NVIDIA GeForce 4XX Series Discussion

[Lionheart]

Its funny how some things can turn out. Couple yrs back ATI were struggling against nvidia, then ATI surprised them with the HD4000 series at a great price performance ratio and now ATI are leading with their HD5000 series.

Gotta luv it

[EastCoasthandle]

Wow, that's pretty sneaky, they either:
A. Used 1.0 for the 5870
B. 1.1 offers no improvements for the 5870.







As you can see, the benchmark I just did vs the one in the video are nearly identical. Although it looks like Cat 10.3 is offering a tad bit better performance in some spots (which I assume they didn't use as this is rather old). I'm accumulated enough evidence to officially call it fud now.

[Aleksander]

Your bench starts from 70 and the other one starts from 64-65, i think it is similar but they have lowered it down a bit...

[EastCoasthandle]

Yeah but the valleys are nearly identical though. That is the part were, IMO, aggressive culling would improve performance with 1.1. That's why, in part, I believe either A/B mentioned in my previous post. I do believe that Cat 10.3 did bring some performance improvements though as this benchmark was shown in a slide a while back during one of their events.

You do remember this from last year right?


Well it's a similar (if not the same) graph except that a portion of it was "highlighted".

[20mmrain]

Yeah but the valleys are nearly identical though.

I know my Graph isn't very scientific.... but yes I believe these graphs are somewhat real. I ran my test at 1920x1080 at 850 core speed and 1200 Mem speed. I also used settings of

AF: x1
AA: none
Tessellation: Enabled
Shaders: High
Res: 1920x1080
Full Screen: Yes
My i5 750 is @ 3.8Ghz which is about the same performance as the i7 used in the Nvidia bench.

I took every section and marked the highest FPS in each section of the Benchamrk. Then I took those 25 points and laid it up against Nvidia's Graph here. I took theirs and broke it down into those same 25 points.

So while it is drawn in a art program and it is not scientific at all...... I believe that these are right.

But that is not to say that Nvidia isn't running the 1.1 version with theres. And even still it isn't that impressive.

So here is what I found....



But that picture that EastcostHandle posted..... LOL that is pretty slick there Nvidia..... Use a small section of the benchmark and say it's two times faster.... I would say maybe 15% at tops. I just don't see it happening! In some area's of the benchmark I could see it probable that the 5870 is faster. I would say these cards are closer together..... (so far) Than we think.



Real uncensored Pic Nvidia Doesn't want you to see... I swear..... this isn't doctored see the 5870 is at least 10% better LMFAO

[kid41212003]

Yeah, we will see how it will turn out in 20 days or more.

While your points are quite valid. I will take a look at this graph before saying it through.

http://www.techpowerup.com/reviews/P...S_Plus/29.html

HD5870 PCS+ version is actually only faster than GTX285 22% (1900x1200 mode)

But is it really show the real different between HD5870 and GTX285? No, I would say imo. Is HD5870 looks real impresses in these graph? No, obviously.

Minimum frames are a totally different things though, I believe one of you did mention about minimum frames are more importance than max frames (I didn't say this).

Looking at the base FPS of HD5870 on that graph vs GTX480 base FPS, it really did impressed me.

[mastrdrver]

Who cares how well it does in Heaven? That's like comparing Vanatge P scores. Its not a game and doesn't represent anything out on the market or in the near future.

The Cyrsis bench is interesting if accurate. If the GTX 470 can come that close in most games I think it will hit shelves at $400. With the GTX 480 being supposedly 5870 < 480< 5970, I don't doubt that retail will be $600 minimum. With 5970 going to the sky in prices (have you seen some of the ebay prices that actually sold?!?!) nVidia is going to be pretty free to price it where they want to.

All this means is that I highly doubt 5 series prices are going to move at all.

[20mmrain]

I believe one of you did mention about minimum frames are more importance than max frames (I didn't say this).

Looking at the base FPS of HD5870 on that graph vs GTX480 base FPS, it really did impressed me.

I didn't say it but I agree with you.... Min FPS does matter. I will admit that I think the 5870 and the 5970 even have problems with this...... But I also think it has a lot to do with the drivers for these.
Just from being an owner of the 5870 is why I say that.

I do agree the GTX 480 will be faster!!! But not by 40% to 50% like some people thought originally.
I would guess 15% tops. But that's just my guess and I don't know jack chit!

[Aceman.au]

Until the cards are released don't celebrate the fact that Nvidia beats ATI... ATI did release first and Nvidia had to delay to make their cards better than ATI

[Mussels]

Originally Posted by l33tGaMeR

Until the cards are released don't celebrate the fact that Nvidia beats ATI... ATI did release first and Nvidia had to delay to make their cards better than ATI

and they seem to be struggling to do that.

[20mmrain]

Until the cards are released don't celebrate the fact that Nvidia beats ATI... ATI did release first and Nvidia had to delay to make their cards better than ATI

I much prefer ATI to Nvidia any day!!! Trust me I would love it if ATI schooled Nvidia for a change. This way I can stop hearing about the 2900xt.
But lets face it..... Nvidia did have to wait in order to win. I think it is victory already.

LOL Nevermind that last comment

[a_ump]

thou i did just remember something, they had only ever said expect twice the performance of the GTX 280 right?

well if the GTX 480 is ~20% or more better than the HD 5870 then they met that goal cause it'd perform slightly better than the GTX 295. Plus the fact that the GTX 480, as has been mentioned, wasn't geared just towards gaming like previous desktop cards but as something new with their GPGPU jazz going on. just thought i'd remind everyone

[OnBoard]

They were only claming that GTX480 does tesselation better than HD5870. And tesselated scenes start @ 70s in the Heaven bench. That's the spot with the dragon the first time.

That 115s spot where the graph dips again is dragon spinaround.

http://forums.techpowerup.com/showpo...postcount=1852
EastCoasthandle's own graph still dips below 20FPS in those spots, just like it's shown in the NVIDIA chart. And we can see from the YouTube vide that 480 runs above 40fps in the first dragon spot.

So that 2x faster on DX11 tessalation is correct, isn't it not? They don't say the card is 2x faster in Heaven benchmark which we can clearly see that it's not.

Oh and that nvidia confidential graph has lower performance (under 40fps in 70s), it's the old graph from last year. So either heaven 1.1 is faster, or they have a better drive by now.

[20mmrain]

thou i did just remember something, they had only ever said expect twice the performance of the GTX 280 right?

well if the GTX 480 is ~20% or more better than the HD 5870 then they met that goal cause it'd perform slightly better than the GTX 295. Plus the fact that the GTX 480, as has been mentioned, wasn't geared just towards gaming like previous desktop cards but as something new with their GPGPU jazz going on. just thought i'd remind everyone
__________________

Yeah but If I remember correctly and I could be wrong you were one a long while ago that was arguing the fact that it would be 40% more powerful. Maybe it wasn't you but I thought it was

Either way you will still have a great card it looks like.

In reality I am happy with my purchase and I think anyone who wants to buy a Ferbi I mena a Fermi will be happy too!!!

[kid41212003]

Lol, you said it like he stated a fact. He was saying his thoughts about it just like everyone esle, 'cause it's not even released yet.

[20mmrain]

Lol, you said it like he stated a fact. He was saying his thoughts about it just like everyone esle, 'cause it's not even released yet.

Well I didn't mean it that way.... we all know it is speculation..... I just more brought it up for the plane and simple fact of showing how fun speculation is...... One minute things couldn't be going better and then the next rumor comes out and things couldn't be going worse.

That's the only reason.... Just like the rumors I fell in trap for with the 5800 sereis was about to be released...... If I remember correctly I think I believed a comment about the 5870 being 20% more powerful than a GTX 295 Boy was I wrong!

[a_ump]

haha u must have me confused, i thought back in january or so that fermi could be a creature, but after i read more "rumors' bout manufacturing probs/unmanufacturable, clock issues i changed my mind. Very true i did def think the GTX 470 would be more on par with the HD 5870 and the GTX 480 nicely in between but it seems the GTX 470 will supposedly debut between the HD 5850/70 and 480 between HD 5870/970 but close to the 5870. Haha yea u had to of confused me with someone(dunno how, name is 1 o kind) cause i don't usually give percentages, cept for recently with the rumored benchies.

i'm dissapointed in nvidia since it took em so long to get some screenshots and "rumored" benches out, but they did meet their claims....even if those claims came 7months late.

[EastCoasthandle]

Originally Posted by OnBoard

They were only claming that GTX480 does tesselation better than HD5870. And tesselated scenes start @ 70s in the Heaven bench. That's the spot with the dragon the first time.

That 115s spot where the graph dips again is dragon spinaround.

http://forums.techpowerup.com/showpo...postcount=1852
EastCoasthandle's own graph still dips below 20FPS in those spots, just like it's shown in the NVIDIA chart. And we can see from the YouTube vide that 480 runs above 40fps in the first dragon spot.

So that 2x faster on DX11 tessalation is correct, isn't it not? They don't say the card is 2x faster in Heaven benchmark which we can clearly see that it's not.

Oh and that nvidia confidential graph has lower performance (under 40fps in 70s), it's the old graph from last year. So either heaven 1.1 is faster, or they have a better drive by now.

I don't think you understand. Let me try it again:
The developers of Heaven's Benchmark have optimized the benchmark via aggressive culling in what is soon to be released Heaven Benchmark 1.1 (noting 30% improvement in performance). This should increase performance not only for 480 which is shown in the graph you mention but also the 5000 series as well. However, for whatever reason there is no performance improvement using the 5870 even though the video clearly show them using Heaven Benchmark 1.1.

Problem is:
A. They used Heaven Benchmark 1.0 for 5870 and Heaven Benchmark 1.1 for 480
B. The Heaven Benchmark 1.1 optimizations are only 400 series centric

In either cases invalidates the charts as there should be performance improvements for the 5870 if they used Heaven Benchmark 1.1. However, my graph shows that's not the case at all.

[OnBoard]

^^ Yeah but the grapht is old from last year. Like I tired to say the system he is benching performs better than shows in the graphs and the graphs are identical to the "2x more tesselation".

The performance increase in the video might be from the 1.1 version of the bench, but that's not in the graph. Maybe they didn't feel like rebenching them ?) Unless a new graph is released somewhere, like he promised in the video.

When they get the Heaven 1.1 out you can test if there is the promised performance increase and compare that to the first dragon part.

[phanbuey]

Its been said SOOOOO many times before, but every day it's more and more true -

late, yield issues, power hungry, reduced clockspeeds, completely new architecture paradigm, massive theoretical perf. but lack of real world perf due to design flaws, drivers tweaked for a specific, fixed bench (2900xt was 3dm06 champ), which then failed to perform nearly as well in real games. IT REEKS OF 2900XT.

one company took the plunge for dx 10 (ATI and VLIW, cluster shaders, tesselation, etc etc) while the other adopted a very 1900XT-type architecture, and held back dx10 in the name of performance. the company that took the plunge ended up losing massively... this time its nv taking the plunge to a completely new gpu design and ATi playing it safe (and benefiting from their previous plunge) in the name of performance.

I expect the Gf100 to be exactly a 2900XT with better marketing... everything feels the same. Even the leaked benches show a similar trend: Low performance in all but one benchmark and fanboys screaming "ITs THE DRIVERS WAIT FOR TEH VLIW... err.... 4 ISSUE GPU THINGY DRIVERS!!1!11!"

Not gonna happen... this gen will be a D O G. It may be a nice product for the price, and I am 80% sure I will buy one, because I am a slave to nvidia, but unseat a 5970 it will not. (IMO)

EDIT: @ EastCoasthandle thats a great point about the bench - the unigine definitely looks tweaked. Point is a fixed bench is a fixed bench... anyone can tweak a driver to do well in a FIXED bench bc they know exactly what conditions the card will face and when... much like the 2900XT just ROCKED at 3dm06 but got beat even by a 1900XTX in some cases in the real world. That heaven benchie looks nothing more than a driver hack.

[EastCoasthandle]

Originally Posted by OnBoard

^^ Yeah but the grapht is old from last year. Like I tired to say the system he is benching performs better than shows in the graphs and the graphs are identical to the "2x more tesselation".

The performance increase in the video might be from the 1.1 version of the bench, but that's not in the graph. Maybe they didn't feel like rebenching them ?) Unless a new graph is released somewhere, like he promised in the video.

When they get the Heaven 1.1 out you can test if there is the promised performance increase and compare that to the first dragon part.

If you watched the video they are using a similar graph. Look at the pics a few posts up. The graph in the video looks the same as the graph from last year, lol! Indicating they've had Heaven's BM 1.1 for a while now. It's not something they "just got a hold of". But in the end we need to see results from an unbiased reviewer at this point. Way to many holes in that video at this time.

[Steevo]

Worst 23FPS on the dragon. yes I have my fan at 100%, but it will be even cooler when I decide to move my block, and no fan.


So how often do you tessellate that much in game? Not very often. I'm getting 60 as a minimum FPS in Dirt 2 at native resolution and ultra settings and 4X.


So for right now, ATI still has the fastest available single GPU on the market, and will have for at least another month.

[20mmrain]

haha u must have me confused, i thought back in january or so that fermi could be a creature, but after i read more "rumors' bout manufacturing probs/unmanufacturable, clock issues i changed my mind. Very true i did def think the GTX 470 would be more on par with the HD 5870 and the GTX 480 nicely in between but it seems the GTX 470 will supposedly debut between the HD 5850/70 and 480 between HD 5870/970 but close to the 5870. Haha yea u had to of confused me with someone(dunno how, name is 1 o kind) cause i don't usually give percentages, cept for recently with the rumored benchies.

i'm dissapointed in nvidia since it took em so long to get some screenshots and "rumored" benches out, but they did meet their claims....even if those claims came 7months late.

Damn I thought it was you who I had that conversation with..... Too bad LOL It was a fun conversation.
Oh well ..... totally my fault!

Worst 23FPS on the dragon. yes I have my fan at 100%, but it will be even cooler when I decide to move my block, and no fan.


So how often do you tessellate that much in game? Not very often. I'm getting 60 as a minimum FPS in Dirt 2 at native resolution and ultra settings and 4X.


So for right now, ATI still has the fastest available single GPU on the market, and will have for at least another month.

I won't lie though I wish they did use more tessellation in games.... It would be a big step forward to making the games look a tone more smooth.
But you are right it is not used very much!

[Lionheart]

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!

[Mussels]

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