Valve Could Ditch Using a Custom Processor for Steam Deck 2 in Favor of One That Might Outperform the PlayStation 6/Xbox Project Helix

Peter_Brosdahl

Moderator
Staff member
Joined
May 28, 2019
Messages
10,990
Reaction score
7,778
It’s now rumored that Valve may choose to use an off-the-shelf AMD processor to power the Steam Deck 2 for a greater generational leap. Valve has previously confirmed that the Steam Deck 2 is still probably years away, something rumors continue to support, with release happening sometime in 2028 or possibly 2029. Pierre-Loup Griffais, lead […]

See full article...
 
I mean, anything is possible really. Current Steam Deck was custom but it wasn't a huge deviation from anything you could buy off the shelf.
 
I mean, anything is possible really. Current Steam Deck was custom but it wasn't a huge deviation from anything you could buy off the shelf.
Eggxactly.
Im sure by the time they are deciding, theyll probably do custom anyway, unless theres too much issues with the # of units they need to pay out of the gate or something.
 
Makes way more sense to use off the shelf.
 
I get the feeling that valve could go for qualcomm apu fabbed on samsung

AMD fabbed on TSMC would be unaffordable, imo

Intel 18a is an option, but Intel needs to demontrate GPU capabilities on their fabs

outside chance is AMD zen 7 fabbed on samsung !!
 
I think Steam is more interested in making the best gaming potato they can. Meaning it will be capable but won't need a long lifecycle in the market. They are creating a market much like they did with linux based hand held gaming devices. They prove the market. Let everyone else jump in with 'better' devices and then sit back and reap their 30% of every game sale.
 
Last edited:
Lord Gaben as they call him, and Valve/Steam, are probably the most favorably viewed leader and company in the gaming space. Having their own branded hardware and OS is working out. Never underestimate brand power.
 
Im getting a feeling there wont be any Steam Deck built anywhere in asia, we may get a deck in 2037 possibly made in the USA.
 
I think Steam is more interested in making the best gaming potato they can.
I don't think it's a bad strategy honestly. Graphics and FX can only get you so far, it's the games (and the communities that spring up around the games) that drive revenue and engagement.

I mean, honestly... who would have thought the top money-making entertainment IP in the world ... not Star Wars. Not Grand Theft Auto. Not Call of Duty or Madden franchise.... Not Mickey Mouse. Not DC Universe or Marvel. But Pokemon. By a long long ways too. And that has always run on a potato.

Roblox looks like Minecraft vomited up a bag of Skittles. And both of those titles are huge money makers. If I go to Steam right now and look at Top Sellers... the one on top right now is CS2, which is ... almost 14 years old now? And almost every other game on that list is pixel or mobile-quality graphics, and the rest of what are left over are mostly capable of being tuned down to run on a potato if necessary.

Your hardware only needs to be good enough... It's only people like us, on these forums, that really chase those last few frames or must check every last box in the graphics options - and at great expense
 
It's only people like us, on these forums, that really chase those last few frames or must check every last box in the graphics options - and at great expense
Lets be real it's always the people willing to spend for the best be that tv's, refrigerator's, or PC gaming that drive the rest of that segment of the industry forward.

With the abandonment of the DIY space advancement for home compute will slowly grind to a halt until money shifts.
 
Lets be real it's always the people willing to spend for the best be that tv's, refrigerator's, or PC gaming that drive the rest of that segment of the industry forward.

With the abandonment of the DIY space advancement for home compute will slowly grind to a halt until money shifts.
You are not wrong.

I predict, if it gets that bad, the market will turn parallel to the mobile phone market - where they are just upgrading the camera for some rediculous reason and nothing else ever really changes generation over generation.

Sure, probably won't be a camera, but it'll be something... I mean, the consumer PC space is already mostly that way - where they will prominently display whatever the latest generation CPU is, and it's almost always some budget crap that's slow as molasses without nearly enough RAM and some tiny drive space so it can hit that magic price point for Wal Mart.
 
Lets be real it's always the people willing to spend for the best be that tv's, refrigerator's, or PC gaming that drive the rest of that segment of the industry forward.
It was indeed the case going way back. But that was then, this is now. Enthusiast whale spending power has become a joke. Kaijus drive everything. Before things get better, if they do, we are going to get thoroughly ROFLstomped Godzilla style. As I would not expect DRAM and NAND prices to stop inflating anytime soon.
 
It was indeed the case going way back. But that was then, this is now. Enthusiast whale spending power has become a joke. Kaijus drive everything. Before things get better, if they do, we are going to get thoroughly ROFLstomped Godzilla style. As I would not expect DRAM and NAND prices to stop inflating anytime soon.
Yea, I mean it's crazy even for enterprise. A server that would have cost 10k after discounts with a vendor is now costing 30k. And lead time is projected at one FLIPPING year. It's just STUPID.
 
zen 7 snowmass releasing in 2028 to have upto 384 cores

this is in addition to steamboat which releases later

 
this is in addition to steamboat which releases later
The seventh-generation EPYC processors, codenamed "Florence," are where AMD's new die-stacking strategy truly shines. Florence apparently tops out at an immense 288 cores, achieved by combining two high-end "Dwarka" I/O dies, dual "Mathura" memory controller dies, and eight unique "Steamboat" CCDs that are described as "C3Ds".

Steamboat is a radical departure that has no L3 cache on the CCD itself.

Instead, it pairs the "C3D" with a dedicated 3D-stacked cache die underneath that provides a massive 7MB of L3 per core, allowing top-tier EPYC configurations to wield nearly 2GB (2016 MB) of total L3 cache.

https://hothardware.com/news/amd-zen-7-leak-reveals-major-ipc-gains-and-huge-cache-upgrades
 
The seventh-generation EPYC processors, codenamed "Florence," are where AMD's new die-stacking strategy truly shines. Florence apparently tops out at an immense 288 cores, achieved by combining two high-end "Dwarka" I/O dies, dual "Mathura" memory controller dies, and eight unique "Steamboat" CCDs that are described as "C3Ds".

Steamboat is a radical departure that has no L3 cache on the CCD itself.

Instead, it pairs the "C3D" with a dedicated 3D-stacked cache die underneath that provides a massive 7MB of L3 per core, allowing top-tier EPYC configurations to wield nearly 2GB (2016 MB) of total L3 cache.

https://hothardware.com/news/amd-zen-7-leak-reveals-major-ipc-gains-and-huge-cache-upgrades
I see this and I know it is awesome... but from a licensing and use case standpoint.. I mean MAYBE if you're running a cloud datacenter and need just insane amounts of density to bill customers with and have more and more capacity... but...

As a standalone business not hosting for others... What is the use case? Does anyone know of one outside of research or running AI I guess?

I just try and fathom a single or even a pair of servers running Dual 288 core CPU's for something like SQL Enterprise.. The licensing alone would make the hardware purchase feel cheap. (15 per two cores... so at 288 cores... I don't even want to do the math. No... I am NOT doing the math.)

Does anyone here have an use case for this dense of a compute solution? I'd love to hear it.
 
I see this and I know it is awesome... but from a licensing and use case standpoint.. I mean MAYBE if you're running a cloud datacenter and need just insane amounts of density to bill customers with and have more and more capacity... but...

As a standalone business not hosting for others... What is the use case? Does anyone know of one outside of research or running AI I guess?

I just try and fathom a single or even a pair of servers running Dual 288 core CPU's for something like SQL Enterprise.. The licensing alone would make the hardware purchase feel cheap. (15 per two cores... so at 288 cores... I don't even want to do the math. No... I am NOT doing the math.)

Does anyone here have an use case for this dense of a compute solution? I'd love to hear it.
agree. doesn't make sense for stand alone business

AMD is primarily looking at rack scale infrastructure now.
commitment to iterate & advance on Helios rack scale platform every year
 
Become a Patron!
Back
Top