This is a potential billion dollar idea with under 200 views. I personally download my games so I can play offline, but I can see how this would be worlds superior to running the game on the cloud...
This is a potential billion dollar idea with under 200 views. I personally download my games so I can play offline, but I can see how this would be worlds superior to running the game on the cloud for very large games. Essentially running the game locally after it downloads the needed data. Kudos to the guy for writing the virtual file system and drivers to make this possible. I'm a software engineer but that is extremely impressive. I could see an issue with installing games on a drive smaller than the two games that run on the same drive. You'd have to redownload the previously cached data if it was overwritten, which is more likely on a smaller drive shared by multiple large games. This is probably a non-issue for most people though, because gamers usually play 1 game at a a time for multiple sessions.
This technology is super cool, but i'm skeptical it's really a billion dollar idea. Game developers usually treat the current crop of consoles as their performance floor. Since the Xbox Series and...
This technology is super cool, but i'm skeptical it's really a billion dollar idea.
Game developers usually treat the current crop of consoles as their performance floor. Since the Xbox Series and PlayStation 5 consoles both sport high speed SSDs, games are being built assuming they will always run on storage that has at least that level of performance. People who have gigiabit internet service are lucky to have gigabit internet service, and that service still only offers a little over half the bandwidth you can expect from spinning rust. Games like Spider-Man 2 would be out of the question entirely.
So it's stuck in a place where if a user's internet is fast enough to make this viable, it is also fast enough that waiting for the full download is a mild inconvenience at worst.
On top of this, SSD prices have been cratering. I recently added an 8TB SATA SSD to my PC for about $300. At that price I doubt it has the write endurance of higher end drives, but it's basically perfect to act as bulk storage for largely static non-critical data like a steam library.
It looks like this dude made it by himself and I'm incredibly impressed. Your assessment about this being a billion dollar idea may not actually be that far off. This, with some proper support,...
It looks like this dude made it by himself and I'm incredibly impressed. Your assessment about this being a billion dollar idea may not actually be that far off. This, with some proper support, will be game changing.
You can probably also increase the cache size a bit to accommodate a player playing multiple games. It's not like storage is expensive, it's just that you no longer need oodles of it. A single drive with a 500GB cache will play anything and 500GB costs as little as € 0,047 per GB.
I think he said the cache is overwritten if it's too small. So the larger disk cache that you choose, the less re-downloading you will have to do if you're switching games. Smaller caches will...
I think he said the cache is overwritten if it's too small. So the larger disk cache that you choose, the less re-downloading you will have to do if you're switching games. Smaller caches will still work, but you'll be re-downloading data if your older cache gets overwritten.
Part of this issue of selling this will be explaining it to the public. The closest marketing summary I can think of, would be 'on-disk buffer' but that isn't entirely accurate and probably still too technical. For me, the most compelling aspect of this is, if you play games where lag, ping, and twitch reflexes are important, other services are inadequate, while this is viable, because you are playing the game data that exists on your local machine. I am still a bit skeptical that this is equivalent to playing a fully downloaded game in terms of loading in assets at runtime, but at least all of your inputs will not have lag on the leg between your controller/keyboard and the running game. I could see this being successful with twitch streamers as well, because they would not have to download anything in advance of streaming if they are starting a new game.
Steam spends... what, $600M/yr on bandwidth? Even assuming this would cut it in half, and probably much more, that's worth a billion in 3 years. And since the solution would likely stay viable...
Steam spends... what, $600M/yr on bandwidth? Even assuming this would cut it in half, and probably much more, that's worth a billion in 3 years. And since the solution would likely stay viable longer...
This is using the network as a disk drive, which does often work pretty well if you have a good enough network. It’s fairly common in a datacenter to have machines with no disks of their own. But...
This is using the network as a disk drive, which does often work pretty well if you have a good enough network. It’s fairly common in a datacenter to have machines with no disks of their own.
But for games, the problem isn’t just playing offline. Network connections vary widely. Someone with a laptop that they take with them will be switching networks a lot, and will get varying performance depending on where they are.
Right, I do think he could have demoed a more impactful 'worst-case scenario' and dived into the constraints of this service a bit more. In the demo it looked fine for Halo. I'm assuming the map...
Right, I do think he could have demoed a more impactful 'worst-case scenario' and dived into the constraints of this service a bit more. In the demo it looked fine for Halo. I'm assuming the map is fairly small compared to something like GTA. I'd like to see a sandbox game like GTA throttled and tested on a slow connection. I think there are probably assumptions that the maps will be readily available on disk and things like that may not be accounted for by the developers. I'm assuming it will create some pop-in, where assets appear a bit late on the edge of the map, but I'd like to see a stress test of this.
Be wary of "I wouldn't want this, but I could see how other people would" sorts of arguments. Otherwise you can have an idea that's widely agreed to be popular but that nobody actually uses or...
This is a potential billion dollar idea with under 200 views. I personally download my games so I can play offline, but I can see how
Be wary of "I wouldn't want this, but I could see how other people would" sorts of arguments. Otherwise you can have an idea that's widely agreed to be popular but that nobody actually uses or wants. Like "Facebook for dog owners", to use a cliche.
Also, ballooning game download sizes aren't due to essential complexity, they're due to developers more-or-less just not caring about download sizes - at least, not when they're in a rush to ship, and then once it ships they don't come back to fix it because that game is shipped and working just fine. Because, like you said, this is a non-issue for most people. I personally hate it, but at the end of the day I don't look up how much disk space a game uses before I buy it, and most people are less anal than me about that stuff already. I chuck it onto my "reasons why all software should be open-source" list and move on.
...also I think Assassin's Creed tried this idea a few years ago? AC3, maybe? The whole "streaming download" thing. It's not widespread nowadays though, which doesn't bode well for this implementation of the idea.
I didn't qualify what I said, but may I should have. I don't usually play AAA games so I don't run into this issue very often. However I do have friends which are avid EA play pass enjoyers which...
I didn't qualify what I said, but may I should have. I don't usually play AAA games so I don't run into this issue very often. However I do have friends which are avid EA play pass enjoyers which have multiple, large games installed at the same time. If I try EA game pass again, I'll be trying this out for sure, so I can quickly pick up AAA games and play without waiting for a full download. The main draw of this for me is that you can play almost immediately- not so much the fact that it will prevent you from downloading tons of data. If anything, the main flaw of this for me is it's not guaranteed that you will download the same data only once, so if you DO have a data cap or throttling after a certain GB amount this could impact your usage in a negative way. That being said, I think it's mostly a non-issue because it's probably similar to streaming HD video in it's data, or maybe even less after it settles into disk cache.
Edit: NVM, I see the dev stating in the comments this only works on Steam because of reverse-engineering. So EA and Epic game passes, which would be the optimal target for this aren't available yet.
What's the minimum mbps you need for this? In the video he throws around 200mbps as "representative of an average connection" and 120mbps as "low", but some of us don't even have that much...
What's the minimum mbps you need for this? In the video he throws around 200mbps as "representative of an average connection" and 120mbps as "low", but some of us don't even have that much...
Seems like the product page is down right now (is slashdotted still a word? Am I now the old man shouting at a cloud? Is the meme I just referenced also from 20+ years ago? Oh no...), so I ended...
Seems like the product page is down right now (is slashdotted still a word? Am I now the old man shouting at a cloud? Is the meme I just referenced also from 20+ years ago? Oh no...), so I ended up searching the name and stumbling across the guy's LinkedIn. Turns out his background is datacenter networking, which is the kind of thing that simultaneously gives a person the skillset to pull this off and leaves them with no calibration for numbers lower than about 25Gbps.
From the video it does look like it should degrade pretty gracefully though - if it's mocking out the filesystem, the games will already need to have been built to tolerate anything from 5,400rpm spinning rust to a tmpfs ramdisk, which is a good few orders of magnitude speed and latency difference already. Slower connection would likely mean slower load times but not gameplay issues in many cases, I'd expect.
[Edit] That said, it does look like the median connection speed is above 120Mbps for 30+ countries, with major markets like the US and China towards 200Mbps.
I guess it’s probably bimodal to an extent, though? Either you have fiber coming into or very close to the premises, at which point there’s loads of headroom and speed is pretty much a function of ISP price plan, or you don’t, and speed is whatever last drops the technology is able to squeeze out of the line you do have.
It's worth noting that, by definition, half of people will have less than the median speed. The people who are hurt the most by "the 100gb problem" are those with completely trash internet. In the...
That said, it does look like the median connection speed is above 120Mbps for 30+ countries, with major markets like the US and China towards 200Mbps.
It's worth noting that, by definition, half of people will have less than the median speed. The people who are hurt the most by "the 100gb problem" are those with completely trash internet. In the US, I know people with under 15 mbps. Downloading a game that's optimized and compressed properly is still a several hour affair, and them forgetting to install a AAA game or having their computer run into an issue during an overnight install is a death sentence for game night. The people with gigabit can more so uninstall and reinstall stuff on the fly like it's no big deal. These sub 15mbps kind of people are also the ones who'll be left behind by any cloud gaming solution, so this kind of workaround optimized to serve them would make it a true billion dollar idea like the other comments are saying.
dang. I wonder if that's purely based on advertised landline speeds. I "have" a 400Mbps connection, but that's only if I'm right next to my router. I could easily drop down to 30Mbps just from...
it does look like the median connection speed is above 120Mbps for 30+ countries, with major markets like the US and China towards 200Mbps.
dang. I wonder if that's purely based on advertised landline speeds. I "have" a 400Mbps connection, but that's only if I'm right next to my router. I could easily drop down to 30Mbps just from going upstairs to my room (on the same side of the house). signal penetration is hard (and my house only has one ethernet cable. Wonder how common that is)
Now, I spent quite a bit on a mesh network to solve this, but that's not an option for everyone. I work from home so it's important to me and to various pandemic benefits from my workplace pretty much subsidized such a cost because it paid my internet bill for 2 years.
Presumably that data would be advertised speeds from the ISP. The issues you are talking about are often resolvable by the end-user, speeds from the ISP you can't resolve necessarily (obviously if...
Presumably that data would be advertised speeds from the ISP. The issues you are talking about are often resolvable by the end-user, speeds from the ISP you can't resolve necessarily (obviously if they have different speed packages that's one thing, but the actual infrastructure of the ISP isn't something that can be easily changed).
The wifi limitations you are talking about can be resolved by using ethernet or fiber for some people (though quite rare in the home), and some wifi options are better than others too, plus there's MoCA and powerline (very hit or miss).
Yeah, I was strongly considering it and originally the mesh routers I bought were originally just going to be temporary relief until I could comfortably let an electrician in the house. But they...
Yeah, I was strongly considering it and originally the mesh routers I bought were originally just going to be temporary relief until I could comfortably let an electrician in the house. But they ended up working so well I ended up investing in some higher quality routers later on.
Yeah his definition of low is definitiely pretty skewed, it would depend on the game of course but I imagine it needs to be relatively high in general (unless you're using it for very small games...
Yeah his definition of low is definitiely pretty skewed, it would depend on the game of course but I imagine it needs to be relatively high in general (unless you're using it for very small games which kind of defeats the point). Why not give it a try yourself and see how it performs?
Edit: Apparently the average download speed in the US (based on speedtest.net results) is ~213Mbps, so maybe not as skewed of a number as I initially thought.
Oh my god, this is actually so cool. I might give this a shot tbh. Edit: I'm literally just flipping through games in my steam library like a fucking magazine this is incredible. This is probably...
Oh my god, this is actually so cool. I might give this a shot tbh.
Edit: I'm literally just flipping through games in my steam library like a fucking magazine this is incredible. This is probably the coolest tech I've seen in gaming in a long while.
Edit 2: I ran it on a M.2 SATA drive and got downloads of only like 4-6 mB/s, dark souls remastered, borderlands 2, path of exile, trials of mana, monster hunter rise, and basically any indie game all ran well.
To stress test, I didnt expect it to work, but GTA V was basically unplayable with hitching and unloaded assets plus unbearably long load times. Counter-Strike 2 took forever to load and had unloaded assets in the main menu and seemed like it was going to be unplayable.
So if its kind of like a AA, older AAA, or indie, and maybe not open world, you can run it on pretty modest drive and connection.
If your connection is amazing and you use like an NVMe or Optane like the developer maybe you can do just about anything with it.
So the use case is either amazing internet and storage constrained, or poor internet and don't want to wait for a download and only playing relatively modest titles in terms of what they might load at a time.
One amazing use case that comes to mind is trying Steam Next Fest demos without downloading them all. Another would be Steam Deck, if the developer gets this running on linux and Deck (currently Windows only) which is pretty storage constrained, during times when you're at home or have access to a fast connection.
All your concerns about data integrity can be pretty much solved (or are already) through the Steam integration. Steam already has an authorative database of how a downloaded game should look like...
All your concerns about data integrity can be pretty much solved (or are already) through the Steam integration. Steam already has an authorative database of how a downloaded game should look like for each version, including the cases where it only gives you a third party launcher which actually downloads the game - as far as Steam is concerned the launcher is the game, and the same of course applies to via. Thankfully this practice is rare. Steam also already knows if a game is running and doesn't touch its files as long as it does.
I'm pretty sure via wants to be bought out by Steam and not stay an independent service, which would also address your concerns about another party being able to insert malicious code.
And assuming that Valve is both willing and able to implement all good ideas like this strikes me as a bit optimistic, by the way. Writing kernel mode drivers isn't something you do on a whim.
Not to mention that you will always have your connection maxed out for no reason, instead of downloading once, playing whenever. I'm sure there is probably some way to do this while keeping all...
Not to mention that you will always have your connection maxed out for no reason, instead of downloading once, playing whenever. I'm sure there is probably some way to do this while keeping all the files so they can be transferred out of the Via cache for offline play.
Via is an early access program/service that lets you forego downloading an entire game and instead it downloads parts of it on demand as you play, completely transparently as if you just...
Via is an early access program/service that lets you forego downloading an entire game and instead it downloads parts of it on demand as you play, completely transparently as if you just downloaded and launched the game normally. The game itself still runs locally on your machine so unlike game streaming there shouldn't be any additional latency or other associated problems.
I can't try it out myself yet because I'm on Linux, but this is very cool. Of course, I'd prefer if they stopped making have such large disk space requirements in the first place, but since that won't happen I'm happy that this exists.
Very cool idea. It feels like a lot of games have massive data files, but often don't require them. Think fully voiced dialogue in a dozen languages, likely in uncompressed wav. But this is also...
Very cool idea. It feels like a lot of games have massive data files, but often don't require them. Think fully voiced dialogue in a dozen languages, likely in uncompressed wav. But this is also great for just jumping into a game you feel like dusting off, or playing something new to see if you like it.
Would love to see this with a prefetch API so games could request assets shortly before they're needed. It seems like if developers got behind the concept, some of the rough edges could be smoothed out.
Of course there will be some problems with existing games. If they load large data files (eg. Guild Wars 2), or have timeouts on file accesses, that will not work well. It also requires users have both a speedy (enough) connection, and a gaming PC.
The author seemed a bit hopeful that Valve would start paying them as a CDN. I feel it's more likely Valve would implement this themselves with their own extensive CDNs. An acqui-hire might be a more realistic outcome.
Still though, I think it's a great idea. I'd love to see more exploration of the concept and testing to see which kinds of games work well. Best of luck to the dev.
Not very useful for me but this seems like a much better solution than regular game streaming. I've tried all of the game streaming services and I still find the latency to be unacceptable. Via...
Not very useful for me but this seems like a much better solution than regular game streaming. I've tried all of the game streaming services and I still find the latency to be unacceptable. Via increases load times but it doesn't seem to add any latency to the gameplay which is really nice.
Maybe it already does this but it would be cool if the downloaded files remained stored locally so you'd have the full game downloaded after a while (as an option).
It does. In the video he rebooted the game and Via to "simulate coming back the next day" and it downloaded next to nothing. If you were to allocate say 500gb of your drive, it might keep about 5...
Maybe it already does this but it would be cool if the downloaded files remained stored locally so you'd have the full game downloaded after a while (as an option).
It does. In the video he rebooted the game and Via to "simulate coming back the next day" and it downloaded next to nothing. If you were to allocate say 500gb of your drive, it might keep about 5 or 6 large games cached at any one time (or less because you never ended up needing the entire file system of the games) and he said it would then remove the older stuff when you start go past that.
Though he also said you could run these AAA games on a 20gb allocation, if you so desired. I personally don't really see the point in that when storage is so cheap these days.
It seems like a pretty cool service, and it would be really neat if something like this was integrated with Steam, even if I personally would never use it due to network constraints.
That's the dream, right now I have lots of games on the show hdd and very few on the fast nvme, and I found no good way to automatically shift the games I'm actually playing right now to fast...
That's the dream, right now I have lots of games on the show hdd and very few on the fast nvme, and I found no good way to automatically shift the games I'm actually playing right now to fast storage. I tend to hop around a lot, and basically end up staring at load screens more than I'd have to.
This seems cool... but I'm curious. One problem I have with huge games outside of the practical aspect is the energy consumption from hundreds of thousands of people downloading all that data...
This seems cool... but I'm curious. One problem I have with huge games outside of the practical aspect is the energy consumption from hundreds of thousands of people downloading all that data worldwide. Would this not result in more data being downloaded over time?
Possibly, without more details it might be hard to say. It's possible that it could reduce the data being downloaded because it only downloads what your game needs/requests. So if you play a game...
One problem I have with huge games outside of the practical aspect is the energy consumption from hundreds of thousands of people downloading all that data worldwide. Would this not result in more data being downloaded over time?
Possibly, without more details it might be hard to say.
It's possible that it could reduce the data being downloaded because it only downloads what your game needs/requests. So if you play a game for the multiplayer only, but normally you would download the entire game including the single player campaign, you might end up downloading significantly less assets if they're not needed for multiplayer gameplay. Because it stores the data as cache locally on your drive, it's not necessarily re-downloading the same data over and over again. Or say you download a huge open world game, but you end up not liking it after an hour of gameplay. Well the traditional way, you still downloaded the whole thing, but with this method, you only downloaded whatever was needed to play that one hour of gameplay.
Of course the other aspect to answering the question you asked is how the cache works. He mentions in the video that a single game could in theory be larger than the amount of space reserved for cache, which means that it will overwrite data for that game which you haven't recently accessed for data you need now. So if you're playing through the campaign of a game, data you needed in the first part of the campaign but not in the later stages might get overwritten, that means if you play the campaign again, it may end up downloading that data again. If the cache size is sufficiently large, you might be able to play the whole game multiple times repeatedly only downloading once, but if you decide to play another game and then come back to the prior game later, you might end up downloading it all over again, all very dependent on what the cache size is and the size of the games you're playing.
The other thing to consider in what you are asking is whether that is even a significant amount of energy consumption, even if the bandwidth usage in this method ended up being more. How much energy does it really cost to transmit X amount of TB or PB etc. over the internet, depending on the scale of what you're talking about here of course. There's other things to factor into this too, which is that this method in theory requires lower amounts of local storage, so you're using less energy to power local drives among any other facets of traditional local storage bound gameplay. That could also include energy required to manufacture the drives you buy to put in your PC to have the space to download all these games to begin with. Also in the video they're comparing it to cloud gaming, so seemingly they think they're possibly targeting a few markets with this, but the appeal of cloud gaming to me is largely not about the storage savings but more about the overall hardware needed to run a lot of these games and this doesn't solve the latter problem. You still need a proper GPU to play the game, you just don't necessarily need the same amount of storage space.
It's probably no worse than high quality streaming. YouTube in 4k streams at 20GB per hour and since Via seems to cache the data it doesn't need to download everything continuously.
It's probably no worse than high quality streaming. YouTube in 4k streams at 20GB per hour and since Via seems to cache the data it doesn't need to download everything continuously.
In all likelihood it will be, but there's no telling with software this early in development. I don't have the numbers so I'm hesitant to be absolute. My assumption is that it will reduce overall...
In all likelihood it will be, but there's no telling with software this early in development. I don't have the numbers so I'm hesitant to be absolute.
My assumption is that it will reduce overall data transfers because, like @Grumble4681 said, parts of the game may not have to be downloaded at all while the rest will be cached for multiple uses. For all intents and purposes it will download something once as necessary and then unlike YouTube, stops downloading data at all unless you wipe the cache or start playing something else. With a large enough cache size (not difficult with the current storage prices) you'll likely rotate through data slowly rather than quickly.
I don't think this tech is meant for everyone, nor does it intend to be, but it has its use cases where overall reliance on storage can be severely diminished to the point someone needs perhaps a drive or two at best (boot + storage) rather than having a bunch of disks in their PC to accommodate their entire games library. And this is just games. It could be applied to anything really. Having fewer drives manufactured and shipped may even offset the impact caused by the downloading.
Lastly, I don't think that's a very high bar. Streaming is relatively light on the climate. IEA brings up a good point that e-waste is a much larger problem which this tool directly impacts by driving down storage consumption. I'm using this article as reference, although I must admit I haven't fully checked the veracity of their claims so please allow me to ask for some preemptive slack-cutting as I could be wrong.
You can get 1tb SSD's for 60ish bucks 8TB HDD for 100ish, why go through all of this every time rather than setup a local steam cache? Or put a large drive in your PC and always have them local...
You can get 1tb SSD's for 60ish bucks 8TB HDD for 100ish, why go through all of this every time rather than setup a local steam cache? Or put a large drive in your PC and always have them local anyway. Cloud gaming has the advantage of having high powered hardware making it a good fit for someone with a low spec PC. This on the other hand still requires you to run it locally which means you would need to have a rig capable of running whatever game. Since you already have the hardware I don't see how it wouldn't be easier and better to just spend another 100 bucks for a decent sized drive.
An advantage is that you only download the files shortly before they are needed, so 100+GB games are instantly playable and big patches aren't gonna cause any significant downtime.
An advantage is that you only download the files shortly before they are needed, so 100+GB games are instantly playable and big patches aren't gonna cause any significant downtime.
Yeah I'm with you on this, I think games are going in a different direction where this solution may become less viable. If the future of gaming still looked like it lived on HDDs then this has a...
Yeah I'm with you on this, I think games are going in a different direction where this solution may become less viable. If the future of gaming still looked like it lived on HDDs then this has a lot of potential, but we're seeing game design move away from that. You already had Starfield that stated the minimum requirement was an SSD, and as you mentioned DirectStorage is a more recent introduction (this isn't something like Linux where people have said it was going to take over soon for the past couple decades).
I think a lot of people are focusing on just the bandwidth, but there's also latency that comes into play. I don't know what performance demands are in DirectStorage type games or other games designed around using an SSD, but obviously the latency between the connection of your local disk through your motherboard to your GPU or whatever is going to be quite less than the latency between a server multiple states or multiple countries away from you. The guy in this video even highlights latency as an aspect of consideration. Are games calling for various assets that expect to get those assets in less than 15 milliseconds? In this way, even if the only asset required is a 5mb file, you could have 500mbps connection or 1gbps connection or 10gbps connection and none of that matters if the latency is too high. If the game needs that 5mb file in 1 millisecond but it takes 10 milliseconds for you to request it and 10 milliseconds for the asset to get sent to you, then you're 20x slower than what the game expected.
I think if Steam bought this they would individually certify games for it. Then you’d do a little speed test after buying a game and determine if Via can let you play right away. The best solution...
I think if Steam bought this they would individually certify games for it. Then you’d do a little speed test after buying a game and determine if Via can let you play right away.
The best solution would also passively download the files in the background, ordered by frequency of access.
First time hearing about DirectStorage, very interesting. It's really a matter of how much ahead of time the game asks for the asset, I suppose. Something like DirectStorage would decrease load...
First time hearing about DirectStorage, very interesting. It's really a matter of how much ahead of time the game asks for the asset, I suppose. Something like DirectStorage would decrease load times which means you don't have to ask for that data as early, but you would still be asking for it before the asset is needed, so it's a question of if your network speed can still keep up with that, so it's difficult to know if it will still work (well) for games using DirectStorage. However if DirectStorage is really as fast as is claimed, then you should also be able to apply much higher rates of compression to data, resulting in smaller files and thus less time spent downloading the data.
In my very short amount of research after you mentioned it, part of the benefit of sending it directly to the GPU (besides the fact that you'll want it there in the end anyway) is that since it's...
In my very short amount of research after you mentioned it, part of the benefit of sending it directly to the GPU (besides the fact that you'll want it there in the end anyway) is that since it's being decompressed on the GPU in can be used immediately, whereas typically you're processing the data twice, first decompressing it on the CPU then sending the decompressed data to the GPU. Assuming it's possible to more thoroughly compress data without scaling up the amount of time decompression takes, then you can reduce file size further which means the game overall will take up less space and you have less bandwith usage to the GPU which could help even more with load times.
Very cool! My only concern here is actually illustrated perfectly by his choice of game in the video. Halo Infinite already has a long section of match loading where you wait forever on an "other...
Very cool! My only concern here is actually illustrated perfectly by his choice of game in the video. Halo Infinite already has a long section of match loading where you wait forever on an "other players loading" bar. What happens when half the players are using this and you constantly get stuck in a queue with some rando still running a 28k modem and loading this map for the first time? It seems like it would degrade the multiplayer experience by forcing everyone who already has the game downloaded to wait for anyone who didn't. In most games this wouldn't be a problem I suppose, and it solves itself on an individual level over time at least, but I imagine online matchmaking would run into a whole lot more timeouts and need to refill queues a lot more often.
If someone's connection is that bad then I imagine this probably wouldn't be something they would want to use anyway, since it would be noticeably worse.
If someone's connection is that bad then I imagine this probably wouldn't be something they would want to use anyway, since it would be noticeably worse.
This is neat, but this means games have to preferably be split up in as many files as possible.. which is usually not how games are packaged because that's much slower to download and install in...
This is neat, but this means games have to preferably be split up in as many files as possible.. which is usually not how games are packaged because that's much slower to download and install in the traditional sense.
This is frankly amazing - huge props to the developer. I hope a steamdeck / linux native client (might actually be easier to implement in linux to be honest) is in the works.
This is frankly amazing - huge props to the developer. I hope a steamdeck / linux native client (might actually be easier to implement in linux to be honest) is in the works.
has anyone kept up with this project/ could let us know how its been? i was considering using it but it felt a little too sketchy putting my password into a 100kb program. from the looks of the...
has anyone kept up with this project/ could let us know how its been?
i was considering using it but it felt a little too sketchy putting my password into a 100kb program. from the looks of the guys twitter, he's kept up with working on it but the website is still not too trust inspiring being so bare.
I've been checking on it every few weeks, but basically my thoughts are the same. It's not that I distrust the project, but it's way too small at this time for me to trust my Steam login to. If...
I've been checking on it every few weeks, but basically my thoughts are the same. It's not that I distrust the project, but it's way too small at this time for me to trust my Steam login to. If they could find a way to operate with OAuth, or if there were some major backers that could vouch for it, I'd feel better about it.
It's still a very cool technology though, which is why I keep checking back. I hope to see some more movement in the future.
This is a potential billion dollar idea with under 200 views. I personally download my games so I can play offline, but I can see how this would be worlds superior to running the game on the cloud for very large games. Essentially running the game locally after it downloads the needed data. Kudos to the guy for writing the virtual file system and drivers to make this possible. I'm a software engineer but that is extremely impressive. I could see an issue with installing games on a drive smaller than the two games that run on the same drive. You'd have to redownload the previously cached data if it was overwritten, which is more likely on a smaller drive shared by multiple large games. This is probably a non-issue for most people though, because gamers usually play 1 game at a a time for multiple sessions.
This technology is super cool, but i'm skeptical it's really a billion dollar idea.
Game developers usually treat the current crop of consoles as their performance floor. Since the Xbox Series and PlayStation 5 consoles both sport high speed SSDs, games are being built assuming they will always run on storage that has at least that level of performance. People who have gigiabit internet service are lucky to have gigabit internet service, and that service still only offers a little over half the bandwidth you can expect from spinning rust. Games like Spider-Man 2 would be out of the question entirely.
So it's stuck in a place where if a user's internet is fast enough to make this viable, it is also fast enough that waiting for the full download is a mild inconvenience at worst.
On top of this, SSD prices have been cratering. I recently added an 8TB SATA SSD to my PC for about $300. At that price I doubt it has the write endurance of higher end drives, but it's basically perfect to act as bulk storage for largely static non-critical data like a steam library.
It looks like this dude made it by himself and I'm incredibly impressed. Your assessment about this being a billion dollar idea may not actually be that far off. This, with some proper support, will be game changing.
You can probably also increase the cache size a bit to accommodate a player playing multiple games. It's not like storage is expensive, it's just that you no longer need oodles of it. A single drive with a 500GB cache will play anything and 500GB costs as little as € 0,047 per GB.
I think he said the cache is overwritten if it's too small. So the larger disk cache that you choose, the less re-downloading you will have to do if you're switching games. Smaller caches will still work, but you'll be re-downloading data if your older cache gets overwritten.
Part of this issue of selling this will be explaining it to the public. The closest marketing summary I can think of, would be 'on-disk buffer' but that isn't entirely accurate and probably still too technical. For me, the most compelling aspect of this is, if you play games where lag, ping, and twitch reflexes are important, other services are inadequate, while this is viable, because you are playing the game data that exists on your local machine. I am still a bit skeptical that this is equivalent to playing a fully downloaded game in terms of loading in assets at runtime, but at least all of your inputs will not have lag on the leg between your controller/keyboard and the running game. I could see this being successful with twitch streamers as well, because they would not have to download anything in advance of streaming if they are starting a new game.
Steam spends... what, $600M/yr on bandwidth? Even assuming this would cut it in half, and probably much more, that's worth a billion in 3 years. And since the solution would likely stay viable longer...
I would hope Steam makes it so this is integrated in the way a remote library works locally.
Being able to run this at home would be even cooler imo.
This is using the network as a disk drive, which does often work pretty well if you have a good enough network. It’s fairly common in a datacenter to have machines with no disks of their own.
But for games, the problem isn’t just playing offline. Network connections vary widely. Someone with a laptop that they take with them will be switching networks a lot, and will get varying performance depending on where they are.
Right, I do think he could have demoed a more impactful 'worst-case scenario' and dived into the constraints of this service a bit more. In the demo it looked fine for Halo. I'm assuming the map is fairly small compared to something like GTA. I'd like to see a sandbox game like GTA throttled and tested on a slow connection. I think there are probably assumptions that the maps will be readily available on disk and things like that may not be accounted for by the developers. I'm assuming it will create some pop-in, where assets appear a bit late on the edge of the map, but I'd like to see a stress test of this.
Be wary of "I wouldn't want this, but I could see how other people would" sorts of arguments. Otherwise you can have an idea that's widely agreed to be popular but that nobody actually uses or wants. Like "Facebook for dog owners", to use a cliche.
Also, ballooning game download sizes aren't due to essential complexity, they're due to developers more-or-less just not caring about download sizes - at least, not when they're in a rush to ship, and then once it ships they don't come back to fix it because that game is shipped and working just fine. Because, like you said, this is a non-issue for most people. I personally hate it, but at the end of the day I don't look up how much disk space a game uses before I buy it, and most people are less anal than me about that stuff already. I chuck it onto my "reasons why all software should be open-source" list and move on.
...also I think Assassin's Creed tried this idea a few years ago? AC3, maybe? The whole "streaming download" thing. It's not widespread nowadays though, which doesn't bode well for this implementation of the idea.
I didn't qualify what I said, but may I should have. I don't usually play AAA games so I don't run into this issue very often. However I do have friends which are avid EA play pass enjoyers which have multiple, large games installed at the same time. If I try EA game pass again, I'll be trying this out for sure, so I can quickly pick up AAA games and play without waiting for a full download. The main draw of this for me is that you can play almost immediately- not so much the fact that it will prevent you from downloading tons of data. If anything, the main flaw of this for me is it's not guaranteed that you will download the same data only once, so if you DO have a data cap or throttling after a certain GB amount this could impact your usage in a negative way. That being said, I think it's mostly a non-issue because it's probably similar to streaming HD video in it's data, or maybe even less after it settles into disk cache.
Edit: NVM, I see the dev stating in the comments this only works on Steam because of reverse-engineering. So EA and Epic game passes, which would be the optimal target for this aren't available yet.
I was thinking that if I was an investor in tech I’d be sending over a term sheet right now.
What's the minimum mbps you need for this? In the video he throws around 200mbps as "representative of an average connection" and 120mbps as "low", but some of us don't even have that much...
Seems like the product page is down right now (is slashdotted still a word? Am I now the old man shouting at a cloud? Is the meme I just referenced also from 20+ years ago? Oh no...), so I ended up searching the name and stumbling across the guy's LinkedIn. Turns out his background is datacenter networking, which is the kind of thing that simultaneously gives a person the skillset to pull this off and leaves them with no calibration for numbers lower than about 25Gbps.
From the video it does look like it should degrade pretty gracefully though - if it's mocking out the filesystem, the games will already need to have been built to tolerate anything from 5,400rpm spinning rust to a tmpfs ramdisk, which is a good few orders of magnitude speed and latency difference already. Slower connection would likely mean slower load times but not gameplay issues in many cases, I'd expect.
[Edit] That said, it does look like the median connection speed is above 120Mbps for 30+ countries, with major markets like the US and China towards 200Mbps.
I guess it’s probably bimodal to an extent, though? Either you have fiber coming into or very close to the premises, at which point there’s loads of headroom and speed is pretty much a function of ISP price plan, or you don’t, and speed is whatever last drops the technology is able to squeeze out of the line you do have.
It's worth noting that, by definition, half of people will have less than the median speed. The people who are hurt the most by "the 100gb problem" are those with completely trash internet. In the US, I know people with under 15 mbps. Downloading a game that's optimized and compressed properly is still a several hour affair, and them forgetting to install a AAA game or having their computer run into an issue during an overnight install is a death sentence for game night. The people with gigabit can more so uninstall and reinstall stuff on the fly like it's no big deal. These sub 15mbps kind of people are also the ones who'll be left behind by any cloud gaming solution, so this kind of workaround optimized to serve them would make it a true billion dollar idea like the other comments are saying.
dang. I wonder if that's purely based on advertised landline speeds. I "have" a 400Mbps connection, but that's only if I'm right next to my router. I could easily drop down to 30Mbps just from going upstairs to my room (on the same side of the house). signal penetration is hard (and my house only has one ethernet cable. Wonder how common that is)
Now, I spent quite a bit on a mesh network to solve this, but that's not an option for everyone. I work from home so it's important to me and to various pandemic benefits from my workplace pretty much subsidized such a cost because it paid my internet bill for 2 years.
Presumably that data would be advertised speeds from the ISP. The issues you are talking about are often resolvable by the end-user, speeds from the ISP you can't resolve necessarily (obviously if they have different speed packages that's one thing, but the actual infrastructure of the ISP isn't something that can be easily changed).
The wifi limitations you are talking about can be resolved by using ethernet or fiber for some people (though quite rare in the home), and some wifi options are better than others too, plus there's MoCA and powerline (very hit or miss).
Even just moving the router upstairs (if possible) would yield better wifi through the whole home.
Sounds like you should start running Cat6 in your walls!
Yeah, I was strongly considering it and originally the mesh routers I bought were originally just going to be temporary relief until I could comfortably let an electrician in the house. But they ended up working so well I ended up investing in some higher quality routers later on.
Yeah his definition of low is definitiely pretty skewed, it would depend on the game of course but I imagine it needs to be relatively high in general (unless you're using it for very small games which kind of defeats the point). Why not give it a try yourself and see how it performs?
Edit: Apparently the average download speed in the US (based on speedtest.net results) is ~213Mbps, so maybe not as skewed of a number as I initially thought.
Oh my god, this is actually so cool. I might give this a shot tbh.
Edit: I'm literally just flipping through games in my steam library like a fucking magazine this is incredible. This is probably the coolest tech I've seen in gaming in a long while.
Edit 2: I ran it on a M.2 SATA drive and got downloads of only like 4-6 mB/s, dark souls remastered, borderlands 2, path of exile, trials of mana, monster hunter rise, and basically any indie game all ran well.
To stress test, I didnt expect it to work, but GTA V was basically unplayable with hitching and unloaded assets plus unbearably long load times. Counter-Strike 2 took forever to load and had unloaded assets in the main menu and seemed like it was going to be unplayable.
So if its kind of like a AA, older AAA, or indie, and maybe not open world, you can run it on pretty modest drive and connection.
If your connection is amazing and you use like an NVMe or Optane like the developer maybe you can do just about anything with it.
So the use case is either amazing internet and storage constrained, or poor internet and don't want to wait for a download and only playing relatively modest titles in terms of what they might load at a time.
One amazing use case that comes to mind is trying Steam Next Fest demos without downloading them all. Another would be Steam Deck, if the developer gets this running on linux and Deck (currently Windows only) which is pretty storage constrained, during times when you're at home or have access to a fast connection.
All your concerns about data integrity can be pretty much solved (or are already) through the Steam integration. Steam already has an authorative database of how a downloaded game should look like for each version, including the cases where it only gives you a third party launcher which actually downloads the game - as far as Steam is concerned the launcher is the game, and the same of course applies to via. Thankfully this practice is rare. Steam also already knows if a game is running and doesn't touch its files as long as it does.
I'm pretty sure via wants to be bought out by Steam and not stay an independent service, which would also address your concerns about another party being able to insert malicious code.
And assuming that Valve is both willing and able to implement all good ideas like this strikes me as a bit optimistic, by the way. Writing kernel mode drivers isn't something you do on a whim.
Not to mention that you will always have your connection maxed out for no reason, instead of downloading once, playing whenever. I'm sure there is probably some way to do this while keeping all the files so they can be transferred out of the Via cache for offline play.
Via is an early access program/service that lets you forego downloading an entire game and instead it downloads parts of it on demand as you play, completely transparently as if you just downloaded and launched the game normally. The game itself still runs locally on your machine so unlike game streaming there shouldn't be any additional latency or other associated problems.
I can't try it out myself yet because I'm on Linux, but this is very cool. Of course, I'd prefer if they stopped making have such large disk space requirements in the first place, but since that won't happen I'm happy that this exists.
Very cool idea. It feels like a lot of games have massive data files, but often don't require them. Think fully voiced dialogue in a dozen languages, likely in uncompressed wav. But this is also great for just jumping into a game you feel like dusting off, or playing something new to see if you like it.
Would love to see this with a prefetch API so games could request assets shortly before they're needed. It seems like if developers got behind the concept, some of the rough edges could be smoothed out.
Of course there will be some problems with existing games. If they load large data files (eg. Guild Wars 2), or have timeouts on file accesses, that will not work well. It also requires users have both a speedy (enough) connection, and a gaming PC.
The author seemed a bit hopeful that Valve would start paying them as a CDN. I feel it's more likely Valve would implement this themselves with their own extensive CDNs. An acqui-hire might be a more realistic outcome.
Still though, I think it's a great idea. I'd love to see more exploration of the concept and testing to see which kinds of games work well. Best of luck to the dev.
Not very useful for me but this seems like a much better solution than regular game streaming. I've tried all of the game streaming services and I still find the latency to be unacceptable. Via increases load times but it doesn't seem to add any latency to the gameplay which is really nice.
Maybe it already does this but it would be cool if the downloaded files remained stored locally so you'd have the full game downloaded after a while (as an option).
It does. In the video he rebooted the game and Via to "simulate coming back the next day" and it downloaded next to nothing. If you were to allocate say 500gb of your drive, it might keep about 5 or 6 large games cached at any one time (or less because you never ended up needing the entire file system of the games) and he said it would then remove the older stuff when you start go past that.
Though he also said you could run these AAA games on a 20gb allocation, if you so desired. I personally don't really see the point in that when storage is so cheap these days.
It seems like a pretty cool service, and it would be really neat if something like this was integrated with Steam, even if I personally would never use it due to network constraints.
That's the dream, right now I have lots of games on the show hdd and very few on the fast nvme, and I found no good way to automatically shift the games I'm actually playing right now to fast storage. I tend to hop around a lot, and basically end up staring at load screens more than I'd have to.
I don't mean as a cache but as a permanently installed game.
This seems cool... but I'm curious. One problem I have with huge games outside of the practical aspect is the energy consumption from hundreds of thousands of people downloading all that data worldwide. Would this not result in more data being downloaded over time?
Also, what about non-Workshop mods?
Possibly, without more details it might be hard to say.
It's possible that it could reduce the data being downloaded because it only downloads what your game needs/requests. So if you play a game for the multiplayer only, but normally you would download the entire game including the single player campaign, you might end up downloading significantly less assets if they're not needed for multiplayer gameplay. Because it stores the data as cache locally on your drive, it's not necessarily re-downloading the same data over and over again. Or say you download a huge open world game, but you end up not liking it after an hour of gameplay. Well the traditional way, you still downloaded the whole thing, but with this method, you only downloaded whatever was needed to play that one hour of gameplay.
Of course the other aspect to answering the question you asked is how the cache works. He mentions in the video that a single game could in theory be larger than the amount of space reserved for cache, which means that it will overwrite data for that game which you haven't recently accessed for data you need now. So if you're playing through the campaign of a game, data you needed in the first part of the campaign but not in the later stages might get overwritten, that means if you play the campaign again, it may end up downloading that data again. If the cache size is sufficiently large, you might be able to play the whole game multiple times repeatedly only downloading once, but if you decide to play another game and then come back to the prior game later, you might end up downloading it all over again, all very dependent on what the cache size is and the size of the games you're playing.
The other thing to consider in what you are asking is whether that is even a significant amount of energy consumption, even if the bandwidth usage in this method ended up being more. How much energy does it really cost to transmit X amount of TB or PB etc. over the internet, depending on the scale of what you're talking about here of course. There's other things to factor into this too, which is that this method in theory requires lower amounts of local storage, so you're using less energy to power local drives among any other facets of traditional local storage bound gameplay. That could also include energy required to manufacture the drives you buy to put in your PC to have the space to download all these games to begin with. Also in the video they're comparing it to cloud gaming, so seemingly they think they're possibly targeting a few markets with this, but the appeal of cloud gaming to me is largely not about the storage savings but more about the overall hardware needed to run a lot of these games and this doesn't solve the latter problem. You still need a proper GPU to play the game, you just don't necessarily need the same amount of storage space.
It's probably no worse than high quality streaming. YouTube in 4k streams at 20GB per hour and since Via seems to cache the data it doesn't need to download everything continuously.
I mean, that's a HIGH bar. It's exactly the kind of consumption we need to reduce, so I'd hope it's not 'no worse' but better.
In all likelihood it will be, but there's no telling with software this early in development. I don't have the numbers so I'm hesitant to be absolute.
My assumption is that it will reduce overall data transfers because, like @Grumble4681 said, parts of the game may not have to be downloaded at all while the rest will be cached for multiple uses. For all intents and purposes it will download something once as necessary and then unlike YouTube, stops downloading data at all unless you wipe the cache or start playing something else. With a large enough cache size (not difficult with the current storage prices) you'll likely rotate through data slowly rather than quickly.
I don't think this tech is meant for everyone, nor does it intend to be, but it has its use cases where overall reliance on storage can be severely diminished to the point someone needs perhaps a drive or two at best (boot + storage) rather than having a bunch of disks in their PC to accommodate their entire games library. And this is just games. It could be applied to anything really. Having fewer drives manufactured and shipped may even offset the impact caused by the downloading.
Lastly, I don't think that's a very high bar. Streaming is relatively light on the climate. IEA brings up a good point that e-waste is a much larger problem which this tool directly impacts by driving down storage consumption. I'm using this article as reference, although I must admit I haven't fully checked the veracity of their claims so please allow me to ask for some preemptive slack-cutting as I could be wrong.
You can get 1tb SSD's for 60ish bucks 8TB HDD for 100ish, why go through all of this every time rather than setup a local steam cache? Or put a large drive in your PC and always have them local anyway. Cloud gaming has the advantage of having high powered hardware making it a good fit for someone with a low spec PC. This on the other hand still requires you to run it locally which means you would need to have a rig capable of running whatever game. Since you already have the hardware I don't see how it wouldn't be easier and better to just spend another 100 bucks for a decent sized drive.
An advantage is that you only download the files shortly before they are needed, so 100+GB games are instantly playable and big patches aren't gonna cause any significant downtime.
Exactly, it's such a huge user experience improvement.
You could already have them downloaded and have them immediately when you need them. Having to wait is not an advantage.
Yeah I'm with you on this, I think games are going in a different direction where this solution may become less viable. If the future of gaming still looked like it lived on HDDs then this has a lot of potential, but we're seeing game design move away from that. You already had Starfield that stated the minimum requirement was an SSD, and as you mentioned DirectStorage is a more recent introduction (this isn't something like Linux where people have said it was going to take over soon for the past couple decades).
I think a lot of people are focusing on just the bandwidth, but there's also latency that comes into play. I don't know what performance demands are in DirectStorage type games or other games designed around using an SSD, but obviously the latency between the connection of your local disk through your motherboard to your GPU or whatever is going to be quite less than the latency between a server multiple states or multiple countries away from you. The guy in this video even highlights latency as an aspect of consideration. Are games calling for various assets that expect to get those assets in less than 15 milliseconds? In this way, even if the only asset required is a 5mb file, you could have 500mbps connection or 1gbps connection or 10gbps connection and none of that matters if the latency is too high. If the game needs that 5mb file in 1 millisecond but it takes 10 milliseconds for you to request it and 10 milliseconds for the asset to get sent to you, then you're 20x slower than what the game expected.
I think if Steam bought this they would individually certify games for it. Then you’d do a little speed test after buying a game and determine if Via can let you play right away.
The best solution would also passively download the files in the background, ordered by frequency of access.
First time hearing about DirectStorage, very interesting. It's really a matter of how much ahead of time the game asks for the asset, I suppose. Something like DirectStorage would decrease load times which means you don't have to ask for that data as early, but you would still be asking for it before the asset is needed, so it's a question of if your network speed can still keep up with that, so it's difficult to know if it will still work (well) for games using DirectStorage. However if DirectStorage is really as fast as is claimed, then you should also be able to apply much higher rates of compression to data, resulting in smaller files and thus less time spent downloading the data.
In my very short amount of research after you mentioned it, part of the benefit of sending it directly to the GPU (besides the fact that you'll want it there in the end anyway) is that since it's being decompressed on the GPU in can be used immediately, whereas typically you're processing the data twice, first decompressing it on the CPU then sending the decompressed data to the GPU. Assuming it's possible to more thoroughly compress data without scaling up the amount of time decompression takes, then you can reduce file size further which means the game overall will take up less space and you have less bandwith usage to the GPU which could help even more with load times.
Very cool. I hope all the big stores license this technology, if it can be applied to something other than Steam.
Very cool! My only concern here is actually illustrated perfectly by his choice of game in the video. Halo Infinite already has a long section of match loading where you wait forever on an "other players loading" bar. What happens when half the players are using this and you constantly get stuck in a queue with some rando still running a 28k modem and loading this map for the first time? It seems like it would degrade the multiplayer experience by forcing everyone who already has the game downloaded to wait for anyone who didn't. In most games this wouldn't be a problem I suppose, and it solves itself on an individual level over time at least, but I imagine online matchmaking would run into a whole lot more timeouts and need to refill queues a lot more often.
If someone's connection is that bad then I imagine this probably wouldn't be something they would want to use anyway, since it would be noticeably worse.
This is neat, but this means games have to preferably be split up in as many files as possible.. which is usually not how games are packaged because that's much slower to download and install in the traditional sense.
We don't know how it works. It could cache larger chunks of data than necessary as an optimization that it's likely files are stored sequentially.
This is frankly amazing - huge props to the developer. I hope a steamdeck / linux native client (might actually be easier to implement in linux to be honest) is in the works.
The dev has expressed some interest in that. He even said the same thing actually, about how it would probably be easier.
good to hear - i'll be following closely.
has anyone kept up with this project/ could let us know how its been?
i was considering using it but it felt a little too sketchy putting my password into a 100kb program. from the looks of the guys twitter, he's kept up with working on it but the website is still not too trust inspiring being so bare.
I've been checking on it every few weeks, but basically my thoughts are the same. It's not that I distrust the project, but it's way too small at this time for me to trust my Steam login to. If they could find a way to operate with OAuth, or if there were some major backers that could vouch for it, I'd feel better about it.
It's still a very cool technology though, which is why I keep checking back. I hope to see some more movement in the future.