NV30 : not so fast?

Not that it really matters, but I work for LeadTek, and that means that I am somewhat briefed on this account. Flame all you want. I have a very thick skin.

 

TG, is leadtek doing the reference board for NVIDIA?

Cheers,
Darkman

 

i believe MSI did the reference boards for Geforce4 TI series.

 

Is this the same Umbra that is used in EverQuest and doesn't seem to do a damn thing to improve performance or reduce the drawing of occluded polygons?

 

To be fair the everquest graphics engine is an absolute mess.
Because of the way they break the terrain up on cell boundaries, it pushes a lot more polys than it ever looks like it's doing.
I doubt any graphics engine could do much with the everquest datasets.

 

EQ1 is NOT using Umbra. Umbra wasn't even around when EQ1 came out. I believe EQ2 has licensed it however.

Many of the problems with the EQ1 engine I believe lies in the fact that it wasn't designed for what EQ1 tries to do.

 

I thought VisionTek has been doing them since the GF3 or GF Ti series?

 

not at all, your just another person posting another story. Not to offend, but its fairly obvious that nobody except those under NDA know whats going on. And I don't expect NDA'ers to talk.

 

Actually doing a quick search proves that EQ is using Umbra, as I thought it was. It was introduced with the Shadows of Luclin, and on top of having problems with AMD processors in beta testing, nobody I know noticed that it really helped at all.

As for the EverQuest engine being a pile of dog poo, you'll get no arguments from me there. But that's all the more reason Umbra should show a significant improvement, IMHO.

 

multigl2,

Do you _really_ believe that? Do you honestly believe that those under NDA don't talk? Think about all the things that have been under NDA for the last 2 years...think about all the leaks (not just 3D chips) that surfaced.

The only exception to this is Matrox, and we all know they are a different breed. I'm not saying one way or the other, as far as this individual is concerend...I'm just realistic to understand that, despite the NDA's, people will still talk.

 

Off topic, but getting back to Everquest and Umbra...

Everquest in it's current version (Shadows of Luclin and above DX8 engine revision) does indeed use the Umbra engine, as licensed. But it is not a good example of Umbra's power.

They didnt redesign/retool any of the game assets to be "Umbra friendly"- but instead just converted their old, pre-glide assets/textures/models to work with it using a conversion tool. This takes a serious hit in itself as most assets convert to "worst case" in terms of performance.

Umbra has also been crippled dramatically as their userbase is so varied. Umbra offers brute-force, software occlusion routines but these were discovered to not work with K6/K6-II processors on Luclin launch, so these routines were totally scrapped from the engine. With 400,000 active subscribers (they hit 100K online at the same time recently), you would be surprised how many of them are K6/K6-II and/or integrated video owners. Furthermore, they had to kludge scene object sorting to get around bugs with older Rage and S3 videocard drivers, so overdraw is about 10x worse than it should be.

The moment the can tap 1/100th of Umbra's ability, they will wind up with at least 10,000 angry customers... like they did on the launch of Luclin.

 

Considering they're playing a pathetic game like LevelQuest, I wouldn't be surprised if 95% of them had K6 processors. When (if) they graduate to real games, they'll all have to upgrade to halfway decent systems. You have to admit these online non-RPGs are making a killing with the clueless, skilless casual gamer crowd though.

I guess I probably should have said all that, now someone who is an EQ fan will flame me for sure.

 

Shortly about dPVS, Hybrid's guy said in the presentation : "10 gigahertz demos with 1 gigahertz hw"....

I have tried the dPVS demos and with GF2 and gigahertz Athlon performance was ok.

From one of the dPVS demos abstracts :

" The database in the picture above contains two million objects. All models have 108 triangles and 258 vertices, thus pumping the total triangle count to 216M. The database file requires 130Mb of disk space. The memory used by the visible portion of the database (and other dPVS allocations) is in the order of a few megabytes. The database is not static: 120,000 of the objects are moving around every frame. The startup time of the demo is a couple of seconds. "

So, they got something there 8)

 

I can assure you that I did not break my NDA by giving that minimal piece of info. It is quite a pain if you break an NDA, even accidentally. As to the reference board, it may very well be .You`ll have to wait and see. And to people that think that it is absolutely impossible to have hard info on the NV30, think again. All official partners of nVidia have been briefed and are up to date to the status of this chip. Do you think that a marketing campaign can be concocted in the short time from the FINAL tapeout to availability on the market?

 

Leadtek have an office in the UK? I thought they just sold through a third party in the UK, Tecrep or something like that?

 

Are you related to Biggussss... Dickussss?
He has a wife you know.....Incontinentia Buttocks

sorry, just couldnt resist

 

i dunno how some people can claim there's
nv30 beta boards around when the CEO say
the nv30 has NOT been tape-out???

oh great,what's next? they gonna argue for
the next 18 pages that those pictures were
actually rendered on a nv30 beta boards? ???

 

Before they tape-out the final version they have made a few production samples.

 

Is it possible that although the final single chip NV30 hasn't yet taped out that they may well have had a multi-chip tape out to test the hardware design on?

I can imagine a whole stack of bread boards stuck together with bits of wire running in the labs.

 

i don't think there's such thing as :

"Before they tape-out the final version
they have made a few production samples."

you're talking about revisions not tape-out