Bookmark and Share


Chief executive officer of Nvidia Corp. said that besides continuously increasing capital expenditures that the company ran into in the recent months will be accompanied by lower than expected gross margins in the forthcoming quarter. The company blames low yields of the next-generation code-named Kepler graphics chips that are made at TSMC’s 28nm node.

“Decline [of gross margin] in Q1 is expected to be due to the hard disk drive shortage continuing, as well as a shortage of 28nm wafers. We are ramping our Kepler generation very hard, and we could use more wafers. The gross margin decline is contributed almost entirely to the yields of 28nm being lower than expected. That is, I guess, unsurprising at this point,” said Jen-Hsun Huang, chief executive officer of Nvidia, during a conference call with financial analysts.

Nvidia’s operating expenses have been increasing for about a year now: from $329.6 million in Q1 FY2012 to $367.7 million in Q4 FY2012 and expects OpEx to be around $383 million in the ongoing Q1 FY2013. At the same time, the company expects its gross margins in Q1 FY2013 to decline below 50% for the first time in many quarters to 49.2%.

Nvidia has very high expectations for its Kepler generation of graphics processing units (GPUs). The company claims that it had signed contracts to supply mobile versions of GeForce “Kepler” chips with every single PC OEM in the world. In fact, Nvidia says Kepler is the best graphics processor ever designed by the company.

[With Kepler, we] won design wins at virtually every single PC OEM in the world. So, this is probably the best GPU we have ever built and the performance and power efficiency is surely the best that we have ever created,” said Mr. Huang.

Unfortunately for Nvidia, yields of Kepler are lower than the company originally anticipated and therefore their costs are high. Chief exec of Nvidia remains optimistic and claims that the situation with Fermi ramp up was ever worse than that.

“We use wafer-based pricing now, when the yield is lower, our cost is higher. We have transitioned to a wafer-based pricing for some time and our expectation, of course, is that the yields will improve as they have in the previous generation nodes, and as the yields improve, our output would increase and our costs will decline,” stated the head of Nvidia.

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. Entry-level chips may not get all the features that Kepler architecture will have to often.

Tags: Nvidia, Kepler, 28nm, TSMC, Geforce, Tesla, Quadro


Comments currently: 19
Discussion started: 02/16/12 01:28:11 PM
Latest comment: 02/18/12 07:21:50 PM
Expand all threads | Collapse all threads


Blaming TSMC for yield problems is pretty lame when ATI, Qualcomm and others report no such problems. Looks more like Nvidia is to blame for designing a chip that does not work well on TSMC's 28nm process.
6 4 [Posted by: Memristor  | Date: 02/16/12 01:28:11 PM]
- collapse thread

I dont think its a chip that doesnt work well on 28nm i just think its a chip thats is so large and complexe that its more prone to faults and as a result bad yelds... The smaller mobile and low end parts on 28nm wouldnt bad as bad.
4 2 [Posted by: vid_ghost  | Date: 02/16/12 02:26:26 PM]
show the post
0 4 [Posted by: madooo12  | Date: 02/16/12 04:17:31 PM]
If the defect area is ~0.1% of the wafer (example, it's far less), a hypothetical chip the size of the entire wafer is essentially guaranteed to be defective. Meaning that despite 0.1% defect rate, you would expect nearly a 100% failure rate of the resulting chip. As the chip is reduced in size then the failure rate will converge on the defect rate (0.1%). So yes chip size is absolutely critical for yields and Nvidia's problems largely stem from having a tendency towards super-massive processors rather than engineering stupidities. It's not something they have not experienced before, they just have to wait for TSMC to reduce defect rates.
3 0 [Posted by: sanity  | Date: 02/16/12 04:55:16 PM]
Wafer constraints for TSMC have actually been widely documented by multiple sources (even prior to JHH commenting on them), based on analysts' direct discussions with other companies who have ordered 28nm wafers from TSMC.

1) Mike Bryant, technology analyst with Future Horizons Ltd. has said that foundry Taiwan Semiconductor Manufacturing Co. Ltd. is in trouble with its 28-nm manufacturing process technologies, which are not yet yielding well. Bryant referenced un-named contacts made with multiple companies waiting for designs to be produced by TSMC on 28-nm processes.

2) Bryant’s comments echo those made by Bob Johnson, research vice president at market research firm Gartner in November 2011.
“In 2012, total 28-nm HKMG shipments will not exceed 200,000 300-mm wafers; or less than 4 percent of foundry’ revenue. Shipments won’t start until 2012 and its coming slower than people had previously thought,” Johnson said.–say-execs-

3) TSMC decided to produce chips using 28nm technology process exclusivelly at Fab 15, which will have capacity of more than 100 thousand 300mm wafers per month when fully operational. TSMC began equipment move-in for the phase 1 facility in June 2011 and started volume production of 28nm technology products for customers in October, 2011. The phase 2 of Fab 15 is projected to be completed in Q4 2012 and this will boost total 28nm capacity to reach about 50 thousand wafers per month.

Based on assumption that a 300mm wafer processed using 28nm manufacturing technology costs from $4000 to $5000, the fact that TSMC expects 28nm wafers to account for 1% of its revenue in Q4 2011.

TSMC is currently 28nm wafer constrained and will be until at least Q4 2012 due to high demand for wafers and it experiencing lack of sufficient manufacturing facilities/capacity.

4)TSMC is completely booked - wafer constrained

According to Digitimes TSMC has received enough orders from its major clients such as AMD, Broadcom, MediaTek, MStar Semiconductor, Nvidia and Qualcomm to completely book the foundry’s production capacity for advanced sub-40nm manufacturing processes.

In other words, there is definitely some truth to what JHH has said in the financial call.

That's 4 unrelated sources with similar information about 28nm slow wafer manufacturing ramp-up.

How do yields relate to wafers? Consider that NV generally produces 500-550+ mm^2 high-end GPU chips. Large die sizes tend to have lower yields since they are more complex to manufacture on a new node process, especially in trying to achieve their targeted power consumption, all functional/workable units and clock speeds.

For example, you might get GK104 and GK110 chips with:

1) All functional units are working but unusually high power consumption for a large % of parts;

2) Targeted clock speeds and power consumption, but some broken functional units (i.e., 480 SP GTX480 GF100 vs. a fully functional GF110).

3) Fully workable GK104 and GK110 chips with all functional units, desired clocks, but they may require much higher voltage than you had planned to hit expected clock speeds.

Sure, you could up the GPU voltage to hit your target speeds and sell these cards, but it would skew your power consumption for the final working GPUs if you were to resell it (i.e., and have very high power consumption GT480 style).

In simplest terms, you can definitely see how Nvidia can actually have "yield" problems simply due to the nature of their design methodologies, while companies such as Apple, Qualcomm or AMD wouldn't experience those yield problems on the same 28nm wafers. What many people here perceive as "yield" problems may not be as simple as # of workable chips per wafer. But JHH isn't going to start going into engineering and physics details with analysts. For example, if HD7970 had a targeted clock speed of 1250mhz, but AMD might have decided to launch first and improve yields by lowering the clocks to 925mhz. Well, Nvidia might want the performance crown at all costs and management might not want to lower clocks by 20% to hit targeted yields. They might choose to launch a quarter later at the originally intended clocks. Just a different business strategy vs. first to market strategy of AMD (but only 20-25% faster than GTX580).

Now since Nvidia pays per wafer, lower yields and high wafer costs at the moment make a strategic sense for the company to wait until production & yields improve so that it becomes more cost-efficient to mass produce.
2 1 [Posted by: BestJinjo  | Date: 02/16/12 06:33:19 PM]

"the best GPU we have ever built and the performance and power efficiency is surely the best that we have ever created,” said Mr. Huang." ... YES! and the next GPU they make will again be the best(better) then this one.. the statement is kind of moot
7 0 [Posted by: vid_ghost  | Date: 02/16/12 02:21:54 PM]

Edit: ^ ah you beat me to it because I had to sign up for an account.
Original message:
Best GPU you've ever built you say? No really? I expected the new generation to be worse...
6 0 [Posted by: mokahless  | Date: 02/16/12 02:24:47 PM]
- collapse thread

You signed up ayy .. ahh welcomes to your dooommddednesss... And like Mr. Huang would of said... mmmeh its not bad, or its exeptable or any other demeanor that doesnt promote the product.
0 0 [Posted by: vid_ghost  | Date: 02/16/12 02:29:17 PM]

kepler will rule. ramp ups on new nodes ALWAYS improve over time. Much ado over nothing.
1 2 [Posted by: beck2448  | Date: 02/16/12 03:42:16 PM]
- collapse thread

Im really looking forward to seeing how the second gen 28nm parts from AMD and Nvidia turn out... by that point 28nm would have matured and the designs would be more capable of taking advantage of the shrink
1 0 [Posted by: vid_ghost  | Date: 02/16/12 03:53:19 PM]
By the time it ramp up on production, AMD has already release a new generation of cards.
3 1 [Posted by: xybit  | Date: 02/16/12 05:09:39 PM]
HD8000 series is scheduled for 2013 based on AMD's roadmap. It's more likely that AMD will refresh HD7970 but there will not be any new generation of AMD cards in 2012.
1 2 [Posted by: BestJinjo  | Date: 02/18/12 07:25:11 AM]
I got downrated for stating facts? That's a new one.

No HD8000 series (i.e., Sea-Islands) until 2013.
2 0 [Posted by: BestJinjo  | Date: 02/18/12 07:21:50 PM]

nVidia's performance doesn't impress me
5 4 [Posted by: madooo12  | Date: 02/16/12 04:15:31 PM]
- collapse thread

This is super funny if you meant what I though you meant!
2 3 [Posted by: jonup  | Date: 02/16/12 05:23:45 PM]
I did if you thought I was repeating what the nVidia financial guy said about AMD

if it isn't then what did you think I meant
1 3 [Posted by: madooo12  | Date: 02/16/12 05:27:31 PM]

Does it worth to wait? It looks like it's the classic tale of the man standing under the apple tree waiting for the apple to fall so he can eat it.
1 0 [Posted by: xybit  | Date: 02/16/12 05:12:17 PM]

Oh buoy, Fermi's story all over again...
3 5 [Posted by: TAViX  | Date: 02/17/12 08:01:59 AM]

Big complex chips on a immature 28nm process is a recipe for trouble, I mean 500-550mm chips seriously gtfo!
2 4 [Posted by: redeemer  | Date: 02/17/12 03:24:55 PM]


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