report from event

I'm curious, when was the FP standard changed to FP32 as minimum for SM 3.0 ?

Cause i thought that it was verified that it was FP24 .

 

One of the demo system of MSI is using a 6800 but it had only one dvi(maybe because they don't use reference Dual dvi). Interesting thing about the card was that it used a dual slot cooling that exhausted the hot air to outside of the casing, pretty much like the 5950 ultra.

 

can someone explain me why the 9800xt only has 19195 on aquamark3

http://www.hardwareanalysis.com/content/article/1708.5/

 

here is another review of the nv40 from fiingsquad found it at nvnews forum

--------------------------------------------------------------------------------

the 15,000 3dmark03 score was run at 800X600, instead of the default 1024 X 768. when run with the latter rez, it scores 12,000.

edit-here's the link: http://www.firingsquad.com/hardware...ltra/page10.asp

but anyway, DAMN, 6800ultra is fast!

I must say, nvidia's done a good job with this card so far

 

Perhaps the "move" to IBM was a good one.

 

can someone explain me why the 9800xt only has 19195 on aquamark3

http://www.hardwareanalysis.com/content/article/1708.5/

 

Awesome stuff

Is there a video of the SM3.0 Far Cry???

 

http://www.jamesbambury.pwp.blueyonder.co.uk/unreal3_0002.wmv

finally




I am slighy disappointed on the AA, I wasn't expecting gamma correction but I was thinking that they may go for an 8x multisampling mode. Still 4x is finally rgaa and thats the most important issue.

I'm now wondering if r420 will have the same angle dependant AF or not as r3xx.

 

I'm surprised we didn't see NVIDIA's "Oh, oh, GC is bad! It's not really accurate! It's only a hack! Wait till our next gen to see the REAL thing" comedy yet - maybe they're reserving that for the R420 launch

Although I seriously don't see what you mean above - NVIDIA *does* support 8x MSAA in the NV40. BTW, does anyone know if 8x MSAA is RG or not? I've only seen 4x RGMSAA patterns in NVIDIA's docs right now, so that's kind of suspicious in my book, hehe.


Uttar

 

Thanks, DC & Gunmetal......
I'm very interested to see how much heat the 6800U dumped into the box.... although knowing that at least one manufacturer has a version that vents out the back is a very good thing...... well, not for me, I'm water cooled!

Also glad to hear it's not exruciatingly loud - again, not a real problem for me.

Color me impressed. I'd have no problem with an ultra in my gaming box. Now let's see what ATI has up their sleeve......

One question (I think we all know the answer to this one ) notice that the synthetic benchmarks are way higher than ingame from the first reviews.

Now, if the gamemakers would just let me increase the text size when going above 1024x768........ even on my 21" Trinitron I can't see so good! Age has it's drawbacks......

 

I'd assumed from the hardocp shots leaked that the 8xAA was a mixed mode which is why it didn't have as good edge smoothing (for near horizontal/vertical anyway) as r360 6xAA.

Would seem a bit odd to have a 8xMS and not have it as a sparse pattern after making 4x one. unless the grid they can position within the pixel is only 4x4.

 

Adding yet another mirror to the list
http://sanbjoe.home.online.no/nv40/unreal3_0002.wmv

And wow. That was impressive.

So far it looks like nvidia has learned a thing or two from the NV30 debacle. If they're not adding cheats for the NV4x to the drivers I might even buy one.

 

xslynux

The exceptionally low score on Aquamark3 for the 9800XT is because they didn't install the AGP northbridge chipset driver before they installed Directx 9.0b.

If they were to do a

1. Re-Install AGP chipset driver
2. Re-install Directx 9.0b
3. Re-install Catalyst 4.4

You would expect a score around 50K or higher on that rig.

My little 2.2GHz Athlon and 9800 Pro scores around 44.5K on Win2000 in AM3. A regular 5950U scores around 55K as it uses _pp to lower the workload alot relative to a 9800 Pro / XT chip - so don't be fooled into thinging its a straight apples vs apples comparision. The 5950U scores would be 40% lower it it were forced to use high precision mode like teh ATi cards.

 

Around August, one of my ex-sources, now in another technology sector (which is, amusingly, twice as lucrative as the GPU one, looking at his new salary hehe) told me NV36 was the smoothest chip in all of NVIDIA's history. The tape-out went great, yields were pretty good, and they managed to beat ALL records when it comes to how little time they needed to get it up & running in Q3 debug (barely took them a few hours to get it running after they got the silicon back from IBM!!!)

As to whether yields were good in the long run, and whether NV40 yields are good, I've got no clue, frankly.


Uttar

 

They've revised the price down to $500 -- from the conclusion:

"If you can afford it, as suggested retail pricing is around $500, you’ve got yourself a graphics card that’ll chew up anything you throw at it with performance to spare."

Pharma

 

thanks for the heads up, 600 dollars would have made it a helluva expensive card.
later,
epic

 

$500 is still hella expensive for a card for some of us!

Thanks again DemoCoder for the excellent event coverage and goodies, it was faboo! Thanks to everyone for the mirrors and contributions to the thread, it was a great way to wake up and get right back into the swing of things.

Now that I got me appetite all wetted, I'm gonna go read me a few of those reviews. 8)

Sounds like the 6800U is delivering on a lot of nVidia's promises, this is going to be a fun match-up!