I commented about this here: http://news.ycombinator.com/item?id=4886028 the other day.

Essentially, how often will Valve upgrade its Steam Box? Certainly more than every 6-7 years. For those of you thinking right now: "you can ugprade it yourself!", well, yes you could , but that defeats the whole purpose of having one standardized spec for developers to target. Throw in a new graphics card here and there, and you've suddenly splintered your user base into those with PCs, Official Steam Boxes from Valve, and Official Steam Boxes from Valve that have been upgrade by the users. And now, certain games will only work on some combination of those three.

I'm a fan of Valve, their games, and their culture, and want to see this succeed. I'm sure they've put much thought into scenarios like this and trust them to make the right decision.

I image the Official Steam Box from Valve (OSBfV) wouldn't actually be that upgradable. I would want something that would fit aesthetically with my TV, XBox, etc. Trying to make the OSBfV cheap enough to entice consumers, yet pack enough of a punch for "gamers" (I really dislike that stereotype, but that isn't the point) would, I imagine, require not quite interchangeable hardware, much like the XBox. But, don't discount that Steam could easily release a OSBfV1, 2, 3 etc every other year with the latest and greatest (that fits in the form factor and budget constraints).

The other possibility is that their Steam Box fits with their "games as a service" model. Imagine if they charged a monthly/yearly fee and occasionally shipped out hardware upgrades to all Steam Box owners.

We'll be doing that model in 2013 with a B2B SaaS service that benefits from an onsite edge node to reduce latency, among other things.

The hardware isn't sold, it's leased, but the effect is the same: we manage the onsite hardware, fix it when it breaks (next day replacement), and send out new hardware every few years. The hardware is literally plug and play, and works correctly behind a NAT router with DHCP.

We actually decide what hardware a customer gets (it's based on load -- standard provisioning stuff), although customers can upgrade to a system they don't actually need for a fee.

I think this approach is way better for customers, and the response so far has been extremely positive. No longer do they have to worry about provisioning and maintaining hardware, or even deciding what to buy. Of course they don't know the best way to provision hardware for our software. Why would they?

It'd be really interesting to see this approach done in the consumer space.

If they enter the hardware market I suspect they will make different Steam boxes and have them meet certain specifications. Like Steambox A to run a certain list of games, and Steambox B to run the latest games at full performance. With such clusters of performance they could have pre-defined scripts to tweak the performance so that even a older Steambox can run the latest titles decently (make at lower details/resolution) or drop compatibility altogether if a decent performance cannot be achieved. That would be an "organized" segementation of the PC gaming scene.

This being said I am not sure this is really going to be necessary because there has not been much of race for graphics in game recently on PC - I still have a GPU from 2 years ago and it still runs all games very well, even the latest ones. This may change with the introduction of new consoles, but I am not too sure about that yet.

This isn't actually a big problem. Comparing to the current situation, Valve will just release their own hardware with a massive user base, and have that platform be the target spec for PC games. That is the stock Steam Box and Valve with developers will make sure all games run perfectly on that.

All custom specs will be handled like they are now, expect now there is a single target spec everyone agrees on.

> Throw in a new graphics card here and there, and you've suddenly splintered your user base...

Given how aggressively they've pushed to make their games work on Mac and Linux (going against the grain of the industry), and given the years of massive user configuration surveys, and the fact that they've essentially run a fractured ecosystem for years, I'd bet they have something up their sleeve here... or they have a reason why this isn't as big a problem as it's been in the past.

One problem I see with low cost custom-built gaming boxes is that the Windows licence takes up a massive share of the price. I hope that Valve’s push to Linux will lead developers there as well but I don’t see the big publishers jumping on that in a hurry.

I'm not sure criticisms around cost are as relevant as they used to be. Windows 8 Pro currently sells for $40 on Microsoft's website, and yes, that key can be used to perform a clean install. If you paid $40 to Microsoft every three years when they released a new Windows version, that would work out to $13.33/year.

That isn’t exactly expensive... $13.33/year is basically a rounding error compared to the costs of console gaming.

But you pay that 40 dollars at once, and it's on the price tag of the SteamBox you would buy (if they go for a Windows box). So, 40 dollars on a console price is actually a lot of money, since most console hardware is sold at 300-400 dollars point.

The xbox360 core launched at $300. Assuming prices of the next generation are fairly inline with last generation, then yeah that's a ton. I have no idea what the cost to Microsoft on those units where, but I remember here they were either sold at a lose or with almost no profit. 13% on the BOM is more than significant especially when there is a viable alternative.

But if you're competing with a $500 PS4, you're going to have problems. Even when you throw in price savings with scaling and a colossal subsidy, that's the retail equivalent of building a PC which competes with the next-gen consoles for $800 or so. So a marginal $40 is a lot of money in that.

The crazy thing is that hardware compatibility might sort of solve itself. Developers already have to deal with all the different hardware permutations; making sure that a single specific one, with a known software configuration to boot, should be pretty easy.

They'll still have to convince people to develop a Linux version in the first place, though...

Wouldn't they only need to convince people to develop a Wine compatible version?

No. The hassle of getting that to work is probably way harder than just getting things to work on Linux in the first place.

I think that Valve will eventually put out a libvalve which can be used to target a number of linux specific integration points for typical windows programs.

Wine already runs many windows binaries, without modification, on linux. Often times when wine does not work on a given binary, the issue can be resolved by patching wine, again without touching the binary. It seems like it should be relatively easy for someone who can modify the binary source code to resolve any bugs that wine runs into (probably exposes subtle flaws in the game code in the prosses). The other issue I can think of is performance, but my sense is that the win32 API pipe is not a bottleneck in almost any program, so that is likely not an issue.

Nothing is preventing them from doing both. Dealing with nonworking wine games is incredibly frustrating (and often fruitless).

But I can see it working. In addition to being able to fix issues in the software itself, they are also working with a known hardware configuration and they can sandbox each wine instance.

PlayOnLinux already helps doing the latter, to the point where it lets you use separate wine versions for each sandbox. Much better than using synaptic to jump between different package versions.

I hope they'll stick with native versions, though.

The hardware they will launch on is probably going to be more powerful and with support for full HSA (heterogenous computing) and with DDR4 RAM, if they are going to wait until 2014 to launch it (and assuming they are going to use AMD's delayed and probably modified Kaveri APU).