new NVIDIA GeForce 6 series

[Ckerr812]

Originally posted by eple
Do these offer any significant improvements for content creators or are they mainly aimed at gamers???
[/url]

It's up in the air really, who knows. These cards aren't meant for DCC but they work with it.

The 6800 has an onboard video processor which would be a god send for video encoding and decoding.

I am thinking of buying it too, but it's a risky investment.

If you get the quadro your not risking much, you know it will work and be stable.

You get the geforce, it could work, only thing that stops me from buying it is all the games are DX9 now, and Maya and XSI are pure open GL apps, so my guess is Nvidia is going to put all there brain power and software engineers on optimizing for DX 9 (like ATI did), and Open GL apps (like maya and XSI) could suffer because of it, they might abadon ALL support for geforce in Maya XSI (not that there is much now, but at least they update the odd time for fixes to maya and xsi) and just tell people to get Quadro's....I hope that dosen't happen, but you run the risk it could...

Who knows though, only time will tell.

[dmeyer]

http://www.anandtech.com/video/showdoc.html?i=2023

Wow..look at that cooler....hahaha

:rolleyes:

[3dRacer]

I wonder also how it will work with 3dStudio, I read that someone meant that it would be even better than a quadro 3000! Can that be true???

[iC4]

well....evertime when nvidia starts a new graphiccards line you read alot bullsh... When they introduced the gforce fx you read that it would be 100% faster then ati......but after they released it the fx wasn't that great....

but hey....if it is really that great it would be nice :)

[3Dfx_Sage]

The FX series was a total flop, however the new GF6 series is definitely the top right now. I don't care how optimized the Quadro's are, when you compare the 4 pixel pipelines of the high-end FX series to the 16 of the GF6 who do you think is going to be faster?

I no longer read anandtech for graphics, they are well-known to be nVidia-biased. The only site that I know of that manages to maintain objectivity is beyond3d.com and here's (http://www.beyond3d.com/previews/nvidia/nv40/) the link you want.

[heavyness]

Customers who end up buying this card will most likely need to upgrade their power supply as well.

this thing sucks up more watts then my vacuum cleaner. looks they should package new power supplies w/ the card. but then again, its so damn powerful, who cares. if the computer crashes, you might need to jump it w/ your car's battery to get it going again.

[3Dfx_Sage]

nVidia recommends at least a 480W PSU. If you're running dual CPU's then you can look forward to buying an even bigger one.

[imashination]

Interresting reviews. I've now read 6 review sites which all show identical benchmarks. Are these actual reviews or just the marketting BS that nvidia sent them?

[3Dfx_Sage]

truth be told, most reviewers don't "waste" their time running benches themselves. instead, they just take PR's numbers and repeat them. Look at the B3D article, I know Dave did his own testing.

[stephen2002]

I find it funny that Guru3D.com, Beyond3d.com and NVIDIA's N-Zone are all down/running slow today. NVIDIA did a good job with the hype this time around :)

The video processor is very intriguing as I do a lot of video editing. My system (dual-CPU) can currently handle video encoding with most formats in real-time and sometimes greater though. But it would be great to be able to have a video encode going on the GPU in real time and still be able to work with the system.

I look foreward to more benchmarks. The numbers so far are pretty impressive, with a pretty good margin (30% or so at the least) above ATI's 9800XT. The resolutions that the card can spit out at playable framerates are very impressive.

From a preformance perspective it is going to be even more interesting when ATI's next gen-stuff comes out.

As for the cooler, NVIDIAs marketing pics show a single-slot design. The huge thing is just an engineering sample.

Based on the preformance in OpenGL games NVIDIA is still on top when it comes to OpenGL support. Supposedly the card is 100% OpenGL 1.5 compliant.

[3Dfx_Sage]

Originally posted by stephen2002 I look foreward to more benchmarks. The numbers so far are pretty impressive, with a pretty good margin (30% or so at the least) above ATI's 9800XT. From a preformance perspective it is going to be even more interesting when ATI's next gen-stuff comes out. you really shouldn't be comparing the GF6 to a Radeon 9xxx and nor should reviewers. The R3x0 is a previous generation product not the GF6's competition, and doesn't belong there.

As for the cooler, NVIDIAs marketing pics show a single-slot design. The huge thing is just an engineering sample.actually, the "huge thing" is not an engineering sample, it's the reference board. The engineering samples (the ones that were circulating at CeBit) were running at 475MHz and the reference boards are running at only 400. Of couse, the boards that actually get shipped from some manufacturers will likely be fitted with several different (though similar) cooling solutions, just as with the FX series. I suggest you read the B3D article as is explains all of that rather well.

[CgFX]

Originally posted by 3Dfx_Sage
The FX series was a total flop
You consider greater than 65% market share for the DX9 class market segment a total flop?

Pretty big complement to nVidia that when they don't get 80 to 100% marketshare for a new generation it is a failure.

What will be interesting to see is how much of a hit ATI takes for their next card only being the same (partial) Shader Model 2.0 architecture versus a next generation Shader Model 3.0 architecture.

ATI's focus was to release a new R3x0 chip with some tuning and a native PCI-E interface. I think that may end up costing them. Their only hope is if R420 is much faster with older and current DX9 Shader Model 2.0 games so they can focus on that and overpower this not being a new architecture and the fact that they have no support for DX9 Shader Model 3.0 features coming out.

[CgFX]

Another interesting point is that R420 was to have a large chunk of its pipes disabled and hidden as they felt they could beat NV40 with 8 or 12 pipes. After Game Developers Conference it came out that they were going to light up all 16 pipes from the get-go.

Clearly an "uh oh!" move based on developer feedback as they were hoping to reserve those pipes for a later fight with NV45 or later but need them now to compete with NV40.

Should be interesting.

[Novakog]

http://www.overclockercafe.com/

Have a pretty big set of reviews there, a few (such as the bit-tech one, as well as other places on the web) show the retail version with a cooler that only needs one slot.

you really shouldn't be comparing the GF6 to a Radeon 9xxx and nor should reviewers.

What else can they (reviewers) compare it to right now? They're obviously not saying it will be better than the R420, because they don't know for sure, just for a comparison of speed increase between generations. They have to compare it to something, or else a lot of people won't know whether the numbers they're giving are fast or not.

Most of the results show it seeming to be more than 50% better when it comes to 4x AAing, 8x (or 16x) AFing, and 1600x1200 res than the 9800 XT (again, just for reference). Unfortunately, as it uses [what is it called, super sampling?], the 8x AAing makes a huge performance hit.

What I'm really waiting for is seeing some benchmarks of the 6800U running in Doom 3 or something, in "32x0" mode.

I'm also concerned about nVidia stopping support of Maya/XSI with gamer cards, but it seems unlikely they'd take it to the extent that ATI has (in consideration that, supposedly, even the FireGLs suck in Maya and XSI).

And the million dollar question, how long till the NV40 based Quadro cards?

[gigantor]

Damn that Nvidia card is amazing, it scored over 14k in 3dmark03 live in front of an audience.

[Shinova]

So, would getting this 6800 make any difference in 3d work or is 3d still mostly dependant on cpu and ram?

[DeViSoR]

and check this out.. http://www.jamesbambury.pwp.blueyonder.co....nreal3_0002.wmv

think I'lll get that card..

damn happy that I get a lot back on my taxes ;)

new system og lw8 ;)

[imashination]

Originally posted by DeViSoR
and check this out.. http://www.jamesbambury.pwp.blueyonder.co....nreal3_0002.wmv

think I'lll get that card..

A real link would be a tad more useful.

[DeViSoR]

sorry

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

[3Dfx_Sage]

Originally posted by Novakog
What else can they (reviewers) compare it to right now? They're obviously not saying it will be better than the R420, because they don't know for sure, just for a comparison of speed increase between generations. They have to compare it to something, or else a lot of people won't know whether the numbers they're giving are fast or not.they should be comparing it to the GeForceFX's, just as B3D did. Why? Because that's how you're going to see how much of an improvement nVidia has made. The FX was a pretty horrible series of chips and everyone was afraid that the GF6 would end up the same fate. The GF6 will not be competing against the Radeon 9xxx so it's misleading to compare the two. We can clearly see that the GF6 is a HUGE improvement over the FX.

Most of the results show it seeming to be more than 50% better when it comes to 4x AAing, 8x (or 16x) AFing, and 1600x1200 res than the 9800 XT (again, just for reference). Unfortunately, as it uses [what is it called, super sampling?], the 8x AAing makes a huge performance hit.well, ever since the NV20 nVidia's GPU's have been capable of 2 multisaples per pipe per clock. However, they are (purposely) hard-limited to one loopback which means the highest MSAA they can use is 2x. In order to get higher sampling levels you must use a combination of MSAA/SSAA and SSAA eats bandwidth which is usually the limiting factor on the 6800. They did employ a rotated grid scheme like 3dfx introduced with the Voodoo5 which looks much better, however they still do not have gamma corected AA like ATi has.

I'm really disappointed with the AF scheme they have employed this time. Their AF used to applly the full level of AF at all angles, however ATi used and angle-dependant approach that used lower number of samnples at angles where it was less needed. It doesn't make a huge IQ difference, but you can see that the ATi method is inferior. nVidia have adopted ATi's method with the GF6.

[Ckerr812]

Originally posted by Novakog


And the million dollar question, how long till the NV40 based Quadro cards?

*rumours I read on the net*

Yes, a new Quadro card will be out in the spring. Based on the NV40Gl. Followed by the nv41, nv45 and nv46. which will all be pci express cards. The card at the top range will be based on the nv45gl chip, and have a transistor count much much higher then the chip made by IBM (sorry Modders).

Whats the hold up?

Apparently Nvidia and Intel are in cahots!, to synchronize the launch with the PCI express professional unveiling.

*rumor mode off*

Personally I am just going to take a wait and see approach until nvidia tells us all what exactly their plans are, waiting a couple months to upgrade could be a wise decision.

[Novakog]

Originally posted by 3Dfx_Sage
they should be comparing it to the GeForceFX's, just as B3D did. Why? Because that's how you're going to see how much of an improvement nVidia has made. The FX was a pretty horrible series of chips and everyone was afraid that the GF6 would end up the same fate. The GF6 will not be competing against the Radeon 9xxx so it's misleading to compare the two. We can clearly see that the GF6 is a HUGE improvement over the FX.


Oh yeah, that's what I meant. Compare it to the GFX and the 9800XT - the current generation. Sorry for the misunderstanding.

[dmonk]

When will these be available to purchase?

[3Dfx_Sage]

Originally posted by Ckerr812
Apparently Nvidia and Intel are in cahots!, to synchronize the launch with the PCI express professional unveiling.
very unlikely. nVidia may be synchronizing their pro cards with PCI-E launch but Intel isn't too fond of nVidia and would not plan something as major as it's PCI-E launch around the schedule of childish little nVidia.

[Ckerr812]

Originally posted by 3Dfx_Sage
very unlikely. nVidia may be synchronizing their pro cards with PCI-E launch but Intel isn't too fond of nVidia and would not plan something as major as it's PCI-E launch around the schedule of childish little nVidia.

Well I think your just plain wrong, probably right though about nvidia planing for the launch.point being they will be around the sametime... :)

http://www.nvidia.com/object/IO_12607.html

http://www.nvidia.com/object/IO_12748.html

I would imagine in a likely hood it could be announced at one of these venues.

[dmeyer]

Originally posted by dmonk
When will these be available to purchase?

If memory serves the Anandtech article quoted ~45 days.

[Novakog]

Originally posted by DeViSoR
sorry

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

DAMN that's sexy. More more!

Makes me think of this: http://www.penny-arcade.com/view.php3?date=1999-07-23&res=l

[Tommy5547]

My "old" 9800 pro is still good enough for me.
Nvidia have a problem with image quality i think.
They have done it better with the 6-series though.

http://www.tomshardware.com/graphic/20040414/index.html

I don't like Nvidia anymore since they tried to cheat in 3dmark and so on...

Don't start the Nvidia vs. ATi war now hehehe

:thumbsup:

[Damm]

it definately looks cool but i would wait till pci-express is out and mebbe even wait a year or so till motherboards with DDR2 support is out... Save my money and buy a whole new comp in other words :)

[loop29]

OMG, awesome job. That video definetely showed some awesome effects. That´s just beautiful, let´s see with what ATI is coming up. Great times for graphics fanatics : )

regards

[CgFX]

Originally posted by Tommy5547
I don't like Nvidia anymore since they tried to cheat in 3dmark and so on...

You understand that ATI admitted to cheating on 3dmark too, right? And the makers of Halo are saying that ATI is still cheating with their game (cutting shaders short below a certain intensity and not rendering the detail textures at all).

They all cheat but nVidia is tighter with the game developers. That is all that matters to me for my home system.

[3Dfx_Sage]

That UE3 demo is most impressive. One thing that will really make games start looking much MUCH better is the "virtual displacement mapping" otherwise known as offset mapping. The first offset mapping demo (that I know of) was done by Humus of beyond3d.com. You can check out that demo here (http://esprit.campus.luth.se/~humus/) .

[imashination]

Originally posted by CgFX
You understand that ATI admitted to cheating on 3dmark too, right? And the makers of Halo are saying that ATI is still cheating with their game (cutting shaders short below a certain intensity and not rendering the detail textures at all).

Yes, but the halo 3D engine is so painfully slow that it needs all the help it can get :-)

Cheating in benchmarks, that should be frowned upon, but cheating in badly programmed games which run slowly.... Im not complianing.

[DaForce]

Originally posted by imashination
Yes, but the halo 3D engine is so painfully slow that it needs all the help it can get :-)

Cheating in benchmarks, that should be frowned upon, but cheating in badly programmed games which run slowly.... Im not complianing.

Amen to that!!

[3Dfx_Sage]

Originally posted by imashination
Yes, but the halo 3D engine is so painfully slow that it needs all the help it can get :-)

Cheating in benchmarks, that should be frowned upon, but cheating in badly programmed games which run slowly.... Im not complianing. interesting that CgFX would complain about ATi not rendering everything in Halo when nVidia is also not rendering everything... not only that, but nVidia's problems are many more than ATi's. But, really, you can't blame either IHV because it's just a very poorly ported game.

Now, making a presentation on how much better PS 3.0 looks and comparing a PS 3.0 screenshot on Farcry with, and it was actually labelled PS 2.0, one with no shaders and all detail turned down to minimum (later they went back and said it was actually PS 1.x but that's still a lie, my Voodoo5 looks better).... well, now, that's some serious BS.

[Ckerr812]

Well for people who follow these sort of things,

The Quadro FX 4000 will be introduced at NAB.

Quote:

Nvidia debuts its latest graphics powerhouse card built around its speediest GPU ever. The Quadro FX 4000 includes throughput speed-ups of more than 5X the performance over previous generation graphics systems, according to the Santa Clara, Calif.-based company. The FX 4000 also offers the industry's “only true 128-bit floating-point graphics pipeline” for millions more colors in a broader dynamic range.

Santa Clara, Calif.; 408-486-2000;
www.nvidia.com

[elvis]

Sounds like good news for the "Maya Hardware Assisted Renderer" users.

[Ckerr812]

Originally posted by elvis
Sounds like good news for the "Maya Hardware Assisted Renderer" users.

From what I have heard...the former smartie pants at Exluna/Entropy have settled with pixar and are now working with them on a new hardware renderer for maya called Gelato.

According to NVIDIA, Gelato will support scanline rendering, raytracing, antialiasing without artifacts, and an interface to Gelato. All using th new line of Quadro cards.

Nab should be pretty informative this year :)

[elvis]

That sounds mighty tasty!

It was inevitable that renderers would one day utilise hardware video GPUs to some extent. Here's hoping the trend continues and we see some real improvement in what can be done with add-on hardware in terms of both quality and speed.

[Aegis Prime]

Man, I'd love to see a 3D package take advantage of the GF 6800's sub-surface scattering :drool:

[stephen2002]

Originally posted by Aegis Prime
Man, I'd love to see a 3D package take advantage of the GF 6800's sub-surface scattering :drool:

It's just a shader, probably PS3.0 In other words it is not something speical built into the hardware, you could create a shader like it yourself in your 3D package (depending on support) for non-real-time renders.

However the quality achived with this card, espeically the hair in the Naul demo, is getting pretty close to what can be done with standard CG rendering techniques. It is quite impressive.

[Vushvush]

Just look at the power of the current GPUs. I mean, transistor wise, haven't they surpassed CPUs?

[3Dfx_Sage]

Originally posted by Vushvush
Just look at the power of the current GPUs. I mean, transistor wise, haven't they surpassed CPUs?
if I recall correctly, the P4 lies somewhere in the 60M range. the NV40 (GF 6800) has 222M and it's rumored that the Quadro version will have even more.

[3Dfx_Sage]

Originally posted by stephen2002
It's just a shader, probably PS3.0 In other words it is not something speical built into the hardware, you could create a shader like it yourself in your 3D package (depending on support) for non-real-time renders.

However the quality achived with this card, espeically the hair in the Naul demo, is getting pretty close to what can be done with standard CG rendering techniques. It is quite impressive.
don't forget that this is still just DX9. Wait until NV50 / R500 and DirectX Next.... will make DX9 cards look like childs toys :drool:

[Novakog]

Originally posted by 3Dfx_Sage
don't forget that this is still just DX9. Wait until NV50 / R500 and DirectX Next.... will make DX9 cards look like childs toys :drool:

Yeah, but... anything 18 months from now will make the current stuff look like child's toys (processors, maybe LCDs, HDDVDs, DDR2 or even DDR3, who knows?), it's kind of a fairly generalized statement isn't it?

Not to say that the GF 6800U doesn't make the 5950U and 9800XT look like child's toys.

[Tommy5547]

Originally posted by CgFX
You understand that ATI admitted to cheating on 3dmark too, right? And the makers of Halo are saying that ATI is still cheating with their game (cutting shaders short below a certain intensity and not rendering the detail textures at all).

They all cheat but nVidia is tighter with the game developers. That is all that matters to me for my home system.

Ouch! Didnt know that.. I better stick with my old 3dfx Voodoo...:)

[3Dfx_Sage]

Originally posted by Tommy5547
Ouch! Didnt know that.. I better stick with my old 3dfx Voodoo...:)
heh actually there's a big difference between what ATi was doing and what nVidia was doing. ATi was simply reordering PS instructions which resulted in a 1.4 percent performance increase. nVidia was (and still is) replacing shaders with their own ones which were not mathematically equivalent. They were also using hand-coded clip planes so that the card was only drawing a small portion of what's supposed to be draws (ie what all other cards were drawing). nVidia never admitted to cheating either and had their lawyers hound 3dmark into retracting their statemens about nVidia's cheats. Also, nVidia has done a wonderful job of alienating developers (little things like lieing to them about future products, current chip specs [they still insist that the NV30/35/36 have 8 pixel pipelines] and performance, not to mention doing app-detection in the drivers to reduce image quality on some games even though the game requests higher) over this past year, a LOT of devs are getting royally pissed at nVidia. Now, if all you buy is the top-level TWIWMTBP games then, yeah, nVidia is the best choice.... however TWIWMTBP-status on most games are merely paid advertising for nVidia.

[Novakog]

I can see how one would be mad at nVidia for that - in fact it somewhat turns me off to them (which I already was from the 9800 being so much faster)...

But I still think - get the most bang for buck (where bang would be speed, stability, support, and features). I honestly don't care which company cheats, as long as I know they are, or their "cheats" are something which will help me (some consider optimizations to be synonomous with cheats).

[rabidtongue]

will the non-ultra Geforce 6800 outperform the FX 5950?

[loop29]

Sure, with 16 pipes it will almost outpace any other card available of earlier generations, even with 12 pipes it should be fast enough for high resolution and FSAA and AF. But this is true for the X800XT too besides it only supports fp24 precision.

[3Dfx_Sage]

Originally posted by loop29
Sure, with 16 pipes it will almost outpace any other card available of earlier generations, even with 12 pipes it should be fast enough for high resolution and FSAA and AF. But this is true for the X800XT too besides it only supports fp24 precision.

yes, the 6800 NU (which only has 12 pipelines) and the X800 Pro should both easily out-perform the 5950.

It is kind of funny that you mention the X800XT "only" supports fp24 precision because it's a step up from the 5950. Sure, theoretically the 5950 can do fp32 but in reality it's simply too slow to use (if you use more than 2 registers in the PS your throughputs gets cut in half, vs FP 16 where you can use 4... which is still a ludicrous limit that coders have to do fancy little code dances to get around) and in some driver versions fp32 is disabled alltogether. Of course, let's not forget how hard nVidia was pushing the use of not fp16 or fp32, but int12 with the GeForceFX 5800 (pushing the use of PS 1.4 instead of PS 2.0 because PS 2.0 was waaay too slow). Now, that doesnt mean that int12 or even fp16 are really all you need but lets get some perspective here and realize that nVidia's previous generation did have lower precision. I'm a bit miffed that ATi didnt come out with a part capable of fp32 and vs3.0, but it is a step above the 5950 for sure.

[rabidtongue]

yes, the 6800 NU (which only has 12 pipelines) and the X800 Pro should both easily out-perform the 5950.

so is it safe to say that having more pipelines plays a bigger part compared to the graphics chip and memory? because from what i read, the 6800 non-ultra will be clocked at 375MHz and 700Mhz for the memory...

the 6800 non-ultra will only be using GDDR memory is that correct? is there a big difference between GDDR and GDDR3?

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.