Discussion Steam Deck 2 handheld ?

[marees]

AMD recently revealed FSR 4 & touted power savings for handheld gaming.

AMD plans for FSR4 to be fully AI-based — designed to improve quality and maximize power efficiency

The team in charge has already been working on this for 9–12 months.

www.tomshardware.com

Zen 5 reviews mentioned a generational advance in power modes (especially for handhelds)

https://twitter.com/x/status/1823422603643527174

RDNA 3.5 is supposed to be more efficient in low power scenarios & incorporates lessons learnt from Samsung's mobile RDNA 2

AMD's tweaked RDNA 3.5 GPU is solely focused on improving mobile gaming performance

There are plenty of little changes under the hood to help the graphics processor do more for less power.

www.pcgamer.com

The current strix point is unlikely to be used for handheld gaming due to latency between zen 5 ccd & zen 5c ccd

Upcoming Kracken / kraken / krackan point APU has a single ccx & 8 cores in total ( 3+ 5 or 4 + 4 ) plus 8 or 12 CUs.

AMD to Launch Affordable 'Kraken Point' Zen 5 Mobile in 2025

The company is targeting Intel's Lunar Lake and Qualcomm's eight-core Snapdragon X chips.

www.extremetech.com

Putting all these together the above could be an ideal APU for next steam handheld. Will valve bite though ?

The Steam Deck 2 might be significantly faster than its predecessor.
Valve has clarified that the Steam Deck 2 will likely not be hitting shelves anytime soon until a "generational leap in compute" takes place, as reported in an interview with Reviews.org.

https://www.tomshardware.com/video-...tm_campaign=socialflow&utm_source=twitter.com

 

[marees]

Latest rumour on CPU cores for Kraken Point

https://twitter.com/x/status/1838246179261583440

 

[Saylick]

The Steam Deck 2 might be significantly faster than its predecessor.
Valve has clarified that the Steam Deck 2 will likely not be hitting shelves anytime soon until a "generational leap in compute" takes place, as reported in an interview with Reviews.org.

Given that Valve wants a generational leap in hardware, I think it's safe to say that the Steam Deck 2 will be significantly faster. It will likely take N4P coming down in price a little bit more before it happens, though. I'm thinking it will be another 2 years before we get even a whiff of the Steam Deck 2.

It should have 4 Zen5c cores and 8 RDNA 4 CUs at a minimum. Not sure if they will try to cram in 8 cores, but if the die space had to be spent it may be more prudent to spend it on Infinity Cache rather than 4 more CPU cores.

 

[marees]

Given that Valve wants a generational leap in hardware, I think it's safe to say that the Steam Deck 2 will be significantly faster. It will likely take N4P coming down in price a little bit more before it happens, though. I'm thinking it will be another 2 years before we get even a whiff of the Steam Deck 2.

It should have 4 Zen5c cores and 8 RDNA 4 CUs at a minimum. Not sure if they will try to cram in 8 cores, but if the die space had to be spent it may be more prudent to spend it on Infinity Cache rather than 4 more CPU cores.

I don't expect RDNA 4 to be in any igpu

Valve might as well jump to RDNA 5 + Zen 5. The advantage could be that they share an APU with Microsoft who would be looking for hardware that supports Direct ML & advanced machine learning etc.

8 CUs is too low imo. 12 or even better, 16, would be ideal

 

[Saylick]

I don't expect RDNA 4 to be in any igpu

Valve might as well jump to RDNA 5 + Zen 5. The advantage could be that they share an APU with Microsoft who would be looking for hardware that supports Direct ML & advanced machine learning etc.

8 CUs is too low imo. 12 or even better, 16, would be ideal

Ehh, I wouldn't count on RDNA 5. That's like a 2026 product, and consoles generally aren't on the cutting edge architecture.

12 or 16 CUs sounds nice, but how are they going to feed it bandwidth and power? I don't think there's enough silicon or wattage to make it worthwhile.

 

[marees]

Valve says its 'not really fair to your customers' to create yearly iterations of something like the Steam Deck, instead it's waiting 'for a generational leap in compute without sacrificing battery life'​

News
By Dave James
published 21 hours ago
Valve may not be creating yearly iterations of the Steam Deck, but it's having to give the same answer every year.

This almost precisely echoes what Yang told me around the launch of the OLED version last year: "It needs to be the right time," he tells me. "And we have to have the right parts for it. So we really want there to be a generational leap in performance for us to be able to comfortably call it a Steam Deck 2.

"We're keeping an eye on chips and APU that are available. The things that are available right now are not right yet. But maybe in two to three years, there will be something that is good in the same way that our current APU is, in terms of the power envelope, with performance, with battery draw. All of those things have to be in the sweet spot for a real move to a new chip."

"For us to make a second version," Aldehayyat then chimes in, "we will be able to have a substantial performance improvement while sticking to a similar kind of power range and weight to battery life. And that's not going to happen next year or the year after that. It's probably going to be more than that."

Valve says its 'not really fair to your customers' to create yearly iterations of something like the Steam Deck, instead it's waiting 'for a generational leap in compute without sacrificing battery life'

Valve may not be creating yearly iterations of the Steam Deck, but it's having to give the same answer every year.

www.pcgamer.com

 

[Saylick]

The more I think about it, the more I start to think that N3E or even N3P may be the node they go with. N6 to N4P only gets you to 50% more performance/watt, which is a nice bump but also closer to what a mid-cycle refresh will offer these days (I'm looking at you, PS5 Pro). N6 to N3E/P would get them closer to a true doubling.

 

[Mopetar]

Valve might as well jump to RDNA 5 + Zen 5.

Is there even going to be an RDNA5? I'd heard that AMD is unifying RDNA and CDNA so they don't have separate architectures to produce. Maybe that's overkill for an APU, but so is doing separate GPU architecture design for an APU if it's not going to be high volume.

There's also nothing stopping Valve from releasing a mid-generation "pro" model akin to what console manufacturers have been doing for a while now.

 

[marees]

Is there even going to be an RDNA5? I'd heard that AMD is unifying RDNA and CDNA so they don't have separate architectures to produce. Maybe that's overkill for an APU, but so is doing separate GPU architecture design for an APU if it's not going to be high volume.

There's also nothing stopping Valve from releasing a mid-generation "pro" model akin to what console manufacturers have been doing for a while now.

I am guessing RDNA 5 would be renamed as UDNA 1

Seems like Valve is reluctant to release a 4nm refresh.

Probably waiting for Microsoft's handheld plans & reuse the same just like they did with Van Gogh

 

[marees]

Given that Valve wants a generational leap in hardware, I think it's safe to say that the Steam Deck 2 will be significantly faster. It will likely take N4P coming down in price a little bit more before it happens, though. I'm thinking it will be another 2 years before we get even a whiff of the Steam Deck 2.

It should have 4 Zen5c cores and 8 RDNA 4 CUs at a minimum. Not sure if they will try to cram in 8 cores, but if the die space had to be spent it may be more prudent to spend it on Infinity Cache rather than 4 more CPU cores.

Given the zen 6 rumours by MLID / Kepler / Adroc, it seems there is no suitable monolithic zen 6 APU for steam deck 2

The choices are

1. Cut down zen 6 medusa premium — this would be costly, maybe in 2028
2. Monolithic zen 7 apu — assuming nv & AMD convince MS to drop NPU tax (this would be 2029 or 2030)

So assuming an even number of WGPs, a monolithic laptop zen 7 APU (with RDNA 5+) should have atleast 4 WGP ?

So if we assume like 3 monolithic chips like zen 6 medusa point, then we could have (in 2029-2030):

1. Point 1 — 12 or 8 WGPs
2. Point 2 — 6 or 8 WGPs
3. Point 3 — 4 WGPs

I think this is what I remembered:


Still leaves a lot of open questions. For one are NPUs deprecated nextgen in all Zen 6 + RDNA 5 based products or later. If it applies to mobile too how will AMD adress the concerns regarding time to first token (execution latency) and power efficiency, because getting an inferior solution in terms of battery life on a brand new product vs last gen is just unacceptable.
Are we talking about customizations to ML and core in RDNA 5 to effectively emulate an "NPU mode" to save on power? Could that perhaps be very fine-grained power gating, architectural changes to ML HW and even special modes of operation, and in general massive architectural changes to cachemem and data locality? Just a bit of spitballing.

This processing in cache patent should increase ML and RT performance sizeably: https://patents.google.com/patent/US20240264942A1
Really any branchy (PT for example) or mem hungry workload should benefit, as long as BW heavy instructions are offloaded to CCUs. Perhaps this is the Processing-in-Cache patent that was mentioned January?


Obviously no confirmation but might be reasonable to expect this is roughly how the NPU in Mediatek Dimensity 9500 SoC achieves CIM. IIRC CIM is touted as one of the big reasons for why the NPU is so power efficient.

 

[MrMPFR]

@marees that Co-compute unit patent was already shot down by @Kepler_L2. Not in RDNA 5.
As for later µarchs, who knows.

 

[marees]

@marees that Co-compute unit patent was already shot down by @Kepler_L2. Not in RDNA 5.
As for later µarchs, who knows.

The broader question remains. Does RDNA 5 igpu do away with a need for an NPU ?

Because an NPU would add unnecessary cost to steam deck

 

[marees]

Still the Same story

Valve Says It Has a 'Pretty Good Idea' of What Steam Deck 2 Is Going to Be, Explains Why It's Holding Off for Now​

Frame cap in hand.​

the same things we've said in the past where we're really interested to work on what's next for Steam Deck… the thing we're making sure of is that it's a worthwhile enough performance upgrade to make sense as a standalone product,” Griffais explained.

“We're not interested in getting to a point where it's 20 or 30 or even 50% more performance at the same battery life. We want something a little bit more demarcated than that. So we've been working back from silicon advancements and architectural improvements, and I think we have a pretty good idea of what the next version of Steam Deck is going to be, but right now there's no offerings in that landscape, in the SoC [System on a Chip] landscape, that we think would truly be a next-gen performance Steam Deck.”

https://in.ign.com/steam-deck/24653...-2-is-going-to-be-explains-why-its-holding-of

 

[DAPUNISHER]

I missed this post, don't dip into the console forum much anymore.

 

[marees]

My guess. You may have to wait until 2029 for a steam deck 2


Until then, Valve might keep rollling out steamOS for handhelds manufactured by others (such as the Lenovo legion go etc.)

My guesstimate is zen 7 & rdna 5+ monolithic apu that replaces krackan. 2029 at the earliest

 

[marees]

Created a poll. Please vote to register your guesses for the steam deck 2 hardware

The announcement of qualcomm arm based steam frame VR headset running steam OS on FEX translation layer has set the tongues wagging

What is your opinion

Question - What will be the chip used in Steam Deck 2

The announcement of qualcomm arm based steam frame VR headset running steam OS on FEX translation layer has set the tongues wagging What is your opinion

forums.anandtech.com

 

[marees]

Valve engineer on ARM

Valve won't breathe a word about the Steam Deck 2, but they're singing a different tune about Arm processors. Fresh off announcing their Arm-powered VR headset that can run Android apps, the company is already sketching out a much broader silicon strategy that could reshape portable gaming. The timing isn't coincidental.

As Nvidia reportedly preps an Arm gaming laptop and Qualcomm courts major OEMs like Razer, Valve appears ready to ride the wave rather than fight it. "I think that it paves the way for a bunch of different, maybe ultraportables, maybe more powerful laptops being Arm-based," Griffais told The Verge during a recent interview. "Handhelds, there's a lot of potential for Arm, of course, and one might see desktop chips as well at some point in the Arm world." The statement marks a significant shift for a company that built its portable gaming empire on x86 architecture.

But it's not just wishful thinking - companies are already knocking on Valve's door. Griffais confirmed that handheld manufacturers have reached out about SteamOS partnerships, with OneNetbook among those experimenting with powerful Arm chips for gaming devices.


For Valve, this represents more than just following trends - it's about expanding SteamOS beyond its current x86 limitations. "We'll keep greasing the wheels, so to speak, so that SteamOS can work on a wider variety of Arm devices, but also so that the catalog becomes more reliable there in terms of compatibility and performance," Griffais explained.


current Arm chips might not pack enough punch for Valve's flagship ambitions. When pressed about the Steam Deck 2's "generational leap in compute without sacrificing battery life," Griffais suggested today's Arm offerings compete mainly in lower-power segments. "When you get into lower power, anything lower than Steam Deck, I think you'll find that there's an Arm chip that maybe is competitive with x86 offerings in that segment," he noted with what the interviewer imagined was a wink.

https://www.techbuzz.ai/articles/valve-signals-arm-future-for-steamos-beyond-vr-headsets

The Steam Frame (and its controllers) are designed to play VR titles, as well as traditional PC and mobile games in a resizable in-headset window that actually felt like a big-screen experience during my hands-on time with the headset. Valve engineers told me the company thinks of the Steam Frame less as a VR headset and more as “a new way to play your entire Steam library.” The Frame can do this both by streaming titles wirelessly from your PC, or running them internally on its built-in Arm-based Qualcomm Snapdragon 8 Gen 3 chip.

the company uses the Fex (officially stylized as FEX) software emulation layer that brings SteamOS to the Arm instruction set, which certainly has implications beyond this device.

Fex brings Steam games to Arm​

But that’s not the only way the Steam Frame can game. The company also showed off the x86 version of Hades 2 running standalone (as in not streaming from a PC) on the Steam Frame. And the game ran just fine and looked good at what Valve reps told me was 1400p in a window inside the headset, which I could actually resize to something that filled a large part of my field of view.

“The magic trick is that the game doesn’t know it’s running on an Arm chip,” designer Lawrence Yang told me. The game may be designed for a Windows PC, but “it’s actually running on Linux, running on Arm.”

That happens thanks to Fex, which is an emulation layer, so that will almost certainly mean increased power consumption / shorter battery life. Valve isn’t saying anything about battery life yet (except for the 40-hour per AA controller claim), as they continue to work on the software for now. But like the Steam Deck, this means your gaming time on the Steam Frame is going to vary dramatically, depending on whether you’re streaming from a PC or playing something directly on the device.

Valve’s software developer, Jeremy Selan, tried to assuage some of my fears about what Fex might do to battery life, saying the device uses Vulkan, that many games now are natively Vulkan, and that as soon as the game engine makes API calls, they start running in on “natively compiled Arm processing code.”

Similarly, on the CPU side, Selan says it’s mostly things like the user interactions themselves (what happens when you pull a trigger or hit a jump button) that go through the Fex layer. He says there is about a 10-20% performance hit with the emulation, but that it only applies to specific aspects of the code.

“When we set out, we didn’t know quite how far this would get,” said Selan. “But on games like Hades 2, where we are seeing 1440p at 90 Hz, while running the full VR stack. We’re really very happy with it.”

For games installed and running on the headset itself, Valve tells me there will be a "Verified" game program similar to what already exists for the Steam Deck, where the company is going to test titles in the Steam catalog and provide guidance.

More software tweaks to come​

Selan also says the Valve team is working on a way to pre-cache the CPU shaders, in a similar way that the company already pre-caches GPU shaders ahead of time on the Steam Deck. This should further reduce the overhead of the Fex emulation layer, but it’s not shipping yet.

Hands-on with Valve's new Steam Frame headset — Arm-powered, mixed-mode device uses Fex translation layer for traditional x86 games

Valve doubles down on headset gaming, but it isn’t all about VR.

www.tomshardware.com

 

[marees]

Valve lays the foundations for an Arm-based gaming handheld future with first Snapdragon-powered SteamOS VR headset​

News
By Zac Bowden published 2 days ago
The new Steam Frame is coming next year and is Valve's first all-in-one headset running SteamOS, powered by a Snapdragon chip.

Now that SteamOS officially supports Arm via the Snapdragon XR SoC, it's only a matter of time before we see SteamOS running on a gaming handheld that's also powered by a Snapdragon SoC.

It's quite unlikely that this will be the only Arm-based SteamOS device going forward. With NVIDIA rumored to be launching its N1X chips next year, it's reasonable to expect we'll see NVIDIA-powered Arm-based gaming handhelds in the near future too, and SteamOS will be well prepared to support it.

Valve lays the foundations for an Arm-based gaming handheld future with first Snapdragon-powered SteamOS VR headset

The new Steam Frame is coming next year and is Valve's first all-in-one headset running SteamOS, powered by a Snapdragon chip.

www.windowscentral.com

 

[marees]

Looks like steam deck 2 is 2030

https://twitter.com/x/status/1993349207856492554