Install Steam
login
|
language
简体中文 (Simplified Chinese)
繁體中文 (Traditional Chinese)
日本語 (Japanese)
한국어 (Korean)
ไทย (Thai)
Български (Bulgarian)
Čeština (Czech)
Dansk (Danish)
Deutsch (German)
Español - España (Spanish - Spain)
Español - Latinoamérica (Spanish - Latin America)
Ελληνικά (Greek)
Français (French)
Italiano (Italian)
Bahasa Indonesia (Indonesian)
Magyar (Hungarian)
Nederlands (Dutch)
Norsk (Norwegian)
Polski (Polish)
Português (Portuguese - Portugal)
Português - Brasil (Portuguese - Brazil)
Română (Romanian)
Русский (Russian)
Suomi (Finnish)
Svenska (Swedish)
Türkçe (Turkish)
Tiếng Việt (Vietnamese)
Українська (Ukrainian)
Report a translation problem
http://steamcommunity.com/groups/homestream/discussions/0/540734423623592629/?tscn=1393068956
I think with very light applications it might work, or with a slightly faster proc than I had.
I guess with Bioshock Infinite you were using DX11? I know that DX11 titles generally perform much better than DX9 as Steam will use asyncronous capture. I wonder how much that had to do with it?
Looks like the Athlon performs a few percent better, but not nearly as much as the clock speed would suggest. You have to give Intel the gold star for instructions per clock and lower power performance.
I can't look at the steam hardware survey right now (evil Flash!) but t think something like 40% of machines are two core and probably iffy as hosts unless they're desktops with a hardware encoder card. I'm really interested to see how hardware support mixes up the minimum requirements (although that won't affect your laptop performance).
I can only imagine a dual core struggling unless it has a high clock speed and hyperthreading maybe? I have another laptop, an i5 - maybe I'll give that a try.
The nvidia 9600M GT GPU in my laptop can do hardware encoding in splashtop, so there's hope yet for the other one.
I have a lot of trouble understanding GPU model numbers :(
I finally got around to doing some perfmon on the 2 core test box and it was a bit weird.
If I play Bioshock directly on the host it runs reasonably well (again, very low settings) and caps out the two cores. If I run Bioshock streamed it stutters a bit but the two cores (which should be doing more) only get up to 70-80% (when I expected it to be more the other way around).
If I underclock my Phenom II x4 to 1.6GHz and set the memory down to 800MHz (just to try like for like on CPU speed and memory) I find that Bioshock uses 60-70% CPU local, and when I stream it uses a bit more (80%ish). I didn't swap out video cards, so maybe that or the L3 cache accounts for the difference in overall CPU use.
In any case, the quad core seemed to make more use of the cores when streaming, and the dual seemed to use less. Maybe it is something about process switching, or the processes blocking eachother on dual core systems.
Oh, and Dirt3 seemed fine (although still 90-100 ms latency) streaming from the quad. No nasty spikes during play.
As for splashtop, I have no idea what it is actually doing. From memory though, the CPU drain is far lower hosting from the laptop than it is from any of my other boxes.
It might just be some driver-level support for the capture, not the encoding?
EDIT: I went and took some measurements with Splashtop to reassure myself it wasn't all just placebo.
With the laptop as host (nvidia) it will stream an idle desktop using only 10% CPU. With the desktop as host (AMD) it will stream a similar idle desktop at the same resolution using 30-40% CPU. When playing a movie, the client framerate is 30 with the laptop as a host, just 20 with the desktop as host.
So they're not selling snake oil, it is really doing *something*. What they are doing I don't know, I just hope that Valve can take advantage of the same technology.
Either way I have my eye on a new Maxwell GPU to replace my 6850 in the desktop :-)
[Keyboard fail gave me a chance to review that Wikipedia page on CUDA]
CUDA looks to allow programmers to write code that execute on dedicated data processing chunks of the GPU, so instead of saying "Take this chunk of data and perform your XYZ process on it" it turns into "Take this chunk of data and run this program I give you" - and that sort of power can be leveraged to do video encoding (or, more commonly these days, Bitcoin mining).
A quick search points at at least one program that leverages CUDA for H.264 encoding (http://www.mainconcept.com/products/sdks/gpu-acceleration/cuda-h264avc.html), but it may be a non-trivial job to get it going in streaming (depending on what's already out there and how it's licensed).
The Toshiba laptop with the i5 430M CPU (2.26ghz dual core + hyperthreading) didn't drop a single frame when streaming 480p @ 30fps from the video capture stick to the Macbook Pro.
When set up the other way around, with the Intel Core 2 Duo T9600 Macbook Pro (2.8ghz) as host, there were fairly frequent stutters.
The log shows that encoding on the Toshiba was 1.5x faster (10ms average encoding time vs 15ms). I don't know if that's down to the newer processor or the hyperthreading, but I really didn't expect that.
Anyway, now I can stream from my Xbox 360 next to the TV across the living room to my laptop and play on the couch while the wife watches TV. Quality no worse than plugging the composite output of the Xbox into a 15" telly, and around 25-30ms latency.
Here's how its done:-
- Xbox 360 in PAL60 mode + set to widescreen mode connected by the composite output to an...
- Avermedia 830M USB TV stick plugged into a...
- Toshiba laptop with i5 430M CPU connected by wifi running...
- VLC Media Player in capture mode with a bit of cropping and yadif deinterlacing which was...
- added to Steam as a non-steam game and...
- streamed to a Macbook Pro connected via wifi to the same...
- D-Link DWR-921 router as the Toshiba
The range of the Xbox wireless controllers was more than enough for this, so no ghetto hacking was needed to get the controllers to work, but I understand from other posts that this can be done.
Once I get a faster HTPC I won't even need to use two laptops and can have a more permanent setup too.
I set up the Avermedia input in VLC manually, then saved that config to a playlist file. Then I have an AutoHotKey script that loads up VLC with command-line arguments to set up the effects and start the playlist. When I hit escape the AHK script closes VLC automatically and then exits.
I did it this way because Steam has a limited length for the command line parameters of a non-steam game and the VLC command line settings are pretty long.
Capturing from VLC is still a bit rough round the edges but there's still no better software I've found for displaying the output from a capture card.