Slashdot Mirror

Valve has been getting faster FPS (but not by a huge amount) in Linux than in Windows.

Windows (DirectX): 270.6 FPS
Windows (OpenGL): 303.4 FPS
Linux: 315 FPS

I'm really impressed with Valve right now.

While the Valve model is a part of it, you should be directing your praises specifically at Michael Abrash.

Abrash is a long-time graphics and optimization guru (author of Zen of Assembly Language, Zen of Graphics Programming, and two legendary Dr Dobbs series of articles, one titled Ramblings In Realtime and the other Graphics Programming Black Book) that Valve has been trying to hire for a very long time.

This is the guy who single-handedly made the Quake rendering engine, with its software-based perspective-correct texture mapping and lighting, a possibility at the time that it was released. Valve finally succeeded in landing him about a year ago, and he has been investigating the practicality of Virtual and Augmented Reality ever since.

He even writes about some of his findings in his blog, Ramblings in Valve Time

The complaint isn't Steam startup time, it's the startup time of the individual programs that are launched from Steam.

What Steam is doing is checking the central server to see whether you actually own the game as well as checking if there are any updates to it. The latter is necessary for certain multiplayer games that update a lot as Steam doesn't always immediately pick up that there are updates to them if they were updated after Steam was started. This is most noticeable with Team Fortress 2 but also Half-Life 2: Deathmatch, Day of Defeat: Source, and Counter-Strike: Source as they all use the same game engine (CS:S's updates are delayed, though). Likely CS:GO and DOTA2 also have this problem as both are still in beta testing with CS:GO launching in 12 days.

Running the game in offline mode should make that screen a lot shorter, but also disables achievements.

(When it sits there for a few seconds going "Preparing to launch Portal" or whatever, and you want to tell it what's with all the preparing and to stop preparing and just go. :-))

Unfortunately, the Internet doesn't run at Ludicrous Speed.

On the referenced blog, I asked whether they'd repeated the test for a 64-bit Linux distro to directly compare to the 64-bit Windows installation they used. Unfortunately, my comment there got deleted. Does anyone have any insight as to what effect switching to a 64-bit distro might have? On one hand, x86-64 has a reputation for being more compiler-friendly than x86-32, what with more explicitly-named registers and all the other goodness. On the other hand, it'd have to sling around longer pointers (and possibly waste more space on 8-byte-aligned data structures? Is that true?). What would the net result likely be?

Put another way, I wish they'd eliminated that rather large test environment variable before publishing their numbers.

Not necessarily better than OpenGL, but better than 270.6 fps.

Valve's blog post, near the bottom, indicates that they plan on fixing the hang-up with Direct3D, now that they know that the hardware can do better than 270 fps.

The test setup is provided on their blog: http://blogs.valvesoftware.com/linux/

Here is an excellent blog post by a Valve Software employee about the potential of augmented reality. Basically, the real thing like what you see in the video above while the guy is cutting the cucumber is very hard. Things like perfect motion tracking, contextual awareness, seamless overlays are science fiction at this point. But this is a very compelling scenario and very smart people are working on it so sooner or later it's happening. Hopefully Google Glass will get us one step closer. Ironically, one of the best uses for it is real life ad block. Imagine riding down the freeway and every billboard is replaced by a giant sequoia. Or a mushroom Smurf house.

If you're curious, you can actually read the Valve Employee Handbook at their site:

http://www.valvesoftware.com/company/Valve_Handbook_LowRes.pdf

From the handbook and other things I've read, I think nobody at Valve is told what to work on... period. They work on whatever they want / think will be valuable. Valve sets the hiring bar so high that this hasn't been a problem. And, even if it was, they do periodic peer reviews that would expose the truly weak links.

It's a really, *really* interesting model. Valve, having had the huge success that is Steam, is in the relatively unique position of having loads of cash and operating in an open-ended market that rewards creativity. I sometimes wonder if it could work in more traditional companies / businesses. I imagine it could work at some place like Microsoft or Goole that's flush with cash (if they weren't public companies, that is). I doubt it would work well at a smaller company whose life depends on executing well on a very narrow strategy.

To put it more succinctly:

Valve understands that a *fun* game will be fun. As long as the graphics are good enough to support the gameplay, the ame will be fun whether you're running it at 2006-era graphics or at 2016-era graphics.

Valve understands this. They make a fun game, then make it run on the lowest hardware they expect will be commonplace. They design their system to be scalable. They allow features to be disabled, have an extensive set of shader fallbacks. Examine this somewhat-outdated wiki page detailing the features enabled and disabled for each DirectX level in the original Half-Life 2. That's no longer current, I believe - they patched it to use a newer engine revision that I think dropped support for some of the lower levels, and I know it added higher ones.

I have played that game many times on many different computers. It was fun on my Athlon 3000, Radeon X700 build. It's fun on my dual-Xeon, Radeon X1900 rig. It was fun on my Core 2 Duo, GeForce 9600M laptop. It was fun on my Phenom II X3, Radeon 4830 build. It would probably be fun on this new Core i7, GeForce 660M laptop, but I haven't replayed it yet on this.

The only machine it wasn't fun on? My ancient Pentium II, Rage Pro laptop, and that was because it glitched like crazy - corrupted textures, BSOD after a few minutes. The machine just could not handle some of the things that were actually necessary for gameplay - the Havok physics (used in puzzles), the fade-in shaders (used for one-way gates), the dynamic lights (used to highlight gunfire). Remove those, and it wouldn't have been a fun game, so Valve just didn't remove it.

But the rest? Water refract/reflect shaders? Rim lighting? Normal maps? Soft shadows? Turn them off if necessary. They don't make the game less fun. Less immersive, perhaps - that's why they have them as an option - but the fun doesn't change.

And the fun is what is important.

Pssh, digital hats are for the Steam servers for the rumored Steam box, not Ouya.

(For the record, Valve launched a Linux blog yesterday, don't think /. mentioned that anywhere)

Go ahead and go over to the valve linux blog and email them from there. They really seem interested in getting input.
http://blogs.valvesoftware.com/linux/steamd-penguins/

The mail address is in the text and on the sidebar on the right.

Don't believe anything from Valve with regards to dates, even if I comes from Gabe himself.

Hey! Get a life, a Half-Life. Work for Valve.

"That means at some point in the next 7 months anyone running Linux will be able to download Steam and start playing a number of games" you are clearly unfamiliar with the concept of https://developer.valvesoftware.com/wiki/Valve_TimeValve Time

Yes, Source uses scripting. I think you're confused between Source and most Valve games, which tend to use scripting very lightly, and usually not during gameplay. They prefer to use the game object I/O system for performance reasons.

Having said that, the AC you're referring to is 100% correct that the lion's share of porting Source games (which doesn't include Half-Life, of course) is porting Source, in particular, the stuff to do with infrastructure, packaging, and low-level I/O. The hard parts have already been done, since Source already works on XBox, PlayStation, PC and Mac. The game object layer, where most of the custom C++ stuff for each individual game lies, is already portable.

...given programming legend Michael Abrash (now currently at Valve Software) just announced that he's currently researching wearable computing more or less as a direct result of Neal Stephenson's book Snow Crash!

His post on the Valve blog is really interesting and worth reading.

What I mean is that for things like the "pellet launcher" and "pellet target" rather than placing a single object you have to build it up out of many peices in the map editor.

http://developer.valvesoftware.com/wiki/Creating_an_energy_ball_launcher_and_catcher

9 steps to built a launcher and 25 steps to build a target!

Linux has no games? Nice try, Troll. Obviously, you've never looked in the repositories - there is a nice little list of games in mine. Just as obviously, you've never explored the possibilities of Wine. Literally thousands of Windows games will load via Wine. Then there are the commercial tools like Cedega. They load most of the "latest and greatest" games commercially available. Not to mention that any online games that rely on Java or Flash runs great on Linux. Oh yeah - Steam for Linux. I don't subscribe to Steam, so I googled to be sure. http://developer.valvesoftware.com/wiki/Steam_under_Linux

"Linux has no games" is nothing more than trolling. Go find something better to troll.

You are correct. Quoting Erik Johnson at Valve:

When we were getting very close to releasing Half-Life 1 (less than a week or so), we found there were already some projects that we needed to start working on, but we couldn't risk checking in code to the shipping version of the game. At that point we forked off the code in VSS to be both $/Goldsrc and /$Src. Over the next few years, we used these terms internally as "Goldsource" and "Source". At least initially, the Goldsrc branch of code referred to the codebase that was currently released, and Src referred to the next set of more risky technology that we were working on. When it came down to show Half-Life 2 for the first time at E3, it was part of our internal communication to refer to the "Source" engine vs. the "Goldsource" engine, and the name stuck.

Now you can get a version of the engine for nothing

Alien swaem has been free for a while and if you own any ATI or nvidia card you've been able to get half life 2 deathmatch and half life 2 lost cost free (the nvidia version of the offer also includes a demo of portal and another game i've never heard of) too though this isn't widely advertised (I found the page below while reading up on half life modding). There was also a free portal offer a while back.

http://developer.valvesoftware.com/wiki/Half-Life_2:_Deathmatch

but these freebies and other promotional offers did not include the SDK, to get that you have to actually buy a game.

Now you can get a version of the engine for nothing (TF2) the SDK is free too.

If you read TFA this was prompted by someone asking if a purchase from the mann co store (which according to the previous /. article is how you get an uncrippled copy of TF2 now a crippled copy is free) would enable the SDK. It would appear that while they answered yes to his question they also decided to make it free for everyone (though they haven't actually done so yet).

They're working on it. It'll be ready in late 2011

How do we now describe the expected delivery date of vaporware

"Valve Time"

The technique for Phong Shading was introduced in 1973 as an improvement to Gouraud Shading, but was too computationally intensive to be used for graphics back then. This is no longer the case.

It was too computationally intensive for *realtime* rendering in 1973, but clearly not out of reach for the kind of modeling software NASA people were using ...

Also, it should be noted that realtime phong shading was already common in demos/intros running on 33 MHz 386 CPUs back in the 90s

The technique for Phong Shading was introduced in 1973 as an improvement to Gouraud Shading, but was too computationally intensive to be used for graphics back then. This is no longer the case.

Yo dawg, I heard you like cars, so we put a car in your computer so you can drive while you compute!

But does this car have valves or is it steam powered? What about the car's specific impulse?

People write bots for FPS games all the time.

They write aimbots, not player bots. You cant just run a bot, wlak away, and be level 50 like you could with Glider.

Um, no. Hell, just google "Call of Duty bot" and check the results.

Bots are fairly common for FPS games. It's just that most of them are intended to be NPCs and show up in games as such. Engines like Source even have AI Node files for maps so that the bot AIs know how the map is traversed. These same types of things are built into the Unreal map format.

Player bots would just be external programs that operate on the same principle, but they would pretend to be a specific player instead of identifying as a bot.

Since FPS games have keybinds, a bot can just emulate these keybinds rather running through the program. The problem is making the AI smart enough to properly interpret what the host computer sends back to it and respond to it appropriately.

On a similar note, I've heard that Glider does much more than just kill enemies. For instance, it can also follow flights paths back to a capital city, go into an auction house and list things up for sale automatically. It can also likely sell junk items to vendors, repair armor, run back to the corpse if the player somehow died (and resurrect).

It likely can also attempt to engage PvP players on PvP WoW servers... you did know WoW has PvP servers?

Blizzard could single handedly make linux a gaming platform. They already release OpenGL versions for the Mac so technologically, they are a short hop from a linux client rather than a giant leap

The Source engine does OpenGL on the Mac now too, so Valve is in the same position. The Steam client partially runs on Linux (natively, that is, not under Wine) too, although Valve is denying there's an actual Linux client for end users now. (Of course, Michael Larabel of Phoronix claims otherwise.)

Time will tell if it will come to pass or not. In the meantime, write an email and express your interest.
http://us.blizzard.com/en-us/company/about/contact.html
http://www.valvesoftware.com/contact/
Protip: writing a physical letter carries a whole lot more weight. Do that if you can. ;)

Valve changed its tune at E3 2010, announcing that their first official PS3 game is Portal 2, due out in February*.

*June or July if you take Valve Time into account.

If you think June or July you obviously don't have enough experience with Valve. Christmas may be a long shot.

The version of the Source engine that TF2 ran on and the version that CSS ran on were not the same until about a month ago. Valve released several changes to the Source engine to go along with the Orange Box. CSS wasn't updated to run on that version of the engine until recently. Benefit the changes were probably not to drastic so the port work to either update CSS or port the old engine shouldn' t have been to extreme, but it would not be trivial.

The version HL2 ran on until 6 months ago was the same as the one CS:S ran on, and that one runs just fine on the PS3.

No, the real reason is because Valve hated the PS3. Valve's CEO Gabe Newell said that the PS3 is "a total disaster" in 2007. Orange Box PS3 was ported from the 360 version by EA.

Valve changed its tune at E3 2010, announcing that their first official PS3 game is Portal 2, due out in February*.

*June or July if you take Valve Time into account.

Well they deny it but they are hiring a Software Engineer whose role will include "Port Windows-based games to the Linux platform" http://www.valvesoftware.com/job-SenSoftEngineer.html I am sure they just wont announce anything until they have it ready

I came across this job post at Valve just recently. To save you from having to follow the link, it includes the item
"Port Windows-based games to the Linux platform".

Just sayin'.

So, if now is 5 days from now... maybe they're releasing it in 4 days! http://developer.valvesoftware.com/wiki/Valve_Time

February, they say? Here's a guide to help you translate from what Valve says to what Valve actually means: http://developer.valvesoftware.com/wiki/Valve_Time

The Release Date is a Lie! (I just had to ;)

More likely than I'd like to admit, unfortunately.

IT'll be interesting to see if they make it.

Multiplayer games however, this could work. I find: ...

Most of the MP games I've got really into have stagnated from lack of fresh content as the game gets "old". Often these games go on for years longer thanks to some good modding, though fan made maps rarely fare so well.

Strangely, Valve has combated this in Team Fortress 2, but Valve hasn't try to monetize it. The latest (11th major) update came out last week with 4 new maps, 4 new Engineer weapons, and 38 new Engineer achievements. This is the last of TF2's nine class updates. Also, fan made maps are quite popular, and some even make it into the base game distribution during updates (ctf_turbine, cp_fastlane, cp_egypt, cp_junction, arena_watchtower, pl_hoodoo, cp_frieght, and cp_coldfront are the ones I can think of).

- MP games often come out with too much content for people to get properly into, resulting in a long lead time of people being inexperienced with the levels.

TF2 only started with only 7 maps comprising 3 (4 actually) game types, and no unlockable weapons or items.

TF2 today has (if I'm counting right) 34 maps (26 Valve-created maps, 8 community-created maps) comprising 7 (8 actually) game types, 34 unlockable weapons (27 Valve-created replacement weapons, 7 community-created replacement weapons (Medic and Spy don't have any yet)), and 49 hats/misc items (30 Valve-created hats/misc items (3 per class, and 3 generic), 19 community-created hats/misc items (2 per class, but Medic has 3)). Note: I'm ignoring the 10 specialty hats/misc items and 2 reskinned weapons that aren't randomly dropped.

Valve is also planning on adding the winners of the Polycount Contest to the game... they were supposed to announce the winners sometime this week, but that announcement was subject to Valve Time.

- related to above, many people tend to pick a few favourites and just ignore other maps, even if they're still quite good. These maps may offer more value if introduced when they are adding freshness as the old favourites are getting a bit tired.

Certain maps in TF2 are disliked. tc_hydro seems like a good map on the surface, and is one of three maps that has a developer commentary. Valve clearly put a lot of effort into it. However, it is easily the most stalemate-prone map in the game, which in turn makes it unpopular.

As for the new maps, the people on the OCRTF2 servers, which I'm an admin on, have already chosen maps they like and maps they don't. For instance, plr_hightower is disliked by some... it's a relatively small map and has this tendency for one team to steam-roll the other. pl_upward seems to be well liked. cp_coldfront is a map that we already had on our servers in its release candidates (Valve adds community maps in some updates, cp_coldfront was added in this update), and it... can be good or bad, depending on the teams. pl_thundermountain, I'm not sure about as we don't seem to play it as often as the others; I thought it was interesting, though, even if the map does sometimes get stopped before it reaches the final stage.

- the high initial price puts people off because MP games are "high risk" - good balance is hard to achieve.

TF2 has the advantage of being part of the Orange Box. OB had an MSRP of $50 at launch in late 2007, and has an MSRP of $30 today. It also includes all of HL2 (original plus both episodes), Portal, and TF2. On Steam, TF2 alone sells for $20... but a boxed copy from Amazon sells for $9.99. The boxed copy needs to be registered to a

I am aware of narbacular drop (the original game). However the fact that not many other people do kind of shows my point. It isn't well known and it wasn't nearly as good because while the mechanics were similar and engaging, the graphics and writing was meh. Before the portal project Valve hired 2 (or 3, I forget) writers and basically let them do whatever the hell they wanted for a while until the portal project started and they decided to work on it. That's one of Valve's strengths. They let people do whatever they want. You want to spend 10 years making a game? You can! That's how valve time happens. They don't force deadlines because they're so far in the black they don't have to. A smaller company can't do that sort of thing because they need a product on the market before they all starve to death.

Valve can afford to hire people they don't even need yet because they think they might be useful later. Their Cabal system of game development requires ridiculous amounts of time and money to implement properly but it makes games like HL, HL2, TF2 and Portal.

But enough of my flagrant Valve Fanboying.

Yes special effects will get cheaper, but a movie is more than special effects. You need costumes, you need sets, you need film equipment and most importantly, you need actors. no matter how cheap equipment gets, people don't. Extras, crew, actors etc. For a full scale movie you're probably going to need a significant financial investment. Plus marketing costs (viral will only get you so far) food for the actors your already paying if you need to travel anywhere. Travel costs etc. You can't use JUST cgi. Lucas tried that. It was terrible.

Or simply get a copy of Portal, which is available for free until the 24:th.
All Source games gives you access to the Source SDK base, which includes all models and similar content from Half-Life 2. (Their wiki states that this excludes some free promo games, but I doubt the Portal one's part of it.)
Gives you access to most community mods too, as an added bonus.

Is there even a market, let alone a convention, for selling game concepts?

Nope. Quite frankly, the only way its going to get made is if you do it yourself. I'd suggest using an established engine to cut development time/cost to a minimum and going with a digital distribution service like Steam to bring the product to market.

Possible engines include the Blender one mentioned, and http://www.ogre3d.org/ since they are both free and open source

Is there even a market, let alone a convention, for selling game concepts?

Nope. Quite frankly, the only way its going to get made is if you do it yourself. I'd suggest using an established engine to cut development time/cost to a minimum and going with a digital distribution service like Steam to bring the product to market.

http://www.valvesoftware.com/job-SenSoftEngineer.html

Well, colour me impressed: so it does. Good on Valve for grasping the nettle.

Steaming from Valve Software's?

thats not what Valve said

related

http://developer.valvesoftware.com/wiki/Valve_Time

There are no tools for the server admin to make things like md5 or cvar checks

MD5 checks are built into the game. This is controlled through the sv_pure cvar and the pure_server_whitelist.txt file (when sv_pure 1 is set, this file is used to determine which files the client can use that don't match the files on the server).

As for cvars, some are already checked and disallowed (anything marked as a cheat in the game).

Also, the Sourcemod plugin "Kigen's Anti-Cheat" has more rigorous cvar checks.

no screenshot facility to check players

Spectate, left or right-click until you find the problem player, then hit F5 to take screenshots (at least in TF2). Or even better, spectate and use the client-side record command to record a demo, which you can later play back. I believe SourceTV can be used to record demos automatically, but I don't think it records one for each player.

or even the ability to kick a player

OK, now you're just being ignorant. Source servers have built-in commands to kick and ban people through either the server console or through the game using the RCON commands (which require a password that you set in the main server configuration).

This is in addition to addons such as Sourcemod, which include kick and ban abilities for any users matching specific Steam IDs, and the optional Sourcebans module, which allows bans to be synchronized between servers.

There are no tools for the server admin to make things like md5 or cvar checks

MD5 checks are built into the game. This is controlled through the sv_pure cvar and the pure_server_whitelist.txt file (when sv_pure 1 is set, this file is used to determine which files the client can use that don't match the files on the server).

As for cvars, some are already checked and disallowed (anything marked as a cheat in the game).

Also, the Sourcemod plugin "Kigen's Anti-Cheat" has more rigorous cvar checks.

no screenshot facility to check players

Spectate, left or right-click until you find the problem player, then hit F5 to take screenshots (at least in TF2). Or even better, spectate and use the client-side record command to record a demo, which you can later play back. I believe SourceTV can be used to record demos automatically, but I don't think it records one for each player.

or even the ability to kick a player

OK, now you're just being ignorant. Source servers have built-in commands to kick and ban people through either the server console or through the game using the RCON commands (which require a password that you set in the main server configuration).

This is in addition to addons such as Sourcemod, which include kick and ban abilities for any users matching specific Steam IDs, and the optional Sourcebans module, which allows bans to be synchronized between servers.

Half-Life did use the Q2 engine.

It actually used the Quake engine which was modified with some parts of the Quake2 codebase. The modified Quake engine was known as Quakeworld.

http://developer.valvesoftware.com/wiki/Quake_Engine_Hierarchy

Admittedly I haven't tried doing the conversion recently, but in the past one would have to solve a large number of compilation issues related to compiler bugs and newer versions of libraries. Looking at the developer wiki article about compiling under 2008, this process seems easier than it used to be, but it's still extra unnecessary steps.

Upgrading projects is not equal to the change in the compliers. Here's the guide on getting the Source SDK working in VS2008: Compiling under VS2008.
Actually, upgrading from 2003 to 2005 was far worse, as it required changes to individual files. Fortunately Valve cleaned most of these problems up, making 2005 to 2008 easier.

There is much that can be done with a clever client to hide and compensate for latency, see:

http://developer.valvesoftware.com/wiki/Latency_Compensating_Methods_in_Client/Server_In-game_Protocol_Design_and_Optimization