Is Geforce 2 TnL useless in the future??

[bananaboy]

I am interested in upgrading my GeforceMX to a Geforce2Pro but considering a Kyro 2 based on this information. I read a review that showed in Aquanox that all the Geforce 2 level cards performed on par with the Kyro2 a card that has no onboard hardware TnL support. I dont know if any of you have read a similar review, I just did a extensive search trying to find what review it was because alot of you will not believe me otherwise. I think the review was off of Toms Hardware.

I assume this is because Geforce 2 based boards don't support the programmable TnL that is required to fully take advantage of future TnL games. I also take into consideration that future games based off of DX8 will only use this programmable TnL not the Geforce2s implementation.
This has been the major factor in my decision of Geforce2Pro or Kyro2... and should probley also be for anyone else looking to upgrade until more tile based renderers are available.

I feel its odvious that immediate mode renderers are out of air and the Geforce3/3Ultra will mark the end of the era, thats why I'm leaning towards the Kyro2 but curious on what feelings are on Geforce2 TnL.

 

[bananaboy]

I found it! It was Anandtechs very own review of the Geforce 3..

"It isn't a surprise that the GeForce3 comes out on top again, what is interesting is that the three GeForce2 cards and the Kyro II are all capable of the same real world T&L power. Keep in mind that the Kyro II has no hardware T&L, meaning that all four of the cards are relying on the host CPU for the T&L processing.

If this is an indication of what can be expected from future titles, are GeForce2 owners left in the lurch with a hard-wired T&L unit that will yield no tangible performance improvements in future games? If developers all move to support programmable T&L like that on the GeForce3, which they most likely will, will the T&L units on the GeForce2 series of cards be rendered completely useless?" Link
BTW my system is a Athlon 700/100mhz fsb @ 733/133fsb w/alpha heatsink/fan, 256mb pc133 cl2 crucial, abit kt7a, evga geforce mx if this information might help if someone wants to make a suggestion one way or the other. i might overclock the cpu more sometime but i dont feel a need to, the 266mhz fsb helps enough for now.

 

[vss1980]

As processors get faster, they will eventually be able to not just run the game but process polygons faster also.

The GF2 T&L unit was as fast as it needed to be over 1 year ago (yes it is that old).

One question you have to ask yourself, has the T&L unit been of any use anyway (as in, how many games do you have that use it)?

Edit: Even todays CPU's (top of the line ones like 1.5GHz Palamino's) must be getting to the point where they could literally outperform the old Voodoo1's.

 

[TravisBickle]

not true actually. I saw an article on the voodoo 1 of years ago. software mode was not as fast as hardware rendering, and moreover software mode was cheating a great deal, as you could see by the image quality. projected CPU speed to truly do what a video chip does is absolutely enormous. i'll be damned, can't find the article again.

 

[vss1980]

Travis, I know what you are talking about. There are a few problems with the analogy (comparison was Quake2 hardware and software rendering) as far as I saw it.

Firstly, SSE and 3DNow! instructions are much faster than CPU maths abilities when used properly. Quake2 knows nothing about them and considering that the PII was new technology when it was made so expect only MMX optimisations.

DirectX is actually able to do a complete software Direct3D, but it is very slow (compared to a 3D card). I am willing to hedge a bet that if there was a highly optimised software 3D renderer (and lets face it, hardware ones are the height of optimisation) I think the 1500MHz Athlon 4 should be able to beat the Voodoo1.

 

[Archknight]

I have to ask you a question, how long do you plan to keep the card? If not too long even up to a year both will do find. You will see improvement since coming from a MX. I would pick the KyroII for the price.

Just remember that Aquanox was tailor to show off Geforce3's strength. I guess Nvidia slip this time, they worked so closely with Aqunox on the GF3 and forgot about their own T&L on the Geforce2 series.

Eventually the GF2 T&L will left alone yes. But how long maybe only the game developer can tell. But ask yourselves that how many games really take real advantages of the T&L now. Even in the infamous Nvidia PDF the games listed were limited, (they even include Diablo2).

Since the price of CPU is dropping and with the Palomino out, the price would drop even further. Save some cash from the video card and add more to it to get a better CPU. Worry free.

 

[DaveB3D]

For any app that uses vertex shaders, the GF2 T&L unit shuts off... So as those apps become more and more common, it will be useless. That is why we were waiting to do T&L at 3dfx. Just do a vertex shader and not have a useless T&L unit.

 

[BFG10K]

No, I don't think it's going to be useless. Even if the programmable stuff can't be done, you can still do the repetitive hardwired calculations as normal. Also you can use SIMD optimisations in the drivers to help the CPU perform the programmable calculations.

Honestly, your card will probably run out of fillrate before its T&L unit will become obsolete. By that time everyone will have a card with programmable T&L, so the issue will be closed.

 

[jbirney]

Useless? Maybe, maybe not? Will developers still program for a static TnL? Maybe, but only for backwards compatability...

I still feel that 3DFX was right in their blief that Hard Ware TnL was not needed back then. Tell me honsely how many games were greatly improved thanks to hardware TnL? A few. However we are almost 2 years into having consumer hardware TnL and we only have a small handfull of games that use TnL. If the Aquanox is a valid bench mark of things to come, then it looks like Gary T was right as a hardware TnL may become a bottle neck. We shall have to wait and see..

 

[mchammer187]

I think the GF2 T&L will eventually become useless but it will last you well into the future. As for Kyro II vs GF2 Pro I dont think it really matters because you win either way if you buy these cards.

 

[RobsTV]

If the T&L programing is done right, it is a huge benefit.
A good example of this is the new DX8 sim, Nascar Racing 4.
It allows you to choose D3D or D3D with T&L, if your video card supports it.
With a GF2 GTS, I get about a whopping 50% boost using D3D with T&L, so it IS working fine!
This on a 1GHz AMD CPU, not a low end CPU.
I can only guess that as more DX8 games come out, it will be used more, not less.

 

[bananaboy]

It looks like I'll order the Gforce 2Pro because I can get it for $140 with tv out and to get tv out on a kyro2 its about the same to get tv out. I couldnt give the same for a card that will never beat the geforce pro but its a good technology to watch I hope they get tile based rendering on a faster chip and memory then I think it will be all over for anything that isn't tilebased.

 

[DaveB3D]

Rob,

You are confusing fixed function and vertex shaders. GF2 is fixed function. As vertex shaders become more common, fixed function T&L will get less and less use. When a fixed function pipeline is there and no vertex shader and shaders are used, the fixed function unit shuts off and goes to the CPU.

 

[LuciferHaze]

Absolutely not. Game developers are, just recently, writing T&L code so those with GeForce2's will be in the driver seat. Game developers aren't gonna alienate the vast majority of gamers who still have GeForce2s by using the GeForce3's programmable T&L.

 

[dfloyd]

I think the point is that Anand did a test on one of these soon to be released games that you are referring too and found that a card with no T&L performed exactly the same as the Geforce II. So what if they are writing it for the card it is already becoming limited as the CPU is doing it faster.

 

[TravisBickle]

unless you program your own games, T+L would not have helped you much, because only a small industry minority supported it.
we are over 18 months past the original Geforce, and we're still debating if T+L was useful, or is going to be useful. I never played a game where I was deprived because I wasn't using T+L. T+L is about triangles, and the real problem with video cards currently is bandwidth... haven't you read the news?