Hey, all. It's me, ASaiyan, and as you may or may not know I have been an incessant skeptic of game streaming technology, and the insistence of certain industry members (such as Yves Guillemot) that it represents "the future of gaming". I have never doubted that the tech is possible, but I have always maintained that it would be impractical in the face of certain external realities – i.e, the relatively slow average internet speeds in key markets like the U.S. I failed to see how the average American consumer could achieve acceptable IQ and latency from such a service in the near future. I've articulated this view in virtually every Era thread created on the subject.
So when Google announced a beta for its "Project Stream" last week, promising I could stream the latest Assassin's Creed game right through Chrome, I signed up just for shits and giggles. I wanted to see how bad it would be. To my surprise, they got this program up and running in very short order and I found an access email in my inbox today. I decided I'd give it a go as soon as I got home tonight.
I played about an hour of Assassin's Creed: Odyssey through my web browser, and while I'm not gonna pretend it was "perfect" or "indistinguishable" from running it locally, I must say I feel I've been proven quite wrong about where this technology is at.
First, the key question: What is my internet speed? The "minimum speed" suggested for Project Stream is 10Mbps, and the "recommended speed" is between 20 and 25Mbps. Surprisingly, I initially found it quite difficult to figure this out, because ISPs are known to detect and cheat speedtests and the figures I kept getting back were way above what I knew I was actually getting. Thankfully, JershJopstin suggested me a simple and foolproof test: just open Task Manager during gameplay and see how much bandwidth the service is actually getting. So I did, and here are the results:
That looks pretty accurate to me. So, for the below observations, assume an internet speed of 23 Mbps – within the "recommended" range for Project Stream.
Now that's out of the way, here are my initial observations on some key factors:
IQ/Resolution: Unsurprisingly, this was pretty good. Keep in mind, however, that I'm only using a modest 1080p60 display. But just like Netflix always brings me the pretty pixels, Google brought me to Ancient Greece just fine. The Project Stream version appears to be using a 'High' graphical preset (i.e., as opposed to 'Ultra'), with no graphics settings available to change on the user end, and the level of detail is very respectable. In 'rural' areas the game seems to maintain a solid 1080p (though I didn't do any pixel-counting). In intensive areas, however, such as the busy village of Sami, I did notice some 'buffering'/'artifacting' of the image down to something like 900p, or even 800p, for extended stretches of a minute or more. Weirdly, I only noticed this in a second session. Did I just not see it the first time? Or was my connection actually slightly worse the second time? A strange new conundrum brought about by streaming technology.
Framerate: This was pretty much perfectly stable and smooth. In the hour I played, I only had one noticeable frame dip, when panning the camera through a particularly busy village dock full of ships and merchants. Unfortunately, I couldn't get RivaTuner to work with Chrome for me to accurately count the frames. Additionally, since this is my first-ever Assassin's Creed game, and the combat is rather slow, I have to admit I had a hard time judging the framerate just by eyeballing it. However, from professional outlet impressions there seems to be some consensus that the Project Stream version of Odyssey is running at 30fps. Which makes the most sense, given that the PC version is incredibly taxing even on high-end PCs.
Load times: They're pleasantly fast! Granted, I dunno how long it takes the game to load locally (and even then you've got HDD vs SSD loads, etc). But the "startup load" experienced when beginning a new session from the title screen lasts only around 10 to 15 seconds or so, and the game doesn't seem to do much loading at all after that. Pretty good for a massive open-world game!
Latency: This is, of course, the make-or-break factor for game streaming technology. Obviously I played this with a wired, Ethernet connection. I did, however, decide to perform two different input tests – one with my trusty Xbone controller connected wirelessly, and one with it directly wired.
With a wireless controller, input lag is noticeable, but definitely playable. To my surprise, even with this deliberately-inhibitive setup, we are talking about only milliseconds of time between my pressing the attack button (RB) and the animation starting, and not actual seconds. Being as these are subseconds, and I'm no Digital Foundry-level expert, this was very hard for me to try and time – though I did my best. I want to stress that these are rounded estimates for a service that is still in beta, and reflect just one person's experience. All that being said, with a wireless input, the latency I found was around 50ms.
With a wired controller, input lag was more than halved. Shocking, I know. This made it even harder for me to time; but it was certainly still there, and above the 7-9ms Google says is due to my ISP. Latency with this optimal input method was somewhere between 20 and 25ms.
Multiple people suggested I use gcping.com to more accurately measure the latency of my connection to Google's servers and corroborate my own findings. So I did:
So, latency from my PC to the nearest Google servers is somewhere around 20 to 30ms. I was honestly surprised that my initial amateur estimates were so on-the-nose, which makes me fairly confident in them. To summarize, when using a controller, the overall input latency of the beta was somewhere between 20 and 50ms in my experience. Not bad at all.
In conclusion, Google really taught me a thing or two about where streaming tech can be in 2018. The latency still needs to be improved, of course, but I have to admit it's way better than I thought it would be, and definitely in a very playable state – especially for a "slow combat" game like AC. Before today I was convinced that if game streaming tech was going to become a real pillar it would do so in that vague future horizon of "5 to 10 years from now". But from what I've just experienced, I feel like this could become a viable product as early as 2020. Microsoft's next-gen "Scarlett Cloud" box no longer sounds so silly.
Now of course, there are still some legitimate concerns to be had here. After all, it may have worked well on my internet, but does that mean the average American household can support this kind of service? Well as it turns out, according to an Akamai report from 2017, the average American internet speed these days is around 19 Mbps – well above the 10Mbps minimum speed for Project Stream, and right up against the low-20s recommended speed. So we actually seem to be okay there. The other big bugbear, of course, is data caps. I can't find any solid data on what percentage of households are affected by this awful practice these days, and I'm fortunate enough to say I've never experienced a cap on my home internet. But how much data would a service like this use? Well, let's assume for the sake of worst-case estimate that the game is using your full 20Mbps connection every second of streaming. That's 9GB of data per hour of gameplay, which stacks up pretty fast; so unfortunately, those of you suffering under caps might still have a real problem with something like this.
Anyway, that was a loooooong write-up, which presumably many of you didn't read it in its entirety (if you did feel free to start your reply with 'Bananas', and remember that we are on the honor system here :P). But that about sums up my new views on this subject. But what about you, Era? Anyone else currently involved in the Project Stream beta had a similar, or completely different experience? Have you read any write-ups from your favorite outlets/journalists that corroborate or reject my own impressions? Are you living in Nowhereburg with a 500kbps connection and feeling really left out by this whole business? Please discuss. I've gotten a very different perspective today, and I'd like to hear yours.
These are just the first three examples I could find, lol.
So when Google announced a beta for its "Project Stream" last week, promising I could stream the latest Assassin's Creed game right through Chrome, I signed up just for shits and giggles. I wanted to see how bad it would be. To my surprise, they got this program up and running in very short order and I found an access email in my inbox today. I decided I'd give it a go as soon as I got home tonight.
I played about an hour of Assassin's Creed: Odyssey through my web browser, and while I'm not gonna pretend it was "perfect" or "indistinguishable" from running it locally, I must say I feel I've been proven quite wrong about where this technology is at.
First, the key question: What is my internet speed? The "minimum speed" suggested for Project Stream is 10Mbps, and the "recommended speed" is between 20 and 25Mbps. Surprisingly, I initially found it quite difficult to figure this out, because ISPs are known to detect and cheat speedtests and the figures I kept getting back were way above what I knew I was actually getting. Thankfully, JershJopstin suggested me a simple and foolproof test: just open Task Manager during gameplay and see how much bandwidth the service is actually getting. So I did, and here are the results:
That looks pretty accurate to me. So, for the below observations, assume an internet speed of 23 Mbps – within the "recommended" range for Project Stream.
Now that's out of the way, here are my initial observations on some key factors:
IQ/Resolution: Unsurprisingly, this was pretty good. Keep in mind, however, that I'm only using a modest 1080p60 display. But just like Netflix always brings me the pretty pixels, Google brought me to Ancient Greece just fine. The Project Stream version appears to be using a 'High' graphical preset (i.e., as opposed to 'Ultra'), with no graphics settings available to change on the user end, and the level of detail is very respectable. In 'rural' areas the game seems to maintain a solid 1080p (though I didn't do any pixel-counting). In intensive areas, however, such as the busy village of Sami, I did notice some 'buffering'/'artifacting' of the image down to something like 900p, or even 800p, for extended stretches of a minute or more. Weirdly, I only noticed this in a second session. Did I just not see it the first time? Or was my connection actually slightly worse the second time? A strange new conundrum brought about by streaming technology.
Framerate: This was pretty much perfectly stable and smooth. In the hour I played, I only had one noticeable frame dip, when panning the camera through a particularly busy village dock full of ships and merchants. Unfortunately, I couldn't get RivaTuner to work with Chrome for me to accurately count the frames. Additionally, since this is my first-ever Assassin's Creed game, and the combat is rather slow, I have to admit I had a hard time judging the framerate just by eyeballing it. However, from professional outlet impressions there seems to be some consensus that the Project Stream version of Odyssey is running at 30fps. Which makes the most sense, given that the PC version is incredibly taxing even on high-end PCs.
Load times: They're pleasantly fast! Granted, I dunno how long it takes the game to load locally (and even then you've got HDD vs SSD loads, etc). But the "startup load" experienced when beginning a new session from the title screen lasts only around 10 to 15 seconds or so, and the game doesn't seem to do much loading at all after that. Pretty good for a massive open-world game!
Latency: This is, of course, the make-or-break factor for game streaming technology. Obviously I played this with a wired, Ethernet connection. I did, however, decide to perform two different input tests – one with my trusty Xbone controller connected wirelessly, and one with it directly wired.
With a wireless controller, input lag is noticeable, but definitely playable. To my surprise, even with this deliberately-inhibitive setup, we are talking about only milliseconds of time between my pressing the attack button (RB) and the animation starting, and not actual seconds. Being as these are subseconds, and I'm no Digital Foundry-level expert, this was very hard for me to try and time – though I did my best. I want to stress that these are rounded estimates for a service that is still in beta, and reflect just one person's experience. All that being said, with a wireless input, the latency I found was around 50ms.
With a wired controller, input lag was more than halved. Shocking, I know. This made it even harder for me to time; but it was certainly still there, and above the 7-9ms Google says is due to my ISP. Latency with this optimal input method was somewhere between 20 and 25ms.
Multiple people suggested I use gcping.com to more accurately measure the latency of my connection to Google's servers and corroborate my own findings. So I did:
So, latency from my PC to the nearest Google servers is somewhere around 20 to 30ms. I was honestly surprised that my initial amateur estimates were so on-the-nose, which makes me fairly confident in them. To summarize, when using a controller, the overall input latency of the beta was somewhere between 20 and 50ms in my experience. Not bad at all.
In conclusion, Google really taught me a thing or two about where streaming tech can be in 2018. The latency still needs to be improved, of course, but I have to admit it's way better than I thought it would be, and definitely in a very playable state – especially for a "slow combat" game like AC. Before today I was convinced that if game streaming tech was going to become a real pillar it would do so in that vague future horizon of "5 to 10 years from now". But from what I've just experienced, I feel like this could become a viable product as early as 2020. Microsoft's next-gen "Scarlett Cloud" box no longer sounds so silly.
Now of course, there are still some legitimate concerns to be had here. After all, it may have worked well on my internet, but does that mean the average American household can support this kind of service? Well as it turns out, according to an Akamai report from 2017, the average American internet speed these days is around 19 Mbps – well above the 10Mbps minimum speed for Project Stream, and right up against the low-20s recommended speed. So we actually seem to be okay there. The other big bugbear, of course, is data caps. I can't find any solid data on what percentage of households are affected by this awful practice these days, and I'm fortunate enough to say I've never experienced a cap on my home internet. But how much data would a service like this use? Well, let's assume for the sake of worst-case estimate that the game is using your full 20Mbps connection every second of streaming. That's 9GB of data per hour of gameplay, which stacks up pretty fast; so unfortunately, those of you suffering under caps might still have a real problem with something like this.
Anyway, that was a loooooong write-up, which presumably many of you didn't read it in its entirety (if you did feel free to start your reply with 'Bananas', and remember that we are on the honor system here :P). But that about sums up my new views on this subject. But what about you, Era? Anyone else currently involved in the Project Stream beta had a similar, or completely different experience? Have you read any write-ups from your favorite outlets/journalists that corroborate or reject my own impressions? Are you living in Nowhereburg with a 500kbps connection and feeling really left out by this whole business? Please discuss. I've gotten a very different perspective today, and I'd like to hear yours.
Last edited:
