Radeon 9700 Or NV30?

[HellPreacher]

Thanks
That's what I thought

In my last post I said that it isn't worth it to buy an all-mighty radeon 9700 when you already have a g4 4400/4600 and good machine.
And it's not, because all games within the next few months will be played with frame rates above 90 fps (underlined above)

When games turn out to be played slow, go to the stores and you'll have radeon9700 and NV30 plus their families
Also with the proper hardware to give them the space to explode.

 

[Tarmax]

*sigh*

Poor guys. You do realize that games won't be able to take advantage of DirectX 9.x until, I'd say July of next year at the earliest. Hell, games that take advantage of DX8.x are just starting to trickle out. There's really no point in spending ~$400 on a video card that has technology that won't be taken advantage of until a year later (in games). It's pointless.

The only point that I can see to picking up one of these 2 cards is for bragging rights. That includes benchmarks too. I mean, come on, do people really care that much that you can get more than 20,000 3DMarks in 3DMark 2k1?? Are the other players in that CS pickup game really that cynical when even they can't tell what framerate (over 100) that poor DX5/6 game is running at???

Hell, UT2k3 is basically a DX7 game with DX8.x effects thrown in here and there to make it look more attractive. Look at America's Army. The game looks amazing, and it was just released. The game is based on the Unreal Engine (same version that's being used in UT2k3), and it's only using DX7 effects (it requires a GeForce2 card to be played at playable framerates.. ~30).

In the end, it's not worth the money unless you run 3DMark 24/7. Hell, the human eye can barely tell the difference between 60 and 100fps.. do you really think your eyes are going to care if you're running the game at 300fps??

Also: Doom3 (which is actually just a remake of Doom1, with a slightly different story line... it's not a continuation of the 'saga') is only a DX8.x generation game, and it won't be released until Xmas, if that. WarCraft 3, possibly the BEST RTS yet, is only using DX7 'technology,' again, if that. It may even be using DX6 as a baseline..

And just to reiterate one of my many points: until at least this time next year, the only "games" that you'll be able to play that use DX9 effects (Vertex/Pixel Shader 2.0) will be 3DMark2k2/3 and whichever tech demos ATI and nVidia release.


But until that time comes, I'm going to buy a Radeon 9700, and go render Final Fantasy: The Spirits Within in real time at 2fps.. ;P

Oh yes, and from what I gather.. the NV30 will only be able to render it at ~3fps in real time Unfortunately, the NV30 is not the Holy Grail that nVidia is trying to make it appear as..


The only thing that will differentiate the NV30 from the Radeon 9700 (in terms of hardware capabilities) is the core/memory frequencies, and the ability for the NV30 to execute an ungodly number of instructions. But answer me this: what good is it if you can only execute 65k-instruction vertex shader at 3fps???

Not exactly worth the 3-month delay, IMO. You're better off going with a Radeon 9700 or something DX8-ish. Ti 4x00, Radeon 8500, Trident XP4, etc..

-Tarmax

And please, no flames. I didn't type all that just so some 12 year old who knows nothing about this industry can tell me I'm an idiot. Talk about the pot calling the kettle black.. =P

If you disagree with my opinion, well, that's nice. I respect your decision, as I expect the same from you.
We're all entitled to our own opinions.

And lastly, thank you for taking your time to read this ungodly long post I'm sorry if I've wasted your time

 

[WyteWatt]

Tarmax i am not saying you are wrong or anything because anything could happen but what if the N30 does what the ATI 9700 did to the GF4 ti 4600 ? Or maybe even faster? Then would you surprised? In a way i would be surprised but in one way not really.

 

[Uttar]

Originally posted by: Tarmax
*sigh*
The only point that I can see to picking up one of these 2 cards is for bragging rights. That includes benchmarks too. I mean, come on, do people really care that much that you can get more than 20,000 3DMarks in 3DMark 2k1?? Are the other players in that CS pickup game really that cynical when even they can't tell what framerate (over 100) that poor DX5/6 game is running at???

After reading your whole post, i've got the following to say:
I agree with you that i see no use at all for 100FPS in any game and that DX9 effects won't be coming for new games so soon. But there is a DX8 game i'd really want a NV30 for - Morrowind.
There are some very nice configurable options in morrowind.ini to increase the quality of the graphics, mostly the pixel shading of the water ( an area where the NV30 is supposed to be great )
Also, i like AA and Aniso with Morrowind - it greatly increases the quality. And let's not speak about max view distance.
Conclusion? In Morrowind, i barely get 7-14 FPS with those settings, and even less if i change morrowind.ini a little to have better quality graphics.

Yes, Morrowind isn't really a greatly optimized game - but i doubt they'll optimize it much more. So if i want to see all of the great graphics of it with AA and Aniso, i need a NV30. So it'll still have a use for some of us, but not for much of us


Uttar

 

[Tarmax]

Originally posted by: imtim83
Tarmax i am not saying you are wrong or anything because anything could happen but what if the N30 does what the ATI 9700 did to the GF4 ti 4600 ? Or maybe even faster? Then would you surprised? In a way i would be surprised but in one way not really.

Well Tim, if that does happen -which I seriously doubt it will- I'll eat a broom =/

LoL Uttar. Yeah, Morrowind is one beast of a game. I'm suprised it even runs on the Xbox.


Other than that, kinda correcting myself: I'm obviously not taking the effects that Cg and RenderMonkey will create into account. Hell, that could be why UT2k3 has been delayed so much.. maybe Epic got their hands on a copy of Cg or RM

But yeah, Cg and RM could have an astounding effect on the PC gaming industry. For example, instead of waiting ~1 year to see current technology appear in games, those two programs (if you would call them that) could significantly shorten that wait down to a few months, or even less, depening on how good ATI/nVidia's developer relations are. But I think I can safely say we won't see any real DX9 games until at least Q2 2003, if that; but like I just said. Cg and RM will significantly alter that...

-Tarmax

 

[WyteWatt]

Tarmax plus i guess if we wait for the N30 then the R300 may be alittle cheaper at least! I want it to go down to at least $250. Or even if we waited a while for the N30 to go down too. I really hate having to pay $400 for a video card when it first comes out. Thats mostly why i brought a Leadtek GF4 ti 4400 video card for around $250 and overclocked it to GF4 ti 4600 speeds for $100 less. I just can't see spending much more on a video card. Heck $250 seemed a lot but it was not as bad. Thats the most i ever spent so far. Well it was really $263 but thats pretty close to $250 hehe.

 

[ant80]

Patience, my friends, patience. There is an Indian folk tale that my mother used to tell me. There was a king who wanted a flying horse. He ordered his minister to create one. The minister agreed, but asked for a five year period. His wife asked him the reason and he said, "A lot of things may happen in five years. The horse might die, and I might another five years. Or I could die, and that would spare me death from the king. The king could die. And if none of those happens, I am sure the horse will fly".

So hang on folks, the horse might just fly.
ant

and please dont flame me. I am only trying to help.

 

[AnAndAustin]

Whoa ... this thread's really getting some response!

GrimReefer, I did have a quick look for Rad9700 scores in 3Dmark, but they are still very thin on the ground (like 5 in total) ... are there any final release retail cards available yet? IN any case here's what I can find on the Rad9700 and how it seems to compare...

AthlonXP@1600mhz:
Radn9700: 11500, 47, 122, 58, 93
GF4TI4600: 11600, 56, 119, 70, 67
Rad8500: 10000, 56, 92, 68, 53

P4@2.7ghz:
Radn9700: 15000, 73, 150, 90, 99
GF4TI4600: 14000, 74, 132, 92, 77
Rad8500: 11000, 68, 95, 81, 54

Of course without many Rad9700 it is very difficult to comment on the overall accuracy of this, but this is precisely what I have currently found on 3Dmark2001.

 

[AnAndAustin]

Tarmax, you make a lot of very good points, both DX9 and AGP8x are pretty pointless until games are out to take advantage, and when that happens there will be both faster and cheaper DX9 cards anyway. However, Rad9700 does show excellent promise in terms of eye candy and pure 3D speed, anybody looking for the best and willing to pay for it will not be disappointed. For those of use with a GF3, GF4, Rad8500 or Rad9000 there really isn't any point in upgrading until the cheaper versions of the new DX9 cards are released.

For those who have less than a DX8 card, upgrade to one of these fine cards, they're a steal at $90-150! With Rad9700 I don't think it's so much of wanting to play Quake3 at 300FPS but about playing even new games in 1280x1024x32 with max Aniso and AA settings and still achieve 100FPS. For those with the money, why not go for it? You know it sure would be cool to know what nVidia know about the NV30, are they quietly confident or shaking in their shoes?

 

[GrimReefer]

YOU STUPID BASTARDS, GO PREORDER YOUR 9700 RIGHT NOW YOU GOD DAMN COMMUNISTS!

 

[sxr7171]

Judging by the article now on Anandtech, I think Nvidia screwed up for the first time since their inception. They took long to get the NV30 taped, and they made took a risky decision to introduce the NV30 as a 13 micron part that turned out to be a bad decision. I suppose that in this industry companies have to make risky decisions and sometimes they end up on a wrong side of a decision. Things could have worked the other way around as well with TMSC 13 micron yields rivalling Intel's and the silicon taped around the same time as the R300. Maybe in the long run it won't make much of difference maybe it will?

As for the NV30, I picture it as a faster R300 since for all practical purposes they have the same architechture mandated by DX9 compatibility. Given that they will be 4 months late, I don't see that as any improvement on the R300 in this industry as timing decides the value of your product. I expect ATI to release a 13 micron 9700 maybe a month after NV30 hits the streets and it wouldn't surprise me if it turns out to be faster than the NV30 given its slightly smaller feature set. To me it seems that we have to a point where both ATI and Nvidia have basically been asked to make the same part by Microsoft to be compatible with DX9. So it seems that both companies are in the same place technology wise and going forward we may each company branching off with variations on the floating point graphics processor theme.

 

[AnAndAustin]

Just to comment on the 3Dmark Rad9700 results listed above...

The AthlonXP@1600mhz obviously refers to the AthlonXP1900+.

IMHO Rad9700 is better than these results show, 3Dmark2001 is really showing the probs of '1 standard benchmark for all' and 1024x768x32 without any candy is really not showing the diffs in these cards, these cards can't truly get going. The latest cards, even GF4TI are not showing there true potential until they are set at 1280x960x32 with some degree of Aniso and/or AA.

 

[AnAndAustin]

Regarding 3Dmark2003, I know it isn't always this easy or simple, consistency in AA and Aniso is a complex area, but I think in addition to DX9 funcs they should also impliment 3 distinct standard benchmarks like so:

Lower Tier directed at DX7 cards like GF2TI, Rad7500 etc and should run at 1024x768x32.
Medium Tier directed at DX8 cards like GF3, GF4, Rad8500, Rad9000 etc and should run at 1280x960x32 AA & low Aniso.
High Tier directed at DX9 cards like Rad9700, NV30 etc and should run at 1600x1200x32 AA & high Aniso.

 

[WyteWatt]

Wow the ATI 9700 is not very fast without having FSAA or AF on maximum I don't even use FSAA or AF because i never notice a difference in any games or resolution. I tried racing games, plane games, RTCW, quake 3 arena, SOF2, etc with FSAA and AF to maximum and so no difference at all. But then again thats my opinion. If the ATI 9700 is around the same performence of the GF4 ti 4600 with no FSAA and AF on then i know its not for me. Too bad its like that. I wonder if the ATI 9700 makes any games like Quake 3 arean, RTCW, nascar racing games, plane games, etc look more movie like with no FSAA or AF on? I have a feeling no. Too bad. I wonder if the FSAA and AF is more notice of a difference in picture quality that i could see with the ATI 9700 video card than other video cards. I have a feeling no on that too. But i could be wrong. I just wish in a way i could tell the difference when FSAA and AF is on maximum.

 

[jbond04]

Tarmax, you forgot one other use that one might have for the NV30...one area where it's ungodly instruction length for both pixel and vertex shaders will come in handy. That use is the replacement of renderfarms. I plan on purchasing an NV30 not just because it will kick ass in games, but also because it will speed up the rendering of complex 3ds Max scenes by several orders of magnitude. With such an increase in productivity, spending $400 or even $500 on such a card is more than worth it. One area I think that ATi has not taken into account (or, at least, I have not heard of) is the creation of "plug-ins" for 3D animation/visualization software. nVIDIA is creating a plug in for several of these packages (i.e. Renderman, Maya, and 3ds Max).

DX9 graphics cards will become more than tools to accelerate games...they will become truly workstation-capable (current workstation cards are pretty much a joke) graphics cards that will greatly reduce the cost and sheer difficulty of rendering large, ultra realistic 3D animations.

(And BTW, 3fps vs. 2fps is a huge difference in 3D animation, where you have tens of thousands of frames )

 

[AnAndAustin]

sxr7171, just read that article. I'd take the vast majority with a pinch of salt, nVidia are going to experience the obvious probs by going 0.13mu, I hope for their sake they are finding 0.13mu brings the clocks etc that it should, unlike AMD's AthlonXP T.bred core (oops!).

It seems clear nVidia aren't in the position they would like, NV30 IS late and they no longer have the top product. It will take a lot more than a late release and teething probs to knock nVidia, they still have the biggest share of the market, and with GF4 cards have happily (for them) saturated the market with their own products, before thinking of all the people they got to pay $350+ for the GF4TI4600 over the last 6 months! IMHO ATI should capitalise, soak up a few weeks of Rad9700 at $400, then stick a $200 Rad9500 out there to really give nVidia some headaches. Much the way nVidia did with GF4TI4400 and 4600 cards before quietly sticking the excellent 4200 out to there after 8 weeks of silence.

 

[AnAndAustin]

Now if ATI can only get their global pricing sorted!

imtim83, AA is little use when running in a high res in relation to your monitor (IMHO 15" at 800x600, 17" at 1024x768) as the squares which make up the image are far less noticable. With Aniso, you should notice a big diff, very evident in the sharpness of textures, esp distant textures, it is great with AA. Radeon cards have fast but relatively poor Aniso, while GF3 & GF4TI have great Aniso quality but double the hit. Of course the impact of these enhancements does depend largely on the game's engine and implimentations. Beauty is in the eye of the beholder with these things but I would suggest trying these settings on a GF4TI card:

800x600x32
800x600x32 2xAA (only 10% perf hit)
800x600x32 QxAA (only 10% perf hit)
800x600x32 QxAA & low Aniso (about 20% perf hit)
800x600x32 QxAA & high Aniso (about 30% perf hit)
800x600x32 QxS AA (only 40% perf hit)