Overclock.net banner

1 - 20 of 23 Posts

·
Premium Member
Joined
·
2,328 Posts
Discussion Starter #1
Quote:


NVIDIA needs zero defects from its foundry partners, particularly in the vias on its leading-edge graphics processors, said John Chen, vice president of technology and foundry operations at the GPU powerhouse. With 3.2 billion transistors on its 40 nm graphics processor now coming on the market, the 7.2 billion vias have become a source of problems that the industry must learn to deal with, Chen said in a keynote speech at IEDM.

As you may know, NVIDIA’s upcoming Fermi GPU has over 3 billion transistors and they are looking to their own future with even more massive GPUs. The entire speech is fascinating and we found this particularly interesting because we predicted it in September, 2008:

SOURCE
 

·
Premium Member
Joined
·
65,162 Posts
More interesting... 3D chips.
 

·
Registered
Joined
·
509 Posts
Quote:

Originally Posted by Licht View Post
Zero defects... if that is their standard then the chip's yields will be nonexistent.
lol this

more OT: I remember reading about ibm's plans for 3d chips and i really fail to see how anyone is going to push them into reality.
honestly we have problems trying to cool 2d chips right now, even though the fab is getting smaller and smaller the coolers seem to be getting bigger and bigger, just look at the 5870.....
Only way i see this happening is if we make extreme advances in vapor cooling methods and stack them in layers between the chips...maybe they can use those ever resourceful carbon nano tubes to do the job
 

·
Registered
Joined
·
142 Posts
I predict that this will be very trouble-some for NVIDIA and will be a test for them.

Succeed: become top dog again
Fail: end to fermi and nvidia
 

·
Registered
Joined
·
4,845 Posts
Quote:

Originally Posted by cypher\his View Post
lol this

more OT: I remember reading about ibm's plans for 3d chips and i really fail to see how anyone is going to push them into reality.
honestly we have problems trying to cool 2d chips right now, even though the fab is getting smaller and smaller the coolers seem to be getting bigger and bigger, just look at the 5870.....
Only way i see this happening is if we make extreme advances in vapor cooling methods and stack them in layers between the chips...maybe they can use those ever resourceful carbon nano tubes to do the job

What are you talking about?


Quote:

Originally Posted by stargate125645 View Post
OK, what the heck is a "via?"
http://en.wikipedia.org/wiki/Via_%28electronics%29

Quote:
In printed circuit board design, vias are pads with plated holes that provide electrical connections between copper traces on different layers of the board. The holes are made conductive by electroplating, or are filled with annular rings or small rivets. High-density multi-layer PCBs may have microvias: blind vias are exposed only on one side of the board, while buried vias connect internal layers without being exposed on either surface. Thermal vias carry heat away from power devices. They are typically used in arrays of about a dozen vias.

In integrated circuit design, a via is a small opening in an insulating oxide layer that allows a conductive connection between different layers. A via on an integrated circuit is often called a through-chip via. A via connecting the lowest layer of metal to diffusion or poly is typically called a "contact".
 

·
Bundling Bundler
Joined
·
1,881 Posts
Even if all does not go well with GF100 yields for nVidia, I highly doubt that it would be the end of the company. They would just have to completely restructure their business model and really start to innovate. Look what happened with ATI. They almost bit the dust and came back strong. They are still digging themselves out at this point but at least they have firm base on which to further build their success.

As must as I dislike nVidia's business practices and holier than thou attitude, I would never want to see them disappear from the GPU scene. No competition means bad things to the consumer market. Look at ATI's current pricing of the 58XX series. Imagine if nVidia simply couldn't get fermi to market, that would almost guarantee no price drop. ATI could then charge whatever premium they want for their new lineup. I'm not saying they would, but you never know.
 

·
Premium Member
Joined
·
5,898 Posts
That would be pretty amazing. If they can make a chip like this and eliminate the need for internal buses, then it open the way for actually much smaller VRAM needs and incrediable increases in performance for otherwise a smaller cost. I imagine a VRAM size of 1gb being far more then enough for generations if the memory can be loaded in fractions of a milisecond with much reduced latencies. Not to mention this would be a god send for Tesla cards, becuase they could afford to allocate more traffic to the already existing VRAM chips without the need to instead pack more densitiy into cache hierarchy. I wish Nvidia the best of luck with this one.

Are they planning this technology for the version of Fermi to be released early next year, or is this a produce goal for future GPU?
 

·
Premium Member
Joined
·
4,026 Posts
Quote:


Originally Posted by Pen
View Post

I predict that this will be very trouble-some for NVIDIA and will be a test for them.

Succeed: become top dog again
Fail: end to fermi and nvidia

NVIDIA wouldn't be the top dog. Sure they'd have the two most powerful cards, but what good is that when ATI are pretty much better in every other segment of the market?
 

·
Premium Member
Joined
·
5,898 Posts
Quote:


Originally Posted by MrAlex
View Post

NVIDIA wouldn't be the top dog. Sure they'd have the two most powerful cards, but what good is that when ATI are pretty much better in every other segment of the market?

In the HPC market they would be unchallenged, and with software and compiler development could start to force both intel and AMD out of servers. I would say that Nvidia's goal seem pretty clear relative to AMD/ATI's. This technology would also have a lot of impact on gaming, VRAM has to be refreshed many, many times at any given moment. This advancement would be like packing GDDR10 on to the board in many ways, but with much less latency and still incredible bandwidth.
 

·
Premium Member
Joined
·
4,026 Posts
Quote:


Originally Posted by mothergoose729
View Post

In the HPC market they would be unchallenged, and with software and compiler development could start to force both intel and AMD out of servers. I would say that Nvidia's goal seem pretty clear relative to AMD/ATI's. This technology would also have a lot of impact on gaming, VRAM has to be refreshed many, many times at any given moment. This advancement would be like packing GDDR10 on to the board in many ways, but with much less latency and still incredible bandwidth.

Don't get me wrong, this is going to be one hell of a thing, but I'm talking about the company as a whole. ATI control pretty much every other segment of the market. Every $10/20 there's a different card.
 

·
Registered
Joined
·
2,889 Posts
Quote:


Originally Posted by stargate125645
View Post

OK, what the heck is a "via?" Or does the article mean the silicon company? If so, poorly edited article.

A via is a vertical interconnect for two or more horizontal interconnects. Look at any PCB and look for the tons of small little 'holes' in/at end of traces. Those are vias. The same tech is used for silicon, interconnecting copper traces on the silicon substrate.
 

·
Premium Member
Joined
·
65,162 Posts
Quote:


Originally Posted by AMD_Brewer
View Post

Hrm Zero Defects? Obviously the author did not speak with any Material Engineer. There will always be defects in the production of any semiconductor.

Of course, but some parts are more prone to defects than others.

With enough VIAs and smarts, any minor defect can be rerouted around (like existing semiconductor tricks to improve yields).
 

·
Premium Member
Joined
·
2,328 Posts
Discussion Starter #18
Quote:


Originally Posted by AMD_Brewer
View Post

Hrm Zero Defects? Obviously the author did not speak with any Material Engineer. There will always be defects in the production of any semiconductor.

read the article again. the author didn't say that, Nvidia's Chen said that he wants that.
 

·
Premium Member
Joined
·
4,026 Posts
Quote:


Originally Posted by fraudbrand
View Post

If Nvidia fails...then Intel will gobble them up.
So lets all hope for the best. It's bad enough that AMD owns ATI.

Sorry but that's rubbish. It's because of AMD that ATI is where it is now.
 

·
Registered
Joined
·
1,077 Posts
Quote:

Originally Posted by MrAlex View Post
Sorry but that's rubbish. It's because of AMD that ATI is where it is now.
Not do shatter your illusion or anything, but ATI had HD3 series in the works long before they merged with AMD. Since 4th and 5th series are just more of everything (in a case of 5th series dx11 support plus new tech node are minor tweaks).

ATI's next gen will probably the first to be developed after merger with AMD given the several year chip design cycle.
 
1 - 20 of 23 Posts
Top