[Opinion Topic] Demo2 was not ready (and other thoughts...)

Anything and everything

Moderators: jelco, bert_the_turtle, Chris

User avatar
Miah
level5
level5
Posts: 3155
Joined: Mon Oct 13, 2003 7:28 pm
Contact:

[Opinion Topic] Demo2 was not ready (and other thoughts...)

Postby Miah » Mon Sep 26, 2005 4:43 pm

I was going to post this in the Beta Tester's section, but since the Demo has been released, everyone would now be able to comment on this. So...

As some of you know, I am rather disappointed that Demo2 has been released in the current state it's in, at least from my own perspective. As I am seeing from this post and this post, I am not the only person who feels this way.

In my personal opinion, it is because of a few timing problems in the code, making it such that scripts are not running when they should be. I am told by other beta testers that they do not see this issue, even when they intentionally bog the game down.

However, many other beta testers, such as NeoThermic (just an example, nothing against him), have decent CPUs and GPUs. The computer I was testing on, however, was one I made to be dispensable. This would represent your average person, who would be playing Darwinia on a family computer; not meant for performance, but to serve a general (not a specific) purpose.

This computer serves the purpose of showing what someone on a particularly tight budget would get: a Celeron (shudder) processor, a GeForce4 MX, and 1 GB of RAM. The entire setup would be about 400 dollars. Reasonable.

However, this computer had some major issues with the scripting. Every time that the Doctor said the Red Darwinians were attacking (since beta2, beta1 was fine), there were none to be seen for almost 90 seconds. Had I not tested the first beta, my initial reaction would have been "WTF!" As it was, I was already wondering what the hell had just happened.

But in honesty, scripting probelm may very well be the least of Darwinia's problems. I have worked with snoonan (he is, in fact, the assistant manager/developer of Onlink) to find out what the source of his major crash problem is and we have traced it as far as we are reasonably able.

As it would appear, Darwinia is passing something that iglicd32.dll doesn't like. This is a necessary DLL for the graphics card he has. Most people would readily assume that the driver is too old/corrupt and should be updated/reinstalled. We checked both of these. The driver was at the most current version, and he reinstalled regardless, to be certain it was not simply corrupt. It certainly wasn't. Darwinia still bombed on startup.

Normally, I wouldn't even be posting about this issue, namely because it's not mine, and it's already been stated. Except that his problem should not be ignored. As I will quote from him, in a private conversation:
snoonan wrote:Any Sony Vaio PCG-TR series notebook would have this issue, along with about any other brand, unless they specifically bought the computer with a nice graphics card. Laptops have a tendency to come with Intel graphics chipsets.

Is he right? I'm not one to say, but someone else with an "Intel graphics chipset" should be someone to test this. Likewise, my bug report which I stated was in fact a new low-FPS bug cannot be tested just by someone who bogs down their performance. NeoThermic did this (kudos to him for his effort) and was not able to replicate this problem.

Besides this, I can also see a major issue that is also bogging down every other computer as well.

First, I would like to point out that we constructed a letter to Introversion regarding partial register stalls. These are performance hindering ASM-level bugs that, when hit, forces the CPU to take several cycles to clear the stall. We counted only SBB EAX,EAX stalls. We found there were as many as 27. At least 16.

After this letter, we received no reply from anyone, and after inquiring Tom about if the message was received, I got as a reply that Chris and Andy are looking into it, but are more concerned about graphics handling.

Now, I don't have a program debugging database available to trace what functions have these stalls, but I can probably tell you where one is. It's glaring.

Get a mass of a thousand of so Darwinians from a Biosphere-type incubator and have them all follow an officer. After a small time, you will notice that your FPS goes to the shitter. Sure, there are a lot of Darwinians to process. But now throw a grenade to the center of them. Watch how fast they scatter. The FPS picks up with the AI response change. My guess is that the AI for following an officer has one of these partial stalls. Since it's called hundreds of times per frame in some situations, you can imagine was an effect it would have on the game play. There's your graphics handling. :/

These stalls can also reside in places we cannot stress-test, such as sky rendering code or airstrike formations, but we have no way of knowing without a .pdb.

Now, we know that Introversion is suffering some fiscal woes, and that this new Demo was made to impress the public and to put Darwinia back in the spotlight. We even offered help hunting down the stalls (since we wouldn't even need the source for that, but instead with a .pdb to find what functions were throwing them).

We would have loved to help, and still are willing, but cannot. And to us, the new demo, made to impress, does not.

I'm sorry if this seems blunt. But I feel that this new demo was released prior to being truly ready.
User avatar
prophile
level5
level5
Posts: 1541
Joined: Fri Feb 18, 2005 4:17 pm
Location: Southampton, UK
Contact:

Postby prophile » Mon Sep 26, 2005 4:47 pm

Introversion released Darwinia before it was fully ready... and now demo2 as well. I think IV needs to think more about quality control in this case, because if it doesn't work the first time people are less likely to come back and try it again (and therefore less likely to pay for it)

So basically I agree with everything Kyuuketsuki is saying here :|


Also, I disagree with what TGF is about to post.
Steven
level2
level2
Posts: 140
Joined: Mon Jun 20, 2005 4:16 am
Contact:

Postby Steven » Mon Sep 26, 2005 6:05 pm

I agree with Miah, everything he has stated is correct. I'd like to explain what iglicd32.dll is. It's basically the entire Intel "Extreme Graphics 2" chipset driver. Yes, the Intel graphics chipset really sucks to begin with, but I can play Darwinia 1.2x at 1280x768 (wide screen) with 30 to 50 FPS no sweat. So this chipset should work with Darwinia 1.3, and it does... However, the chipset does not work with Demo2. Peculiar? I think so.

Users who go out and purchase a laptop/notebook probably don't know straight off what the best graphics card is, or why they'd need it. Most users regret this error, as I did. (Not to fear, my next notebook will be as powerful as I can get.) Though Dell doesn't even say what graphics card their machines have unless it's something spactacular, most of them use the Intel "Extreme Graphics" series. IBM says straight away on most of their machines that it has an Intel graphics chipset. Sony does the same. And because more and more people will show up with this issue, it is not minor, and it really needs to be dealt with.

There are even some oddity bugs that have been around since v1.0, and apparently have never been touched. One example I will provide here is the Mine level. If you set terrain detail to 'Medium' or 'Low' (I believe this is partly the cause), the moving mine carts will go through the ground when they are supposed to move through the trunk port (heh, a very Stargate-esque trunk port).

I fear that Introversion is losing their touch in a sense: They seem to sacrifice quality for a timely release now. IV doesn't DO deadlines, right?

Miah and I have offered to help multiple times, but we haven't received responses. IV seems to be hesitant to let us help. We would rather offer real help for free than to worry about watching IV go for bust and not be able to do anything about it.
User avatar
The GoldFish
level5
level5
Posts: 3961
Joined: Fri Mar 01, 2002 9:01 pm
Location: Bowl / South UK
Contact:

Postby The GoldFish » Mon Sep 26, 2005 6:23 pm

I disagree that IV released Darwinia before it was 'ready', since it depends how you define ready - I was however surprised with the single day's worth of final version testing, in which a potentially game crippling bug was fixed. I do however think that if they want the game to perform faster they need to optimise it in some way CPU wise, since this is and always has been the thing that affects Darwinia performance the most after having a reasonably capable graphics card.

Regardless of if they're fixed or not, the stalls really should be pin pointed so they're better understood.

Suggesting that IV needs to think more about QC is all well and good, what are your ideas about how they can do that? I mean, they got all the beta testers to try and run it, and only one guy came back saying it had a hard crash on start. Now, from what I can tell, it would be quite hard to diagnose the exact problem from that information (all credit to snoonan, that's pretty much as much information as he can give) - unless I'm horribly wrong, they'd need a far more debuggy version of Darwinia which could ascertain the problem. And knowing IV, that'll create more bugs than it will help fix.

The only reason I would be weary of having Kyuu do anything is because he tends to keep things to himself too much, or at least, away from me :P. That aside, I don't deny he can probably help with Darwinia's performance alot. I expect other people can probably help as well, in a good lot of different areas. But we're never asked, so we don't.

I think that IV isn't losing their touch at all; I'd say that in fact, very little has changed.

(edit - got snoonan's name wrong :shock:)
Last edited by The GoldFish on Mon Sep 26, 2005 6:45 pm, edited 1 time in total.
Steven
level2
level2
Posts: 140
Joined: Mon Jun 20, 2005 4:16 am
Contact:

Postby Steven » Mon Sep 26, 2005 6:38 pm

The GoldFish wrote:I disagree that IV released Darwinia before it was 'ready', since it depends how you define ready - I was however surprised with the single day's worth of final version testing, in which a potentially game crippling bug was fixed. I do however think that if they want the game to perform faster they need to optimise it in some way CPU wise, since this is and always has been the thing that affects Darwinia performance the most after having a reasonably capable graphics card.


Thankfully I never said Darwinia wasn't ready, that was Prophile. I said that v1.3 and Demo2 weren't ready.

The GoldFish wrote:Regardless of if they're fixed or not, the stalls really should be pin pointed so they're better understood.


I'll post a link to my article about partial register stalls shortly.

edit: Here it is! Steven's PC Tips and Secrets

The GoldFish wrote:Suggesting that IV needs to think more about QC is all well and good, what are your ideas about how they can do that?


Mainly, they could listen to bug reports more readily (I posted mine a couple weeks before 1.3's Beta2 was 'released', and they haven't replied to it, nor fixed it). Additionally, they could extend the beta cycle. They put the release build up within 2 days of saying they'd hit 'release', and in spite of my bug report, still no reply. Miah's bugs still remain unfixed as well.

The GoldFish wrote:I mean, they got all the beta testers to try and run it, and only one guy came back saying it had a hard crash on start.


Wouldn't it make sense that they want to find out why? I'd say such a crash is a severe issue, especially since the bug occurs only in Beta2 (not in v1.3). Secondly, as I just said, Miah's bugs are still unfixed as well.

The GoldFish wrote:Now, from what I can tell, it would be quite hard to diagnose the exact problem from that information (all credit to snooman, that's pretty much as much information as he can give) - unless I'm horribly wrong, they'd need a far more debuggy version of Darwinia which could ascertain the problem. And knowing IV, that'll create more bugs than it will help fix.


I've submitted such bug reports to people such as the guy who makes Cain with that information and he successfully tracked down the bugs I've submitted. With the information I've given, it shows the exact address in the binary where the crash occurred. By using a .pdb or .map file, one can easily discover which function is causing it.

The GoldFish wrote:I think that IV isn't losing their touch at all; I'd say that in fact, very little has changed.


I guess we'll have to wait and see who's right. I certainly hope you are correct, because I really enjoy IV's work.
User avatar
Chris
Introversion Staff
Introversion Staff
Posts: 1172
Joined: Sat Nov 25, 2000 7:28 pm
Location: Cambridge, UK
Contact:

Postby Chris » Mon Sep 26, 2005 6:51 pm

I think we need to answer a couple of points here.

- Beta testing on the new Darwinia demo began on August 22nd with 100 testers, and was expanded to 200 a couple of weeks later. In total we spent a whole month looking at the reported bugs and fixing them, with 200 beta testers and everyone at Introversion playing as much as possible. This is not to say the finished version is without a bug - no release is perfect - but we did what we could, and to spend any longer than a month would be madness. The full game was beta tested from scratch in just a couple of months!

- To suggest that IV's standards have dropped recently is madness. Have any of you played Uplink v1.0? We've come on in leaps and bounds since then. It is simply impossible to release a product without bugs when you only have 1 programmer - we try the best we can but invariably things get through. That's why we're up to v1.3 now with the Darwinia patch, and undoubtedly there will be more. The new demo is certainly not perfect, but its a huge step up from the first demo, which I might remind you didn't even have a tutorial. How many companies do you think would still be patching and promoting their game full-time, six months after launch?

- 'They seem to sacrifice quality for a timely release now' is one of the funniest statements i've yet read. This demo has taken nearly five months of our time. Even by IV standards this is insanity, with us spending way more time than planned making the demo as cool as we could.

- The 'potentially game cripping bug' fixed on the final day (I presume you mean the performance problem) was extremely simple - in fact it was a bug we've seen before, and accidentally re-introduced. The fix was trivial and obvious and we knew it didn't need extensive testing.

- I am sorry to be blunt, but the Intel Graphics Chipset is a crock of shit. Darwinia will run on pretty much anything (I once saw it running on a Voodoo3 graphics card, albeit at 5fps), but we don't promise good framerates if you don't have the hardware. If the new demo does not run on that chipset then thats a bug we need to trace, but the framerate will always be low.

- Finally, we have read your emails about the register stalling issue. Andy plans to look at it at some point, but he no longer works for us fulltime. We agree it needs investigating, but the performance gains will be tiny. Darwinia is primarily 'Fill Rate Limited' - which means the speed at which your graphics card can fill polygons on screen largely determines the speed at which darwinia runs. This is why Geforce4 ti4600's are excellent (tons of graphics bandwidth and fast fill rates), but Geforce4 MX's are terrible (the opposite). The next most important factor is cpu speed, which only really comes into play when there are large battles involving Darwinians. Our assessment (from knowing the code base pretty well) is that the lost cpu time due to register stalls is very small, compared to (say) the Darwinians code for running away from threats. It's all about priorities, and even if we were working on performance enhancements, register stalls would be low on the list.

- Having said all of that, we genuinely do appreciate your efforts to help us, and I apologise that we didn't get around to writing a reply to you since you've obviously put a lot of effort into tracking that issue down. You genuinely wouldn't believe how rushed we are during big product launches.
User avatar
The GoldFish
level5
level5
Posts: 3961
Joined: Fri Mar 01, 2002 9:01 pm
Location: Bowl / South UK
Contact:

Postby The GoldFish » Mon Sep 26, 2005 7:08 pm

You seem uh, very confrontational, it really is all about you :P (this is RE Snoonan's post before, not Chris', who posted while I was making mine)

By 'better understood', I mean, are we losing 5, 50 or 500 cycles, not 'what is a register stall?'. That first step is by far the most important.

You know as well as I do that not all bugs are the same - I'm pretty sure I said it would be hard to fix with only that information. Just because you know which function is causing the problem, doesn't mean you know (as it has been indicated would be important) exactly what the data being passed which is causing the problem is, and why it only causes a problem for you. Unless I'm horribly wrong. I think that would be the hard part, maybe you disagree, I'm just trying to be clear on what I mean.

I can quote you a bunch of bugs that I know of that haven't been fixed for several versions. Maybe they're not as crippling as yours or Miahs are/can be, but there are plenty of bugs that don't get touched, so don't feel particularly unique. I'm sure people are aware of the unfixed bugs, so please realise that my opinion is in spite of the existing problems, not ignorent of them.

And I think it's quite possible you didn't completely comprehend what I meant by my final statement :P

edit RE the 'simple bug', yes, I'm sure it was quite simple otherwise you wouldn't have put it out a day afterwards :P - I was just surprised is all.
elDiablo
level5
level5
Posts: 3111
Joined: Thu Mar 14, 2002 12:23 pm
Location: London, UK

Postby elDiablo » Mon Sep 26, 2005 7:28 pm

Chris wrote:...
- I am sorry to be blunt, but the Intel Graphics Chipset is a crock of shit...


This has to be quoted. Simply...

In response, I think the demo was pretty damn good (in gameplay and lack of bugs). By the time I got each demo (excluding the RC beta, which I read the day AFTER the demo was released), each bug I found had already been reported... And the demo runs fine for me now =/

Anyway. Good quote :D
We dont stop playing cos we get old... We get old cos we stop playing.
User avatar
Dev
level1
level1
Posts: 67
Joined: Mon Sep 12, 2005 9:41 pm
Location: Poole
Contact:

Postby Dev » Mon Sep 26, 2005 7:42 pm

Chris wrote:...
- I am sorry to be blunt, but the Intel Graphics Chipset is a crock of shit...


That truley is so
-------------------------------------------------------
Darwinia - Realtime Metaphysics Engine
Dev - Realtime Messmaking Engine
Devwinia - Realtime Physicsmessing engine
-------------------------------------------------------
User avatar
Josh mc
level3
level3
Posts: 465
Joined: Fri May 02, 2003 8:30 pm
Location: :noitacoL . .///////|\\\\\\\ . . . . . . |..---...,...---..| . |........|........| . |...\__.__/...|
Contact:

Postby Josh mc » Mon Sep 26, 2005 7:46 pm

...I am sorry to be blunt, but the Intel Graphics Chipset is a crock of shit...


I wouldn't trust the handily named "Extreme" Chipset to render Ms Word with ease. Not many people with that chipset can expect to play new games. The Sims 2, the most widely known game to the kind of people who buy things with "Extreme" graphics, won't run on them. They are so poor as to be unbelievable. At my school some PCs run them and some have no discernable graphics card at all, and there is literally no performance gap outside of benchmarking.
User avatar
edd8990
level5
level5
Posts: 1738
Joined: Wed Apr 23, 2003 8:26 pm
Location: Crewe, Cheshire, England
Contact:

Postby edd8990 » Mon Sep 26, 2005 7:54 pm

Complaining that Darwinia wont run on an intel Extreme graphics card is like complaining that it wont run on my 10 year old machine under my bed.

If you want to play new games, you need decent hardware. Don't try and play it on hardware that can't cope.
User avatar
prophile
level5
level5
Posts: 1541
Joined: Fri Feb 18, 2005 4:17 pm
Location: Southampton, UK
Contact:

Postby prophile » Mon Sep 26, 2005 8:01 pm

Chris wrote:How many companies do you think would still be patching and promoting their game full-time, six months after launch?


Blizzard :)
Riddla
level3
level3
Posts: 262
Joined: Sat Feb 28, 2004 6:21 pm
Location: Sooomewheaaaaarrre over the raiaiiiiiiinbooow...
Contact:

Hi slash Bye.

Postby Riddla » Mon Sep 26, 2005 10:46 pm

NCSoft! Granted those are mostly MMORPG's, but even stuff like Splinter Cell: Chaos Theory is still getting patched.

>.>

SAVE THE DARWINIANS. Traces of the viral infection have penetrated the digital worlds operating system.

Even though people like Ubisoft have much bigger teams they are releasing games that need patching after 2 years or more of development. Ubisoft didn't make Darwinia. For me Darwinia and Uplink didn't even redefine genres, they've redefined the meaning of the words "computer game".

Personally I say put the time back into promoting Darwinia and get yourselves a nice fat publishing contract or something, then maybe IV can get the boost it needs to maybe hire more coders or join with another development team and become a true force to be reckoned with. Chris can't be expected to sit, chained to a steel office desk, bread and water in a dish by the side of his plastic coated keyboard, forced to code day and night. o.O;

The F.E.A.R MP Beta lasted for a few weeks before they stopped handing out clients, but I think they'll be testing until release in October. The Single player beta however I couldn't comment on, but if they're spending up until pretty much release week testing the less-focused on half of their game, the single player being the main focus.. well, I think they'll still need to patch the damn thing up, to be honest :/ but if not only can you make games that make people sit up and pay attention (and you do! You do!) but can also minimize possibility of crashes and errors while at the SAME TIME making the smallest yet most addictive and downright niftiest (probably not a real word xD) games on the shelves, it'd put that extra gleam in the IV portfolio. Darwinia combat > F.E.A.R combat anyway <_<

Take it easy, order extra reserves of spit, and get a-shining up. One man can't rule the world. ;)

</cameo>
*Riddla+*, when mind-bogglingly insane just isn't enough, they come to me.
Image
User avatar
prophile
level5
level5
Posts: 1541
Joined: Fri Feb 18, 2005 4:17 pm
Location: Southampton, UK
Contact:

Postby prophile » Mon Sep 26, 2005 10:55 pm

*blinks*

Riddla?

Also, I agree mostly with Riddla.
User avatar
Miah
level5
level5
Posts: 3155
Joined: Mon Oct 13, 2003 7:28 pm
Contact:

Postby Miah » Tue Sep 27, 2005 12:09 am

Chris wrote:Our assessment (from knowing the code base pretty well) is that the lost cpu time due to register stalls is very small, compared to (say) the Darwinians code for running away from threats.

Don't mean to offend, so apoligies if I do.

However, I stated (perhaps in far too many words) on my first post the exact opposite. When Darwinians are to amass near an officer, an AI function takes over (I would assume) to tell them where to go. As one would imagine, the performance bogs down. It would have to, the more darwinians there are.

However, from my *own* personal experience (now, this just might be the rather crap hardware, but maybe not), I find that when I throw a random grenade in the middle of the mass of them, the performance actually picks up, as they run away.

If there is a stall here, which there may or may not be, it would be looping the stall. In large fights, the stalls become dangerously risky. This is, of course, provided they exist there.

Basically, we only wanted to see if that was the case. And offer a quick solution, if available.

Return to “General”

Who is online

Users browsing this forum: No registered users and 8 guests