Bookmark and Share


Nvidia Corp. said it had learnt the lessons with 40nm process technology and would not repeat its mistakes with the 28nm fabrication process. The company spent time on learning the peculiarities of the 28nm manufacturing technology from Taiwan Semiconductor Manufacturing Company and already has working 28nm chips.

"We are far better prepared for 28nm than we were for 40nm. Because we took it so much more seriously. We were successful on so many different nodes for so long that we all collectively, as an industry, forgot how hard it is. So, one of the things that we did this time around was to set up an entire organization that is dedicated to advanced nodes. We have had many, many tests chips run on 28nm, we have working silicon," said Jen-Hsun Huang, chief executive officer of Nvidia.

TSMC's 40nm manufacturing process was plagued with low yields conditioned by problems with the technology itself, manufacturing issues and design errors. Nvidia started to develop its 40nm lineup without knowing about the potential issues and when it ran into a set of unexpected problems it had to delay launch of its key-products based on Fermi architecture. With 28nm process technology, the company seems to be better prepared.

"[Our experience with 28nm] is looking really good, it is looking much better than our experience with 40nm. It is just a comprehensive, across-the-board engagement between TSMC and ourselves making sure that we are ready for production ramp when the time comes. So I feel really good about 28nm," said Jen-Hsun Huang.

Nvidia said earlier this month that it plans to test-drive its next-generation Kepler graphics processing unit (GPU) this year and introduce the new chips commercially in 2012.

Kepler is Nvidia's next-generation graphics processor architecture that is projected to bring considerable performance improvements and will likely make the GPU more flexible in terms of programmability, which will speed up development of applications that take advantage of GPGPU (general purpose processing on GPU) technologies. Some of the technologies that Nvidia promised to introduce in Kepler and Maxwell (the architecture that will succeed Kepler) include virtual memory space (which will allow CPUs and GPUs to use the "unified" virtual memory), pre-emption, enhance the ability of GPU to autonomously process the data without the help of CPU and so on.

Tags: Nvidia, 28nm, Kepler, Geforce, Tesla, Tegra, Semiconductor


Comments currently: 21
Discussion started: 08/12/11 05:01:08 AM
Latest comment: 08/16/11 09:45:37 AM
Expand all threads | Collapse all threads


show the post
7 10 [Posted by: Memristor  | Date: 08/12/11 05:01:08 AM]
- collapse thread

Really now? Despite Fermi launching 6 months after HD58xx series, it was faster by a good 15%. Which was the best bang for the buck card all of last summer? GTX460.

Obviously when AMD is giving away HD6950s which unlock into 6970s, how can you expect to compete with such a loss leader strategy? NV is laughing to the bank with 51% profit margins while AMD's division continues to barely stay afloat with R&D investment taking them into the red. Unlike AMD, NV is actually able to release competitive products and still make $$$. It's pretty easy to be "competitive" when you are selling your $400 GPUs for $250 and making no profits!!

NV still retained 60% market share lead on the desktop despite being late with Fermi. GTX570 is as good as the HD6970 at 1080P, GTX560 Ti easily competes with the 6950 and yet NV's architecture is far more forward looking in both GPGPU and Tessellation performance. NV has so far produced a faster single GPU in each of the last 3 consecutive generations:

1) 8800GTX > HD2900XT
2) GTX280/285 > HD4870/4890
3) GTX480/580 > HD5870/6970

Also, considering major GPU releases have slowed down significantly (i.e., we haven't had a large performance increase since HD5870 in September of 2009), being 4-5 months late is nothing. AMD and NV both have confirmed longer GPU cycles (18-24 months from the previous 12-15 month cycles). In other words, future generation GPUs will continue to sell for another 1.5-2 years. So being first by 3-4 months isn't as critical today as it was say 3-5 years ago when GPU cycles were much shorter.

Maybe you should talk about the last time AMD had a faster single GPU product - X1950XTX launched on August 23, 2006.

Let's not even bring up the fact that AMD is ditching VLIW-4 in favor of a scalar architecture which Fermi has already adopted. It's pretty clear that NV had the right vision of GPGPU architectures being the future all along.
3 4 [Posted by: BestJinjo  | Date: 08/12/11 09:49:55 PM]
show the post
3 8 [Posted by: Memristor  | Date: 08/13/11 07:35:57 AM]
show the post
0 4 [Posted by: TAViX  | Date: 08/13/11 12:17:12 PM]
Fermi has 3x the die size, and ~20% more power usage.It is on average 5-10% faster, but it depends on the game. They suffered huge yield issues with chips that cost far more to make. I really don't see how they made any money selling at similar prices. For fastest card, they lost indisputably (with 590 and 6990), due to having to severely under-clock the cores due to insane power usage. They tried to say they were faster, but everyone knew they were lying.
1 2 [Posted by: xeridea  | Date: 08/13/11 04:50:47 PM]
Dude, cut back on the coffee!! To begin with, nV is not producing 51% profit margins. You may want to read their P&L statements. Secondly, nV wanted to sell their products (read: 480) b/w $600-$650 in order to make some money, but instead, took a loss selling their 480s due to cost and yield. Your statements about the AMD line are simply untrue. AMD made GPUs that the consumer/gamer wanted; not some obscure GPU programmer making tiny little niche programs that most people cannot use. Thirdly, their so called GPU power has yet to be realized in the real world. Intel's quicksync obliterates nVs video encoding with smaller instruction sets with better results (i.e. superior video quality). Moreover, when it comes to bitcoin GPU programming, AMD GPUs are 500% faster than nV GPUs. In fact, a 5870 is faster in bitcoin GPU performance than a quad GPU 590 setup!!

Yes, nV had some interesting ideas, but they were poorly executed that didn't really benefit the consumer. The original 480 was a mess - plain and simple. It was hot, used an obscene amount of power and sounded like a shopvac. These products, of what's left, are selling for $200 in retail stores. Do you think retailers and nV are still making money? nV made their money on their low end GPUs and ultra high end professional cards. They took a hit with their flagship products.

nV no doubt is an innovator, but "has the potential" is not the same as "can do now". To think that AMD cannot bring/design a product with unbelievable GPU compute power to market quickly is naive. You're dealing with a company that is the marriage of CPU and GPU technologies; something that nV could only dream of, and as of the outcome of the previous law suit, nV no longer will ever have access to.

The bottom line is, nV needs to convince people that their GPU design is beneficial to them. Intel has quashed the need for video GPU transcoding, and there are few truly useful nV GPU compute programs available to the end user. It's not by serendipity that nV is clinging on to ARM as a way out. PS - just an aside. How long has nV not had a sophisticated multi-monitor setup? Ans. They still don't.
1 1 [Posted by: KittyMonster  | Date: 08/16/11 09:45:37 AM]

show the post
2 13 [Posted by: Memristor  | Date: 08/12/11 05:01:45 AM]

show the post
0 15 [Posted by: Memristor  | Date: 08/12/11 05:01:57 AM]

show the post
2 5 [Posted by: BernardP  | Date: 08/12/11 06:07:25 PM]
- collapse thread

Maybe in price/performance and efficiency. 2GB of VRAM was also a great move. But GTX470 > HD5850 in performance and GTX480 > HD5870 in performance. Similarly NV is making bank with the $450-500 GTX580 while GTX570 is easily able to hang with AMD's flagship HD6970. Last generation GTX275 was easily competitive with the 4890 and GTX260 216 wasn't any slower than the HD4870. I would say NV has been in there all along, while offering the superior single GPU card at the top if you were willing to pay the price for it. Also, the GTX590 is in stock and doesn't sound like a space shuttle launch, unlike the HD6990 which has been MIA most of its life.

AMD's strategy continues to be undercutting NV which is great for gamers, but is awful financially for the firm. AMD's stock performance over the last 5 years (since purchase of ATI) is a -68%. No wonder when you are selling your HD6970 "repackaged" as a $225-250 HD6950 2GB.... How do you then expect to sell your flagship 6970 for $325-350?? lol.

Sure, they might sell the HD7970 for $349 again vs. a $500 GTX680. 2-3 more of such generations and soon AMD's graphics division will be deep in the red with such a strategy.
1 0 [Posted by: BestJinjo  | Date: 08/12/11 10:06:49 PM]

I'd love to see one of these companies release a 28 nm GPU with 1 GHz stock clock speed... then again, we still don't have consumer CPUs that run at 4 GHz stock...
4 0 [Posted by: DirectXtreme  | Date: 08/12/11 06:46:23 PM]

They actually can make cheaper cards because Nvidia chips have a larger die area than ATI chips 389 mm² for the HD 6970 versus 520 mm² for the GTX 570/580.

And yes the purchase of ATI was expensive and yes it had to be written of substantially. But this was all done with long-term goals in mind. Such as fusion where we've seen the first proof of concept in form of Llano.
1 0 [Posted by: Tinus  | Date: 08/13/11 02:24:29 AM]

Bro, are you 17-20 years old? In all your comments you seem to be very childish, you are insulting people over GPUs...WTF?! Get a girlfriend, hang out with your friends (if you have any), go out in to nature, or something, but stop being so goddamn immature, it's just simple hardware we are talking about.
3 5 [Posted by: 63jax  | Date: 08/13/11 08:07:44 AM]
- collapse thread

show the post
3 6 [Posted by: Filiprino  | Date: 08/13/11 09:51:21 AM]

8. the end will matter only the company that will have the best price/performance/consumption card. Wee need strong competition in order to have a price war. To bad we don't have a 3rd competitor. God, I miss 3DFX, Matrox, S3 and PowerVR...
2 1 [Posted by: TAViX  | Date: 08/13/11 12:22:50 PM]
- collapse thread

Miss 3DFX ads!

But they deserved to die for arrogantly thinking they knew better than everyone else, standards and consumers alike. I definitely won't be looking forward to a return of GLide vs OpenGL.
0 0 [Posted by: jihadjoe  | Date: 08/13/11 03:52:19 PM]

ATI is evil, NVidia is good. I matters not what you think about ATI products, if you just want to be evil.
0 2 [Posted by: TeemuMilto  | Date: 08/16/11 08:41:51 AM]


Add your Comment

Related news

Latest News

Wednesday, November 5, 2014

10:48 pm | LG’s Unique Ultra-Wide Curved 34” Display Finally Hits the Market. LG 34UC97 Available in the U.S. and the U.K.

Wednesday, October 8, 2014

12:52 pm | Lisa Su Appointed as New CEO of Advanced Micro Devices. Rory Read Steps Down, Lisa Su Becomes New CEO of AMD

Thursday, August 28, 2014

4:22 am | AMD Has No Plans to Reconsider Recommended Prices of Radeon R9 Graphics Cards. AMD Will Not Lower Recommended Prices of Radeon R9 Graphics Solutions

Wednesday, August 27, 2014

1:09 pm | Samsung Begins to Produce 2.13GHz 64GB DDR4 Memory Modules. Samsung Uses TSV DRAMs for 64GB DDR4 RDIMMs

Tuesday, August 26, 2014

10:41 am | AMD Quietly Reveals Third Iteration of GCN Architecture with Tonga GPU. AMD Unleashes Radeon R9 285 Graphics Cards, Tonga GPU, GCN 1.2 Architecture