→ peepers:幹你媽的腦殘 寫甚麼教學 07/23 18:41
→ peepers:對不起..........我在測試.....跑錯地方~"~ 07/23 18:41
好像是很特殊的架構....
http://www.theinquirer.net/?article=14878
Nvidia's NV40 mimics a 16x1-pipeline architecture with just eight pipelines
Exclusive: Vertex Shader 3.0 workaround
By Fuad Abazovic: 星期一 22 三月 2004, 12:54
INVESTIGATIVE JOURNALISM is not an easy thing. But we do what we have to do ju
st as Nvidia and ATI need to make their chips. Our job is to post news as exclu
sive as possible, this is occasionally at the risk of jeopardising communicatio
ns with the company that we write about. Still we have to do it.
Anyhow, we have something that you all have been waiting for.
It was easy enough to state that Nvidia's NV40 was an eight-times-two pipeline
architecture before and now claim it's 16x1, but it was very hard to find out h
ow this chip actually works without being able to speak with the company's Davi
d Kirk and ask him all the nasty questions. We don’t sign NDAs, that’s our pr
oblem. If we did, we'd know, but we wouldn't be able to tell you. As it is, we
can post this stuff as soon as we get it.
So, the claim that we made more than once that NV40 is 16x1 chip is partially t
rue and but the claim that I made previously that the chip is 8x2 is also true.
How's that possible?
NV40 will score magnificent 6500 Mtexels per second in 3Dmark01 whereas ATI's R
adeon 9800PRO will give you around half that, but there is a catch here.
We understand that the NV40 will actually only have eight physical pipelines, b
ut these will appear act like 16 in certain games and, indeed, in 3DMark 2001 N
vidia was telling people how 3DMark01 is a very nice benchmark since they can r
ender 16 textures per pass in it, using only eight pipelines. What Nvidia is us
ing is the ability of the Vertex Shader model 3.0 (known as PS 3.0 or VS 3.0) w
here the Shader can actually render textures as a virtual pipeline but can only
render them without filtering information.
But, in modern games you always apply Bilinear, Trilinear or Anisotropic filter
ing to textures that you render, otherwise they will have some visual malformat
ions. This is why, in normal games, eight pipelines will be able to render eigh
t textures in single texturing pass and possibly even 16 if you render both tex
tures in same pass using an 8x2 architecture.
What we understand Nvidia is doing is processing eight textures by pipelines wh
ile the other eight textures get processed by the Vertex Shader 3.0. If they de
cide to use this in modern games this would roughly mean that they can render e
very second texture properly, which would mean that you would be able to notice
the difference. However, this can be tweaked in the driver if they decide to g
o that way, but image quality would have to be sacrificed for that.
To get into more details, the Vertex Shader model 3.0 feature has the ability t
o fetch textures but you could use your VS 3.0 only where you don’t have any g
eometry - when you have to process just pixels. This cannot be used in modern g
ames as you process just point sample textures. In 3DM2001SE you get such high
numbers because you deliver 16 pixels per clock.
We asked Nvidia to comment but since they haven’t announced NV40 they sdaid th
ey can't comment on "unannounced products". But we expect they will be able to
tell us more then they launch on 14th in the old continent.
ATI's R420 offering remains at 12x1 architecture which makes this games-running
business very complicated, since you cannot bring any final judgment about who
will win the chip speed race.
However, we expect NV40 to deliver 16x1-like performance in Doom 3 (or should w
e say 16x0, which is how people call Nvidia's approach) where they render all t
he information except Z or colour features. This will be used by Doom III and g
ames based on this engine.
It's likely that Nvidia will launch up to two weeks prior to R420's introductio
n. μ
※ 引述《jhs1213 (...)》之銘言：
: http://www.nvidia.com/object/nzone_geforcelanparty_home.html
: GeForce LAN Party
: You're invited to the launch of our next-generation GPU, and the first ever
: GeForce LAN Party, in San Francisco!
: Are you thirsty for a new level of graphics technology? Tuesday, April 13th,
: NVIDIA will be hosting an unprecedented all-day LAN party in conjunction with
: the introduction of our next-generation GeForce GPU. Play Unreal Tournament
: 2004 and Battlefield Vietnam the way they're meant to be played, on custom
: NVIDIA gaming machines powered by our next-generation GeForce GPU!
: In addition, you can join NVIDIA executives and industry luminaries for the
: launch event that evening. Whether you want to play in the LAN party, or just
: want to attend the NVIDIA launch event, be sure to register now for your chance
: to win cash and prizes!
: Win a trip for two to the CPL!
: All attendees at NVIDIA's GeForce LAN will be entered in a raffle for a special
: giveaway from the CPL. The raffle winner will receive two airline tickets and
: a hotel room, as well as 2 spectator passes to the Cyberathlete Extreme World
: Championships this summer in Dallas, Texas. This prize has a combined value of
: $1,500!
: ※ 引述《jhs1213 (...)》之銘言：
: : 需要外接兩個4-pin電源@@
: : http://www.theinquirer.net/?article=14827
: : First hands-on look at NV40
: : CeBIT 2004 Shocking power requirements
: : By Paul Dutton: 星期五 19 三月 2004, 16:33
: : THE INQUIRER has managed to get hold of one NV40 sample.
: : We agreed not to photograph what we saw, but can confirm the following.
: : When removing the heatsink, we noted that the NV40 GPU itself was, in
: : comparison to what we have seen before, quite large.
: : However whilst our earlier information indicated that it comprised of 205 -210
: : million transistors, our sources currently say that what we were holding was a
: : 175 million transistor part.
: : With the heatsink removed we could see that a small formed surrounding was
: : employed to assist leveling of the heatsink.
: : Unlike the ill fated GeForce FX 5800 Ultra (NV30) where a heatsink on the back
: : of the PCB could also interface with a small section of the PCB directly behind
: : the GPU as well as the memory modules, additional circuitry in this area on the
: : NV40 prohibits this.
: : Whilst a heatsink still seems necessary to assist thermal control of the new
: : lower voltage memory modules, the heatsink on the front of the GPU will have to
: : do all the work.
: : Eight 32MB video memory modules are installed – four on each side of the PCB
: : and as we first reported from Computex in Taiwan – these are of the new GDDR3
: : type.
: : However these were not from Micron, which at that time seemed to be both ATi
: : and Nvidia’s GDDR3 development partner. Instead, Samsung seems to be the
: : preferred supplier of the 256MB we saw installed.
: : The primary shocker was that the board requires two large 4-pin power
: : connectors as opposed to the single power connector on current high-end products.
: : Perhaps a lot of end-users who have been thinking that they might wait and
: : upgrade to NV40 when it becomes available, should start saving their sheckles
: : to upgrade to a new power supply unit as well.
: : One final thing, contrary to previous speculation, we think that ‘NV40’ that
: : will come to market may be what was originally planned, as NV45 is, we’re told
: : , the PCI Express part.
: : Whether that has an AGP to PCI Express bridge chip remains to be seen.
--
※ 發信站: 批踢踢實業坊(ptt.cc)
◆ From: 202.178.160.225