NVIDIA GeForce 4XX Series Discussion

[Benetanegia]

Hmm ok 4x AF and no AA. That's it. That's the difference.

So yeah, I admit it, I was wrong on that detail. The HD5770 CAN do more than 30 fps on heaven. And I'm soooo wrong on the point I was making because Heaven at 1024x768 no AA 4x AF does more tan 30 fps ona heavily OCed HD5770 and everyody knows that every new DX11 game will be expected to be run at 1024x768 no AA 4xAF and 30 fps...

Why not be less short sighted and see what my point was instead of trying to discredit my legit claim with poitless "facts".

 

[Binge]

Hmm ok 4x AF and no AA. That's it. That's the difference.

So yeah, I admit it, I was wrong on that detail. The HD5770 CAN do more than 30 fps on heaven. And I'm soooo wrong on the point I was making because Heaven at 1024x768 no AA 4x AF does more tan 30 fps ona heavily OCed HD5770 and everyody knows that every new DX11 game will be expected to be run at 1024x768 no AA 4xAF and 30 fps...

Why not be less short sighted and see what my point was instead of trying to discredit my legit claim with poitless "facts".

Because these facts have merit on TPU. I consider the average consumer who knows jack-sh!t about computers to be short sighted. Maybe that's short sighted. I'm no expert on sight.

 

[Benetanegia]

Because these facts have merit on TPU. I consider the average consumer who knows jack-sh!t about computers to be short sighted. Maybe that's short sighted. I'm no expert on sight.

You fail to impress me with that.

What has that to do with the fact that the HD5770 can't run tesselation properly? I've been here in TPU for long and since when 1024x768 is enough? Since when is 0xAA enough? Sinc when is 4x AF enough? Since when is 30 fps enough???? All I see is two people arguing with my statement, just for the sake of doing it. If anyone had said 1024x768 0XAA 4xAF 30 fps in any other context, EVERYONE would have said, UNPLAYABLE, UPGRADE. But yeah shhot me down as much as you want guys. Come on.

EDIT: OMG. I'm going to backpedal that statement, because I've realised something that's going to make me happy after so many time of little frustration. The thing is that my 8800 GT can run Crysis just fine!!! And by comparison my brothers 9800 GTX+ runs it wonderfully!!!

Yes. I'm being sarcastic.

 

[Cool Vibrations]

Triangle handled differently? no WAI! Almost like tesselation but not quite like the tesselation that ATI has had on the previous three generations of cards? You know, the ones that Nvidia took a stand against and told us all that it was worthless. Like that? But better since it is green vapor?

Spot on.

This is the most drawn out paper release ever. I just want my fricken benchmarks so I can choose my next videocard, a 5970 or a GF100. If this ends up like the 2900XT, I'm going to be so sad. Hype is so retarded sometimes.

 

[Binge]

You fail to impress me with that.

What has that to do with the fact that the HD5770 can't run tesselation properly? I've been here in TPU for long and since when 1024x768 is enough? Since when is 0xAA enough? Sinc when is 4x AF enough? Since when is 30 fps enough???? All I see is two people arguing with my statement, just for the sake of doing it. If anyone had said 1024x768 0XAA 4xAF 30 fps in any other context, EVERYONE would have said, UNPLAYABLE, UPGRADE. But yeah shhot me down as much as you want guys. Come on.

Changing the subject more often so we can have a more poignant discussion. You said at first that the 5870/5970 are the only card that can run DX11+moderate/high tessellation. You were met with a counter argument. Then the subject changed to 95% of end users do not overclock to get results. This is different than your original statement, and you are met with an argument that has the same focus. There are lower end cards that can perform at the same levels as high end models with overclocking/crossfire, and there's even a screenshot that shows that 1680x1050 has a very slight performance hit from the 1028x764 benchmark run. Then you called a number of people liars. You were met with an argument that showed an even more successful run with a 5770, and the guy even used a decent OC. DAMNED that guy to hell. His benchmark was whack, right? According to you this is short sighted for us to think differently and find legitimate information to back up our initial claim. The 5870/5970s of the world aren't the only cards that can run tessellation at playable frames. It takes a smidge of OCing, and that's what we do on TPU. You're not impressed? Apparently not impressed that we don't tell every Jon Doe to sodd off and buy new components.

Hey Bene, 5850s have OCed higher than most 5870s ever will and at those OCs beat them in any/all game or benchmark to date. Tessellation in games/apps is entirely possible.

 

[qubit]

At last:

As for the game benchmark, the card was running a Far Cry 2 at 1920x1200 resolution with 4xAA and Ultra High settings and compare it to some sort of reference card. The GF100 was capable of getting up to 84 frames per second. The reference card, rumoured to be a HD 5870, hoover around 50 frames per second at the same settings.

Looks like it'll spank a 5870, but not a 5890.

Fudzilla

EDIT: I meant 5970, not 5890.

 

[Binge]

At last:



Looks like it'll spank a 5870, but not a 5890.

Fudzilla

What 5890?

 

[qubit]

What 5890?

Sorry, 5970 dual chip card.

 

[Binge]

Sorry, 5970 dual chip card.

Ahhh, well that's always a toss-up. A lot of people prefer the single gpu to dual gpu solutions. I know I'm going to be hard pressed come release time because I REALLY like the 5850, but I can't help but love NV's drivers. With those drivers I might even miss the energy I'm saving.

 

[Benetanegia]

Changing the subject more often so we can have a more poignant discussion. You said at first that the 5870/5970 are the only card that can run DX11+moderate/high tessellation. You were met with a counter argument. Then the subject changed to 95% of end users do not overclock to get results. This is different than your original statement, and you are met with an argument that has the same focus. There are lower end cards that can perform at the same levels as high end models with overclocking/crossfire, and there's even a screenshot that shows that 1680x1050 has a very slight performance hit from the 1028x764 benchmark run. Then you called a number of people liars. You were met with an argument that showed an even more successful run with a 5770, and the guy even used a decent OC. DAMNED that guy to hell. His benchmark was whack, right? According to you this is short sighted for us to think differently and find legitimate information to back up our initial claim. The 5870/5970s of the world aren't the only cards that can run tessellation at playable frames. It takes a smidge of OCing, and that's what we do on TPU. You're not impressed? Apparently not impressed that we don't tell every Jon Doe to sodd off and buy new components.

Hey Bene, 5850s have OCed higher than most 5870s ever will and at those OCs beat them in any/all game or benchmark to date. Tessellation in games/apps is entirely possible.

First, when I'm talking about playing something, I'm definately not talking about playing it at those settings, and if you try to make me believe that you game at anything close to that, yes, I'm going to call you a liar. I have not played at 1024x768 since 1999 I think, 11 years and again I'm going to have a hard time believing you have do it either.

Second, I'm not calling the guys liars, but the runs could have been made without tesselation, MANY people have posted results without tesselation. So in effect they would be false, as they failed to prove the point he was making. But as I explain above, I was not taking into account they had no AA and were 4x AF. I was comparing with 4xAA 16xAF runs that I had seen, because that's the normal settings that cards are benched. Once again shoot me for adopting the standards

Third, I have never changed the subject. I was talking about game developers and them making games implementing tesselation. Those will NOT make games thinking on enthusiasts, because enthusiasts represent a very little audience believe it or not . When I make a point about a card being able to run something I talk about stock, because that's what most people will be using and the game HAS to be made with that in mind. I could care less about what an OCed card can do in that regards. One tweaked Ford Fiesta (search it) can do 300 km/h but I would never say the Fiesta can do 300km/h just because I was presented with one that can. Period.

Finally, it's short sighted to concentrate in ONE statement I did and not trying to understand the overall meaning of what I was trying to say. You can bend it as much as you can, wee can discuss if 30 fps is playable, we can discuss if playing with no AA in 2010 is desirable, we can discuss a lot of things, but at the end of the day, if you were honest you would have to agree that the HD5770 is not really running the Heaven benchmark, it's not really capable of doing tesselation to that level. And definately the HD2xxx/3xxx series couldn't do tesselation, which was my original claim regarding tesselation. Furthermore Heaven is a benchmark, an interactive game will always be slower if it has the same level of detail.

Please don't continue acting as if I had said something over the top, because if 30fps at those settings is playable, I'd like you to come with me and revisit all those "Crysis sucks, is an unoptimiced POS" because Crysis does run perfectly even on a 8600 GT or X1900, even so much better than the HD5770 runs Heaven. So please be coherent and accept the posibility that my claim was correct in the context I was speaking and that Iwas not trying to make any harm. If you don't want to concede anything, I'm done with this.

 

[Binge]

No harm, no foul. I never in any of my days said that about Crysis. People who whined about the game are high or something.

 

[Benetanegia]

No harm, no foul. I never in any of my days said that about Crysis. People who whined about the game are high or something.

Well most people want 30 fps minimum. I'm ok with 25 fps minimum, but the HD5770 does 30 fps average, wich would mean extremely unacceptable lows if it was a game. Similarly the HD5850 would suffer of the same, 50 average means 25 lows probably, so I said "barely playable". That has always been my point, if it can run the damned benchmark at some settings is irrelevant. It can't do it properly at 1680x1200 4xAA and beyond and that's enough for me call it unplayable. Games, DX11 games are going to be played at those higher settings, I don't see how someone would enable tesselation, if it had to play at 1024 in order to do so.

 

[KainXS]

The performance is pretty much what I expected solidly 60% faster than the HD5870 and I say solidly because in some situations it might dip far under or higher than that, with an estimated price tag(which I think is almost correct) or $499 to $549 and two SKU's at launch the GTX360 and GTX380.

and its a little bit smaller than the 8800GTX die

http://www.brightsideofnews.com/new...100-architecture-alea-iacta-est.aspx?pageid=2

a really cool feature seems to be that you can remove the top shrout and main heatsink from the heatsink assembly itself and be left with only the part cooling the memory and vrm's allowing you to use the bottom base for watercooling without risking not cooling the vrms and memory, it also has the same mounting holes as the GTX2XX cards so current aftermarket heatsinks will fit them.

 

[crazyeyesreaper]

well for the price a single GTX380 if thats the one featured above costs more then my 2 5850s and i get better performance.. im okay with that to those that get gt300 cards also

 

[bobzilla2009]

The performance is pretty much what I expected solidly 60% faster than the HD5870 and I say solidly because in some situations it might dip far under or higher than that, with an estimated price tag(which I think is almost correct) or $499 to $549 and two SKU's at launch the GTX360 and GTX380.

and its a little bit smaller than the 8800GTX die

http://www.brightsideofnews.com/new...100-architecture-alea-iacta-est.aspx?pageid=2

Maybe i'm just being stupid, but those results don't match up to what i've been getting with my stock hd5870, 4xAA 1920x1200 results in about 30fps average, 8xAA being less than stated. I'm happy to assume the gtx380 stats are correct, but until i see what resolution they're running those tests at...

regardless, with OC i tend to reach 36fps @ 1920x1200 4xAA, so a boost up to 43 is a pretty nice 20% over my OC'd card if that is the same settings. Now to see if the price is any good! (doubtful ) but if the $499 price tag is correct, nvidia can bugger off! The gtx360 selling at (around) £400 and the gtx380 at £450-500 (Yes, i know that's not the direct conversion from $'s but no-one in the Uk will deny a £400 price tag will happen) when the hd5870 will probably drop to £200 is not going to go down well.

Still, that's still some pretty nice results, but i want a real review out first.

EDIT: just noticed that it's just the dragon part covered in the test now to try and match up the perf to my hd5870 ^^

Well most people want 30 fps minimum. I'm ok with 25 fps minimum, but the HD5770 does 30 fps average, wich would mean extremely unacceptable lows if it was a game. Similarly the HD5850 would suffer of the same, 50 average means 25 lows probably, so I said "barely playable". That has always been my point, if it can run the damned benchmark at some settings is irrelevant. It can't do it properly at 1680x1200 4xAA and beyond and that's enough for me call it unplayable. Games, DX11 games are going to be played at those higher settings, I don't see how someone would enable tesselation, if it had to play at 1024 in order to do so.

I agree, i'm just as picky, and minimum fps are generally far more important for me overall.

All in all, decent performance as far as i can see, but the price is too high. Wait for hd6 series before buying methinks ^^ Or see if any games actually need any more than a hd5870 for a while

NIGGLING FEELING ALERT! Why are nvidia JUST showing this sequence? why not the whole thing? is it because the nvidia cards are better at this section of the demo? and that the ATi cards actually pull back some frames at different points? Hmmmm, i don't like it, we need a full heaven benchmark run. Especially since the ATi cards are FAR closer at the beginning (20% to gf100) and end (~33% to gf100) than the 1.6x better would make you believe. Overall i think it may be more like a 25-40% boost over a hd5870 than 60%.

In fact, most of that '1.6' was gathered in the section of 20-30s, where the frames are nearly double. The rest is more like 1.3-4. Also, the minimum is sensationalist statistics on nvidias part. The framerate delta between max and lowest is actually HIGHER for nvidia. Also, any idiot knows that a 15fps boost when the lower number is lower anyway is always bigger. This thing reeks of hype the more i look at it

EDIT: ahhh, all is clear, nvidia is far better at tessellation, hence why they are using a tessellation benchmarks to show off the gf100! Which is great, but i want to know how it stands up where tessellation isn't the limiting factor in frames But hey, at least the dirt 2 main screen will run faster on the gtx380 for sure ^^

 

[phobias23]

Ive just read Fermi' s news,and its very impressive all what they are saying,but THEY are the ones who are posting that info not a single person from this forum or other site.Im not Ati or Nvidia fanboy,i read a lot your articles,cuz and a pretty noob in this matter,but i not such an idiot,i compare everyones opinions and give a lot of credibility to the ones who really test the cards.So its Nvidia really getting this benchmark over Ati,this is my opinion.No ones has the HOT$ card yet to test it.so im gonna wait and see what really is Nvidia card at the end.

 

[bobzilla2009]

indeed, although i DO expect a 20-30% increase over the hd5870. This benchmark has in fact made that seem more the case. Heaven is a huge tessellation benchmark. Nvidia have amazing tessellation horsepower. But when there isn't a lot of it going on (which is the point just at the beginning of the graph), the difference is around 10-20%.

Overall, the card is looking promising, but this graph is not representative of real world gameplay. Unless all games start using tessellation as extensively as the heaven demo (which i would LOVE to see). But even then the hd5870 ain't too bad, i feel that the card is hard done by in this 'benchmark' (EDIT: Well, OF COURSE it is ^^, this is nvidia propaganda atm), because i've never had frames that low. the dragon part (the lowest point) sits at 23fps for me @ stock, not 20fps.

Still, I wish it would define which card the gf100 is. I would assume the gtx380. Which would mean that this generation is almost a copy paste of the last in terms of what lines up against what.

 

[kid41212003]

Laugh my fucking ass off, laugh my fucking brain out.

HD5850 is slower than HD5870 about ~11% @1280x1024 according to this chart
http://www.techpowerup.com/reviews/ATI/Radeon_HD_5870/30.html NONE DX11 test.

When you look at this DX11 bench chart:
http://forums.techpowerup.com/showthread.php?t=112995

What do you see?

wolf-----Sapphire 5870 950/1250 1018 i7 920@4.2ghz
jjFarking------XFX 5850 850/1200 910 X4 965@3932

Oh yes! The HD5870 is faster than HD5850 (obviously) in DX11 bench with tessellation ON! And it's ~10% faster!

What does this mean? It's mean with or without tessellation the performance between the cards are the same.

"This graph doesn't represent gameplay... ect.. crab", oh yeah sure, then we don't really have to buy DX11 cards, 'cause all the current games doesn't use enough tessellation to make it actually look noticeable different.

Furme is suck anyway, but fuck it, im fucking going to buy it anyway, probably will put it on my desk for masturbation purpose.

 

[Binge]

Match the clock speeds and the results would be the same Kid.

2 Boneface sapphire 5870 1000/1300 1044 i5 750 @ 3.6GHz
3 BiNGE Sapphire 5850 1010/1200 1034 i7 920 @ 4.2GHz

 

[TheMailMan78]

Hmm ok 4x AF and no AA. That's it. That's the difference.

So yeah, I admit it, I was wrong on that detail. The HD5770 CAN do more than 30 fps on heaven. And I'm soooo wrong on the point I was making because Heaven at 1024x768 no AA 4x AF does more tan 30 fps ona heavily OCed HD5770 and everyody knows that every new DX11 game will be expected to be run at 1024x768 no AA 4xAF and 30 fps...

Why not be less short sighted and see what my point was instead of trying to discredit my legit claim with poitless "facts".

I know I know. All those pointless facts get in the way of meaningful guessing.

 

[subhendu]

guys any idea about gf 100 ----> models and prices ?

 

[Goodman]

http://store.steampowered.com/hwsurvey/

45% of steam users are now on DX10

considering ~10% are SM2.0 or lower, its pretty equal between DX9 and DX10 systems these days.

45%? hmmm! with only 39 DX10 games out there... & out of those 39 games 5 of them needed a patch to use Dx10 ...
http://en.wikipedia.org/wiki/List_of_games_with_DirectX_10_support

I'll say most people game with DX9 since 3/4 of the games out there are DX9 ..

 

[Benetanegia]

I know I know. All those pointless facts get in the way of meaningful guessing.

Man just leave it. 25 fps average is nowhere near playable and that's at 1680x1050 without antialiasing. 34 fps is still borderline playable and that's at 1024x768. That's simply not playable, no one is going to play at 1024 on a new game on a new API on a new card. When they have bought a new card they will have surely bought a new monitor in at least, I don't know 10 years maybe??

And just to justify why I don't believe everyresult that is posted in a forum, by someone that I don't know, just because it has been posted... there you have, my Heaven DX11 result:

It's 100% legit. I have opened up Heaven, selected DX11 and run the benchmark. But obviously it's not doing tesselation nor any other DX11 feature since it's a 8800 GT, but how can you differentiate this result from the ones you posted? You simply can't.

 

[TheMailMan78]

Man just leave it. 25 fps average is nowhere near playable and that's at 1680x1050 without antialiasing. 34 fps is still borderline playable and that's at 1024x768. That's simply not playable, no one is going to play at 1024 on a new game on a new API on a new card. When they have bought a new card they will have surely bought a new monitor in at least, I don't know 10 years maybe??

And just to justify why I don't believe everyresult that is posted in a forum, by someone that I don't know, just because it has been posted... there you have, my Heaven DX11 result:

http://img.techpowerup.org/100118/heavendx11.jpg

It's 100% legit. I have opened up Heaven, selected DX11 and run the benchmark. But obviously it's not doing tesselation nor any other DX11 feature since it's a 8800 GT, but how can you differentiate this result from the ones you posted? You simply can't.

The fact they are 5700 series cards and if they didn't use translation the FPS would be much higher? Also you go on and on about "Nobody games at 1024" when in fact the majority of the public does. Just look at the Steam survey.

Another thing you need to remember is the 57xx and 58xx drivers are still very immature and will get better over time. We saw some HUGE jumps during the 48xx era just with driver updates. The FACT of the matter is all we have are facts when it comes to the 5xxx series and guesses when it comes to Fermi. When you mix the two you get misinformation and that is not what TPU is about.

 

[Benetanegia]

The fact they are 5700 series cards and if they didn't use translation the FPS would be much higher? Also you go on and on about "Nobody games at 1024" when in fact the majority of the public does. Just look at the Steam survey.

Another thing you need to remember is the 57xx and 58xx drivers are still very immature and will get better over time. We saw some HUGE jumps during the 48xx era just with driver updates. The FACT of the matter is all we have are facts when it comes to the 5xxx series and guesses when it comes to Fermi. When you mix the two you get misinformation and that is not what TPU is about.

Erm,

1. As I stated, I wouldnt say my 8800GT is running this benchmark at all. It's doing 40 fps on a low resolution and no AA and 4xAF. I would NEVER, I repeat, NEVER play at those settings if I can avoid it. Even if I had to run at those settings at that performance (i.e Crysis high/very high XP tweak), I would never be satisfied and I would definately never say I'm playing it just fine. That is just NOT playing it fine. Period.

2. My point was that the HD5770 can't do the tesselation present on Heaven on a game at the settings most people play. For the record only 13% of people play at 1024 and there's much more people using DX9, meaning that most DX10/11 people HAVE to be playing on
higher settings. Those 2 stats just have to overlap. there's simply no way people have DX11 hardware on a 1024x768 screen or playing a 6600GT on a 1920x1200 screen.

3. Show me a HD5770 doing 30+ fps at 1680x1050 4xAA and 16xAF and you have a point. Until you do that, just shut up, because my point will still hold true.