that to consider, investing in iGPU is exactly the correct move to make. It’s also why nVidia themselves are getting ready to jump into the SOC market. That, and, the very real threat that Intel and AMD could simply choose to lock nVidia out of their laptop ecosystems at any point in the future (Apple, is that you?).

 

[Random_User]

Let me put some input...

AMD mobile dGPUs are irrelevant. There were so few of them, and so weak performance. So this is completely fine, to drop these, since the laptop makers will stick with intel/nVidia, AMD/nVidia anyway. What they better to focus on, is power efficient laptops, with powerful iGPU instead, further improving them. And release iGPUs with RDNA5 iGPUs ASAP along with their newest CPU parts, instead of sticking with outdated solutions, like RDNA3, and RDNA3, like they did with Vega, with Zen 3.

After all, the Strix Halo, IMO, is the beginning of "fruition" of "fusion" idea, they were bearing all these years, and why they've bought ATi, back in the day. AMD simply has no other way, but focus on increasing the amount of these things, right now, and to make sure it has the sane price. Either to make off the money, they've put into it's R&D, and also to their overall "fusion" idea was correct. AMD has went a long way, to prove the APU idea that is capable of gaming is possible.
Now, here comes the most important part: they just have to get enough allocation, and price them properly, to outsell every low end mobile dGPUs, and even desktop dGPUs. Otherwise, they will lose, not only the high and mid end dGPU space, but the entire iGPU market as well. Because AMD's presence in both mobile and APU space is just barely existant.
This is where, IMHO, AMD should put a lot of efforts and money. Because dGPUs would inevitably sink into oblivion, as more compact solutions will come from all GPU and CPU vendors. APU is the future. Better to keep, and increase/improve this advantage, they have now. As the competition will be dire. Especially. the moment, intel get their issues sorted out with Battlemage, and nVidia will release their very own custom ARM-based CPU (IMHO, Mediatek joint is just temporary solution).

 

[trsttte]

I’ve heard some speculate it to start around $2200 for 32GB RAM

That's just the price of the Asus ROG Flow tablet/convertible. A regular laptop, especially one without the ROG branding, will be considerably cheaper.

Their dGPUs just aren’t competitive in the laptop sector and haven’t been for over a decade if we’re being honest

That doesn't tell the hole story, if we go back a decade neither were their CPU's and the company was almost wiped out (kind like Intel now come to think of it, except AMD didn't have other assets like fabs and existing market and mind share to lean on). There are also a ton of back door agreements and supply constraints preventing AMD solutions - be it CPU or GPU - for making a dent in the laptop market.

Intel is nipping at their heels with Battlemage. It’s crazy to think that Intel are presenting a serious threat to AMD in the GPU space

Except they aren't really, Battlemage was well received and offered a decent improvement over Alchemist but it's not even competing at mid higher range, they went straight for the lower end of the market to try to gain market share to justify the investment because anything higher was not worth it. Which is what AMD will do with RDNA4 as well except that they can target more demanding consumers in the 500-750$ space

Strix/Lunar Halo are the new/old approach of tech consolidation: cramming more performance and functionality into a single chip. From the Intel/AMD perspective, it allows them to advertise those SOCs in higher-end markets. From the OEM perspective, it presents the attractive proposition of reduced complexity vs having to design a notebook with a dGPU. So maybe they don’t quite achieve 4070M performance. But who cares? It turns out, x70 and higher dGPUs represent an increasingly small size of the market anyway! It’s the x60 segment and below where most customers are, and that’s exactly what needs to be targeted for any real market share to be recovered from nVidia.

Both Strix Halo and Lunar Lake are modular. I think they're simply finally designing a solution to deliver efficient laptops with decent GPU raw power which the market has been begging for since forever. Discrete GPUs on laptops are always very inneficient but previous APUs never offered particularly high GPU performance. This bridges that gap, between an ultrabook with ok battery but bad performance, and bigger laptop with more performance but loud hot and with low battery.

A 4070 laptop might be cheaper and might even win on some tests but it won't be as small, as quiet and it's battery won't last half as long. That's what this new crop of APU's is aiming to solve.

 

[hsew]

That's just the price of the Asus ROG Flow tablet/convertible. A regular laptop, especially one without the ROG branding, will be considerably cheaper.



That doesn't tell the hole story, if we go back a decade neither were their CPU's and the company was almost wiped out (kind like Intel now come to think of it, except AMD didn't have other assets like fabs and existing market and mind share to lean on). There are also a ton of back door agreements and supply constraints preventing AMD solutions - be it CPU or GPU - for making a dent in the laptop market.



Except they aren't really, Battlemage was well received and offered a decent improvement over Alchemist but it's not even competing at mid higher range, they went straight for the lower end of the market to try to gain market share to justify the investment because anything higher was not worth it. Which is what AMD will do with RDNA4 as well except that they can target more demanding consumers in the 500-750$ space



Both Strix Halo and Lunar Lake are modular. I think they're simply finally designing a solution to deliver efficient laptops with decent GPU raw power which the market has been begging for since forever. Discrete GPUs on laptops are always very inneficient but previous APUs never offered particularly high GPU performance. This bridges that gap, between an ultrabook with ok battery but bad performance, and bigger laptop with more performance but loud hot and with low battery.

A 4070 laptop might be cheaper and might even win on some tests but it won't be as small, as quiet and it's battery won't last half as long. That's what this new crop of APU's is aiming to solve.

They need to be cheaper. Considerably cheaper. Like starting at $1600.

The difference between Intel and nVidia in the laptop space is, even in the early 2010s, Intel had a history of threatening OEMs who wanted to work with AMD and a history of bullying those OEMs into dedicating all their premium designs to Intel chips. It’s not clear if nVidia was doing the same thing. In fact, they likely weren’t, since even a decade ago they had probably close to 90-95% of the high performance laptop market anyways and saw no real threat from AMD in that segment, and that they would have nothing to gain by restricting OEMs from pairing nVidia dGPUs with AMD CPUs.

The comment I made regarding Battlemage “nipping at AMD’s heels” was specifically regarding the iGPU and low-mid level dGPU markets. I pointed to the fact that even this upcoming ultimate APU from AMD won’t be equipped with their latest RDNA4 architecture, which is a real disappointment. Not only will it not be the best possible performance from AMD, it won’t even be compatible with their soon to be released FSR4. It’s a form of stagnation at the wrong time no matter how you cut it. Their upcoming RDNA4 dGPUs are a different story, or at least are HOPEFULLY a different story if they have the right performance, and efficiency for a good price. But even then they may be too far gone. It’s unfortunate but we’re almost at the point where a 25% performance advantage is necessary to get customers to look away from nVidia.

And finally, the Halo APUs are modular, but in all honesty they are a bigger deal to Intel/AMD than to actual customers because they are basically the only real avenue either has to break into the high performance mobile graphics sector at this point. The manufacturers might also appreciate the reduced complexity of an SOC vs CPU+GPU, too. But nVidia is just too dominant. And you can bet money that they aren’t resting on their laurels. They see the Halo SOCs as a real threat to their dGPU ecosystem, since Intel/AMD could just lock nVidia out of their platforms. That’s why nVidia is working overtime to get into CPU.

As for power consumption, 120 Watts is 120 Watts. It doesn’t matter if it’s a single chip or multiple chips. Battery life will be impacted the same. In 2025, I fully expect nVidia, who have been doing Optimus for a decade and a half now, to have perfected it.

 

[trsttte]

As for power consumption, 120 Watts is 120 Watts. It doesn’t matter if it’s a single chip or multiple chips. Battery life will be impacted the same. In 2025, I fully expect nVidia, who have been doing Optimus for a decade and a half now, to have perfected it.

That hasn't been my experience, Optimus is not optimal at all. To be fair the blame migth be more on windows than on nvidia but I haven't had a laptop where this configuration worked well.

 

[HeggeBegge]

Most of you are missing out on the basics. Whether SoCs like Strix Halo will become more frequent over time, is probably a very good guess, but they will not replace dGPUs in laptops for some time to come. Even the new AI MAX+ 395 APU is still only on half the performance of a real 4070, because it can only operate at 65W total. The 4070 can alone draw 115W, and when taking that into consideration, there is still a long way to go to get to Apple levels of power consumption and performance. The benchmarks from AMD are simply disingenuous. For fans of lightweight designs with a fat budget and only doing light gaming, it is a thing, but not for the wider audience.

And meanwhile, the market for laptop dGPUs (counted in sold units, not value) is now bigger than the desktop market. Leaving walk over on the chance of gaining huge market shares, where nVidia is totally dominant, is just a lousy business decision.

6850M XT and 7900M are both well-performing high-end cards in the last two generations, that just didn't gain traction. Rumor has it that AMD are simply much worse for the laptop manufacturers to work with when integrating their cards. That is something they can improve on, instead of just letting nVidia have it all. Maybe there is malpractice from nVidia to some degree, but we should have heard something if it was widespread.

BTW. Has anybody heard anything about whether Intel will launch a laptop version of their Arc Battlemage? They did with Arc Alchemist and it was a total disaster, so I'm not expecting anything. Still it would be interesting to know with some level of credibility, such as this article.

Edit: I found a source. Intel will not enter the laptop GPU market.

JayKihn’s reply also points to the original poster mixing up “Intel’s commitment to laptop dGPUs and desktop dGPUs”. Intel has reportedly canceled all plans to launch discrete laptop Battlemage cards, and it appears Arc Celestial and Arc Druid will follow the same path.

Intel Arc Battlemage is not the end as Arc Celestial and Arc Druid desktop GPUs are reportedly still under development

Despite many reports to the contrary, Intel is reportedly not giving up on its desktop discrete GPU dreams. The company is gearing up to follow the next-gen Arc Battlemage boards, which will be available to buy soon, with Arc Celestial in 2025 or 2026. The Arc Celestial GPUs are expected to be...

www.notebookcheck.net

 

[hsew]

Most of you are missing out on the basics. Whether SoCs like Strix Halo will become more frequent over time, is probably a very good guess, but they will not replace dGPUs in laptops for some time to come. Even the new AI MAX+ 395 APU is still only on half the performance of a real 4070, because it can only operate at 65W total. The 4070 can alone draw 115W, and when taking that into consideration, there is still a long way to go to get to Apple levels of power consumption and performance. The benchmarks from AMD are simply disingenuous. For fans of lightweight designs with a fat budget and only doing light gaming, it is a thing, but not for the wider audience.

And meanwhile, the market for laptop dGPUs (counted in sold units, not value) is now bigger than the desktop market. Leaving walk over on the chance of gaining huge market shares, where nVidia is totally dominant, is just a lousy business decision.

6850M XT and 7900M are both well-performing high-end cards in the last two generations, that just didn't gain traction. Rumor has it that AMD are simply much worse for the laptop manufacturers to work with when integrating their cards. That is something they can improve on, instead of just letting nVidia have it all. Maybe there is malpractice from nVidia to some degree, but we should have heard something if it was widespread.

Edit: I found a source. Intel will not enter the laptop GPU market.

Sapphire Radeon RX 7700 XT Pulse Review

The Sapphire Radeon RX 7700 XT Pulse is priced at AMD's MSRP of $450, which makes it an attractive choice. You still get a great cooler that's paired with outstanding fan settings that result in a whisper-quiet gaming experience that's quieter than most GeForce cards.

www.techpowerup.com

That's why. Page 38. Basically, 4060, 4070, and 4080 have verifiable efficiency leads over their AMD counterparts. Those power consumption differentials might be excusable on Desktop, for the perf/$ advantage AMD offers, but they are entirely inexcusable in a laptop where every watt counts. And yes, I know those figures can be scaled down to make an RX7000 fit in a laptop, but so can they for RTX40, and nVidia's efficiency lead means they don't have to scale down nearly as much, so there goes AMD's performance advantage. Their VRAM advantage is gone too, since VRAM is expensive, increases complexity, and demands more energy in use. It doesn't make as much sense with a limited power and thermal budget anymore. And AMD have a lot more power consumption work to do in non-gaming workloads to catch up with nVidia too. All of this can also be said for Intel.

As for speculating whether or not the upcoming Halo SOCs will replace Intel/AMD's laptop dGPUs? They already have. It's no longer a question of "if". Whatever narrative you can think of for why (and you mentioned a few viable ones) doesn't really make a difference. And that's okay. The majority of gaming laptops sold have X50 and x60 class dGPUs in them. That's where Intel and AMD are aiming to compete. It's the largest part of the gaming laptop market, so it makes the most sense for them to target it. You have to quit thinking in terms of iGPU share vs dGPU share for laptops, it's just "high performance GPU share" going forward. Don't worry about the high end too much. Sure, there are some huge profit margins there, but at the same time, those are low volume skus compared to x50 and x60 combined.

 

[HeggeBegge]

That's why. Page 38. Basically, 4060, 4070, and 4080 have verifiable efficiency leads over their AMD counterparts. Those power consumption differentials might be excusable on Desktop, for the perf/$ advantage AMD offers, but they are entirely inexcusable in a laptop where every watt counts. And yes, I know those figures can be scaled down to make an RX7000 fit in a laptop, but so can they for RTX40, and nVidia's efficiency lead means they don't have to scale down nearly as much, so there goes AMD's performance advantage. Their VRAM advantage is gone too, since VRAM is expensive, increases complexity, and demands more energy in use. It doesn't make as much sense with a limited power and thermal budget anymore. And AMD have a lot more power consumption work to do in non-gaming workloads to catch up with nVidia too. All of this can also be said for Intel.

I appreciate the input. I guess you may be right on the limitations of AMD laptop GPUs, but it doesn't seem like an insurmountable problem that cannot be remedied over the course of generations. AMD certainly has limitations on desktop GPUs as well, but in the CPU market their chiplet designs have given them an edge. I see no reason as to why they couldn't bridge the gap.

As for speculating whether or not the upcoming Halo SOCs will replace Intel/AMD's laptop dGPUs? They already have. It's no longer a question of "if". Whatever narrative you can think of for why (and you mentioned a few viable ones) doesn't really make a difference. And that's okay. The majority of gaming laptops sold have X50 and x60 class dGPUs in them. That's where Intel and AMD are aiming to compete. It's the largest part of the gaming laptop market, so it makes the most sense for them to target it. You have to quit thinking in terms of iGPU share vs dGPU share for laptops, it's just "high performance GPU share" going forward. Don't worry about the high end too much. Sure, there are some huge profit margins there, but at the same time, those are low volume skus compared to x50 and x60 combined.

No, they haven't already. It is still a niche product, that fills a gap between professionals and gamers who prefer battery life and portability. I can understand that this is their angle of attack. No point in going for laptop dGPUs when you're planning to replace them with APUs. But the basic point remain. The Strix Halo chips (or AI MAX+) are just not up to it yet. When they reach the level of performance and power consumption of a M4 Max chip with a decent price, then we have a market disruptor. We still don't and there's no telling how long that will take.

 

[mama]

This is good news. A focus on PC implementation can only give confidence in the next generation (after the 9000 series) products. I'm waiting for the 1090 (Ti?).