How fast is "fast enough"?

[RoboTECH]

Just outta curiosity. When do you draw the line and say "okay, no more eye candy, I need more fps", and when do you say "okay, this is enough speed, let's start upping the visuals"?

With the emphasis on benchmark framerates these days, I'm just wondering where peeps are drawing the line....

lots of games out there are configurable, i.e. you can crank up more eye candy or turn it down to up your speed.

Let's hear which games you play, and where you "draw the line".

 

[Goi]

For me, I satisfy speed first, then visuals, unless its really crap ugly. If I can get above 60FPS minumum, then I'll start increasing the visuals.

 

[Henry Kuo]

yup, for me 60fps seems to be a good turning point too.

 

[shk]

Robotech, if you remember me, yeah i'd like at 8k fps @512 with 8bit of color and 16bit texture. I sacrifice all the eye candy just to get the speed i cannot handle. hehe..

 

[BFG10K]

I usually start off with the maximum image quality and gradually chip away the things that I don't notice and then move to the things I hardly ever notice.

For example in Q3 I have no problem turning off marks on walls, dynamic lights and ejecting brass since those things are either un-noticeable or add little to the gameplay. I also turn down geometric detail to low as well because I really can't see a a difference while you are playing. I also have no trouble playing at 640 * 480 res, although I would never go lower than this.

However, I am reluctant to lower the texture detail because I always notice the bluriness of the images while I am playing, and in my case it is especially bad anyway because of the V3's small texture size limit. I also get around 10 fps extra for low quality sound, but it just isn't worth it for that scratchy sound you get.

I guess each game has to be judged on what you consider is the best balance between speed and quality. Ideally I would probably start turning up eye candy when I hit my monitor refresh, but in Q3 I never do.

With those settings I get 55 fps on my Celeron 500 with a V3 2000 PCI. I believe a GF2 MX upgrade will give me around 15-20 extra fps so I am looking forward to getting this.

 

[tonyou]

For some reason the framerates being reported by Quake3 doesn't tell the whole story. With a Rage Fury MAXX for example, some stage would appear choppy even though the demo 1 benchmark say it has 80+ FPS. A Matrox G400 MAX on the other hand looks smoother at 60 FPS, so go figure...

Trial and error is how I would do it, if I feel the game is smooth enough I'll start adding details.

Tony

 

[BenSkywalker]

More speed= better visual quality for me.

I like 60FPS average, but if minimum drops under 30 too often(or at all depending) I go lower with quality.

I certainly would enjoy Q3 UHQ 1600x1200 that the speed of the GF2U would offer at playable speeds.

 

[RoboTECH]

these are very very interesting answers. Naturally, this was a completely loaded question with a devious purpose behind it, heh....

Now then, going along with these responses, can someone please explain how, if these are our meager needs, why do we live and die on benchmark scores? If all we care about is "60 fps with good visuals", then why is 115 fps better than 100, when no mention is made of visual quality?

I'll tell you why *I* think super high framerate is necessary, at least in Q3 (I have yet to find another game that really required such high FPS)

My "need for speed" seems far greater than anyone else's here, with the exception of shk, natch In Q3, the physics necessitates a MASSIVE framerate. So here's how *I* test to see if a video card is fast enough:

I do the following for Q3:

/seta com_maxfps 125
/map q3dm13

then I do the megahealth jump. backwards, forwards, strafing left, strafing right, strafing backwards left, strafing backwards right, strafing forwards left, strafing forwards right. If I have ANY problems with ANY of the jumps, I turn down visual details. If I can do them several consecutive times without a problem, I up some visual details (i.e. lodbias, r_picmip, resolution, and various other items) until I can't hit the jump 100% of the time.

This is a *tangible* need, i.e. if I'm racing down into that room in a 1v1 match, and I miss the jump even once, that could be the difference of 2 frags (I get fragged = I'm down by 1, or I get the MH before my opponent and nab a frag = I'm up by 1) in a duel match. That, to me, is important.

Now, the 125 fps was chosen because it has been my observation that some of the weapons simply act "differently" with high framerates, especially the shaft and the rail (and to a certain extent the MG as well). Why? I dunno, but several others have noticed it also.

So again, I have a VERY SPECIFIC and VERY HIGH expectation for framerates. If the card can't do what I want (see above), then I have no use for it.

I recently posted a screenshot to the newsgroup alt.binaries.games.quake. It was a pic overlooking the rocket launcher on q3dm9. It was a pic using my "graphics.cfg" that I had developed. It looked, to be honest, outstanding. and yes, it totally passed my q3dm13 test with flying colors.

EVERY ONE of the peeps thought it was a screenie of my GTS in 32-bit color. i had been bragging about how fast the GTS was (I was one of the first in the alt.games.quake3 newsgroup to pick one up), and had been harassing my buddies about their "slowass" DDr's. D'oh! Everyone knew of my requirements for framerates (I'm a framerate hog, hence the system you see in my sig), so everyone was AMAZED at how good the "GTS" could look while still maintaining my "need for speed"

It was actually a pic using the 5500 I "borrowed" from Electronics Boutique. Interestingly, when I posted the next pic I took (after turning on the framerate counter), the "slowass, dog, POS voodoo" read "140 fps". This was with the highest quality settings in the drivers turned on to make the 22-bit post filter look damn near like 32-bit.

When I switched it to 32-bit color, it read "100 fps". Not bad, right?

That seems pretty fast to me. And to think Quake3 was one of the games that the 5500 "sucks at", and the GTS "blows it away" in.

That's why I question so many of you when you say "the 5500 is slow, it sucks, it's lame, blah blah blah"

My question is always HAVE YOU TRIED IT YOURSELF?

Now, admittedly, this was with an overclocked card. Of course, I'm pretty sure that *most* serious gaming peeps will overclock their CPU's and their video cards every chance they get, assuming 100% stability.

I was getting 130 fps in demo001 and ~115ish in Quaver, so I upped visual quality (using the lodbias slider), and I ended up with ~126 fps and 110ish (honestly can't remember exactly) respectively, and the GTS simply couldn't touch it at that resolution. I had to up the resolution to 1152 to even come close while maintaining framerate (see my "com_maxfps 125" test), but it still didn't look as good. If I jumped up to 32-bit color, the framerate sank too low. I could turn on the TC with the GTS, and it would allow me to use 1280x1024 and maintain my framerate, but the TC turned on proved to be too annoying.

So understand just *why* I bristle at those who read the bull$hit reviews of the video cards and declare themselves experts, when they haven't even tried the card yet. A very reasonable 20 MHz overclock was all it took to get the incredibly high framerates I got with the 5500.

3Dfx fooked up an outstanding card (the 5500) in 4 ways:

1) They released it too late (dumbasses)
2) They released it without a lodbias slider initially - and visual quality looked crappy
3) They released it at a very low default speed (I haven't found too many peeps who couldnt' hit 183 easily)
4) They released it with some stupid default settings in their drivers (lodbias, alpha blending and mip map dithering)

Since EVERY reviewer out there does ALL their tests with basic default configurations, the 5500 ended up looking pretty weak by comparison. Very few reviewers (check www.voodooextreme.com/reverend and www.3dspotlight.com) actually bothered to sift through the drivers and tweek the card. When they did, they were amazed at just how good that card performed.

I got the 5500 as a joke, mainly. I had just sold my 32MB GTS and ordered me a 64MB GTS from DecoY here on Anand's boards. Unfortunately, I was "late" on the list, so it was going to be awhile before I would receive the 64MB GTS beast, so I went down to Electronics Boutique with the plans of exploiting them and their "10-day return, no questions asked" policy. I had my eyes on a Herc Prophet II 64MB'er, when on a whim, I decided to pick up the 5500, just for the halibut. I figured it would tide me over until I got my "real card" in, then I'd take the 5500 back to EB (which I, begrudginly, did)

Funny how well that 5500 performed, and this was in DIRECT COMPARISON to the 32MB GTS, and now, the 64MB GTS.

Anyway, thanks for the responses, and thanks for confirming what I already figured.

 

[RoboTECH]

and, another bigass, longwinded post by yours truly....

still trying to implement text compression in my keyboard's drivers. I can't seem to get it to work

 

[MGallik]

You go Robo..

 

[BFG10K]

then why is 115 fps better than 100, when no mention is made of visual quality?

It is mentioned all of the time. Each set of graphs states the visual quality of the test and each video card has exactly the same settings in the game.

Also you mention the LOD bias on the Voodoo? Remember that the reviews are done with the bias set to zero (blurry image). Sharpening the image is fine and dandy, but remember your fps will take a hit. So in order to sharpen the image, the Voodoo will get even lower score than reported by the benchmarking sites.

My question is always HAVE YOU TRIED IT YOURSELF?

Most people don't have money to buy several video cards just to try them out. However you can read a large number of respectable gaming sites and see if they all argree on each others findings.

After looking at around ten different sites, the consensous is that the V5 simply sucks. I think there was a link posted recently which had some benchmarks and reviews of five different cards and the V5 came last out of the 5 cards.

 

[utopia]

FSAA 2x
32bit
800x600
medium eye candy
60-80FPS

 

[RoboTECH]

BFG:

<< then why is 115 fps better than 100, when no mention is made of visual quality?
It is mentioned all of the time. Each set of graphs states the visual quality of the test and each video card has exactly the same settings in the game. >>


no. each set of graphs mentions resolution and color depth, with Q3 settings. No mention is made of visual quality. How many reviews of the GeForces and GTS cards are out there? Literally, several hundred. How many have even mentioned that they have jacked up texture compression (a handful)? How many have mentioned the fact that the GTS does anisotropic filtering? How many have mentioned what type of framerate hit the GTs takes with anisotropic filtering on? Or how the graphics look with it on? Damn near none. Lame. Mentioning the obvious (default Quake3 settings) does NOT mean they're talking about visual quality. About the only site that I saw even BOTHER with mentioning the GTS's $hitty 2d at high-res was Sharky's, and that was in the 5500 review!


<< Also you mention the LOD bias on the Voodoo? Remember that the reviews are done with the bias set to zero (blurry image). Sharpening the image is fine and dandy, but remember your fps will take a hit. So in order to sharpen the image, the Voodoo will get even lower score than reported by the benchmarking sites. >>

Hahaha...you silly guy! You seem to forget that I USED the damn thing. I know there is a framerate hit. I cranked the lodbias to -0.75 which produced RAZOR sharp images, and lost a grand total of.... < 3 fps. Wow. I went from 130 fps to 127. How can I live with that type of hit? But seriously, there really isn't much of a framerate hit until you get past -1.50 and -1.75, but I didn't find a need to go past that, even @ 4xFSAA. -0.75 was VERY VERY sharp, the textures were very detailed and sharp. The image quality was quite astounding, actually. I really got a very great appreciation for the job Carmack did once I got that going with r_picmip 0 on. Impressive. You get a whole new appreciation for some of the maps. Unfortunately, i kept getting fragged in multiplayer cuz I kept staring at the walls and the ceiling. It looked pretty darn good. <g>


<< Most people don't have money to buy several video cards just to try them out. However you can read a large number of respectable gaming sites and see if they all argree on each others findings. >>

Understood. Now, if you want to rip on the 4500, that's fine, the MX should be superior in almost all situations, as the 4500 is literally HALF the 5500. As far as the reviews are concerned, what's funny is that 1 month ago, I would've agreed with you. After reading so many negative reviews of the 5500, I basically &quot;borrowed&quot; it from EB on their &quot;10-DAY RETURN&quot; plan to tide me over until I picked up the 64MB GTS, mostly as a whim. I had planned on grabbing a Prophet GTS 64MB, but since I had just gotten done flaming some &quot;lame 5500 user&quot; for his $hitty card, I figured I had better do some homework and see what this card could do. I'm going to be quite honest with you, most of the stuff I've seen in the reviews has been bull$hit where the 5500 is concerned. The reviews have NOT paralleled my experience at all. I've seen the poor points of the GTS completely ignored, and the good points of the 5500 not even investigated.

How many reviews out there have nothing more than 3dMark2000 (worthless) and quake3 default settings? TONS!!!!! How many websites actually tested using flight sims, or racing sims, or RPG/RTS games? Very very few. Why? Because they don't have built-in benchmarks. Usually, just a passing mention of &quot;flight sim'ers will never go back to playing wihtout FSAA again&quot;. They say that the 5500 gets &quot;beat by the GTS in Quake3&quot;, but they never say if the 5500 can do Quake3 with great visual quality and playable framerates (it can, or I would hate it)

Sorry, but there are a lot of movies out there that have gotten horrid reviews from Siskel and Ebert, and I loved them. Until you try it, don't knock it, because you obviously don't know what you're talking about.

oh yeah, and please refer me to the website that states that the 5500 &quot;simply sucks&quot;. Thanks.

 

[BFG10K]

oh yeah, and please refer me to the website that states that the 5500 &quot;simply sucks&quot;. Thanks.

No problem. Well they didn't say &quot;simply sucks&quot;, they said it comes last out of the five boards. In my book that simply sucks.

Video Card Shootout

A Comment From the article:
3dfx's current flagship has aged poorly since its introduction three months ago, largely due to the gains made by its competition. The V5-5500 is outgunned by the GeForce2 GTS cards in terms of frame rate, and priced at nearly $125 more than a similar performing GeForce2 MX value card.

 

[hans007]

who came up with this stupid 60fps number anyways? I think as long as in every situation in the game i am above 24fps i will start increasing visual quality. 25fps is european TV. That is the threshold of your eyes being able to notice a difference. I'd say an average in demo001 q3a of 55fps is enough to guarantee 25 fps in any part of that game, and any modern card should be able to do that. personally i play counterstrike most of the time now, at 800x600 4xfsaa and i cant drop below 25fps in that.

 

[tonyou]

I remember reading an article a while back that talks about framerates. The reason why we need 60fps or more for computer games like Quake3 is due to the fact that each frames are perfect stills. TV looks smooth at 24-30fps because it has motion blur, when you freeze a frame, you can see that.

Tony

 

[Zak]

Oh my god! The 60fps issue was beaten to death and explained hundreds of times on different websites and still some people come up with the 25fps idea. Search the web, find out why 25fps is not enough for computer games.

Zak

 

[RoboTECH]

<< oh yeah, and please refer me to the website that states that the 5500 &quot;simply sucks&quot;. Thanks.

No problem. Well they didn't say &quot;simply sucks&quot;, they said it comes last out of the five boards. In my book that simply sucks.

Video Card Shootout

A Comment From the article:
3dfx's current flagship has aged poorly since its introduction three months ago, largely due to the gains made by its competition. The V5-5500 is outgunned by the GeForce2 GTS cards in terms of frame rate, and priced at nearly $125 more than a similar performing GeForce2 MX value card.
>>



well, reading ridiculous articles like that, it's no wonder you have your mind made up. ugh....benchmark 3 games and call it a day. <sigh>

 

[tkopp]

I wholeheartedly agree with you, man. I'm one of those people who kicks up the visual quality to max, and then works their way down. I've found that in Q3A, high-res doesn't mean much to me. 800x600 is fine, or 1024x768, but I want to see the high-res textures and curved surfaces. Otherwise, it just bugs me. It makes me feel like I'm playing smooth Q2. I bought a Radeon for the image quality, and I'll be the first to admit that it's not as fast as a GTS. I just like the rest of the board better. As far as the Voodoo5, it's a great board too. I was going to buy one, but the day I got payed the Radeon reviews hit the web and I decided to try that instead. And for the FPS issue; in single player I'm happy with 45fps average. I'm not all that competitive vs the computer, and it's just for fun. I payed a lot for my video card, and it's going to look good. In multiplayer, I get 95fps in 800x600 with a very high quality setup, and it looks good to me. I'm considering dropping that to 1024 levels, but I just don't play enough Q3 to have that be a necessity.

 

[Wingznut]

I'll have to go along with &quot;turn everything up, and go from there.&quot;

Outside of things like volumetric lighting (which actually inhibits multiplayer gameplay), and 32 bit color (since I only notice it with volumetric lighting on), I insist that all details be set to high. And I have yet to find a game, where FSAA wasn't a requirement for me.

Hey BFG10K, if I show you an article that uses games like flight sims, sports games, and racing sims, when comparing the two cards... and states that the V5 is better all-around, will you say that the GTS sucks?

Yeah, I thought not.

 

[RoboTECH]

<< Hey BFG10K, if I show you an article that uses games like flight sims, sports games, and racing sims, when comparing the two cards... and states that the V5 is better all-around, will you say that the GTS sucks? >>



no, he'll just say &quot;those games are run under glide/aren't fillrate limited/aren't good benchmarks&quot;

he doesn't care about playing games, he only cares about the benchmark scores.

 

[hans007]

if you read the hardocp &quot;voodoo5 vs gf2&quot; comparison, the voodoo5 doesnt come out too bad. Also everyone says that the new detonator 3 drivers are so well optimized . I remember when nvidia came out with the riva128 they had to redo benchmarking by having reviewers check off lists, while watching the benchmark. Zdnet had to do it, because cards were taking &quot;shortcuts&quot; do to things. I seriously dont believe that nvidia can keep coming out with new drivers and improving speed and not screwing up visual quality.

 

[BFG10K]

Hey BFG10K, if I show you an article that uses games like flight sims, sports games, and racing sims, when comparing the two cards... and states that the V5 is better all-around, will you say that the GTS sucks?

Well that depends on how it is better. FSAA, yes OK I guess I'll give you that. Raw fps? No I won't give you that.

no, he'll just say &quot;those games are run under glide/aren't fillrate limited/aren't good benchmarks&quot;

If that's the case I won't give even the FSAA to the Voodoo.

he doesn't care about playing games, he only cares about the benchmark scores.

Not true at all.

 

[RoboTECH]

<< Hey BFG10K, if I show you an article that uses games like flight sims, sports games, and racing sims, when comparing the two cards... and states that the V5 is better all-around, will you say that the GTS sucks?

Well that depends on how it is better. FSAA, yes OK I guess I'll give you that. Raw fps? No I won't give you that.
>>



actually, you're completely wrong on that. There are several flight sims that apparently run faster on the 5500, in RAW framerate category. Unreal/UT are NOT isolated incidents (tho somewhat so)



<< he doesn't care about playing games, he only cares about the benchmark scores

Not true at all. >>

so why the obsession? You have a collection of links to benchmarks that you like to post and refer to. There's more to it than just some numbers y'know.

 

[BFG10K]

actually, you're completely wrong on that. There are several flight sims that apparently run faster on the 5500, in RAW framerate category. Unreal/UT are NOT isolated incidents (tho somewhat so)

I would have to see this to believe it. If it were true (which I doubt) it would probably be because of those games being based around Glide.

Even if they were proper OpenGL/Direct 3D coded with no Glide code-base, I don't play flight sims so it really makes no difference to me. Anyway flight sims don't require high fps like 3D fps do. (How's that for a confusing sentence!)

so why the obsession? You have a collection of links to benchmarks that you like to post and refer to.

What obsession? Posting up links? In that case every single hardware site is &quot;obsessed&quot; and every single person here in this forum is &quot;obsessed&quot;.

There's more to it than just some numbers y'know.

True, but for most people benchmarks and price are the most important thing in considering a card, both of which the websites I link to provide.