Whither Thou, Unreal?
or: Quake2 Killer? Maybe, maybe not.
by Bill J. "crash" McClendon


As you all know, Unreal was released on 23 May 1998, and, like many first-person shooter single-player aficionados, I made it a point to get down to the store as quickly as possible to pick this one up. It's been three or four long years in development and, smack talk aside (probably because I was impatient more than anything else), I was really looking forward to this title. I mean, an immersive single-player experience? A real storyline? Non-linear goals, exciting and strange architecture, environments, and creatures? This is what I'd been waiting for since Quake first came out (and, in some ways, was provided by Quake2), and even though I didn't have a lot of money to spend, I was spending money on this, boy howdy you betcha.

Needless to say, even the stated system requirements couldn't hold me back. A Pentium 166 isn't that far above a 133, so I figured with a little optimization I could enjoy a playable (though slightly less detailed, probably) game, right? I'll just take it home and tweak my way through all the menus until I can get a playable game (defined in this review as 20 frames per second minimum); eye candy is dandy, but gameplay is where a game proves its value.

Now, before I begin the review proper, let me just say this: I run a Quake and Quake2 single player level review page. I've played id games since Doom came out, and I've played literally thousands of user-created levels. I've tried other first-person games (Duke Nukem, MDK, etc) but they haven't appealed to me in the way id games have. This is not to say the other games aren't good, however; this is to say that id games have that something that keeps them replayable for me. This, in some ways, is my disclaimer: that this review is not free of bias, but that I am not an "id evangelist" or whatever. Carmack and crew aren't the second coming of Christ, okay? I just like the games they make and the way they make them. End of disclaimer.

That being said, I was very much looking forward to a first-person shooter with a plot that I could really get into (being jaded by years of first-person shooters), and Unreal looked to be the ticket. When I plunked down my US$40 at the software store, I really thought I'd be buying an experience like no other. In that, I wasn't mistaken. Where I was mistaken was in the type of unique experience this would be.

Okay. Enough blather. Installed the game (cringing at the 450mb full install but being resigned to its necessity... I needed every speed advantage I could get), fired it up, and immediately began tinkering with the configuration menus. (The speed of the initial sequence was rather troubling, however; it was disturbingly slow and choppy... and it was a cinematic!) Very nicely laid out configuration menus, really; everything was easy to find and easy to configure. Got everything fixed the way I liked it, and began to play.

Good Stuff About Unreal

First off, the things that are cool about Unreal that I liked. Number one is that it actually has a story. You're on a crashed ship and you're trying to find a way off this rock you've landed on. The exploration, non-linear goals, and "realistic" setting (the two suns in the sky are trés cool) have been done extremely well. Leaving out such information as how much time you've spent in a level, the number of monsters in a level, the number of secrets, and even where the exit is adds to the immersiveness because, besides the level transitions themselves, you're pretty much on your own. I really felt like I was on some other planet.

The second thing that impressed me was the way sounds and "non-critical" items and situations have been implemented. By this, I mean the circling birds way up high that cry out occasionally, the insects and other creatures making noise, and the apprehension you feel trying to figure out what sounds and creatures constitute a threat and which sounds and creatures do not. Also, having the local flora be beneficial to you is an interesting twist to the "health pack"... funny how the little things sometimes have the most impact.

Next, we have the astounding visual effects. The fog in the duct as you crawl out of the ship, the oft-noted waterfall, the ripples you make jumping into and out of water, the flow and current of the water itself, the incredible sky, and the lighting effects (shadows! and focal flares!) all add up to a somewhat overwhelming amount of new visual enhancements. The surprising thing is how well these effects were blended into the setting as a whole; none appeared to be odd, with perhaps the exception of the lens flare/halo one (because it's backwards from "reality"... so maybe that's another "unreal" part...?)

As I began to play more of the game, I found that the scripting action possibilities are very nice indeed, and, in most places, have been implemented well. The first, and most obvious, scripting example is the fight behind the almost-closed door in the first level; this scene adds quite a bit of drama to the game, as do many of the subsequent scripted scenes. (My favorite has to be the first trap, by the way; most excellently done.) I also liked the translator; having alien script that I could somehow mysteriously read without assistance has always been a logical sticking point for me in the past, but Epic have provided a thematic and consistent method to circumvent this.

The monster AI in Unreal is definitely advanced beyond what is currently and commonly available. This is the first time I've ever seen a monster high-tail it away from me when I was kicking its ass; too bad for the monster that it ran away in such a nice straight line, though. The extra animations that the monsters can perform is surprising at first, and proved to be fatal to me in a number of cases. The last thing I expect a monster to do is dive at me, yet that's exactly what the second Skaarj that I ran into did. They also do a number of gymnastic feats (tumbling, rolling, and jumping) that make combat that much more "real". I didn't see much of the "calling for reinforcements" that we had heard about (or am I getting the AI here confused with Half-Life?), and the monsters didn't run away very often (I can remember it happening, well... once), and their tracking abilities were a bit lacking, but on the whole they're tougher to kill than the average. Location damage is very cool, too, especially when you've got the rifle -- one shot and off with his head.

Speaking of monsters, I simply love the idea of non-combatant monsters running around that you have to avoid shooting. It's also nice that they give you stuff when you don't shoot them, and they have a bunch of mannerisms and activities they do when you're not around. The way they wave their arms to get your attention and they way they kinda jog down corridors is too cool... and the way they panic when you start shooting around them is hilarious. Well, at first, anyway; after a while it gets kind of annoying.

Those things, to me, comprise the bulk of the things that I found enjoyable and interesting and new and exciting about Unreal. If you've been following my reviews for a while, you'll have noted some obvious and glaring omissions from the above listing. There's a reason for that, actually... I like to emphasize the positive as much as possible, and minimize the negative. Unless, of course, the negative outweighs the positive -- in my opinion.

Now, here's the warning:

If you like Unreal, and you don't want to read negative commentary about it, read no further.

Still here? Okay.

Bad Stuff About Unreal

First and foremost, the running speed of Unreal is simply outrageous. The requirements on the box state that the minimum is a Pentium 166 with 16mb of RAM. No mention of MMX, no mention of 3d accelerators; just a P166 with 16. Since Unreal has come out, a number of people have been benchmarking their systems against Unreal's somewhat-demanding requirements, and here is a nice list that I've been using for reference. What I found rather remarkable is that the minimum machine required for playable speeds (remember, that's 20fps) in software at 320x240 (near enough the minimum resolution for my purposes) is a Pentium2-266! That's a bit higher than the minimum specs, I'm thinking... and I begin to wonder how Epic arrived at that P166/16 figure. I'd bet that their criterion for "playable" is a bit lower than mine (or the average gamer, really). Am I bitter here? Maybe a little. But realistically, can you honestly say that their minimum machine requirement is a valid assessment of the hardware required for even adequate game play? (For the record, my P133 with 64mb of RAM and Monster3d, with every added feature turned off, gets 1.77 -- one point seven seven -- frames per second at 512x384. In contrast, I get 42.7 in GLQuake and 22.3 in Quake2... with nearly everything turned on.)

I've got a theory why Unreal is so slow, but since I'm not a programmer or game-engine theorist, I can't tell you if it's true or not. If you're still reading this, and I'm wrong, please let me know where I've guessed wrong. I'd really appreciate it, because I'd like to understand why Unreal runs so horribly.

The theory: Unreal doesn't have a lighting table like Quake2 does; it generates lighting information "on the fly". This is why (ironically) someone with a P133 can compile a map in an obscenely-short period of time (under, say, 15 minutes) yet can't even play their own level. This dynamic lighting is really nifty to play around in, because you can generate real-time shadows and lighting effects (though the players, actors, and most of the objects don't have shadows, oddly enough) that are more visually striking than static lighting (a la Quake and Quake2). With this type of lighting, you'll also be able to create a wider variety of lighting and visual effects. The downside, as I've noted (and as others are still discovering, apparently) is that you practically need a workstation to play at a decent speed. MMX evidently helps this out, because I'm guessing that the lighting stuff has been optimized for MMX-capable systems.

The question: Is an engine with dynamic lighting cool enough to require roughly triple the processor requirements of a non-dynamic lighting engine? My answer is "no"; dynamic lighting is pretty cool, but it's not that cool, and in a trade off between playability and eye candy, I'll take playability just about every time. I don't know if it would be possible with this architecture, but having a choice between processor-intensive dynamic lighting and processor-friendly static lighting would have been a nice thing to do, everything else being equal.

Now, the architecture. Regular visitors to my site know that the architecture in a level is usually one of the first things I comment on, and will have noticed that I have yet to comment on the architecture in Unreal. This is because the architecture in Unreal is, for the most part, bland. The poly counts are very low, there aren't many truly complex structures, and many areas are clearly victims of "copy, paste, and rotate" architecture. My first instinct was to blame the level designers; they made the maps, after all. But in any game such as this, the level designers can only build the complexity allowed them by the engine... and if the engine is 85% full of dynamic lighting calculations, that doesn't leave a lot of processor for polygon calculations and line-of-sight calculations and suchlike. I've a very strong feeling that the level designers were very much bound by the limitations of the engine. (And before you say it, every engine has limitations, okay?) For what it's worth, many of the non-connecting areas (everything but hallways, tunnels, and the like) were very well constructed in the minimalist style, and nearly all of them gave me a clear sense of place... which is extremely hard to do when you're limited in the number of structures you can build. I also noticed that much of the "depth" of the architecture was carried off by the texturing instead of by brushes.

Which leads me to my next point -- textures. A 16-bit color palette is all fine and dandy, but there comes a point of diminishing returns in regards to texturing. Basically, this asks the question: How many colors do you need to put in a piece of wood? How many do you need to represent a piece of rock? I'm all for creativity in art and design, but putting 750 colors in a cliff face I barely glance at as I'm running by seems sort of futile. Yes, that's my fault for not admiring the scenery... but I thought I was playing a game here, not going to a museum. Or climbing to the top of a real mountain. Some people might enjoy having a cornucopia of color in their game to make it feel more "real," but I guess I'm not one of them.

One quick note: in a "real" environment, you'd be able to backtrack whenever you wanted to, barring catastrophic events like cave-ins and the like. 'Nuff said.

The much-touted advanced monster AI has been done fairly well, for the most part. The leaping, dodging, spinning, lunging, and pirouetting Skaarj are definitely disconcerting and difficult to deal with... at first. After about ten encounters, however, their AI is predictable and manipulable; I found I could anticipate their moves roughly 80-90% of the time, and I could nearly always maneuver them into my projectiles by firing at one angle, switching weapons, and firing where I knew they were going to jump. It's definitely improved from Quake2, but not as much as I would have expected (based on both the Reaper Bots for Quake and coupled with the two-year "hang time" between the last version of the Reaper and the release of Unreal). I also didn't care for the "pseudo-AI" I found in a number of places. This is where scripting has been used to simulate AI, making the player think the monsters are being clever when in fact they're just following a script.

What about weapons? With one notable exception (the rifle, of course; I'm a big railgun fan, after all), they seem to be seriously underpowered, even with the secondary firing mode. Maybe this is because I'm used to earth-shattering weaponry; maybe not. Be that as it may, the weapons are seriously weak for the most part, and I'm of the mind that this weakness is to make the monsters seem that much tougher to kill. The effects are pretty cool, though, especially the ringing and tinging of the discharging shells bouncing off the walls and floor. I don't know that actually seeing the casings adds anything, though, because in movies and on television you rarely see ejected casings flying around. Also, one thing about seeing the casings on the floor -- the casings are huge, and you'd think that a shell that large would do some serious damage... but it doesn't.

Player physics are, on the whole, pretty good. The player's very limited ability to jump -- well, you can't really call it jumping; it's more like "hopping" -- was irritating, as was the player's extremely limited ability to climb stairs. With the viewpoint being where it was, you'd expect the player to be able to jump vertically maybe half his/her height, and climb stairs roughly a quarter of his/her height. As it stands, the player can jump maybe a quarter of his/her height, and to have difficulty with some of the stair heights I found, the player's knees would have to be about four inches above his/her ankles. This is definitely one aspect of Unreal that's truly unreal. Oh, no climbing is possible, either, though you can crouch.

Errors that I noted in the finished product stem mostly around clipping and texturing oddities. In way too many places, the engine failed to put a texture on a solid brush... so, for instance, I could see through solid switches and bars without being able to walk through them. And yes, I've got screenshots of this. Also noted were the "hollow Nali" hanging on walls and suchlike -- if the player can run into them, they should be made solid or blocked with clipping brushes (if Unreal has them, of course; I haven't edited it, so I don't know). Here (and if you think I've doctored these to make a point, I can email you the originals... they're 589k each, but I'll save them just in case someone wants them):

shot 1 shot 2
Two views of the same switch. Notice how different textures disappear from different angles.
shot 3
The green dots in this one indicate the disappearing bars.
shot 4
A Nali as seen from the inside; looking up his right leg, to be precise. (Taken in the Water God temple upstairs by the pyramid with the button on top.)

That's about it for the gameplay issues, I think. I'd like to end this part of the review with a list of items or occurrences that bothered me logically and intuitively during the course of game play. For me personally, the hallmark of a quality single-player experience is the ability of the designers and level authors and texture artists and animators to provide the player with a true "suspension of disbelief": does everything fit together, does the actual game play fit and accommodate the story/plot, and does it all make sense when taken as a whole? Previous games in this genre have had a certain amount of verisimilitude, but before now, the focus has mostly been on the game play itself, not so much the setting and characterization and like that. Unreal has purportedly "raised the bar" on the single-play experience; due to this self-administered "raising", it must needs be judged accordingly. Some may call this unfair, but, hey, I didn't make the claim, did I?

Troubling Items In Unreal
(or: Unanswered Questionable Aspects)

If you know the answers to any of these, please let me know; just because I'm confused doesn't mean I'm not curious...

Final Thoughts

Is Unreal a Quake2 killer? The answer is "maybe". If you're tired of the Q2 theme and you've got a workstation-class machine (Pentium2-300 or better, 128mb+ of RAM, and a Voodoo2 card or two) and you don't care about multiplayer, Unreal is definitely up your alley and will probably replace Q2 on your hard drive. It's got the eye candy and the plot and theme, and, if you're willing to not think too deeply about the details, you'll have a great time playing this game.

If you're running a machine anywhere from a 200MMX to a Pentium2-266, you may or may not like this one. The running speed will be variable enough (and slower than Q2's, to be certain) to have you possibly questioning the value of dynamic lighting and all the added goodies that Unreal has that Q2 does not. You'll probably enjoy single player, but once the novelty wears off, it's anybody's guess whether Q2 or Unreal will win out.

If you're running a machine anywhere under a 200MMX, Unreal will resemble Myst with guns for you (at varying speeds, of course; remember, I'm still using 20fps as the playable minimum). Take this with a grain of salt, but I've yet to hear from anyone running a machine under 200MHz that has enjoyed Unreal single player.

For multiplayer, it's anybody's guess. I'm not really qualified to evaluate multiplayer, as my focus is primarily on single player. I've heard rumblings about the weapons being too weak and the gibbage unsatisfying and the sounds being too quiet and suchlike, but as I've not played it, I can't say for sure.

Basically, Epic have attempted to take the gaming genre to the next level with this release, in my opinion... and, in my opinion, they've implemented some great ideas and aspects into the game but the extreme machine requirements needed don't offset the number and quality of ideas implemented. In my opinion, Unreal is simply Quake 2 with a 16-bit palette, scripting, dynamic lighting, slightly improved AI, and a deeper story. Is this increase in quality of game play worth roughly triple the processor requirements? No.

Or, to put it another way: I wish I hadn't lost my receipt so I could exchange Unreal for something I would enjoy on my machine.


Got questions, comments, or clarifications about this review? I bet you do.
Send 'em here.
Note that obnoxious or unintelligible messages may be publicly posted (sans addresses, of course) for the edification of all. I tried to take a well-thought-out view in this review; I'd appreciate the same consideration if you choose to send me mail about this. Thanks.

to my home page

All original material on this page ©1998 by Bill J. "crash" McClendon. No part of this material may be reproduced or utilized in any form or by any means without permission in writing. Short excerpts of this material may be quoted as long as a reference to the full document is given.

Commercial distribution of this material, in whole or in part, requires prior agreement with the author. Commercial distribution includes any means by which the user has to pay either for the support (e.g. book, newsletter or CD-ROM) or for the material itself. Unauthorized commercial distribution is prohibited.

All trademarks, copyrighted names, images, and references to same are the properties of their respective holders.