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

Revs
03-18-10, 04:17 PM
I wonder if this turns out to be true after all :)...

NVIDIA Fun Fact of the Week: GF100 graphics cards will provide hardware support for GPU overvoltaging for extreme overclocking!
December 17, 2009 at 4:06pm
http://www.facebook.com/NVIDIA?v=feed&story_fbid=208604209109#

It's a pretty bold claim :lol:

NIGELG
03-18-10, 05:22 PM
From the first sentence of the FAQ page...



Dip****.

I'm very new here but did you did you just call me a dip**** ????

CaptNKILL
03-18-10, 05:33 PM
:rofl


I think he meant dipstick. :bleh:

Rollo
03-18-10, 05:44 PM
295 watts,3.2 billion transistors and only 5~10% faster overall than the HD5870?

Ouch if it's true.

Some will buy it because it's NVIDIA, and they see that logo on every game they fire up.

Some will buy them because they like PhysX or 3d Vision or 3d Vision Surround.

Some will buy because they only buy the fastest.

Some will buy because they know new games will launch with profiles, and they'll have to wait for patches with CF.

Some will buy because they like CUDA apps like Badaboom.

Mostly people will buy because there's not much reason to buy similarly performing cards with less features.

As much as some say power consumption matters and transistor count matters, most people only care about performance, price and features.

I know lots more than I did before, and I'm feeling good about the whole situation. NVNEWS NVIDIA fans can be hopeful for the next eight days, I think you'll all be happy.

NIGELG
03-18-10, 05:57 PM
:rofl


I think he meant dipstick. :bleh:

Well it doesn't matter which.I didn't come here to be insulted.........I don't deserve that at all......

shadow001
03-18-10, 06:37 PM
Some will buy it because it's NVIDIA, and they see that logo on every game they fire up.

Some will buy them because they like PhysX or 3d Vision or 3d Vision Surround.

Some will buy because they only buy the fastest.

Some will buy because they know new games will launch with profiles, and they'll have to wait for patches with CF.

Some will buy because they like CUDA apps like Badaboom.

Mostly people will buy because there's not much reason to buy similarly performing cards with less features.

As much as some say power consumption matters and transistor count matters, most people only care about performance, price and features.

I know lots more than I did before, and I'm feeling good about the whole situation. NVNEWS NVIDIA fans can be hopeful for the next eight days, I think you'll all be happy.


If we ignore that the HD5970 doesn't exist that is,and has been on the market for 4+ months now....Oh and profiles for crossfire are now a seperate update,intentionally made that way so that there's no need to wait 1 month for the entire driver package once a brand new game has been released,assuming there isn't a profile for it already of course.


So it can't be the fastest if there's something faster on the market already.;)

lee63
03-18-10, 06:38 PM
I'm very disappointed I wont be able to run three 480s...we'll see I guess :thumbdwn:

HYBRID
03-18-10, 06:43 PM
Some will buy it because it's NVIDIA, and they see that logo on every game they fire up.

Some will buy them because they like PhysX or 3d Vision or 3d Vision Surround.

Some will buy because they only buy the fastest.

Some will buy because they know new games will launch with profiles, and they'll have to wait for patches with CF.

Some will buy because they like CUDA apps like Badaboom.

Mostly people will buy because there's not much reason to buy similarly performing cards with less features.

As much as some say power consumption matters and transistor count matters, most people only care about performance, price and features.

I know lots more than I did before, and I'm feeling good about the whole situation. NVNEWS NVIDIA fans can be hopeful for the next eight days, I think you'll all be happy.

Spoken like a true Nvidia Focus Group Member. :rolleyes2 I could give a sh&t about cuda, PhysX or 3d Vision or 3d Vision Surround.

I do care about heat INSIDE MY RIG and fast. By the way the CF profile thing became official with 10.2 cc.

These card (TOP END) have always been about the fastest for gaming all of a sudden because speed is no longer an issue with about 99% except for Cysis we get useless feature IMHO.
I really hope this card is smoking so we have another option, options and lower cost period competition is good .

CaptNKILL
03-18-10, 06:47 PM
I've got my eye on this thread.

shadow001
03-18-10, 07:03 PM
I've got my eye on this thread.


No need to worry,i just set the record straight at least when it comes to raw performance,when it was released,and crossfire profiles.


If the idea is to use Cuda,or PhysX or 3D glasses with a single display,go for Fermi as there's no other choice really.


A thought on using 3 Displays,with or without 3D glasses,is that users have to keep in mind that the cards will work a lot harder driving those 3 displays,and i'll happily admit that even with a pair of HD5970's in quad crossfire,running the latest games at high quality settings with AA kicks the **** out of them,even though it is still playable overall,but i am talking 4 GPU's working together afterall.


So those thinking to buy 2 fermi cards to run 3 displays,if their performance is what it is,might very well have to settle for lower graphics settings to keep the FPS up into the smooth regions.

Revs
03-18-10, 07:51 PM
Just been reading the white paper. I says GF100 has 6x the tesselation power of the 5870 in DX11 apps.

Rollo
03-18-10, 08:17 PM
Just been reading the white paper. I says GF100 has 6x the tesselation power of the 5870 in DX11 apps.

This is one of many reasons anyone thinking of purchasing a high end card before 3/26 should keep their debit card in their wallet.

Some surprises coming, good stuff coming soon!

onmikesline
03-18-10, 08:22 PM
yes in d11 lol, and its going to take a year for developers to have 100% tesselation in games, like now, its only little bits they add, but ill still buy one around x-mas time anyways because i always support nvidia

shadow001
03-18-10, 08:24 PM
So,the HD5870 has close to 2X the shading power of Fermi in single precision math....You don't hear ATI making a big deal out of it.


In the end,what matters is how it performs overall,and that involves all the parts within a GPU,not just the tesselation ability,and any given GPU is only as fast as the slowest part within the architecture,not the fastest one.

Razor1
03-18-10, 09:07 PM
So,the HD5870 has close to 2X the shading power of Fermi in single precision math....You don't hear ATI making a big deal out of it.


In the end,what matters is how it performs overall,and that involves all the parts within a GPU,not just the tesselation ability,and any given GPU is only as fast as the slowest part within the architecture,not the fastest one.

ATi does make a big deal out of it from a marketing stand point all their marketing slides make it a point to include gflops.

http://resources.vr-zone.com/image_deposit/up2/12558204291df7f95851.jpg

http://www.legitreviews.com/article/1080/1/

shadow001
03-18-10, 09:35 PM
True for their marketing slides,but that's to be expected from any company.


There not showing it in synthetic benchmarks that are specifically written to use as much shading power as possible,or creating their own tech demos,and really packing on the shader effects.


Nvidia showed us the heaven benchmark results....It's about tesselation.

Nvidia showed the hair demo....Yup,it's about tesselation + physics.

Nvidia showed the water demo....Bingo,more tesselation.

Nvidia showed that rocket sled demo,and it's also about tesselation + physics.

Nvidia showed that grass demo....Again,more tesselation.


Do you notice a pattern here?.....Real game results are what's needed,not showing us stuff that might get used extensively in 2~3 years to the extremes that their demos show it,which by then,much faster hardware will exist,so no one will care about either fermi or Cypress....They'll be very old news by then.

Xion X2
03-18-10, 10:15 PM
Just been reading the white paper. I says GF100 has 6x the tesselation power of the 5870 in DX11 apps.

6x... the tesselation power.. :lol:

Unigine Heaven DX11 (8xAA)

* ATI Radeon HD 5870: 23 FPS
* GeForce GTX 470: 20 FPS

http://translate.google.com/translate?js=y&prev=_t&hl=fi&ie=UTF-8&layout=1&eotf=1&u=http%3A%2F%2Fwww.heise.de%2Fnewsticker%2Fmeldung %2FNvidias-Fermi-Leistung-der-GeForce-GTX-470-enthuellt-946411.html&sl=de&tl=en

I wonder how often white papers translate to real world performance. Show me the money. This is even better than ATI hyping their shading prowess.

ATI has been doing tesselation since the 2900-series. Even had a demo of Ruby flying over some mountains while showing it off. They're not likely to get caught with their pants down in this area, as the tesselation-based bench above shows.

I'll be glad when the real benchmarks finally come out so we can toss all these theoreticals out the window.

shadow001
03-18-10, 11:09 PM
6x... the tesselation power.. :lol:

Unigine Heaven DX11 (8xAA)

* ATI Radeon HD 5870: 23 FPS
* GeForce GTX 470: 20 FPS

http://translate.google.com/translate?js=y&prev=_t&hl=fi&ie=UTF-8&layout=1&eotf=1&u=http%3A%2F%2Fwww.heise.de%2Fnewsticker%2Fmeldung %2FNvidias-Fermi-Leistung-der-GeForce-GTX-470-enthuellt-946411.html&sl=de&tl=en

I wonder how often white papers translate to real world performance. Show me the money. This is even better than ATI hyping their shading prowess.

ATI has been doing tesselation since the 2900-series. Even had a demo of Ruby flying over some mountains while showing it off. They're not likely to get caught with their pants down in this area, as the tesselation-based bench above shows.

I'll be glad when the real benchmarks finally come out so we can toss all these theoreticals out the window.



In practical terms,the strong points from either architecture,be it shader power on ATI's side,or tesselation ability for Fermi,will never get used fully by developers for practical reasons.

If a game was designed to use extreme amounts of tesselation,to the point where the performance differencial would be substancial between Fermi and cypress,and would be playable on Nvidia's hardware,while it couldn't on ATI's hardware,developers wouldn't even think about it like that,as their primary concern is making their game run on as many different cards as possible,as it means more game sales in the end...Good old fashion greed.


And on the other side of the coin,the same situation applies for Cypress extra shader power....It'll never be exploited fully in a game,if it means that Fermi is struggling to keep up on that end of things,and again,developers want to make their game run on all cards.


So judging either architecture for each one's specific strong point,is an exercise in futility,and it's how the the whole package works in overall efficiency that matters the most in the end.


BTW,it seems unigine are coming out with a revised version of the heaven demo(version 2.0 i think),featuring even more objects on screen and pushing even higher tesselation workloads than the current versions do(1.0 and 1.1).

Enrico_be
03-19-10, 12:51 AM
http://translate.google.com/translate?hl=en&sl=auto&tl=en&u=http%3A%2F%2Fwww.expreview.com%2F9888.html

http://www.we-wi.be/Images/xfx_01.jpg
http://www.we-wi.be/Images/xfx_03.jpg
http://www.we-wi.be/Images/xfx_04.jpg
http://www.we-wi.be/Images/xfx_05.jpg

Sexy cards :captnkill:

Enrico_be
03-19-10, 12:57 AM
[rumor-mode]
If these are true : http://translate.google.be/translate?hl=nl&sl=zh-CN&tl=en&u=http%3A%2F%2Fnews.mydrivers.com%2F1%2F159%2F1590 81.htm

Then,

GTX 480 :
Cuda cores = 480
Core freq = 700MHz
Shader units 1401MHz, 1848MHz memory (equivalent 3696MHz)
bandwidth = 177.4GB /s

GTX 470 :
Cuda cores = 448
Core freq = 607 Mhz
Shader units = 1215/1674MHz (equivalent 3348Mhz)
Bandwidth = van 133.9GB /s
[/rumor-mode]

If true, nice :cool: !
GTX 480 will have double the amount of a GTX280 CUDA cores so they have stick to their promise it seems, I know people expected 512 somehow but hey look @ what nvidia said for months on their official geforce Fermi page :) (-> http://www.nvidia.co.uk/object/gf100_uk.html "Packing in 3 billion transistors, double the CUDA cores of previous generation GPUs")
If i turns out to be 512 afterall, even better but I think that's for the professional market (Tesla)

Xion X2
03-19-10, 01:15 AM
So judging either architecture for each one's specific strong point,is an exercise in futility,and it's how the the whole package works in overall efficiency that matters the most in the end.

Yep.


Nice looking cards. I like the look of the 470 better from those pics (don't care for the external heatpipes.) I'll probably end up purchasing one of them for my renderbox.

shadow001
03-19-10, 01:26 AM
[rumor-mode]
If these are true : http://translate.google.be/translate?hl=nl&sl=zh-CN&tl=en&u=http%3A%2F%2Fnews.mydrivers.com%2F1%2F159%2F1590 81.htm

Then,

GTX 480 :
Cuda cores = 480
Core freq = 700MHz
Shader units 1401MHz, 1848MHz memory (equivalent 3696MHz)
bandwidth = 177.4GB /s

GTX 470 :
Cuda cores = 448
Core freq = 607 Mhz
Shader units = 1215/1674MHz (equivalent 3348Mhz)
Bandwidth = van 133.9GB /s
[/rumor-mode]

If true, nice :cool: !
GTX 480 will have double the amount of a GTX280 CUDA cores so they have stick to their promise it seems, I know people expected 512 somehow but hey look @ what nvidia said for months on their official geforce Fermi page :) (-> http://www.nvidia.co.uk/object/gf100_uk.html "Packing in 3 billion transistors, double the CUDA cores of previous generation GPUs")
If i turns out to be 512 afterall, even better but I think that's for the professional market (Tesla)


Same numbers as in this VR-zone article btw:


http://vr-zone.com/articles/nvidia-geforce-gtx-480-final-specs--pricing-revealed/8635.html


We have received the final specs for GeForce GTX 480 and 470 which are slated for official launch on March 26th 7pm US time. Coincidentally, an official photo of the GTX 480 has been leaked in our forums here : http://forums.vr-zone.com/news-around-the-web/586418-finally-official-picture-geforce-gtx-480-a.html

GeForce GTX 480 : 480 SP, 700/1401/1848MHz core/shader/mem, 384-bit, 1536MB, 295W TDP, US$499

GeForce GTX 470 : 448 SP, 607/1215/1674MHz core/shader/mem, 320-bit, 1280MB, 225W TDP, US$349

* Sorry, it should be 480 SP. My source got a little mixed up. Updated with final clocks too.

Internal benchmarks reveal that GeForce GTX 470 is some 5-10% faster than Radeon HD 5850 and similiar for GeForce GTX 480 over the Radeon HD 5870. Interestingly, the TDP of GeForce GTX 480 is almost similar to Radeon HD 5970 which is a dual GPU card. Interestingly, our sources revealed that there are indeed plans for dual Fermi cards and the TDP of the card is probably gonna be mind blowing.


About the only scary thing is the thermal dissipation,which is listed at 295 watts for the gtx 480,so we know why the cooling is built with those 4 large heatpipes that are partially exposed,but i can only imagine a dual Fermi card being built when Nvidia makes the transition to the 28nm fab process late this year or early next.


Now figure on people putting 2 or 3 of these cards together for dual and triple SLI setups,and it makes sense why Nvidia helped thermaltake design a custom PC case just for it....It's going to get a little toasty in there to say the least.

jlippo
03-19-10, 01:40 AM
6x... the tesselation power.. :lol:

Unigine Heaven DX11 (8xAA)

* ATI Radeon HD 5870: 23 FPS
* GeForce GTX 470: 20 FPS

http://translate.google.com/translate?js=y&prev=_t&hl=fi&ie=UTF-8&layout=1&eotf=1&u=http%3A%2F%2Fwww.heise.de%2Fnewsticker%2Fmeldung %2FNvidias-Fermi-Leistung-der-GeForce-GTX-470-enthuellt-946411.html&sl=de&tl=en

I wonder how often white papers translate to real world performance. Show me the money. This is even better than ATI hyping their shading prowess.
Well, you really wanted to chose a case where you can the shader limitation, that would be it.
If you want to check tesselation limitations heres a nice one.
http://forum.beyond3d.com/showpost.php?p=1404586&postcount=3117
640x400, low shaders with tesselation 68.7fps, there seems to be a limitation somewhere else than shaders this time.
Games which do lots of shadows will have problems with this kind of bottlenecks as you need to render all those tesselated meshes to shadow maps, again and again.

Having a lot of shading power is good, but it doesn't help as you have bottleneck before you get to use that power.

Muppet
03-19-10, 01:53 AM
http://translate.google.com/translate?hl=en&sl=auto&tl=en&u=http%3A%2F%2Fwww.expreview.com%2F9888.html

http://www.we-wi.be/Images/xfx_01.jpg
http://www.we-wi.be/Images/xfx_03.jpg
http://www.we-wi.be/Images/xfx_04.jpg
http://www.we-wi.be/Images/xfx_05.jpg

Sexy cards :captnkill:

Those look pretty cool. God i'm getting excited about this launch. Bring on the cards. (nana2)(nana2)(nana2)

Enrico_be
03-19-10, 03:01 AM
http://www.shanebaxtor.com/2010/03/19/gtx-400-series-nda-looks-set/

Im about to jump on a con call with NVIDIA but I just got the usual brief about the con call and it cleraly says:

NDA embargo is 7:01PM EST 3/26 (=10:01AM Sydney 3/27). No leaking!

So it looks like everything is good to go on the 26th of March.

Its pretty safe to assume though that the information regarding the card hitting retail on April 9th is still correct.