NVIDIA GeForce 4XX Series Discussion

[Lionheart]

20mm... why dont you just use the quote button, so we know who you're quoting?

You talkin to me?

 

[Benetanegia]

Its obvious that the GTX480 will beat the HD5870 in performance but not as much as everyone thought it would be, but I think the GTX480 and GTX470 will perform well in tessellation, I think thats why the design is taking so long, and the power and heat issues and that fabrication crap of course as well, but I think that these cards were mainly designed to perform well in DX11, I guess we will find out soon!

Well they were definately designed with that in mind. 4 triangle setup engines and 16 tesselators testify in favor. And that's probably the problem with non-tesselated, if there really is a problem there. I don't know too much about the issue, but I doubt that current games can operate on more than one triangle per clock, I highly doubt they were designed for that at all. IMO The only time those extra setup engines and rasterizers are used is when tesselation is being done.

I don't know what's wrong with people anyway. If it happens that Fermi is only faster than Evergreen when tesselation is ON, what's wrong with that? I mean, how many times I have seen people recomending a HD5770 over more capable DX10/DX10.1 graphics cards based on the fact that it is DX11? DX11 is mainly and mostly tesselation and HD57xx cards can't render a fully tesselated escene, even the HD5870 falls to 20 fps. Fermi stays on more than compfortable 40 fps, wasn't that the purpose of DX11? I'm confused.

 

[Lionheart]

Well they were definately designed with that in mind. 4 triangle setup engines and 16 tesselators testify in favor. And that's probably the problem with non-tesselated, if there really is a problem there. I don't know too much about the issue, but I doubt that current games can operate on more than one triangle per clock, I highly doubt they were designed for that at all. IMO The only time those extra setup engines and rasterizers are used is when tesselation is being done.

I don't know what's wrong with people anyway. If it happens that Fermi is only faster than Evergreen when tesselation is ON, what's wrong with that? I mean, how many times I have seen people recomending a HD5770 over more capable DX10/DX10.1 graphics cards based on the fact that it is DX11? DX11 is mainly and mostly tesselation and HD57xx cards can't render a fully tesselated escene, even the HD5870 falls to 20 fps. Fermi stays on more than compfortable 40 fps, wasn't that the purpose of DX11? I'm confused.

You couldnt have said it any better

 

[phanbuey]

Well they were definately designed with that in mind. 4 triangle setup engines and 16 tesselators testify in favor. And that's probably the problem with non-tesselated, if there really is a problem there. I don't know too much about the issue, but I doubt that current games can operate on more than one triangle per clock, I highly doubt they were designed for that at all. IMO The only time those extra setup engines and rasterizers are used is when tesselation is being done.

I don't know what's wrong with people anyway. If it happens that Fermi is only faster than Evergreen when tesselation is ON, what's wrong with that? I mean, how many times I have seen people recomending a HD5770 over more capable DX10/DX10.1 graphics cards based on the fact that it is DX11? DX11 is mainly and mostly tesselation and HD57xx cards can't render a fully tesselated escene, even the HD5870 falls to 20 fps. Fermi stays on more than compfortable 40 fps, wasn't that the purpose of DX11? I'm confused.

I see what you did there

 

[Mussels]

You talkin to me?

uhh... no, to 20mmrain.

 

[Lionheart]

uhh... no, to 20mmrain.

My bad!

 

[Benetanegia]

uhh... no, to 20mmrain.

My bad!

Yeah, sure.

Chaos Killa: You talking to me?

Mussels: uhh... (gulp) no, to 20mmrain.

Chaos Killa: Mybad. (I'll let you go... for now)

 

[Lionheart]

Yeah, sure.

Chaos Killa: You talking to me?

Mussels: uhh... (gulp) no, to 20mmrain.

Chaos Killa: Mybad. (I'll let you go... for now)

Hahahaha thats the funniest shit i've seen 2day!

Nice one!

 

[Benetanegia]

Thanks, there's yet more Taxi Driver coming from me.

I think that after all the bad press, expectations and broken expectations, Fermi will actually be succesfull, if not in it's current form, in the refresh that is supposed to be released before the end of this year. I can see Fermi saying at launch:

Fermi Driver: http://www.youtube.com/watch?v=QBvg3PkI-PU

Subtitles: Listen, you fuckers, you screwheads. Here is a chip that would not take it anymore. A chip that stood up against the vias, the yields, the delays, the rumors, the bullshit. Here is a chip that stood up.

 

[Steevo]

No, there is a set of engineers that are going to Vegas for a week when it finally gets released.


I really do want it to perform very well, I'm just tired of all the spin they are putting on this chip. They highlight a area of one benchmark that has been reoptomized for their card, and point out a gain that is marginal when you actually compare the hard numbers. A 30% increase and they are performing 50% with that, so really a 20% increase, but only in a overblown example, much like people cried when Vantage, 3D06, came out.

 

[Benetanegia]

I still believe it could all be a smokescreen to prevent Ati from doing an accurate move with their next release. (and I mean everything, moles can be very helpful when you know you have some under the roof)(everything performance related mind you)

I'm not saying that's my main way of thinking about the situation, but it's still there, occupying like a 10% of my speculation. I can't yet believe that an extra of 272 SPs (that's +113%) and underlaying architecture only yields a 25% increase in performance. Have to see it on release reviews in order to believe that.

 

[EastCoasthandle]

I think the point of all this is very simple:
We know nvidia had heaven 1.1 and more then likely used it to make the graph. It's clearly shown in the video. However, it's not clear how they derived to the 5870 results, which driver they used, which version of heaven they used, etc. So it's clear that one cannot just look at the graph, see the gap and declare that tessellation at certain portion of the graph is higher on one card then another. That's what I call an illusion. We need to clearly see results from both cards using the latest drivers (because per my chart posted earlier in this thread the results are better in some portions of the bm then in the video) in order to see the full picture. This should be a method used not only with heaven bm 1.1 but with any game. Be it they have a performance improving update or otherwise.

 

[kid41212003]

Let's just ignore every benchmarks out there or games that labeled or related to NVIDIA, it obviously will not be fair for AMD 'cause it was optimized for NVIDIA cards.

 

[the54thvoid]

Zealots

I cannot believe that so many people when shown Nvida's own benchmark results are shouting "GTX 480 is God". Have any one of these people any understanding of what marketing actually is?
It is in the inherent financial interest of ANY company to release figures that show their product in a positive light. We all KNOW the 480 will beat the 5870, can we not all agree on that? But the margin is absolutely unknown. Nvidia's cases for 'x'% improvements are absolutely, positively optimised for their product and more than likely when compared with ATI, they are hamstringing the 5870 with less than optimised settings.

If anyone buys that damn card without reading a plethora of reviews they require a brain reboot. Anybody that rushes out and buys a new tech item without researching it first deserves all the grief they get.

Can you imagine the current WHQL driver issue had come out at the end of the month and your already hot Fermi card's fans didnt work? Think about how close NV just came to meltdown. This is one company that courts both greatness and a heavy dose of risk taking.

If W1zzard doesn't get a card for review, somebody ought to kill for one. We deserve that much.

 

[yogurt_21]

Let's just ignore every benchmarks out there or games that labeled or related to NVIDIA, it obviously will not be fair for AMD 'cause it was optimized for NVIDIA cards.

yes lets, and while were at it we'll ignore all other games that ati performs badly in.

lol

at this point I'm glad for any real news on fermi. It seems like an eternity since i last bought a gpu and right now I've very little cash to spend. If fermi isn't competitive I havent' a prayer in the world at getting something actually capable of playing dx11 games at decent details.

and being how long it took me to get this tiny stash of cash together. I'd imgaine I'll be on my next card for a similar amount of time.

here's hoping that the gtx480 beats the 5870 by a decent enough margin so that when both sides launch revision cards, prices will be extremely competitive.

 

[ToTTenTranz]

Let's just ignore every benchmarks out there or games that labeled or related to NVIDIA, it obviously will not be fair for AMD 'cause it was optimized for NVIDIA cards.

That's kinda hard since almost every AAA game out there is TWIMTBP-infected.


nVidia will do what they've been doing for years. They design their GPUs with one architectural advantage over competing ATI cards and then they elevate that one advantage to every AAA game through the TWIMTBP-infection.

So here is the first DX11 TWIMTBP-infected benchmark: Unigine Heaven. "Tesselation is faster on nVidia hardware, so let's push tesselation to unreal levels in a few scenes."


The end result is a bunch of games out there that don't take advantage of most features in most GPUs but just run better with nVidia.
Sad, isn't it?



Proof? Compare the number of games with PhysX to the number of games that support HD2000-HD4000 tesselation.

 

[Melcar]

I don't think you can label Heaven as TWIMTBP. It's merely a tech. demo that emphasizes Tessellation techniques, and just so happens that The Fermi is better in that particular aspect as the competition. AMD was first in line with such tech., so it makes more sense to claim that Heaven is (was) there to showcase ATI hardware.

 

[KainXS]

until we can get our hands on 1.1 or 2.0 of the heaven benchmark thats not true, everything is nothing but opinions until we can get solid info from a reliable source because the main two sources that will lie is AMD and Nvidia, and right now Nvidia is the only one with the info and if you believe what nvidia is saying about something they have to make money off of then . . . . . wow.

 

[ToTTenTranz]

until we can get our hands on 1.1 or 2.0 of the heaven benchmark thats not true, everything is nothing but opinions until we can get solid info from a reliable source because the main two sources that will lie is AMD and Nvidia, and right now Nvidia is the only one with the info and if you believe what nvidia is saying about something they have to make money off of then . . . . . wow.

That's one very big sentence, dude.. I really don't understand what you meant.

 

[KainXS]

basically I'm saying theres no reliable sources right now and since nvidia is the only one with heaven 1.1, those heaven benchmarks aren't reliable.

 

[ToTTenTranz]

basically I'm saying theres no reliable sources right now and since nvidia is the only one with heaven 1.1, those heaven benchmarks aren't reliable.

Yes.. and furthermore, AMD is probably in condition to launch higher-clocked Cypress cards now, with +1GHz core clock. And AMD's tesselator should perform linearly faster with the clock increase.

 

[Melcar]

Well, we do know that "someone" did delay the release of the next Heaven build. The report claimed a "hardware vendor" and no name was given, so it could have been anyone. Maybe it was AMD that asked for the delay because their OpenGL tessellation extension wasn't quite there yet (the next Heaven release is bringing OpenGL based tessellation and Linux compatibility), or nvidia because they wanted extra time until their hardware hits the streets and supporting software is available. In either case, it is safe to assume that both teams have the latest Heaven spins at their disposal, since the Unigine team never did hide the fact that both vendors are involved in the dev. process of the tech. demo.

 

[20mmrain]

20mm... why dont you just use the quote button, so we know who you're quoting?

Don't know..... guess I always did it the other way.... you know how that goes..... you get to be a creature of habit after awhile.

 

[kid41212003]

Proof? Compare the number of games with PhysX to the number of games that support HD2000-HD4000 tesselation.

List the games that support tess?

 

[Mussels]

List the games that support tess?

i think far cry did, the original game (in a patch). Feck all beyond that... because nvidia pressured companies not to support it, since their cards couldnt run it.