• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Valve sucks

Status
Not open for further replies.
optimus. i sir, am happy you continue to post and repaet what you are saying. im with you on this. im also happy i went straight form page 2 to 18 then to the last 3 pages, cuz the ati vs nvidia is a waste of time.

my 160 cents:
i loved nvidia. the fx line was a peice of crap. i considered and purchased a 9600xt. wow it was awesome, better color, as we all know, much better filtering. its a midrange card right? therefore not like i COULD use AA all that much anyhoo. hl2 coupon as well oh snap! but then, what happened? crap ass performance in counterstrike. call of duty bugs out the ass. i was stuck on that doorway to the outside of the train station near the end of CoD for months? ati support? talkin about coming out of dual channel mode on the athlon XP? relax timings? what? wtf? so then i learnt abotu the whole nvidia owning on openGL performance bu the fx line sucking at dx9. what did i do? made a choice

i sold the 96xt with hl2 coupon to a clan member of mine who was gonna stop playin cs and LOVED his dx9 mmorpgs etc. and what did i buy? the best midrange non ati card at the time. MSI 5900xt in feb 04 or something.

points to consider folks:
as a previously very active member of the cs community and living in MN where if you know, has a shit load of hte best teams - most of these guys DID NOT PARTAKE in the survey. all them 5200 owners are the dell computers in schools and shit. there are a ton of us out there with the 5900xts etc. so even tho optimus explained that its 1 outta every 10, ima have to say 1 and a half to 2 outta every 10 are runnin the end-of-gen FX line.

ATI paid crap loads. hl2 cost crap loads. they have lawsuits in the meantime. take the money, release the greatest FPS ever, and continually update it thru steam, further validating the reason use and point of the steam distribution service. why say valve sucks? valve is damn good and human at that. for what they have done i applaud them. optimus said he played thru D3 once? hl2 done and more to come? fookin a, im not even in delta labs in d3 and i really cant care less. very few games ever held my attention and im a beast with the games.... nfsu 1 did, prince of persia did (both of which i played during the gfx card switch back then) and that cool james bond girl 60s fps did. dont play rpg's by choice but i respect them and watch the roomates play (one of which has the exact msi 5900xt and is sad it hurts so much on his system)

must... remember.. other... points:
to ring it home. the fx line was crap we all knew this. i knew it when i bought it. i got it cuz it provided way better midrange card (<$200) performance in open gl (CS!!!!) and given that CS - the FAR frikken majority of the steam userbase - needs performance for CS, it jsut makes you think again... more folks have or had the fx line. unless you owned a 9800xt. and when that came out, every gaming site was flooded with guys complainin about lack of performance. folks were playin at 800x600, turning off everything, 16bit. WITH A $500 CARD?

anyhoo i made a plan a year ago to buy an ew computer cuz of all the announcements of stuff like dual layer burnders, athlon 64's, nv40 and r420... etc. the 9600xt was supposed to be my last purchase before selling the system after the summer to get the new system. sadly i had to jump back to nvidia and i cheered when the nv40 came out with a redesigned everything, filtering that looked good (im a gfx designer by day - i see color) and finally supn to make me say, "yes nvidia, i like you and your forceware!"

im soprry i cant provide the rest of my optimus supporting opinions bu he said it already pretty much. we have hl2. its a frikken great game. mainstream level computer parts are purchased FAR more than exotic cuttin edge and only the RICH enthusiasts can buy "ultra" and "PE" cards. the rest of us wait for it to trickle. dont blame valve, blame your pocket, education, society or dry grass outside. hl2 is great, looks great, will be updated, steam is great, im lucky it never fails me or takes more than a min for any update, location locations? i dunno. and there are a lot of FX owners bu i think it was safe for valve to assume A LOT OF FOLKS PLANNED COMPUTER PURCHAES AROUND THE RELEASE OF HL2.

thanks bye
 
Optimus said:
The funniest thing about this is that the reason I chose the 5900 was that I liked the NV3x design. With the exception of the fact that the actual shader path implementation was a little off (alternating arithmetic and texel operations in order to achieve optimum performance), I thought the idea of having 4 32bit floating point pipelines that can be split into 8 16bit floating point pipelines was an ingenious method of taking full advantage of current technology while providing for future upgradability at a time when their competitors had only just begun to consider dx9... Unfortunately, it took them so long to get the thing out, because of GDDR3 problems, that the 5800 was no longer top tech when it was finally released. Still, just because I can afford to buy a new card right now is no reason to give up on this issue and penalize those with this line that can't, whether or not they realize that they aren't getting all they can get.

And when I find the Holy Grail, I'll share it with the entire world... for $29.99 a sip... ;)


I respect your counter-opinions. However, I have to say that the 9700Pro was already over a year old and fully supported DX9 before the FX5900 was released, and still performed better in most situations. I think Nvidia had some interesting ideas *in theory* but the rubber never met the road quite right.

(I still laugh out loud sometimes when I think about the TRAOD/TWIMTBP debacle)

At the time (before HL2, Doom3, FarCry) when these cards were the latest-greatest, my biggest concern about Nvidia cards was the huge deltas in performance that Nvidia was achieving on a game-by-game basis once they had a chance to optomize in the driver. My feeling was that part of the optomization may have been innocuous such as shader instruction re-ordering, but that part of their optomizations were likely to also be cheating where they sacraficed the IQ of the game by neutering some texture/function/whatever. Besides, who wants to wait 3 months after the release of a game to finally be able to get decent performance? As time has gone by I feel that my initial opinion has proven to be more true than not.

Also, I'm a huge believer in eBay and rapid acceptance of new technology. I was able to sell my 9800XT on eBay for ~$290 the day that I bought my 6800GT for $399. That means I made a generational leap for less than $120. That my friends was the best $120 I spent all year.
 
I haven't bothered to read this entire thing I skimmed thru alot of it as it is nv vs. ati bullshit but parts that I do catch are those who have been promoting ATi left and right are defending valve where as those who promote NV are condemming them...and then you got the guy who owns a 5900 that feels fucked out since valve did not implement something that could have helped his card out some. Myself if I still had my 5900 I would be pissed off about this shit...the game stayed in development longer than hell and they couldn't add partial precision calls? I'm no coder but as a consumer I would be very pissed off.
 
i read this thread when it was 2 pages, came back and now its 29 pages. Valve could of coded a better path for Nvidia cards, but how many programmers out there have been told to do something they knew that it would hurt one type of computers or another? The flaming of the video cards, its like intel vs. amd, except no one care to look at one crucial part. I owned a 9600xt, locked up my system constantlly and NEVER WORKED IN LINUX. I sold it to my friend who is a FS freak. I got a 5900XT, after I took the ATI survey. Nvidia works in windows, and linux, i bought my card for compatability and if some game doesn't run as well on my box as an ATI box, at least I have graphiclly abilitys at ALL TIMES..
 
I failed to consdier Linux compatibility when formulating my various responses to the performance of HL2 on FX cards. I have now considered it, and find that I have nothing to add.

Did I mention recently that I own an Nvidia card? Just not a crappy FX5xxx series. It's not prejudice. It's judgment. I didn't pre-judge anything. I considered the evidence.

Now, what to have for lunch?
 
I suppose I didn't quite explain myself as far as "replacing" my FX5900, though I'm glad to hear that it doesn't have to be quite the expenditure I was expecting... exactly... excruciating... sorry...

I set myself, before I purchased the 5900, to upgrade once every two NVidia development cycles... by that I mean nv3x, nv5x, nv7x, etc. unless something fundamentally changes in the industry, because I expected that by the time I was ready to upgrade, I'd be hitting the end of the video card's natural life cycle. For the most part this is true, except that my current card has horrible dx9 performance unless it is specifically programmed for in a certain way. Granted I knew that this was the case when I bought it, I just hoped that, since the dx9 spec provided for such a possibility, it would never become a problem. Honestly, I'm surprised it hasn't b4 now.

Actually, on that rambling note, I wonder if the TWIMTBP marketing (propaganda) was just a way to encourage the game companies to write their code in such a way as to clean up after their mistake... hmmmmm.... :confused:

So as far as price point, I stand corrected. It is not prohibitively expensive to replace a working Video Card if the buyer doesn't care about this mess.
 
What's this valve sucks bullshit anyways? Fuck that guys, valve does NOT suck. Because you made a bad choice buying a geforce fx card, the company sucks? Not to mention, steam is improved upon every day. Why, just the other day, valve released like 3 patches in one day for hl2 dm. I guess you could argue that it was because it was so shitty it needed patching. Well, who cares, they still put forth the effort to squash the bugs very quickly.
 
w00t69 said:
What's this valve sucks bullshit anyways?
Obviously you didn't even read the first page.

Valve did per card detection for capabilities. When a 5900 card owner selects DX9.0 (-dxlevel 90), they get a bad rendering path. Fooling HL2 into thinking the card is a 9800 by editing the dxsupport.cfg file fixes those rendering errors. For that alone, the title is valid.

Otherwise, I don't think it's unreasonable that Valve dropped the mixed mode path. I said it above and i'm sure others have too: the game was very late and it was taking extra time to do make the mixed mode path.

But the reasons Gabe gave for dropping it were as lame as his lies about the state of HL2 back in Sept 2003 and the causes of the delays through the spring (the stolen code wasn't the reason it didn't lauch last year... it simply was nowhere near complete). Mixed mode would have worked fine for the higher end FX series.

And buried deep in a few pages back I posted a way to run the ATI detection path with DX8.1 mode and world reflections for a really big speed increase on even a FX 5200 card.
 
I stand corrected on another issue... this forum really does need a rename now. It's drawing the attention of people who don't even read the last two pages before posting.... :p
 
I don't mean to be rude pxc, but you missed one of my posts:

Optimus said:
On the ATI ID issue:

A previous post of mine was missed apparently due to being in the midst of the flame war. Here it is in it's entirety:
Optimus said:
I went to http://forums.guru3d.com/showthread.php?threadid=115614
and followed the directions and saw no speed improvement either way with my NVidia GeForce FX 5950 Ultra (Flashed 5900). However I have localized the water fix. On my card water would never show in dx9.

Deleting the line labeled "NoUserClipPlanes" fixed this water problem nicely.

I have not tried limiting to fp16 shader performance yet, but, otherwise, I see no speed benefit to straight up dx9 in HL2 on an FX series video card.

While I was working on this it became relatively apparent how HL2 recognizes your video card. It does so by Vendor ID and Device ID. (Proof: I ran into a problem with this while attempting the fix. There are two listings for the NVidia GeForce FX 5950 Ultra. The first is Device ID 0x310 and the second is 0x333. Mine is the latter but I started with the former. When I changed the settings in the former nothing would happen, but when I changed them in the latter it started working.)

When you follow the instructions specified by http://forums.guru3d.com/showthread.php?threadid=115614, you are not rebadging your video card. In fact, as far as HL2 is concerned, you are deleting it from the list, forcing HL2 to go with the default settings at the very bottom of the file. The settings at the very bottom of the file are generic and thus do not incorporate any specific settings such as "NoUserClipPlanes." This tells HL2 to treat your video card as a standard, dx9 compliant, Nvidia branded video card. It is basically assuming that you have an Nvidia card from the future... as stupid as that sounds.

So the fact of the matter is that Valve did not cripple NVidia video cards. They optimized them for dx8.1, thus crippling them in dx9.0.

It was nothing more than an oversight on their part. It is a forgivable mistake.
--------------------------------------
This sucks.... there's no smiley for taking a bow.
--------------------------------------
 
w00t69 said:
What's this valve sucks bullshit anyways? Fuck that guys, valve does NOT suck. Because you made a bad choice buying a geforce fx card, the company sucks? Not to mention, steam is improved upon every day. Why, just the other day, valve released like 3 patches in one day for hl2 dm. I guess you could argue that it was because it was so shitty it needed patching. Well, who cares, they still put forth the effort to squash the bugs very quickly.


Oh so the fact that a card would be given as a gift from someone who isn't PC litterate is their fault? jesus man use some common sense...not every damn person knows what a card can and cannot do. So yes Valve does indeed suck...with all of the damn delays they could have coded that part into the damn game to help out those who do not own the top of the line graphics cards.
 
If I weren't already going bald I would have already pulled all my hair out. :rolleyes:
 
Visable-assassin said:
Oh so the fact that a card would be given as a gift from someone who isn't PC litterate is their fault? jesus man use some common sense...not every damn person knows what a card can and cannot do.
In this case I would slap the crap out of the person who gave the gift ;)

Visable-assassin said:
So yes Valve does indeed suck...with all of the damn delays they could have coded that part into the damn game to help out those who do not own the top of the line graphics cards.
Valve is the best thing that has happened to PC gaming. Sure the game took too long but it was well worth the wait :) The same coding issue could be said about iD and their Nv favored paths.
 
pxc said:
Obviously you didn't even read the first page.

Valve did per card detection for capabilities. When a 5900 card owner selects DX9.0 (-dxlevel 90), they get a bad rendering path.
False, they gave the rendering path according to the capabilities (who doesn't know the FX line suxors in DX9.0?). If the user doesn't like the patch chosen, they can easily change it. It's not Valves fault the 5900 can't handle FP32 and needs to blend just to perform.
 
R1ckCa1n said:
False, they gave the rendering path according to the capabilities (who doesn't know the FX line suxors in DX9.0?). If the user doesn't like the patch chosen, they can easily change it. It's not Valves fault the 5900 can't handle FP32 and needs to blend just to perform.
See 2 a couple of posts up on how to fix the valve mistake for the 5900 DX9 path. :rolleyes:

Who's talking about a "patch" anyways? And it's not blending, it's partial precision that makes the performance difference. :p
 
Okay, look this is exactly what I was trying to get away from.

Stop acting like little children and discuss the issue like adults...please.

You guys have missed quite a bit, apparently and have not caught up on what you missed.

Optimus said:
The thread topic was intended to get recognition, at which it succeeded, but the point of the thread is to give compelling evidence that a future patch is needed to allow partial precision for at least the FX line of cards, since no others seem to benefit greatly.

If you have constructive counter-evidence, then please stay and submit it.

However, if all you're going to do is bash some poor folks' decisions on hardware purchases, and thereby contribute nothing constructive or helpful, then leave. This includes those who keep saying that Valve or ATI sucks, because that gets us no where pleasant.
 
Facts (gleaned from this thread, or links in this thread):
Partial Precision does not look as good as full precision on all shaders in HL2.

DX8.1 is faster on FX5700-FX5950 cards than DX9.0 is, even when Partial Precision is forced on ALL shaders.

DX8.1 looks pretty damn good.

DX9.0 Looks better.

John Carmack and Gabe Newell both claimed to have spent a LOT of time on the NV30 path.


Facts that are highly relevant... or would be if we new them for certain.:
We don't know how many shaders there are in HL2.

We don't know how many of those shaders look just as good at Partial Precision as they do at Full Precision (and conversly how many look bad).

We don't know how easy it would be to identify which shaders look just as good at PP as they do at FP.

We don't know whether JC and GN really spent lots of time on the NV30 path and then scrapped it or if they are just blatant liars.

We don't know what minimum settings and hardware Valve decided they would use to determine which path, DX8, DX8.1, or DX9.0 was appropriate for each card.

We don't know whether Valve ever completed their mixed-mode precision code and if they did so we don't know why it's not in the game.


Speculation:
My personal speculation....

Valve probably mostly finished the mixed-mode precision code. Then they probably tested an FX5700 on a 1.4ghz CPU with 256mb of RAM at 1024x768 with various combinations of: DX8.0, DX8.1, DX9.0, in game settings, and precision settings. At the end of they day they probably found this setup could get the best overall image quality and speed on the DX8.1 path.... and was completely unplayable on the DX9.0 path unless you dropped to 800x600 or lower... which would have even MORE people bitching. They probably got similar poor results with an FX5950 on a relatively slow PC (MAYBE with it being marginally playable with full PP, but not good looking enough) and so decided to go with DX8.1 for the entire group of faster FX cards.

Companies make decisions like that because they HAVE to program for as many of their customers as possible, and in the PC business that means making a cutting edge engine that requires an uber-PC to max all settings (to make the PC enthusiasts happy) while also making the engine scalable to people who only want to upgrade every few years but still play the more popular games.

On the subject of JC and GN spending lots of time on the NV30 path.... I think they probaby did spend a lot of time on it, but I don't think they scrapped it. I think what is really going on is that they weren't fully precise in describing what they were doing with the NV30 cards and what was taking up all their time. I think they were spending all that time trying to get the NV30 cards to work properly in their games ON the "main path". They weren't writing up tons of special code for the NV30, they were taking the existing code that already worked on other hardware and tweaking it to make it work right on NV30 chips as well as on all the other stuff that had no problems with it to begin with.

Anyway, considering the number of people on this thread who have 5700-5950's and can play the game JUST FINE in DX8.1, I am kind of confused why so many people are upset. You are basically pissed off that the game doesn't look as good as it could on your PC despite the fact that you would take a performance hit to make it better, and despite the fact that ony 2 other games in the entire history of the world look better on your PC (Doom3 and Farcry). Yeah, the game looks a bit better on the "Jones's" PC, but it still looks great on yours and is just as fun.

And to cap it all off...

There is NO proof Valve intentionally tried to screw anyone, so let it GO already.
 
pxc said:
See 2 a couple of posts up on how to fix the valve mistake for the 5900 DX9 path. :rolleyes:
Who said it was a mistake? It was done for a reason..... You think they forgot about 2.55 percent of their users? ;)

pxc said:
Who's talking about a "patch" anyways? And it's not blending, it's partial precision that makes the performance difference. :p
Who said "patch"? Doesn't change the fact the FX sucks unless it can do partial precision. Just happens that Valve didn't code the game that way. For your reference, see below quote regarding my "patch"

R1ckCa1n said:
Valve is the best thing that has happened to PC gaming. Sure the game took too long but it was well worth the wait :) The same coding issue could be said about iD and their Nv favored paths.
 
rcolbert said:
I never thought of quoting myself before.
As long as my quote doesn't quote me, I'm safe. And yeah, don't knock it till you try it!
 
R1ckCa1n said:
False, they gave the rendering path according to the capabilities (who doesn't know the FX line suxors in DX9.0?). If the user doesn't like the patch chosen, they can easily change it. It's not Valves fault the 5900 can't handle FP32 and needs to blend just to perform.

Yes but why does everyone know that the FX line sucks in dx9? because Valve said so...that's some circular logic right there...
 
Actually, the first clue to the FX sucking was Tomb Raider, not Valve, if memory serves me.
And we all remember how the panties at Hardocp got in a twist at the sight of those first FX5800 vs Radeon 9700 benchmarks in 3DMark03.
Even John Carmack said that the 9700 was a faster/better card for Doom3 at the time, unless he used the specific optimized path for NV3x (with slightly lower image quality).
Only after that did Valve speak about their experiences with the FX series.
 
^eMpTy^ said:
Yes but why does everyone know that the FX line sucks in dx9? because Valve said so...that's some circular logic right there...

lol more like tomb raider first showed it...3dmark 2003...halo and every other dx9 game that came out before hl2 and was shown in reviews.....that's some closed minded logic you have there. :rolleyes:
 
My word, what is wrong with you guys? This reminds me of the whole 16 bit versus 32 bit debate back in the days of 3Dfx versus nVidia. So what Valve chose to use FP24. If it runs faster and looks better than FP16, and gets better framerates than FP32 then more power to them. Remember, your precious nvidia chose to use either FP16 or FP32 and not FP24 in their 3D cards, so go blame nVidia for not supporting FP24. Ati chose to support FP24 because apparently they felt the performance penalty was to much with FP32, and that in their opinion it wasn't needed. This isn't Valves fault as far as I can tell. The nVidiots are howling foul play because nVidia chose not to implement FP24 in their 3D cards, so go point fingers that them. Oh wait, you couldn't do that because you would be attacking the very company you worship as being the "best." If HL 2 gets batter framerates at FP24 instead of FP32 and still looks fantastic, that's great. Are you guys so short sited that you fail to realize that nVidia's next gen card will have better performance, and will most likely be able to play HL 2 at FP32 with high framerates?

I do have a bone to pick with ATi for not having decent OpenGL performance. That is shameful. But so is the NV30 series of cards. I feel sorry for the poor saps who bought those worthless cards. Shame on nVidia for that. Both companies have made mistakes, and they continue to make mistakes. The PC industry is brutal. So stop your whining and go play your games.
 
You have a point there, Agent_N... Even with FP16, the FX cards are still significantly slower than the ATi cards... Perhaps Valve chose DX8.1 for FX cards so that at least the framerates would be comparable with ATi... From that angle, Valve actually tried not to put the FX cards at a disadvantage.
 
R1ckCa1n said:
In this case I would slap the crap out of the person who gave the gift ;)


Valve is the best thing that has happened to PC gaming. Sure the game took too long but it was well worth the wait :) The same coding issue could be said about iD and their Nv favored paths.


If you had no fuckin idea about computer hardware you wouldn't so don't be stupid man.
Also what part of NV and ATi use the same paths in Doom3 do you not understand man? Carmack dropped the NV only path a while ago.
 
Visable-assassin said:
If you had no fuckin idea about computer hardware you wouldn't so don't be stupid man.
LOL.... is wittle Visable-ass getting is panties all worked up because the FX line sucks at DX9? Take a chill pill son.


Visable-assassin said:
Also what part of NV and ATi use the same paths in Doom3 do you not understand man? Carmack dropped the NV only path a while ago.
Errrrrrr, who in this world didn't know D3 was written for NV hardware and visversus? Talk about losing it....... Let's try to keep on-topic now.
 
Agent_N said:
The PC industry is brutal. So stop your whining and go play your games.
One that note, this thread should be LOCKED because he just nailed it.
 
R1ckCa1n said:
Errrrrrr, who in this world didn't know D3 was written for NV hardware and visversus? Talk about losing it....... Let's try to keep on-topic now.

Isn't this the other way around - that Nvidia designed their hardware around Carmack's requests?

I'm just not seeing where Nvidia was unreasonable. I mean, they designed hardware around where they thought things were going, based on the requests of arguably the authority in realtime 3D engine design.. Their hardware exceeded the SM2.0 spec. In addition, they supported a feature of the SM2.0 spec to increase performance by using partial precision. It's not a hack - it's a feature defined in the specification, and cards that support mixed precision would see better performance if Valve implemented it.

As for Valve, we have no idea why it wasn't done. Perhaps they were lazy. Perhaps they didn't have time. Perhaps they wrote off that segment of their potential userbase. Perhaps ATI said "No way in hell." We have no idea.

However, it would be best for Valve to go back through and see where partial precision can be used to speed things up. I'm sure Nvidia would have loved to help out their customers by assisting Valve with this, had they been given access.

Agent_N said:
So what Valve chose to use FP24. If it runs faster and looks better than FP16, and gets better framerates than FP32 then more power to them.

And what if it doesn't look any better than FP16? Using higher precision does not, repeat not, repeat NOT automatically make things look better. Not going from FP16 to FP24, not going from FP24 to FP32.

Increasing precision decreases the number of errors due to lack of significant figures. If the errors are large enough in practice to change the value of a pixel, then you need more precision. Otherwise, you don't, and it's a waste of time.

Is anyone keeping track of how many times I've had to explain this?
 
Kevin Lowe said:
Isn't this the other way around - that Nvidia designed their hardware around Carmack's requests?
Requests? If there were requests, the ATI drivers would have been optomized more like they are today right? You could use the Carmack analogy and say NV should have know this too regarding HL2 and it's rendering requirements.

Kevin Lowe said:
Is anyone keeping track of how many times I've had to explain this?
Many times I'm sure, but no one here really wants to pay attention to facts. :)
 
R1ckCa1n said:
Requests? If there were requests, the ATI drivers would have been optomized more like they are today right? You could use the Carmack analogy and say NV should have know this too regarding HL2 and it's rendering requirements.

It's possible Carmack was still a touch upset about the E3 demo leak. I don't think he'd hold it against them though. More likely, Nvidia just built their hardware around what he was asking for, while ATI had to retrofit it into a driver they'd ignored for far too long. (I haven't looked at the results of their efforts; my 9700 made way for a 6800 GT before Doom came out.)

If I had to point to anyone who directly influenced the design of the Source engine, I'd say Microsoft - it seems to be built specifically to take advantage of SM2.0. However, ATI did quite a bit to influence the design of SM2.0 (partly due to arrogance on Nvidia's part) so it's no surprise that Source runs so well on ATI.

Of course, it kicks ass on my 6800 GT as well, so I don't worry too much about it. A patch to take advantage of SM3.0 and partial precision would be great though. Maybe we'll see one around the release of the R5xx.
 
Kevin Lowe said:
I'm just not seeing where Nvidia was unreasonable. I mean, they designed hardware around where they thought things were going, based on the requests of arguably the authority in realtime 3D engine design.. Their hardware exceeded the SM2.0 spec. In addition, they supported a feature of the SM2.0 spec to increase performance by using partial precision. It's not a hack - it's a feature defined in the specification, and cards that support mixed precision would see better performance if Valve implemented it.

You could say that NVIDIA looked at Carmack to predict the future of gaming, where ATi looked at everyone else. In my opinion, Doom3 is not a modern game, since it basically uses the shader hardware to execute simple shading models based on NV1x, only faster. FX cards are very good at this, they're just not very good at more advanced shaders. You could argue that Valve is doing the same by making the FX series run the DX8.1 path. The difference is that they do use the advanced shaders on the Radeons and newer GeForces, while Doom3 doesn't.

And what if it doesn't look any better than FP16? Using higher precision does not, repeat not, repeat NOT automatically make things look better. Not going from FP16 to FP24, not going from FP24 to FP32.

Someone pasted some screenshots of the game running in FP16, and indeed there was significantly less quality in certain materials (especially planar reflecting surfaces like glass and metal), giving ugly banding.
On top of that, at FP16, the game is still significantly slower than in DX8.1 mode (and than the Radeon counterparts at FP24, without the banding issues)... So if you would try to preserve the quality by only using FP16 where there's no visible aliasing, it would only be slower still. I think most people (especially owners of FX5200 and FX5600) would much rather play it in DX8.1 mode, and get decent framerates at decent resolutions, than having slightly more fancy reflections at the cost of playability and resolution.
 
Kevin Lowe said:
Their hardware exceeded the SM2.0 spec. In addition, they supported a feature of the SM2.0 spec to increase performance by using partial precision.

Exceeding a spec isnt really the issue. The issue is how the complied/exceeded the spec.

The FX series shaders are just a pile of crap. They are slow, and look like shit. PP doesnt affect anyone negatively, exept when it actually DOES affect IQ.

As for FP32/24/16 i still would like to remind that ATI hardwar can also run in FP16. But it doesnt seem like they have any problems requiring lowering error margin ;).

by the way : my 6800GT rocks :D (so does my x800xt pe :p )
 
LyCoS said:
As for FP32/24/16 i still would like to remind that ATI hardwar can also run in FP16. But it doesnt seem like they have any problems requiring lowering error margin ;).

No, ATi hardware ignores the _pp hint and simply runs everything in 24 bit. Even fixedfunction is executed with the 24 bit floating point pipeline, it's the only one they have.
That's the trick... On FX, you have the 32 bit pipe, which is too slow, so it also requires a 16 bit mode... Problem with 16 bit is that it lacks the precision to emulate the older shaders, so you need to put in an integer pipeline aswell... so they put in the pipeline from the GF4 series... still it meant that ps1.4 had to run on the 32 bit pipe, which gave them problems in eg 3DMark03... running in 2 passes with ps1.1 forced was faster than a single ps1.4 pass...
All in all, FX is quite a mess, as a result of the choice for 32 bit... If they chose 24 bit like ATi did, all other problems simply would not have existed.
 
LyCoS said:
The FX series shaders are just a pile of crap. They are slow, and look like shit. PP doesnt affect anyone negatively, exept when it actually DOES affect IQ.

Which is why you don't run those shaders in FP16. You run those in FP32 instead. You don't have to run every shader at the same precision; that's the point of hints.
 
Kevin Lowe said:
Which is why you don't run those shaders in FP16. You run those in FP32 instead. You don't have to run every shader at the same precision; that's the point of hints.
Why both with all the tricker to make substandard hardware run? Why not ask why NV released that POS ;) Maybe admit ATI was right in just engineering a core to run pure FP24 fast as hell regardless.

Same thing will be said in a few months about HDR on a 6800 and how it requires you to lower resolution just to keep OK performance........
 
R1ckCa1n said:
Same thing will be said in a few months about HDR on a 6800 and how it requires you to lower resolution just to keep OK performance........

There's various types of HDR and afaik not all incur the same performance hit...
 
R1ckCa1n said:
Many times I'm sure, but no one here really wants to pay attention to facts. :)

You mean facts like "18.93% of the Steam Userbase have NVidia GeForce FX Video Cards?"

Or perhaps facts like "Of the Steam Userbase, 18.93% have NVidia GeForce FX Video Cards?"

Then again there's also the fact that 18.93% of the frelling Steam Userbase have NVidia GeForce FX slagging Video Cards, you pile of steaming Bantha Poudoo!

If any of the previously stated "curse" words are unrecognized... then your lack of nerdiness disturbs me... ;)
 
Status
Not open for further replies.
Back
Top