Nvidia may be in trouble according to this article

[shadow001]

Originally Posted by LydianKnight

Relax... I was just laughing to the Troll2 poster, not your post or Johny C's remark :P

In regards of your quote from the Guru3D's article, well... every of you know I'm not quite valid to say anything valid in what seems to be the eternal 'ATI vs NVIDIA' fight, because, frankly speaking, I don't give a f'ck about it...

512 cores, 550-575MHz, more than enough for me, I just want raw performance to play my current games and some future ones, I don't pretend neither going 100+ FPS in any game or having the best human invention ever since sliced bread...

Anyway, I'm beginning to think: 'if NVIDIA is supposed to have a refresh for fall-winter, and they're going to release that late, what's the real need to spend my hard earned money if 6-8 months later I'll have something even better?'...

Let's hope this 'current' Fermi doesn't dissapoint me, specs-wise, or... I think I'll do what I'm thinking right now...

It was just a general statement anyhow,and i found Guru 3d's comments interesting in the sense that even as the initial performance gap between Cypress and Fermi was supposed to be about 60% faster for Fermi,and as the delays increased with the latter,the gap kept getting smaller and smaller over time,to the point where if we take the comments from the site at face value,and it pretty much matches with charlie's own comments on the matter on his latest article,even for clock speeds,it pretty much implies that it's going to be a competitive fight between Cypress released 6 months ago,and fermi released in a month from now afterall,not the assrape of the century as some here were expecting.


I know i shouldn't say it,but that would make those who waited for Fermi,bought into Nvidia's bull hook,line and sinker,and don't really have a need for it's GP-GPU abilities look quite dumb in the end,and further emphasizes that whoever releases their hardware first usually wins....Just like it happened with Nvidia on many occasions when they were the first out with their hardware,and it was other companies trying to catch up to them.


Release dates matter when there's fierce competition on the market is my main point overall.

[shadow001]

Originally Posted by DarthBeavis

the two companies do not have control over when they can release to market. this should be a lesson on that point. if they have staggered releases that does not hurt the market. people buy what they want at the time they are ready. if you want a single gpu that can run three to six monitors then go AMD. if you want kick arse 3d stereo on larger monitors or multipe monitors go Nvidia. if you want the GP-GPU capabilties go Nvidia.

They do to a certain extent by the actual design of the chip itself in terms of it's transistor budget relative to the actual performance delivered,and even the overall focus of the chip itself,since Nvidia is trying to make the both the best performing GPU for graphics,as well as a much more powerfull GP-GPU to create/run applications using Cuda,and there are differences in the workloads between both,so it's hard to reach a compromise that makes it perform optimally between both,and here's an example:


Fermi supports an ECC memory controler,which is great for GP-GPU applications,specifically mission critical ones(banking,scientific research,etc..),but useless for a gaming GPU,and that eats up extra transistor budget for it inside the die,and there's more examples like that if you read up on the Fermi technical papers.


At the end of the day,we have Cypress at 2.15 billion transistors,giving grief to a Fermi chip packing 3 billion,and most would think that on the transistor budget alone,Cypress shouldn't have a chance in hell,but well it looks like the competition between both is going to be tight afterall,regardless of the 850 million transistor difference.

[shadow001]

Here's another example for Fermi....Memory bandwith.


Fermi uses as 384 bit memory bus,and like ATI,also supports GDDR5 memory,meaning that if both are using memory clocked at the same speed,Fermi always has a 50% memory bandwith advantage right from the start,and that obviously comes in handy in any game situation where it's primarily memory bandwith bound(heavy use of AA for instance)

And again,having that wider bus means adding 2 additional 64 bit memory controlers inside the Fermi die,further increasing the transistor budget,making the chip larger and use more power,but more importantly,how is Cypress able to keep up using only a 256 memory bus,assuming what chalie and Guru 3D said are accurate?.


It shouldn't be possible unless Cypress is exceptionally well focused on primarily running GPU workloads in games as it's highest priority and GP-GPU is a distant second.

[shadow001]

Originally Posted by DarthBeavis

shadow how much of this stuff do you think will make a PERCEPTABLE gaming difference? Nvidia wants to not just focus on pure speed but also additional functionality that they feel will enhance games such as physics and 3d stereo. AMD also clearly cares about added functionality to enhance games as evidenced by their push for multi-monitor support for which they should be commended.

What kind of functionality are you speaking of exactly?....Microsoft controls the DX standard and all the features within DX10.DX10.1 or DX11 are mandatory to support in order to obtain certification...The choices GPU designers make that go above and beyond what DX11 asks for is what i'm talking about,and just how much optimizations did Nvidia do towards GP-GPU rather than straight gaming performance is what i ask myself.


ATI mentioned they can do 3D physics with their hardware just fine,but they won't commit to it until there's a unified standard that works for both,which Nvidia is opposed to and the same goes for 3D displays as well....Even in my case,using my current cards,i can buy the upcoming Acer 23.6 inch 120Hz displays,which support 1920*1080 resolution(Hell even 3 of them) and be ready for 3D glasses once there's a unified standard....I've said it so many times before,but here it goes again:


Unified standards always win in the end....What happened with 3D graphics will also happen with physics and will also happen with 3D display technology,it's not a matter of if,it's a matter of when,and the only advantage with Nvidia is that you can do it right now.

As for the tesselation demo you linked to....Here's my result:




Not exactly a slouch with tesselation and 4X AA on top.

[shadow001]

For those wondering:

http://www.techspot.com/news/37784-a...d-monitor.html


My main reason is to basically use the same display to watch blue ray movies at their native 1080P resolution(1920*1080)as well as gaming and eventually use the 3D glasses for both games and watching movies on the same display.


Unless going for extremely large home theatre panels,we're still limited to 22" monitors for Nvidia's 3D display tech,which display 1650*1080 resolutions.

[shadow001]

Originally Posted by DarthBeavis

those panels work NOW with Nvidia 3d

I'm fully aware of that,but i can still buy 3 of them to enjoy the wider field of view in gaming and a much larger desktop as well,and view blue ray movies at their native resolution,while being ready for a unified 3D display standard,which seems to require 120Hz refresh no matter what the standard ends up beign,which also applies to movies too,not just gaming.


At 399$,they aren't that cheap but that's the price you pay for 120Hz and being the first 1920*1080 models on the market.

[Rollo]

Originally Posted by shadow001

I'm fully aware of that,but i can still buy 3 of them to enjoy the wider field of view in gaming and a much larger desktop as well,and view blue ray movies at their native resolution,while being ready for a unified 3D display standard,which seems to require 120Hz refresh no matter what the standard ends up beign,which also applies to movies too,not just gaming.


At 399$,they aren't that cheap but that's the price you pay for 120Hz and being the first 1920*1080 models on the market.

If you like 3d, three of them is a pretty amazing experience.

As I have a 2233RZ already, I'm going to grab two more. (which will cut my cost of entry into 3d Surround from $1200 to $600)

[shadow001]

Originally Posted by Rollo

If you like 3d, three of them is a pretty amazing experience.

As I have a 2233RZ already, I'm going to grab two more. (which will cut my cost of entry into 3d Surround from $1200 to $600)

The main issue really comes down to the 3D glasses and software support,not just for 3D games,but also viewing 3D blue ray movies at their native resolution,using the same glasses for both.


Nvidia can make sure it'll work for games,but there's still no way to be sure for 3D blue ray movies yet,and i don't feel like buying 2 different sets of glasses + software depending on the application,and at a basic fundamental level,the hardware requirements are the same for both,hence why i'm so insistent on a unified standard.

[turdhat]

I just dont dig 3d in games at this point and abandoned my glasses and setup and purchased a 55" 1080p 240hz (fake 240hz but still cool) Toshiba LCD as my main monitor and use comp from the couch with wireless gear. I have seen most of the different 3d tech and to me it makes cars in driving games look like fake models and people look like action figures. It is neat for a minute but for me it lost its appeal very quickly.

Movies are a whole different deal though. Maybe when graphics improve a bit more I will dig it then and get the needed gadgets.

[Enrico_be]

(Recent Semi-accurate forum topic)
- [Intelligence from China] Fermi-based Tesla cards are very rare, even among AICs.

[shadow001]

Originally Posted by Enrico_be

(Recent Semi-accurate forum topic)
- [Intelligence from China] Fermi-based Tesla cards are very rare, even among AICs.

Ouch if true....Means the lauch at Pax is a paper one basically,and only the GTX470 will be available for the most part.

[Muppet]

Originally Posted by shadow001

The main issue really comes down to the 3D glasses and software support,not just for 3D games,but also viewing 3D blue ray movies at their native resolution,using the same glasses for both.


Nvidia can make sure it'll work for games,but there's still no way to be sure for 3D blue ray movies yet,and i don't feel like buying 2 different sets of glasses + software depending on the application,and at a basic fundamental level,the hardware requirements are the same for both,hence why i'm so insistent on a unified standard.

From what I have read on the Nvidia Forums. All the major software players will support Nvidia's 3d Solution for playing BD movies. IE: Arcsoft TotalMedia Theatre, Corel WinDVD, Cyberlink PowerDVD Ultra and Sonic Roxio CinePlayer BD will all support 3D Vision for Blu-ray 3D playback.