Performance Predictions

[TdB]

ehh!
do you mean next-next-gen dx10 gpus?

[blacklotus]

Re: Performance Predictions

Well, considering the initial reports did mention 20-30% performance increase over a Ti4600, I'll hazard a guess of 17500 in 3D Mark 2001 SE with a P4 2.4GHz 533MHz Northwood-B.

(p.s. I'll be happy if I can just play RTCW or JK2 at 1600x1200 with/without AA at a minimal constant of 60fps.

Originally posted by BroadbandBilly
Thought it might be fun to have a guess at some hyperthetical benchmarks. For example, what do you think the card will score in 3D Mark 2001 SE if it runs on a typical gaming rig, say an Athlon 1800+ with 512MB DDR 2100?

We could predict different scores for the different 'flavours' of card (we're expecting the cards to be specced @ different speeds, etc).

Anyone who guesses spot-on can win.....

A bunny!

Either that, or we assume they've broken their NDA.

[otlg22]

TDB:

Yeah.. I mean dx10 type stuff.. and beyond..

[TdB]

well, i don´t know anything about dx10 and beyond, but i know what i would like to see in future games.

it seems to me that the ultimate goal would be realtime photorealism(or at least computers capable of it), and what we miss in that area, is probably what we will see in the future.

personally i think we need improvement in:

speed: we need more polygons, if we want pretty graphics and some good LOD, so that we don´t waste the polygons,(displacement maps looks promising though).
unless they drop polygons as primitives and use voxels or metaballs or something like that.

better particle acceleration: fog and smoke is still not very convincing, heck! even water could be done with particles.

better lightning: while ray-tracing is a long way into the future, i still think we can do better, bump-mapping is a good start, but "real" shadows is still a challenge.

animation: this could also be improved, most models in games are very stiff, I have never seen a guy in a game that appeared to have muscles, for example.

however i think it all needs to mixed together, before we are really there: lightning should take particles into account, to look great for example, have you ever seen smoke cast shadows in a game?

i haven´t thought very much about this, but this is the areas that need most improvement IMO, will we get it dx10 or beyond?
i don´t know.

[Captain Marvel]

what is the High fidelity video engine and the
VIP 2.0 video interface with the arrow leading to
the camcorder?????

[Captain Marvel]

I guess I closed this thread.....

[superfly]

This actually brings up a good question with regards to parhelia...


With all the emphasis on displacement maps,memory bandwith and vextex shaders,just how powerfull will the lighting engine be?...


I'm asking since proper lighting with usable speed is also one of the most important features and we haven't yet heard anything regarding the card's true capabilities...


Things like max number of simultanious lights it can realistically handle while still maintaining reasonable fps,and of what type as well would be nice to know(diffuse,directional,ray traced,etc...)_

[Liquid Snake]

Stay patient, Tuesday is only a couple of days away. The BBz have said that what IXBT posted isn't even all of Parhelia.

[The_King]

Originally posted by Liquid Snake
Stay patient, Tuesday is only a couple of days away. The BBz have said that what IXBT posted isn't even all of Parhelia.

[Ali]

Im pretty sure the Parhelia will have some sort of bandwidth saving features. They have a block called Depth Accelerator. What could that be except a method to speed up the Z buffer (depth)?

I dont think iXBT got their infor from Matrox, so I would be suprised if they have all the information, or if it is all correct. Eg, the guys at Chip say it does SSAA as well as the FAA-16, but iXBT didnt mention that.

also no mention of the font smoothing stuff.

Whats this great 2D revolution the Parhelia is suppose to have? iXBT didnt mention anything except the good filters. Thats a card based improvement, not a chip based one. There must be something else to this.

I have more questions now then before this (stupid) NDA leak.

I think Ill just wait for the proper reviews before asking them though.

Ali

[Parhelia Pro256]

13,500-19,400

[Dr Mordrid]

Hehehehehe......

Dr. Mordrid

[The_King]

Whats that PP256? 3DMark?

[xortam]

2D

Ali ... how about "Glyph Antialiasing - font hardware border antialiasing and gamma correction." And the iXBT article also mentioned "Adaptive supersampling (not the multisampling!) - 16x Fragment SSAA with up to 16 samples. Activated on polygon borders only."

[Ali]

<b>Xortam</b>...didnt see that. I only read the translated version of the iXBT thing, not the proper English one. By the time I had some spare time to go back and read it, it was gone

Im quite interested in the Glyph AA. Wonder what it does in comparison to MS ClearType?

Isnt Adaptive supersampling what the ATI 8500 was ment to do? Not realy up on AA methods (yet).

Ali