4870 Benchmarks

[CY:G]

This is supposedly the first 4870 benchmark against a 9800GTX


http://i233.photobucket.com/albums/ee121/CYG101/image3php.jpg

http://www.nordichardware.com/news,7854.html

So what do you think?

[Castiel]

The only thing I can say is WOW.
But im still going to stick with NVIDIA.

[P4-630]

As they say:

The Radeon HD 4870 whips 9800GTX by a 40% margin, while 4850 outpaces 8800GT with the same number. :p

AMD evidently planted the stories of 480 shader processors (it's not uncommon for AMD/ATI to plant stuff).

Radeon HD 4850 and 4870
they both sport a whopping 800 shader processors:twitch:

AMD's mainstream chip can do more FLOPS than NVIDIA enthusiast chips, and AMD's coming high-end chips will pack twice the FLOPS...

http://www.nordichardware.com/image3.php?id=5215

This is certainly a :slap: in NVIDIA's face.


http://www.nordichardware.com/news,7841.html


Then AMD-ATI supersedes and :nutkick: NVIDIA

[calvary1980]

it's physically impossible to fit 800 stream processors on a die of that size the technical specifications have been out for an entire month, manufactured and shipped why would they all of a sudden change them from 480 to 800 stream processors? and why is there no AMD link? on top of that someone at XS analyzed the table and the person who created it didn't even use the correct font or spacing.

I also find it hard to believe AMD would create such a blatant benchmark.

- Christine

[P4-630]

Or did they probably mean 800 million transistors....

Anywayz, we will find out very soon:)

[jbunch07]

trying to figure out what they are benching. that chart is weird. what do the numbers on the left side represent.

[generalhoultz]

this benchmark has blew my mind
i have no idea what to expect
the current gen in out for sure
8800gtx/ultra/9800gtx
are nothing compared to these

[ASharp]

trying to figure out what they are benching. that chart is weird. what do the numbers on the left side represent.

The numbers basically mean how many times the performance. The 9800 GTX acts as a baseline (as 1x) and the red bar represents the Radeon HD 4870 consistently around 1.4-1.5x the performance of the 9800GTX.

[jbunch07]

The numbers basically mean how many times the performance. The 9800 GTX acts as a baseline (as 1x) and the red bar represents the Radeon HD 4870 consistently around 1.4-1.5x the performance of the 9800GTX.

oh. ok.
well im not making any speculation until the card is actually out.
but wow if those benchmarks are true!

[magibeg]

The numbers basically mean how many times the performance. The 9800 GTX acts as a baseline (as 1x) and the red bar represents the Radeon HD 4870 consistently around 1.4-1.5x the performance of the 9800GTX.

I feel like I've seen this explanation about 2 dozen times now.

I just want all this speculation to finally end so we can get some hard results and move onto things that really matter. Like facts :P

[sam0t]

If the benchies are true, it would seem that ATI has finally found the sweet spot with their R600 architecture. Performance better, AA fixed, low price and power consumption. What more can a PC gamer want :)

[P4-630]

The numbers in that chart is probably the fan noise level when playing those games...
As you can see the NVIDIA based card has a constant fanspeed and the noise level stays the same, as for the new ati card has a variable fan speed as you can see...

Faster, hotter, means higher fanspeed..
:D

[farlex85]

So much speculation, so little confirmation, just sit tight guys we're almost there.........

[Temps_Riising]

TBH, Shader Processing units dont mean much, for the last 2 years ATI has had considerably more SP's than NVidia and in most cases have trailed painfully behind, ROP's and TMU's are where it's at! If these are legit they are impressive however but of course, impressive with the last generation of NVidia.......cards which are not going to be the HD4000 series competitors does not count for a lot, it's how they perform against NVidia's new offerings that determine if these cards are really going to sell in huge numbers.....everyone knows that in the main, the 9800GTX was in part at least just a bastardisation of the G92.

I do hope finally these nextgen cards from ATI will be more competative and at least be a match for NVidia's new offerings because I like choice and I am sick of the lack of cost effective competative choice but I seem to recall, before the HD3870/3850 and even the 2900XT were released, many people were having very similar conversations so I remain skeptical until these cards hit the shelves.......just my thoughts.

[wolf]

as a 9800GTX owner it doesn't inspire me to buy a 4870....

hec 8800GTX/Ultra owners will probably still be really happy with the investments they made 1-2 years ago!

and Temps_rising, i agree, to an extent..... i do believe that ROPS and TMU's count for alot more than others think. like you say, ATI have been packing way more shaders for ages now, but obviously they're not nearly as good if a 16 ROP, 64 shader card (9600GT) gives the 3870 a damn good run for its money. and even with these new specs the texture processing power will still fall behind a G92 8800.

32 ROPS and 240 nvidian SP's coupled to 1024mb with 140 G/bps of bandwidth should tear the ass out of the 4870, as for the X2, we shall see...

but again as many have stated, not many of these "benchmarks" can be trusted, and we are all eagerly awaiting release dates!

-my 2 cents

-Wolf

[jbunch07]

lets not forget that nv shaders are also clocked allot higher than ati shaders.
afaik the new ati cards are going to have faster shaders that can be oced like nv cards and not linked to the gpu core. but then again like ive said before...only time will tell.

[HTC]

If the 4850 3D06 and Vantage results i've seen are real (i'm assuming they are), even if the 4870 loses in lower resolutions, it will win in higher and / or with more details, assuming it preforms much the same way the 4850 does.

[DaedalusHelios]

To bad all the sources right now can't be trusted. Lets see the hard launch before we say where it fits in the scheme of things. ;)

[HTC]

To bad all the sources right now can't be trusted. Lets see the hard launch before we say where it fits in the scheme of things. ;)

This (http://translate.google.com/translate?u=http%3A%2F%2Fpublish.it168.com%2F2008% 2F0610%2F20080610041602.shtml&sl=zh-CN&tl=en&hl=en&ie=UTF-8) is where i based my previous post (google translated).

[wolf]

lets not forget that nv shaders are also clocked allot higher than ati shaders.
afaik the new ati cards are going to have faster shaders that can be oced like nv cards and not linked to the gpu core. but then again like ive said before...only time will tell.

quick comparison of 9600GT and HD3870 peak computational power.


775mhz x 320sp x 2 (MADD (2 FLOPs)) = 496 G/Flops


1625mhz x 64 sp x 3 (MADD (2 FLOPs)) + (MUL (1 FLOP)) = 312 G/Flops
1625mhz x 64 sp x 2 (MADD (2 FLOPs)) = 208 G/Flops

so even if you add the MUL process (which is unlikely in all scenarios) the 9600GT still loses BADLY to the 3870 in terms of peak shader arithmetic.

hec it even loses in terms of Peak pixel fill rate AND memory bandwidth

HD3870: 12.4 Gpixels/s and 72.0 GB/s
9600GT: 10.4 Gpixels/s and 57.6 GB/s

so if the HD3870 beats the 9600GT's peak shader arithmetic, Pixel fill rate and Memory bandwidth by 37.1%, 16.2% and 20% respectively, then why does the nvidia card come up very close to par (loses some, wins others) ?

it seems the 3870 has a small (yet present) advantage with no AA, but with AA on, they are so close to each other.....

not to mention this is a 65nm part vs a 55nm......

my answer - Nvidias architecture is more efficient.

took comparative scores from this review: http://www.techspot.com/article/88-geforce-9600gt-versus-radeon-hd3870/

-Wolf

[HTC]

quick comparison of 9600GT and HD3870 peak computational power.


775 x 320 x 2 = 496 MADD (2 FLOPs)


1625 x 64 x 3 = 312 (MADD (2 FLOPs) + MUL (1 FLOP))
1625 x 64 x 2 = 208 MADD (2 FLOPs)

so even if you add the MUL process (which is unlikely in all scenarios) the 9600GT still loses BADLY to the 3870 in terms of peak shader arithmetic.

hec it even loses in terms of Peak pixel fill rate AND memory bandwidth

HD3870: 12.4 Gpixels/s and 72.0 GB/s
9600GT: 10.4 Gpixels/s and 57.6 GB/s

so if the HD3870 beats the 9600GT's peak shader arithmetic, Pixel fill rate and Memory bandwidth by 37.1%, 16.2% and 20% respectively, then why does the nvidia card come up very close to par (loses some, wins others) ?

it seems the 3870 has a small (yet present) advantage with no AA, but with AA on, they are so close to each other.....

not to mention this is a 65nm part vs a 55nm......

my answer - Nvidias architecture is more efficient.

took comparative scores from this review: http://www.techspot.com/article/88-geforce-9600gt-versus-radeon-hd3870/

-Wolf

Question: wasn't the 9600 line the one W1zzard found to have a "concealed" OC or something like that? I remember reading something about that!

EDIT

Found the link: http://forums.techpowerup.com/showthread.php?t=53966&highlight=9600

[wolf]

my 9600GT shows the same clocks in GPU-Z as in rivatuner monitor, so im not sure....

[DarkMatter]

Not too impressive IMO considering they are playing in their field. It fits with the rumors of them being 25% faster than their "competition" (IMO are not competition, they will play in a very different price league, HD4850 should have been compared to the 9800GTX). That's why Nvidia didn't put the settings they used, so people like me can't find/guess the weaknesses and advantages out of the chart:

1- 1920x1200 is the resolution that most people upgrading their card are looking for, but is clearly one where G92 suffers as it was never designed to own in that resolution. Ati cards do well here which is good, but doesn't tell the real overall picture. Especially in the case of the 9800 GT, as people buying into that price segment won't (or shouldn't for their sanity) be looking into such high resolution gaming in the future. Anyhow this is a win in performance for Ati, but I don't know if it will translate to a win in sales. If 9800 GT is a lot closer to the HD4850 at 1680@1050 (G92 can keep well until that resolution, then fails at 1920x1200) and is significantly cheaper, and IMO will be, it could be a better deal for people looking into lower-end performance market.

2- The games. Some are old and others are the ones that favore Ati hardware. Then there's Crysis a game where any new card will look a lot better than any older one at that settings. Before anyone thinks of replying to this point, yes, Nvidia does the same in their benchmarks, but I am not talking about that, I am just explaining why I am not too excited about this one.

3- They have used 8xAF, why not x16 as done lately? It's a bit picky but IMO this is why: previous Ati cards have a problem with texturing power, that based on leaked specs has been carried over the new ones too. Twice the texturing power in a card with twice or more shading power, means the same texturing bottleneck. The higher the AF settings the better for Nvidia cards, and IMO that's why they used 8x instead of the usual x16. The difference would not be more than 5% but every bit counts for marketing purposes.

My 2 cents.

[Darkmag]

Not too impressive IMO considering they are playing in their field. It fits with the rumors of them being 25% faster than their "competition" (IMO are not competition, they will play in a very different price league, HD4850 should have been compared to the 9800GTX). That's why Nvidia didn't put the settings they used, so people like me can't find/guess the weaknesses and advantages out of the chart:

1- 1920x1200 is the resolution that most people upgrading their card are looking for, but is clearly one where G92 suffers as it was never designed to own in that resolution. Ati cards do well here which is good, but doesn't tell the real overall picture. Especially in the case of the 9800 GT, as people buying into that price segment won't (or shouldn't for their sanity) be looking into such high resolution gaming in the future. Anyhow this is a win in performance for Ati, but I don't know if it will translate to a win in sales. If 9800 GT is a lot closer to the HD4850 at 1680@1050 (G92 can keep well until that resolution, then fails at 1920x1200) and is significantly cheaper, and IMO will be, it could be a better deal for people looking into lower-end performance market.

2- The games. Some are old and others are the ones that favore Ati hardware. Then there's Crysis a game where any new card will look a lot better than any older one at that settings. Before anyone thinks of replying to this point, yes, Nvidia does the same in their benchmarks, but I am not talking about that, I am just explaining why I am not too excited about this one.

3- They have used 8xAF, why not x16 as done lately? It's a bit picky but IMO this is why: previous Ati cards have a problem with texturing power, that based on leaked specs has been carried over the new ones too. Twice the texturing power in a card with twice or more shading power, means the same texturing bottleneck. The higher the AF settings the better for Nvidia cards, and IMO that's why they used 8x instead of the usual x16. The difference would not be more than 5% but every bit counts for marketing purposes.

My 2 cents.

Well thank you at least now I know your 2 cents aren`t worth much. First of all its rumored 38%-50% faster than current not 25%(Nearly a half the difference)

Games favor ATI?!?! Are you on mad crack? The only gamed optimized for ATI is Call to Juarez. Doom 3 and Quake 4 are both OpenGL and when those games were made it heavily favored nVidia.

And lastly YES YOU ARE BEEN PICKY your looking of any reason to bash this card. if you look hard enough of something you'll find it - English Proverb.

All I want from the next gen is a fast cheap card that can play games at max. Stop been so fanboi-ish if ATI wins good for them if nVidia wins good for them, but I win regardless.

[lemonadesoda]

Twice the texturing power in a card with twice or more shading power, means the same texturing bottleneck
Could you explain that further.

Not true, per se. Depends on where you are scaling.

#1. If you "double" the resolution, then yes what you say is true, but
#2. If you keep the resolution the same, and keep "shader settings" the same, ie. FSAA and other "graphics effects", then you in theory "double" the FPS, and then hit the same bottleneck as you say. (But you DID just double up)
#3. If you double the FSAA or add more effects at the same resolution, the bottleneck has now changed completely. You wont be texture limited as before.

[Temps_Riising]

Well thank you at least now I know your 2 cents aren`t worth much. First of all its rumored 38%-50% faster than current not 25%(Nearly a half the difference)

Games favor ATI?!?! Are you on mad crack? The only gamed optimized for ATI is Call to Juarez. Doom 3 and Quake 4 are both OpenGL and when those games were made it heavily favored nVidia.

And lastly YES YOU ARE BEEN PICKY your looking of any reason to bash this card. if you look hard enough of something you'll find it - English Proverb.

All I want from the next gen is a fast cheap card that can play games at max. Stop been so fanboi-ish if ATI wins good for them if nVidia wins good for them, but I win regardless.

I think he meant......some games architecture favour ATI's GPU architecture. Additionally....I agree that Open GL DID favour NVidia, some newer OpenGL bench software would suggest that in some area's at least, ATI has caught up and even surpassed NVidia in OpenGL performance.

Not trying to take sides here, just my thoughts!

[Darkmag]

I think he meant......some games architecture favour ATI's GPU architecture. Additionally....I agree that Open GL DID favour NVidia, some newer OpenGL bench software would suggest that in some area's at least, ATI has caught up and even surpassed NVidia in OpenGL performance.

Not trying to take sides here, just my thoughts!

I know that why I said when. ATI has made huge strides in the OpenGL performance.

One can have the same it was optimized for nVidia argument with their benchmarks and it will more likely to be true, seeing has how the number of games that has the "The way its meant to be played" logo FAR out numbers that of ATI's.

[DarkMatter]

Could you explain that further.

Not true, per se. Depends on where you are scaling.

#1. If you "double" the resolution, then yes what you say is true, but
#2. If you keep the resolution the same, and keep "shader settings" the same, ie. FSAA and other "graphics effects", then you in theory "double" the FPS, and then hit the same bottleneck as you say. (But you DID just double up)
#3. If you double the FSAA or add more effects at the same resolution, the bottleneck has now changed completely. You wont be texture limited as before.

I'm not going to explain this to you again. Things don't work as you think.

1- Doubling the framerates (increasing performance) has the same effect as doubling the resolution, as long as you don't reach a bottleneck elsewhere. All the card knows and cares about is pixels and the calculations they have to do on them.

2- Indeed. I'm saying that they used 8x AF so that limit is on higher framerates, thus increasing the numbers in comparison to Nvidia's cards that don't have such a texture limitation. Example: HD3870 with texture bottleneck does 30 fps, HD4870 will do 60 fps, because the bottleneck is still there. 9800 GTX with the bottleneck out of the textures does 40 fps. Now lower texturing requirements and 9800GTX will do 40 fps again, because botleneck is not there, but Ati cards will do 40 and 80 fps. Think of it and you will understand what I'm saying.

3- FSAA is already the G92 bottleneck so comparatively is worst for Ati cards to increase the texturing requirements than increasing resolution and FSAA, because of what I said in ·2 Higher AA and res favors Ati cards in comparison to higher textures, that's my point.


DarkMag

The games. Some are old and others are the ones that favore Ati hardware.

Doom3 and Quake 4, => OLD GAMES also FEAR and Farcry

FEAR, Farcry, Call of Juarez, (World in Conflict a bit too) => Favor Ati's architecture, so this chart ONLY shows the best case scenario for Ati's new cards.

EDIT: Oh BTW, congratulations Darkmag. :rockout: since you are the 10th addition to the list of people-who-believe-DarkMatter-is-biased-towards-Nvidia-because-he-doesn't-buy-the-hype-and-looks-closely-to-specs-and-benchmarks-and-makes-his-logical-conclusions
-which-happen-to-not-be-so-favorable-to-Ati-cards-as-Ati-fanbois-would-want-to-be.

[Temps_Riising]

I know that why I said when. ATI has made huge strides in the OpenGL performance.

One can have the same it was optimized for nVidia argument with their benchmarks and it will more likely to be true, seeing has how the number of games that has the "The way its meant to be played" logo FAR out numbers that of ATI's.

I agree but I was'nt sure if that always actually meant that the game developers specifically wrote the code to benefit NVidia's GPU architecure or if in some games the developers simply "software tweaked" the game to respond better to NVidia drivers as in software support rather than hardware support....much like a "software patch" that we see from time to time from NVidia or ATi such as for Crysis........dont profess to be an expert on this!

[Darkmag]

DarkMag



Doom3 and Quake 4, => OLD GAMES also FEAR and Farcry

FEAR, Farcry, Call of Juarez, (World in Conflict a bit too) => Favor Ati's architecture, so this chart ONLY shows the best case scenario for Ati's new cards.

Those old games are still used in almost every benchmark, so its VALID to use them. Furthermore I fail to see how the age of those games can drastically influence the percentile difference of hardware.

And like I said in the other post the optimized argument can be played out from both sides against nVidia and against ATI, thus making it a useless desperate and stupid argument.

[wolf]

its pretty weird to compare the 4870 to a 9800GTX anyway, different generations hello....

while were at it, lets compare the 2900XT to a 7900GTX, oh, whats that? the ATi card won? funny that isn't it.....

sure they might be slated at roughly the same price, but Nvidia wont keep the 9800GTX costing the same if its outperformed, they will price it accordingly, they always do.

[Haytch]

Im not at all impressed with the initial benchmarking results.
From what these people are trying to tell us, is that the 4870 is a meer 50% better then the 9800GTX, which infact was a poor performing gfx card anyways.

The 9800GTX is something you DO NOT compare too. Its not the leading performer to contend against. Compared to the 8800GTX, the 4870 would probably be 15% better performing, and even if it was 25%, it still wouldnt match the 280GTX.

I do see the technological advancement made by AMD/ATi this time round, and the success of the 3870x2 was a lifesaver, but they are gonna have to do better then this to be remotely in the ' high end ' market.

[Darkmag]

its pretty weird to compare the 4870 to a 9800GTX anyway, different generations hello....

while were at it, lets compare the 2900XT to a 7900GTX, oh, whats that? the ATi card won? funny that isn't it.....

While your being an idiot why don`t you compare a 280GTX to a yet to be released in 5 years years ATI card. You see the problem the card doesn`t exist yet hello... So you can`t compare them. Its like obvious why they can`t compare it to the nVidia card, in case you still don`t know its because ATI doesn`t have a GTX280 to compare it too. Thats why were all waiting for reviews.

[DarkMatter]

Those old games are still used in almost every benchmark, so its VALID to use them. Furthermore I fail to see how the age of those games can drastically influence the percentile difference of hardware.

And like I said in the other post the optimized argument can be played out from both sides against nVidia and against ATI, thus making it a useless desperate and stupid argument.

In that chart everything is optimized to favor Ati cards. In Nvidia charts you will see the sdame happening. One thing doesn't anulate the other. REAL OVERALL performance won't be anywhere near what you see on those benchmarks.

Before anyone thinks of replying to this point, yes, Nvidia does the same in their benchmarks, but I am not talking about that, I am just explaining why I am not too excited about this one.


I said it already in my first post, DON'T TRY TO USE THE ARGUMENT THAT THE SAME CAN BE DONE BY NVIDIA, BECAUSE THAT'S OUT OF THE SCOPE OF MY POST. THE CHART IS BIASED AS ANY OTHER CORPORATIVE CHART IN THE WHOLE HISTORY!

Use your brain before trying to claim bias on my part.

[wolf]

While your being an idiot why don`t you compare a 280GTX to a yet to be released in 5 years years ATI card. You see the problem the card doesn`t exist yet hello... So you can`t compare them. Its like obvious why they can`t compare it to the nVidia card, in case you still don`t know its because ATI doesn`t have a GTX280 to compare it too. Thats why were all waiting for reviews.


hmm, well both cards do exist, and they are being compared to each other.

also watch who you call an idiot, you can make your point, and argue your opinion, don't attack people personally.

[lemonadesoda]

I'm not going to explain this to you again. Things don't work as you think.

#1. I'm asking you to explain your point in a way that people who read this thread can actually understand your position.

#2. You seem to agree with my 3 points. So WTF the insult?

#3. In previous thread you bashed the soda very badly for saying more texture units were needed for higher res. You said "no, no, no more shaders is all that's needed, noob". Now you are jumping on the same bandwagon eh? Well there's no room for ya here m8. That spot's taken.

Oh BTW, congratulations Darkmag. :rockout: since you are the 10th addition to the list of people-who-believe-DarkMatter-is-biased-towards-Nvidia-because-he-doesn't-buy-the-hype-and-looks-closely-to-specs-and-benchmarks-and-makes-his-logical-conclusions-which-happen-to-not-be-so-favorable-to-Ati-cards-as-Ati-fanbois-would-want-to-be.
#4. LOL

While your being an idiot why don`t you...

No way to talk to other TPU members. That position's taken by DarkMatter already. LOL

[DarkMatter]

In that chart everything is optimized to favor Ati cards. In Nvidia charts you will see the sdame happening. One thing doesn't anulate the other. REAL OVERALL performance won't be anywhere near what you see on those benchmarks.



I said it already in my first post, DON'T TRY TO USE THE ARGUMENT THAT THE SAME CAN BE DONE BY NVIDIA, BECAUSE THAT'S OUT OF THE SCOPE OF MY POST. THE CHART IS BIASED AS ANY OTHER CORPORATIVE CHART IN THE WHOLE HISTORY!

Use your brain before trying to claim bias on my part.

EDIT: And I haven't seen those games used in reviews for a long time and I read 20 of each new release so...

[wolf]

No way to talk to other TPU members.

absolutly, although personally Darkmatter has never attacked me.

[DarkMatter]

#1. I'm asking you to explain your point in a way that people who read this thread can actually understand your position.

#2. You seem to agree with my 3 points. So WTF the insult?

#3. In previous thread you bashed the soda very badly for saying more texture units were needed for higher res. You said "no, no, no more shaders is all that's needed, noob". Now you are jumping on the same bandwagon eh? Well there's no room for ya here m8. That spot's taken.


#4. LOL

1- It's enough explained for you now?

2- Where's the insult? I have said the same to you 5 times in 3 different threads so far.

3- Ha! Never said that. Not even close to saying that anywhere. More shader's needed for higher resolution. LOL! I will explain this for the 10th time (it must be): you need more of EVERYTHING to get more performance, but performance WILL ONLY INCREASE according to where the bottleneck is. Previous leaked specs suggested similar balance between TMUs and Shader on HD3000 and HD4000 series. Wherever the bottleneck was existent on HD3 it would be on HD4. SPECS HAVE CHANGED leaving the probability of the bottleneck occuring in TMUs as the higher one.
EDIT: Furthermore on those previous threads you were stating that more TMU will anly increase performance at higher resolutions leaving the performance pretty much the same at lower ones.

Anyhow it's kinda funny how in previous threads I WAS DEFENDING the HD4000 series performance against you, and now it seems to be the other way around. :laugh:

[Darkmag]

In that chart everything is optimized to favor Ati cards. In Nvidia charts you will see the sdame happening. One thing doesn't anulate the other. REAL OVERALL performance won't be anywhere near what you see on those benchmarks.


Am not arguing that they weren`t cherry picked but its highly unlikely that the cards performance will be significantly lower than that of a broad selection of test provided that the tests are legit(which we still don`t know).

Techpowerup uses both Farcry and Quake 4 in their benchmarks, I'm literary to lazy to look for more examples.

hmm, well both cards do exist, and they are being compared to each other.

also watch who you call an idiot, you can make your point, and argue your opinion, don't attack people personally.p

I didn`t directly imply that you were an idiot, I just used a metaphor that made your argument seem idiotic, not you. Sorry if you took it the wrong way.

While both card do exist and are been compared to each other, this graph appears to have been made by ATI(And everyone seem to assume this as well) who would have been unable to compare it to a GTX280. Thus the lack of the GTX280 comparison.

[wolf]

I didn`t directly imply that you were an idiot, I just used a metaphor that made your argument seem idiotic, not you. Sorry if you took it the wrong way.

While both card do exist and are been compared to each other, this graph appears to have been made by ATI(And everyone seem to assume this as well) who would have been unable to compare it to a GTX280. Thus the lack of the GTX280 comparison.

thanks for the clarification :)

yeah its annoying, both companies do it, nvidia compares its new cards to the 3870 and X2, which ATi compare to the 8800GT and 9800GTX..... its all marketing at this stage, however, something cool....

GTX 280 drops to $499
GTX 260 to $399

.....interesting indeed......

[lemonadesoda]

...I'm not defending ANY card's performance. It is that there are certain walls that a card hits under different conditions. ATI has scaled very well with shader related effects, but hits major walls when climbing beyond a certain resolution, esp. at 1600x1200 vis-a-vis NV. Since this resolution is now mainstream (or 1680x1080, or 1920x1200) then fixing the resolution bottleneck is the #1 priority or NV will keep winning the resolution benchmarks.

Personally, I dont give a flying fook who has the better card. I dont have shares in either. I'll just buy the better card as a combination of (my preference of) performance/price/power (heat).

Intel is a different matter however. If they dont deliver I'll be raising hell at the next shareholders meeting :-)

5 insults in 3 threads! That must be a record. Do I get a punch-bag prize?

[DarkMatter]

Ok guys it seems I am offensive when I don't know why. Anyway sorry.

But there's one thing I cannot understand: I may have insulted Darkmag because he offended me, but I didn't insult Lemonadesoda*. Darkmag is the only one that has directly insulted a member in the thread and I am flamed for that, while Darkmag remains unpunished? I can't understand.


*I have told you many times the same reasons why what you say is wrong IMO Lemonadesoda. And I only stated that, it's not an insult.

EDIT: And also check your facts because I have never insulted anyone in the forums. Probably you are talking about ChillyMyst. I didn't insult him. I was giving an example of what he was doing that included an insult, but was not meant as an insult (has happened the same to Darkmag if you think of it, but mine was easier to understand, I even explained it in the same post where the insult was!!). I was banned for a week, but mods understood it wasn't an insult. Read the complete story better and you will understand. In the end I think he is permanently banned now, I dunno the reason, but IMO that means a lot to what happened between us.

[flashstar]

Let's stop attacking people here, you don't want someone to get hurt. :laugh:

@ Wolf, where are your sources for the Nvidia price changes. Even if Nvidia manages to get those prices (and loses major $ due to manufacturing costs), ATI will still be the better bargain. The 4870x2 will be $499 and it will beat the 280, probably by at least 15%.

Btw, ATI actually has the better architecture. Just compare the pcb of a Nvidia and ATI card and you will see what I mean. ;)

[DanTheBanjoman]

While your being an idiot ...

Three things:


When calling people idiots get your grammar right.
Don't call people idiots or insult them in any way.
I forgot.

[wolf]

Let's stop attacking people here, you don't want someone to get hurt. :laugh:

@ Wolf, where are your sources for the Nvidia price changes. Even if Nvidia manages to get those prices (and loses major $ due to manufacturing costs), ATI will still be the better bargain. The 4870x2 will be $499 and it will beat the 280, probably by at least 15%.

Btw, ATI actually has the better architecture. Just compare the pcb of a Nvidia and ATI card and you will see what I mean. ;)

Fuzilla said those prices so iunno....i jsut thought it was an interesting tidbit of info....

and as for your architecture comment, obviously this:



quick comparison of 9600GT and HD3870 peak computational power.


775mhz x 320sp x 2 (MADD (2 FLOPs)) = 496 G/Flops


1625mhz x 64 sp x 3 (MADD (2 FLOPs)) + (MUL (1 FLOP)) = 312 G/Flops
1625mhz x 64 sp x 2 (MADD (2 FLOPs)) = 208 G/Flops

so even if you add the MUL process (which is unlikely in all scenarios) the 9600GT still loses BADLY to the 3870 in terms of peak shader arithmetic.

hec it even loses in terms of Peak pixel fill rate AND memory bandwidth

HD3870: 12.4 Gpixels/s and 72.0 GB/s
9600GT: 10.4 Gpixels/s and 57.6 GB/s

so if the HD3870 beats the 9600GT's peak shader arithmetic, Pixel fill rate and Memory bandwidth by 37.1%, 16.2% and 20% respectively, then why does the nvidia card come up very close to par (loses some, wins others) ?

it seems the 3870 has a small (yet present) advantage with no AA, but with AA on, they are so close to each other.....

not to mention this is a 65nm part vs a 55nm......

my answer - Nvidias architecture is more efficient.

took comparative scores from this review: http://www.techspot.com/article/88-geforce-9600gt-versus-radeon-hd3870/

-Wolf

means nothing to you. forgive me but its not entirely obvious from looking at the pcb, that the ATi card has the better architecture, however my figures speak for themselves.

not to mention im comparing ATI's enthusiast level card against nvidia's midrange, they "supposedly" compete in different classes and price segments, but you can see for yourself the results.

I mean a 9600GT, over a whole review testing with and without AA at multiple resolutions and multiple games, is only 0.8% slower than the 3870, thats within the margin of error for testing 2 identical 3870's. come on!

especially since were comparing a 65nm part to a 55nm part. the architecture has to be better. 64 sp's vs 320 is a helluva disadvantage on paper that just doesnt translate into games, and its very obvious that 128 nvidia sp's, and even 92/64 compete easily against 320 ati sp's, even if R700 packs 800 sp's, 240 nvidian ones should still come up trumps, if theyre clocked high enough/have gotten some nice tweaks for this core.

in my opinion, nvidia should concentrate on a tweaked and clocked up 55nm G92b to compete well with RV770 (something in the order of 850core/2200shader and some GDDR5) , and spend more time getting GT200b absolutely right. this would make the gfx stay relatively cheap until we truly see the next gen.

[DarkMatter]

I don't know if those new prices are correct, but they make much more sense to me than the old ones. I never understood how GT200 being twice a G92 in pretty much everything could cost 4 times more and still some sites claimed low profits per card! I always assumed the prices were inflated as with any other previous flagship card from any vendor, and accordingly agreed that the price was "justifiable" but the cost of production couldn't be all that much expensive. I mean yeah the chip has lower yields and thus instead of costing less than $100 per chip the cost is $120-150. Yes, the PCB has to be more complex and could end up being up to 4 times the price of that in G92 (supposedly 2x the bit depth translates to 4x the complexity of making and thus price), but will never be more than $100 IMO as around $25 must be maximum price of the PCB in a card like 9600GT/8800GT/HD3850. Otherwise what's left for the cost of rest of components and profits?

That leaves us with 150 + 100 = $250 versus a final price of $650. That's a great margin for the rest of components and profit, way too much IMO.

[magibeg]

I don't know if those new prices are correct, but they make much more sense to me than the old ones. I never understood how GT200 being twice a G92 in pretty much everything could cost 4 times more and still some sites claimed low profits per card! I always assumed the prices were inflated as with any other previous flagship card from any vendor, and accordingly agreed that the price was "justifiable" but the cost of production couldn't be all that much expensive. I mean yeah the chip has lower yields and thus instead of costing less than $100 per chip the cost is $120-150. Yes, the PCB has to be more complex and could end up being up to 4 times the price of that in G92 (supposedly 2x the bit depth translates to 4x the complexity of making and thus price), but will never be more than $100 IMO as around $25 must be maximum price of the PCB in a card like 9600GT/8800GT/HD3850. Otherwise what's left for the cost of rest of components and profits?

That leaves us with 150 + 100 = $250 versus a final price of $650. That's a great margin for the rest of components and profit, way too much IMO.

AH yes but theres so much more costs to a card after all. Those are the direct costs of making the card theres also the cost of research and development, transporting the cards to the store, the stores need their cut of the profits as well. Theres tons of people to pay along the way.

Also i find it pretty humorous that everyone is already a master of these cards architectures before they've even been released.

[DarkMatter]

AH yes but theres so much more costs to a card after all. Those are the direct costs of making the card theres also the cost of research and development, transporting the cards to the store, the stores need their cut of the profits as well. Theres tons of people to pay along the way.

Also i find it pretty humorous that everyone is already a master of these cards architectures before they've even been released.

All those costs can't be very high per card. Otherwise lower end cards wouldn't be profitable at all.

[wolf]

its all hilarious anyway, because with say 9600GT/8800GT SLi or HD3870CF or X2, you can pretty much play anything out at 1920x1200+ + 4xAA + 16xAF anyway (obvious exception of crysis, but who isn't over that already)

and for that kinda performance were only paying 300-400USD, tats pretty incredible already, i say let the revision wars/die shrinks continue.

[DarkMatter]

its all hilarious anyway, because with say 9600GT/8800GT SLi or HD3870CF or X2, you can pretty much play anything out at 1920x1200+ + 4xAA + 16xAF anyway (obvious exception of crysis, but who isn't over that already)

and for that kinda performance were only paying 300-400USD, tats pretty incredible already, i say let the revision wars/die shrinks continue.

That's because of the consoles. Historically 1-2 years after a "next-gen" console launch, there's been an slow down in games features and requirements advancement and a massive improvement in the 3rd or 4th year. The reason of this is simple: when the average PC user has hardware equal or only a bit better than consoles, even if high-end hardware is a lot faster, developers don't bother to create a much improved version for the PC and "stupidly" increase things like texture resolution and the likes instead for those high-end cards. I say stupidly because increasing texture size when the renderer can't actually take benefit of it is kinda stupid IMO. I expect massive increase on requirements in the (near) future games. Maybe not in the next batch of games (i.e Farcry 2) but surely in the next.

EDIT: Crysis is in fact the only game of the later ones that has followed the curve of the last years.

[batmang]

I'll wait for the REAL reviews from AnandTech, HardOCP and Tom's before I believe anything.

[Megasty]

I'll wait for the REAL reviews from AnandTech, HardOCP and Tom's before I believe anything.

I'm ready for this BS fest to be over too. I can't give credence to any of this stuff anymore. Even if I did, it would still be a different experience for each buyer because of driver issues, system settings, games, stability, cooling, etc. Comparing stupid benchmarks only work when you know the exact system specs. Even that's flawed because if they are testing the cards on the same system, the system's resources will start to falter as the tests continue. I can only see reality now since my salt is gone. Too bad there's no reality with these cards as of yet :p

[J-Man]

ATi will beat NVIDIA this time round. I hope.

[yogurt_21]

ATi will beat NVIDIA this time round. I hope.

I dont' care if they beat them as long as they are competitive. the 8800's rolled over the 2900's they need to bounce back from that.

[imperialreign]

Those old games are still used in almost every benchmark, so its VALID to use them. Furthermore I fail to see how the age of those games can drastically influence the percentile difference of hardware.

And like I said in the other post the optimized argument can be played out from both sides against nVidia and against ATI, thus making it a useless desperate and stupid argument.

jumping in a little late here, but I'm taking DarkMatter's side on those benches

Doom3 ran horrible on ATI cards when it was released, after a CAT update that cleared up an issue with the way the game engine wanted texture rendering handled - ATI cards (of the time) performed as well, if not better, than their competition. The reason Quake4 and ETQW still get pulled in for benching, is that Q4 in itself was more intensive for hardware than D3 was, thanks to the many updates Raven Software did before the game was released.

nVidia did at one point have a major advantage with OGL (thanks to the intellectual properties of 3DFX that they acquired), but that advantage has been practically nil for a few years now. ATI can most definitely hold their own in OGL applications.

Call of Juarez - one of the few games that ATI worked alongside game developers - the performance lead that nVidia typically has over ATI in many games is the same lead we see ATI cards have over nVidia with this game.

FEAR - another game where ATI worked closely with developers, and ATI cards have been running neck and neck with nVidia's in this game with every new hardware release since FEAR was put on the shelf. The only reason this really old title is still included, is that it was one of the most intensive DX9 titles that was released.


Right now, it just looks like both sides are stacking their PR charts in their favor - and until we can actually compare them in hand, all those PR charts really don't mean squat except to fluster our feathers and hype the hardware up.

. . . and, TBH, I think this has been the most hyped up hardware release from both camps since the 1800/1900-7800/7900 releases. I'm surprised users aren't holding release parties for these new cards yet . . .

[TonyStark]

This is supposedly the first 4870 benchmark against a 9800GTX


http://i233.photobucket.com/albums/ee121/CYG101/image3php.jpg

http://www.nordichardware.com/news,7854.html

So what do you think?

Although the pic is fake, it isn't far from the truth (from what I've heard).

[razaron]

when are the next gen graphics cards even coming out?

ps. tesla c870 beats all .......:D

[Nyte]

it's physically impossible to fit 800 stream processors on a die of that size the technical specifications have been out for an entire month, manufactured and shipped why would they all of a sudden change them from 480 to 800 stream processors? and why is there no AMD link? on top of that someone at XS analyzed the table and the person who created it didn't even use the correct font or spacing.

I also find it hard to believe AMD would create such a blatant benchmark.

- Christine

It's physically possible.

[hayder.master]

Wooooooooooooooooooow

[btarunr]

Yay! Finally a single ATI GPU manages to beat the....8800 Ultra :shadedshu

[wiak]

I know that why I said when. ATI has made huge strides in the OpenGL performance.

One can have the same it was optimized for nVidia argument with their benchmarks and it will more likely to be true, seeing has how the number of games that has the "The way its meant to be played" logo FAR out numbers that of ATI's.
+1
even 3dmark vantage should be disqualifyed as it uses "The way its meant to be played" physics

I'll wait for the REAL reviews from AnandTech, HardOCP and Tom's before I believe anything.
dude you forgot something, i dont think w1zzard will like you :P

[wolf]

Yay! Finally a single ATI GPU manages to beat the....8800 Ultra :shadedshu

your point is well amde