PDA

View Full Version : NVIDIA GF100 Previews


Pages : 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 [23] 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52

shadow001
03-12-10, 04:27 PM
[QUOTE=shadow001;2207203]


Its a guess by the author, so I don't know how we can take that seriously since at the time, cards weren't sent out well outside of developers :)


Are you absolutely sure about that?,as it states that nvidia mentioned that in tesselation bound situations,Fermi would be 100% faster than cypress,yet the earlier graph i posted regarding the Heaven demo, fermi doesn't show a 100% advantage even there over cypress.




I do expect stuff like that from the CEO of a company that is making the product, why not?

But seriously though, can't hinge on every word, how bout this from AMD and the r600 being the fastest card they have every produced, so many people twisted that around and stated its going to be the fastest on the market period.


It's still wasn't called an combo of an SR71,toyota prius,100 gallons of moonshine,a light saber,DNA from megan fox(WTF???),fold space and hitting it with millions of lines of code by driver jedi masters.


Gives the impression he wasn't taking to professionals in the industry covering all fields where Fermi can be used,but rather to a bunch of 12 year olds in the audience,and it still doesn't change the fact that it was done 6 months ago,and we've still seen no real independent benchmarks,prices or power consumption,even ignoring what charlie just posted a little earlier about the GTX480 needing 275 watts,which is higher than either of us thought it would be.

Razor1
03-12-10, 04:32 PM
Are you absolutely sure about that?,as it states that nvidia mentioned that in tesselation bound situations,Fermi would be 100% faster than cypress,yet the earlier graph i posted regarding the Heaven demo, fermi doesn't show a 100% advantage even there over cypress.


Yes I am very sure. Look at the date of the article and when the January meeting was


Gives the impression he wasn't taking to professionals in the industry covering all fields where Fermi can be used,but rather a bunch of 12 year olds in the audience,and it still doesn't change the fact that it was done 6 months ago,and we've still seen no real independent benchmarks,prices or power consumption,even ignoring what charlie just posted a little earlier about the GTX480 needing 275 watts,which is higher than either of us thought it would be.


There you go again, taking one person's word for it, ITS A BS ARTICLE ;), look its PR, what was he comparing it to. Last gen cards from nV, this gen's AMD cards? What is the point of view? That is what you, anyone else or I will never know. Have you ever heard, the best kind of lies are lies with truths, guess what, that's the best type of PR.

Everything we have heard about Fermi's power consumption, is a falicy. Its a bit higher then the gtx 280, mark my word on that.

shadow001
03-12-10, 04:36 PM
Yes I am very sure. Look at the date of the article and when the January meeting was


January 19th 2010,about 6~7 weeks ago....So what's changed since then?





There you go again, taking one person's word for it, ITS A BS ARTICLE ;)


That was what Nvidia's CEO directly stated on that stage,in his presentation of fermi 6 months ago...;)


GTX 280 was rated at 236 watts maximum..So define a bit higher here?

Razor1
03-12-10, 04:39 PM
January 19th 2010,about 6~7 weeks ago....So what's changed since then?


When was the deep dive meeting? Notice all the deep dive articles came around the same time as the date of that article......

Correlation is your friend. And just to clarify, some people think your a PR guy for AMD, you just won't cut it, guess what I just switched my jobs to being a Exec Producer for an ad agency, and things that we do are viral marketing, pr, etc. and you wouldn't cut it, because the longer you stick on one point that has no definite meaning, its a moot point, it actually makes it situation much much worse ;)


That was what Nvidia's CEO directly stated on that stage,in his presentation of fermi 6 months ago...;)

Sorry was editing my post,

And these are things that I say are silly to argue about, because there is no point of reference, who the f*ck knows what Jensen was talking about, because he could have been in some alternate realty and comparing it to AMD's r600! which is outlandish but realistically he probably was comparing it to the gt200.

shadow001
03-12-10, 04:49 PM
When was the deep dive meeting? Notice all the deep dive articles came around the same time as the date of that article......

Correlation is your friend.




Sorry was editing my post,

And these are things that I say are silly to argue about, because there is no point of reference, who the f*ck knows what Jensen was talking about, because he could have been in some alternate realty and comparing it to AMD's r600! which is outlandish but realistically he probably was comparing it to the gt200.


And this particular article mentioned the performance figures for real world gaming scenarios(20~50%),and we've seen that compared to that second picture i posted from that Nvidia marketing director,there isn't a 100% performance difference in tesselation performance between fermi and cypress.


You also claimed that it'll use a bit more power than the GTX280,when the GTX280 was rated at 236 watts,while Cypress uses 188 watts.


Fermi also has an additonal 850 million transistors and a 384 bit memory bus,so the single GPU HD5870 shouldn't be able to compete there as well.


2 weeks and counting till we find the truth.


How's that for different points?....;).

Razor1
03-12-10, 04:59 PM
And this particular article mentioned the performance figures for real world gaming scenarios(20~50%),and we've seen that compared to that second picture i posted from that Nvidia marketing director,there isn't a 100% performance difference in tesselation performance between fermi and cypress.

He is guessing that,

Nvidia's early benchmarks has it performing up to twice as fast as the Radeon HD 5870 (ATI's fastest single-chip DX11 graphics card) in some tests. Those are usually directed geometry-heavy benchmarks, though. In real games, it looks like performance will be anywhere from 20-50% faster, depending on the game and settings.

If you are going to hang on every single word, hang on to the red ones, "it looks like" what does that mean, to me when I say something like that, it might be like this because of what I have seen. What did I see, nothing concrete that's why I'm making an educated guess. Doesn't mean this is it.

You also claimed that it'll use a bit more power than the GTX280,when the GTX280 was rated at 236 watts,while Cypress uses 188 watts.


The TDP for the GTX 280 was 280 watts, keep in mind Charlie and ilk has been using TDP numbers as real world power draw.

Fermi also has an additonal 850 million transistors and a 384 bit memory bus,so the single GPU HD5870 shouldn't be able to compete there as well.

Hmm yeah so the r600 was a bigger chip then the g80 what happened to it? Although Fermi isn't an r600 just a hint.

2 weeks and counting till we find the truth.

True


How's that for different points?....;).

Better :)

shadow001
03-12-10, 05:16 PM
Hmm yeah so the r600 was a bigger chip then the g80 what happened to it? Although Fermi isn't an r600 just a hint.




Ahem: http://en.wikipedia.org/wiki/Nvidia_gpus#GeForce_8_.288xxx.29_series


G80 GPU clocked in at 476mm^ and 681 million transistors,with the NVIO chip that controls the DVI ports being a seperate chip.


The Original R600 was built at 80nm from the start and contained about 720 million transistors,for an overall die size of 420mm^.


The transistor budget difference between Cypress and Fermi is larger than the entire transistor budget of either the G80 or R600 or the G92,or the RV670 and not much behind the total budget of the RV770 chip,which clocked in at 956 million transistors total.


You were saying?....;)

XMAN52373
03-12-10, 05:54 PM
Ahem: http://en.wikipedia.org/wiki/Nvidia_gpus#GeForce_8_.288xxx.29_series


G80 GPU clocked in at 476mm^ and 681 million transistors,with the NVIO chip that controls the DVI ports being a seperate chip.


The Original R600 was built at 80nm from the start and contained about 720 million transistors,for an overall die size of 420mm^.


The transistor budget difference between Cypress and Fermi is larger than the entire transistor budget of either the G80 or R600 or the G92,or the RV670 and not much behind the total budget of the RV770 chip,which clocked in at 956 million transistors total.


You were saying?....;)

Put them both on the same process and its a completely different story. G80 was .09nm and R600 was .08nm

shadow001
03-12-10, 06:10 PM
Put them both on the same process and its a completely different story. G80 was .09nm and R600 was .08nm


You could be right,as it would yeild a 504mm^ die if the R600 was built at 90nm instead of 80nm,beating the G80 by 28mm^.


But my main point is that the transistor budget difference between Cypress,at 2.15 billion and Fermi,clocking in a 3 billion,is an 850 million transistor gap,and larger than the entire transistor budget of at least some of the high end GPU's released over the years.


That's 850 million transistors extra to power up relative to Cypress,coming in fact to a 40% increase in transistor budget,so will we see Fermi performing 40% faster than cypress in actual real world game performance?...


2 weeks and counting down to find out for sure if it does.

Rollo
03-12-10, 06:48 PM
You could be right,as it would yeild a 504mm^ die if the R600 was built at 90nm instead of 80nm,beating the G80 by 28mm^.


But my main point is that the transistor budget difference between Cypress,at 2.15 billion and Fermi,clocking in a 3 billion,is an 850 million transistor gap,and larger than the entire transistor budget of at least some of the high end GPU's released over the years.


That's 850 million transistors extra to power up relative to Cypress,coming in fact to a 40% increase in transistor budget,so will we see Fermi performing 40% faster than cypress in actual real world game performance?...


2 weeks and counting down to find out for sure if it does.

Pffft. This sort of point only matters to AMD viral marketers like you.

IRL, no one cares how many transistors are on a chip, how much chips cost to make, or what profits OEMs make. They only care about the benchmarks and the features- and they can't be "fantasy football" features like ATi has- they have to be real features in real games.

Enjoy the last 14 days ATi is relevant. ;)

shadow001
03-12-10, 07:00 PM
Pffft. This sort of point only matters to AMD viral marketers like you.

IRL, no one cares how many transistors are on a chip, how much chips cost to make, or what profits OEMs make. They only care about the benchmarks and the features- and they can't be "fantasy football" features like ATi has- they have to be real features in real games.

Enjoy the last 14 days ATi is relevant. ;)


LOL,actually it's been closer to 180 days since the HD5*** series has been out and Nvidia talking about how great Fermi is,just not showing much,so you're just hoping it's worth all the time waiting.


You better hope it's was worth the 6 month wait,and if it turns out to be only marginally faster,i'm going to laugh so hard and call you an idiot for waiting.


With the focus Nvidia has been putting on GP-GPU performance and professional rendering environments,with features on this chip aimed towards those,it seems like playing games on it is just an extra bonus,not the primary focus of the architecture,unlike Cypress,which is focused towards gaming/multimedia first,and GP-GPU functions/performance coming in a distant second....But that's just my opinion.


For the Nvidia users on this forum who primarily use their cards for gaming,i hope i'm wrong,i really do.

Rollo
03-12-10, 07:15 PM
LOL,actually it's been closer to 180 days since the HD5*** series has been out and Nvidia talking about how great Fermi is,just not showing much,so you're just hoping it's worth all the time waiting.


You better hope it's was worth the 6 month wait,and if it turns out to be only marginally faster,i'm going to laugh so hard and call you an idiot for waiting.


With the focus Nvidia has been putting on GP-GPU performance and professional rendering environments,with features on this chip aimed towards those,it seems like playing games on it is just an extra bonus,not the primary focus of the architecture,unlike Cypress,which is focused towards gaming/multimedia first,and GP-GPU functions/performance coming in second....But that's just my opinion.


For the Nvidia users on this forum who primarily use their cards for gaming,i hope i'm wrong,i really do.

Errrr...yeah. The wait was really tough for me with my GTX295 Quad SLi, I could barely keep my framerates playable.:rolleyes:

I'm already laughing at you with the 3d monitors and 2d video cards. Do you think ATi just can't afford the R&D, or do you think they just can't code the drivers? Tough to say.

shadow001
03-12-10, 07:25 PM
Errrr...yeah. The wait was really tough for me with my GTX295 Quad SLi, I could barely keep my framerates playable.:rolleyes:

I'm already laughing at you with the 3d monitors and 2d video cards. Do you think ATi just can't afford the R&D, or do you think they just can't code the drivers? Tough to say.


This one's just for you


ATI (AMD) and its partners will supposedly soon announce a lineup of 3D products, including a 3D-enabled ATI Eyefinity technology (to counter Nvidia�s 3D Vision Surround), 120Hz 3D-ready displays and notebooks, active shutter glasses and passive polarized ones (have you heard of passive shutter glasses!?), S3D support for DirectX 9, 10 and 11, Quad Buffered OpenGL, Blu-ray 3D support. As you can see from the presentation slide above the 3D-gaming middleware partners are DDD and iZ3D, and for Blu-ray 3D support ArcSoft and CyberLink. ATI is also going to try to work on establishing standards that will help in having compatibility and much wider choices when building your computer and stereo 3D setup if they are adopted by others.


http://vr-zone.com/forums/578728/ati-open-stereo-3d-initiative-at-gdc-2010.html


Oh and despite you having the Quad SLI setup,which i'm sure it's fast,i'm faster still,have support for DX11 and already a few games that use some of the features in DX11, and run 30+ games on 3 monitors without problems either,so as far as dick mesuring contests go,i still win.


:D

Razor1
03-12-10, 07:43 PM
You could be right,as it would yeild a 504mm^ die if the R600 was built at 90nm instead of 80nm,beating the G80 by 28mm^.


But my main point is that the transistor budget difference between Cypress,at 2.15 billion and Fermi,clocking in a 3 billion,is an 850 million transistor gap,and larger than the entire transistor budget of at least some of the high end GPU's released over the years.


That's 850 million transistors extra to power up relative to Cypress,coming in fact to a 40% increase in transistor budget,so will we see Fermi performing 40% faster than cypress in actual real world game performance?...


2 weeks and counting down to find out for sure if it does.

funny you know AMD and nV don't count transistors the same way right :headexplode:

onmikesline
03-12-10, 07:44 PM
its friday night, you guys should be going out

Razor1
03-12-10, 07:50 PM
its friday night, you guys should be going out

been partying all week, new job and all ;) taking it easy today :)

Rollo
03-12-10, 07:53 PM
This one's just for you





http://vr-zone.com/forums/578728/ati-open-stereo-3d-initiative-at-gdc-2010.html


Oh and despite you having the Quad SLI setup,which i'm sure it's fast,i'm faster still,have support for DX11 and already a few games that use some of the features in DX11, and run 30+ games on 3 monitors without problems either,so as far as dick mesuring contests go,i still win.


:D

Oh noes! IZ3d!

http://www.xtremesystems.org/forums/showthread.php?p=4283019
After using a number of technologies I can tell you that iZ3D's is at the bottom of the barrel.

http://www.xbitlabs.com/articles/monitors/display/iz3d_7.html#sect0
Unfortunately, these advantages are all negated by one defect. The left and right pictures are not separated fully, which leads to conspicuous and eye-straining artifacts in every game I have tried.

http://www.extremetech.com/article2/0,2845,2347292,00.asp
The other problem is that there is a serious color shift with the polarized glasses.

But we're still having a hard time recommending this for…well…anyone.

ATi - 15 months late to the 3d game and they trot out crap like IZ3D in hopes potential buyers will be dumb enough to think "ATi has 3D too!".

I love how you point out the polarized glasses as if they're some sort of advantage when the Extreme Tech article goes into great detail about how they screw up color. :D

Your viral tricks won't avail you here.

shadow001
03-12-10, 07:58 PM
funny you know AMD and nV don't count transistors the same way right :headexplode:


So Cypress is 334mm^ at 2.15 billion transistors,so what makes Fermi a much larger chip,since it's also made at the same 40nm TSMC fab process?....Rumors suggest 500mm^+,which given past GPU releases from Nvidia(G80 at 476mm^,GT200 at 576mm^),doesn't seem far fetched in the least.


And in the end,when we start to argue the finer points of how transistors are counted,when both companies revealed the numbers for each architecture,and the gap is large between both any way you count them anyhow,i think bringing up the point of how they're counted really is reaching for straws here.

shadow001
03-12-10, 07:59 PM
Oh noes! IZ3d!

http://www.xtremesystems.org/forums/showthread.php?p=4283019


http://www.xbitlabs.com/articles/monitors/display/iz3d_7.html#sect0


http://www.extremetech.com/article2/0,2845,2347292,00.asp




ATi - 15 months late to the 3d game and they trot out crap like IZ3D in hopes potential buyers will be dumb enough to think "ATi has 3D too!".

I love how you point out the polarized glasses as if they're some sort of advantage when the Extreme Tech article goes into great detail about how they screw up color. :D

Your viral tricks won't avail you here.


Funny how you forgot the part about them also working on unified standards and also supporting shutter glasses and 120 HZ displays....Selective editing much?

K007
03-12-10, 08:06 PM
LOL!!!...love the fights on these forums keep it up rollo.

shadow001
03-12-10, 08:09 PM
LOL!!!...love the fights on these forums keep it up rollo.


He is persistent,i'll give him that much....:D

Johnny C
03-12-10, 08:44 PM
been partying all week, new job and all ;) taking it easy today :)

Same here....

Rollo
03-12-10, 09:15 PM
LOL!!!...love the fights on these forums keep it up rollo.

Whereever ATi's viral minions seek to sell ATi's pipe dreams to the unsuspecting public, I'll be there. ;)

Razor1
03-12-10, 09:17 PM
So Cypress is 334mm^ at 2.15 billion transistors,so what makes Fermi a much larger chip,since it's also made at the same 40nm TSMC fab process?....Rumors suggest 500mm^+,which given past GPU releases from Nvidia(G80 at 476mm^,GT200 at 576mm^),doesn't seem far fetched in the least.


And in the end,when we start to argue the finer points of how transistors are counted,when both companies revealed the numbers for each architecture,and the gap is large between both any way you count them anyhow,i think bringing up the point of how they're counted really is reaching for straws here.



no just sayin :headexplode:

see we still don't know the exact size of fermi, if its ~500mm2 then its transistors are packed just as tight as AMD's, but basing on transistor count can't really do that with performance, cause lets see there are quite a bit different things on that front, which you are just generalizing, sorry to say it.

Let me generalize something for you, and I'm sure you will think this is out in left field, because of Fermi's L2 coherent cache we get 10% extra performance. Makes some sense if that was in a CPU instead of a GPU, but doesn't really make sense over all right? ;) I drew a wrong conclusion by over generalizing because of previous knowledge.

Rollo
03-12-10, 09:21 PM
Funny how you forgot the part about them also working on unified standards and also supporting shutter glasses and 120 HZ displays....Selective editing much?

What's to mention? More ATi related vaporware you're hoping will come to market?

Poor ATi always the follower.