Nvidia Q4313 results

[ShintaiDK]

Using them to destroy your competition, is hardly "zero return".

That is the point of a value added service or item. But the IGP doesnt make money, the CPU part does. Same with the free heatsink you get. That is also simply a value added item in hope that you buy the CPU. And if both add the same value added service, those cancel one another out.

That IGPs on the other hand reduces discrete cards shipped. Thats just something you can exploit if you dont ship discrete cards. While the same IGPs damage your own discrete card sales if you sell such.

 

[AtenRa]

I think you are referring to SNB 4C, which offers *a lot* more CPU performance than Llano but was in a market bracket above it. Llano was in the same market bracket of SNB 2C, which had roughly half of the die area of Llano.

At the same die size, Llano 4C(228mm2) has 50% more iGPU performance than SB 4C(216mm2), SB 4C has 50% more CPU performance than Llano.

http://www.anandtech.com/show/6332/amd-trinity-a10-5800k-a8-5600k-review-part-1

35W with anemic performance. I don't doubt AMD could go all the way down to 10W if they wanted, but that would yield a SKU with subpar performance. They 35W and 17W are just like that. And this assuming they hadn't another 8350 moment when labeling those units.

Again, the 35W Trinity puts to shame the 45W SB in iGPU graphics. You comparing only the CPU performance when one CPU has double the iGPU die space dedicated to graphics and the other Chip has double the CPU die space dedicated to CPU performance.

http://www.anandtech.com/show/5831/amd-trinity-review-a10-4600m-a-new-hope

What's your definition of sell "quite well"? Because Llano didn't improve AMD margins or provided significant added volume.

Mobile Llano c4 + c2 volume in Q3 2011 was higher than Phenom X3/X4 and Athlon put together in Q1 2011. The APU was way better than the Athlon and Phenom lines for Mobile.

Q1 2011 to Q2 2012

Also, BobCat is an APU and it sold millions units and still sell well two years later.

 

[krumme]

The "highly integrated" Kabini advantage goes kaput with Haswell, as mobile SKUs will be SoCs too.

I was under the impression the southbridge wouldnt be integrated before broadwell?

 

[mrmt]

At the same die size, Llano 4C(228mm2) has 50% more iGPU performance than SB 4C(216mm2), SB 4C has 50% more CPU performance than Llano.

(...)

Again, the 35W Trinity puts to shame the 45W SB in iGPU graphics. You comparing only the CPU performance when one CPU has double the iGPU die space dedicated to graphics and the other Chip has double the CPU die space dedicated to CPU performance.

You are confirming what I'm saying: AMD APU concept does not work. They spend half of the die with a GPU but cannot make a single penny out of it, because the price is basically dictated by the performance the CPU, not the GPU.

Mobile Llano c4 + c2 volume in Q3 2011 was higher than Phenom X3/X4 and Athlon put together in Q1 2011.

You cannot draw a straight comparison like that, as the phenom II was launched in 2009 and Llano in 2011, and Llano came to be sold in the same price bracket of the Phenom II, so we had direct internal competition.

If we could find the numbers, I would bet my chips on the Athlon II, because not only AMD had a greater market share at that time, Athlon II had a from launch to EOL some 10 quarters, while Llano had only 6.

he APU was way better than the Athlon and Phenom lines for Mobile.

Of course it was, I already pointed out in previous posts that:


http://forums.anandtech.com/showpost.php?p=34638435&postcount=43

Sure, it opened some new markets for AMD, but did so competing at the very bottom, but that's too few. You don't design and manufacture >220mm^2 chips to sell it only in the $50-$100 market, and this is exactly what AMD got with Llano and more so with Trinity.

As for Brazos, I think you understand that it is a very different beast, not really comparable to Llano. The first thing they took care with Brazos was to keep die area in check and build the entire chip around an efficient core.

 

[mrmt]

I was under the impression the southbridge wouldnt be integrated before broadwell?

http://www.techpowerup.com/177817/Intel-Haswell-and-Broadwell-Silicon-Variants-Detailed.html

...The Haswell-MB consists of mobile chips in the replaceable PGA package, interestingly, Intel includes dual-core "1M" parts. Lastly, there's the SoC package (-ULT extension), which probably is the most expensive to make and sell, since it's a multi-chip module (MCM) of the CPU and PCH (chipset) dies. The package itself shouldn't be much bigger than Haswell-H (BGA), but conserves board footprint for a separate PCH chip, and a ton of wiring on the main board...

 

[krumme]

Thanx. An mcm solution is a good middle step and saves valuable space. Still haswell is another price segment than kabini. Kabini is more like say....richland

 

[mrmt]

Thanx. An mcm solution is a good middle step and saves valuable space. Still haswell is another price segment than kabini. Kabini is more like say....richland

I'm quite baffled by the die size of the quad core Kabini, over 100mm^2. This will put it squarely against IVB 2C, and Probably Haswell 2C, without much of the cost advantage Brazos enjoyed, and if rumors are to be believed, without any power consumption advantage too. I don't think the high end SKUs will have an easy time.

I'm more interested in the bottom 2C SKUs and Temash, because not only they can make nice netbooks, they can make nice, cheap x86 tablets too.

 

[Dribble]

The conundrum posed by the first two points will be solved with Broadwell, as 14nm will give the die area for another descent jump in the die area devoted to the iGPU and whatever bandwidth problems posed by the slow DDR3 memory can be mitigated by a beefed up Crystalwell solution, plus Broadwell should have a reworked GPU architecture, and this is where the things go south for Nvidia.

That never happened in the past - nvidia will be able to create a more powerful gpu too as they will shrink their dies as well. The igp will stay in just the same relative performance bracket as today. Also they can't just solve the memory bandwidth or power usage problems. Memory bandwidth is still going to be limited in comparison to a full gpu, as will thermals - fundamentally it's not cost effective making the gpu too powerful if most of your market don't need that performance.

tbh neither nvidia or Intel are that focused on the old desktop and laptop markets - they aren't where the opportunities lie. It's mobile and gpgpu that both nvidia and Intel will be worrying about.

 

[mrmt]

That never happened in the past - nvidia will be able to create a more powerful gpu too as they will shrink their dies as well. The igp will stay in just the same relative performance bracket as today. Also they can't just solve the memory bandwidth or power usage problems. Memory bandwidth is still going to be limited in comparison to a full gpu, as will thermals - fundamentally it's not cost effective making the gpu too powerful if most of your market don't need that performance.

tbh neither nvidia or Intel are that focused on the old desktop and laptop markets - they aren't where the opportunities lie. It's mobile and gpgpu that both nvidia and Intel will be worrying about.

Intel is shrinking faster than Nvidia and devoting more die area for the GP, meaning that the performance floor is raising faster than the delta between nv generations.

As for the bandwidth issue, why do you think intel is fielding l4 cache?

 

[AtenRa]

You are confirming what I'm saying: AMD APU concept does not work. They spend half of the die with a GPU but cannot make a single penny out of it, because the price is basically dictated by the performance the CPU, not the GPU.

2c SB(148mm2) is only ~34% smaller than Llano(228mm2) and in Mobile, AMD position 4C Llano and Trinity as Core i5 competitor.

You cannot draw a straight comparison like that, as the phenom II was launched in 2009 and Llano in 2011, and Llano came to be sold in the same price bracket of the Phenom II, so we had direct internal competition.

If we could find the numbers, I would bet my chips on the Athlon II, because not only AMD had a greater market share at that time, Athlon II had a from launch to EOL some 10 quarters, while Llano had only 6.

Llano/Trinity came to replace 2/4C Athlon II and 3/4C Phenom II not 6C Phenom II(346mm2). Selling Llano at 4C Phenom prices sustained the same ASP and raised profits due to smaller dies (Llano 228mm2 vs 4C Phenom II 258mm2).

You can see from the table bellow that 4C Llano has higher price than both 4C Phenom in Desktop and Mobile. The FX series replaced Phenom X6.

Q2 2011 ASP(Average Selling Price) (first column units in Thousands)

 

[mrmt]

2c SB(148mm2) is only ~34% smaller than Llano(228mm2) and in Mobile, AMD position 4C Llano and Trinity as Core i5 competitor.

So what? Does it change anything I said before?

Llano/Trinity came to replace 2/4C Athlon II and 3/4C Phenom II not 6C Phenom II(346mm2). Selling Llano at 4C Phenom prices sustained the same ASP and raised profits due to smaller dies (Llano 228mm2 vs 4C Phenom II 258mm2)

1) By your own numbers you know that Athlon volumes are bigger than Phenom, so while AMD *may* have had some improvement going from Phenom II X4 to Llano, they lost a lot by going to Athlon X2 and X4 to Llano, and again by your own numbers, Athlon volumes are far bigger than Phenom.

2) Second you are comparing ASP between almost EOL'ed SKUs, Phenom and Athlon, with a brand new SKU, which means that the former is almost at the bottom of its price curve while the latter is in the top, which means that prices from Athlon and Phenom should have been higher in the past. The Phenom 2 X4 reached retail for $300 in 2009, a price level never reached by Llano SKUs.

3) Third you are disconsidering process. By 2011 45nm was a lot mature, while 32nm was so problematic that GLF had to swallow the bad dies bill from the process. So that goes against Llano.

4) Llano got a 100MM charge against inventory, straight into the project ROI, something that the Athlon and Phenom didn't.

I can discount you on 2, 3 and 4, but there is not a chance in hell that you could have missed 1. You cherry picked the number to give authority for your trolling and you know that. Why did you cherry pick the numbers? why can't you be honest when dealing with AMD data?

And why do you have to twist my words to prove an argument? I didn't mention Llano profits, I did mention ROI, what does minute variations in in ASP have to do with ROI?

 

[shady28]

Got curious and looked at the numbers.

Basically for Year over Year, profit was down. Non-GAAP Net Income went from 734M to 728M. Small drop, but it's there. Nvidia is weird in reporting, comparing prior 3 quarters and then prior 12 months. Usually the report is prior quarter and prior year.

Tegra sales were +90% vs 2011. GPU sales were +7%.

So revenue is up, profits down for the year. Still very profitable though.

Tegra is only around 20% of Nvidia's revenue, very questionable if it is enough to stem any outflux of discrete GPU buyers.

I bet Apple has something to do with strong Q3/Q4 results too. 6770M = out, 650M = in on Macbooks and iMacs for 2012. Strong correlation there.

Still it shows Nvidia strong, no question, just I think people believe they're stronger than they really are.

 

[AtenRa]

I will have to apologize, i got carried away and forgot that this thread is about NV Q4 13

@mrmt, if you like we can continue to an appropriate thread.

 

[krumme]

I'm quite baffled by the die size of the quad core Kabini, over 100mm^2. This will put it squarely against IVB 2C, and Probably Haswell 2C, without much of the cost advantage Brazos enjoyed, and if rumors are to be believed, without any power consumption advantage too. I don't think the high end SKUs will have an easy time.

I'm more interested in the bottom 2C SKUs and Temash, because not only they can make nice netbooks, they can make nice, cheap x86 tablets too.

Naa - its better to compare cost than size. I think it was Idontcare numbers that 28nm is 2500usd a wafer now,and you get aprox 550 4 cores kabini a wafer then its about 4.5 usd sans packaging and logistics per 4 core. Add probably cheap packaging, its going to be dirt cheap in all configurations. I am certain it will hold a small power, - and especially idle - power advantage to the mcm solutions, what really matters most for the casual slow surfer. Then its 4 core - ask Tegra 3 - that is a marketing advantage too.

But yes, 4 core kabini is not interesting, its the 10w and below where the interesting starts. But still, its not in Intel interest to sell 22nm competing with tsmc 28nm in red ocean if it can avoid it. But anyway the cellerons and 2c Richland will keep profit very low for kabini. Probably we will see temash 2c sell for like the double, even though its probably half size.

 

[Ajay]

Geez, looking at the CFO commentary, GPU biz is still 75% of Nvidia business and Tegra is up ~30% (revenue). Double digits for Tegra, while good, doesn't seem to be enough - at that rate it'll take 5 years for Tegra revenues to match GPU. IMhO, Tegra needs a much higher growth rate to become large enough for Nvidia to become a significant player in ARM.

 

[mrmt]

Geez, looking at the CFO commentary, GPU biz is still 75% of Nvidia business and Tegra is up ~30% (revenue). Double digits for Tegra, while good, doesn't seem to be enough - at that rate it'll take 5 years for Tegra revenues to match GPU. IMhO, Tegra needs a much higher growth rate to become large enough for Nvidia to become a significant player in ARM.

NV was still in market share buy mode last FY, when Tegra resulted in operating losses for them. I'm eager to see their annual report to see if they reached break even.

 

[Idontcare]

Geez, looking at the CFO commentary, GPU biz is still 75% of Nvidia business and Tegra is up ~30% (revenue). Double digits for Tegra, while good, doesn't seem to be enough - at that rate it'll take 5 years for Tegra revenues to match GPU. IMhO, Tegra needs a much higher growth rate to become large enough for Nvidia to become a significant player in ARM.

5yrs is the natural "incumbency cycle" or term of incumbency for that segment though. Top dogs rise and fall about twice every decade. So a 5yr "arrival" timeframe is just about right to time Qualcomm's anticipated decline after peaking.

 

[pablo87]

Geez, looking at the CFO commentary, GPU biz is still 75% of Nvidia business and Tegra is up ~30% (revenue). Double digits for Tegra, while good, doesn't seem to be enough - at that rate it'll take 5 years for Tegra revenues to match GPU. IMhO, Tegra needs a much higher growth rate to become large enough for Nvidia to become a significant player in ARM.

EVERYBODY will struggle to maintain revenues in a post PC world. The industry is in a major transition on 4 fronts:

- storage ( solid state vs magnetic)
- cloud computing vs PC
- operating systems
- form factors

Usually transitions result in higher revenues initially until the older technology starts to die. Favorite example floppy drives: 360k, 1.2mb, 1.2mb+1.44, 1.44, -.

Integrated graphics accelerates the PC's demise which is dumb unless you think its inevitable and the immediate goal is the demise of AMD (no anti trust issue here, DOJ already accepted the PC demise argument - see WDC acquiring Hgst, Seagate ditto Samsung HDD, micron ditto elpida) in order to raise x86 cpu prices. Slippery slope though because Nv has deep pockets now, and if it doesn't compete in x86, it certainly will in SOC's. And it will benefit from amd's demise as well.

 

[ShintaiDK]

EVERYBODY will struggle to maintain revenues in a post PC world.

PCs aint going anywhere. They are here to stay.

 

[krumme]

EVERYBODY will struggle to maintain revenues in a post PC world. The industry is in a major transition on 4 fronts:

- storage ( solid state vs magnetic)
- cloud computing vs PC
- operating systems
- form factors

Usually transitions result in higher revenues initially until the older technology starts to die. Favorite example floppy drives: 360k, 1.2mb, 1.2mb+1.44, 1.44, -.

Integrated graphics accelerates the PC's demise which is dumb unless you think its inevitable and the immediate goal is the demise of AMD (no anti trust issue here, DOJ already accepted the PC demise argument - see WDC acquiring Hgst, Seagate ditto Samsung HDD, micron ditto elpida) in order to raise x86 cpu prices. Slippery slope though because Nv has deep pockets now, and if it doesn't compete in x86, it certainly will in SOC's. And it will benefit from amd's demise as well.

I hope the PC industry is separated from the normal macroeconomic situation, because i see exactly the same struggle starting; Dell 13 and 14 is dumped to half price through retail
, IB technology is putt into celeron and send to compete in e450 APU market and so on. I have not seen anything like it for 25 years.

The transistions is different from earlier situations because they all eventuelly leads to lower revenue and profit.
- Integrated SSD in 128GB configurations, will be dirt cheap in 2 or 3 years time and soon be like 16gb micro sd for your phone. The HDD could never go so low in price.
- Cloud is is already free of charge. What is the next step, that you get payed to use it ??
- Nobody pays for the operation system anymore and expects upgrade every ½ year
- And the extreme prices mobile producers charge for low BOM today will soon take a nose dive when the likes of Huawei steps on the pedal, as midrange smartphones is soon good enough, and everyone and his brother have dual core a7 in all their machines.

What the sector needs is some radical innovation, like the ones Apple brought to the table.

 

[mrmt]

Integrated graphics accelerates the PC's demise which is dumb unless you think its inevitable and the immediate goal is the demise of AMD

That's quite ironic, given that iGPU was AMD and NVDA push that started to beef up iGPI. Both dGPU players were responsible for the commoditization of the graphics market, and both are now betting the farm on becoming CPU companies to offset dGPU decline.

 

[Ajay]

PCs aint going anywhere. They are here to stay.

True, but what are 5+ year upgrade cycles (or whatever number it is) going to do to the PS industry? The upgrade cycle is definitely increasing as consumers put their $$s into newer tech, at least for now.

 

[Ajay]

5yrs is the natural "incumbency cycle" or term of incumbency for that segment though. Top dogs rise and fall about twice every decade. So a 5yr "arrival" timeframe is just about right to time Qualcomm's anticipated decline after peaking.

Do you really think Qualcomm is going lose their edge? Seriously, I'm curious what it is that you are seeing.

 

[Idontcare]

Do you really think Qualcomm is going lose their edge? Seriously, I'm curious what it is that you are seeing.

The drive for higher margins and profits combined with complacency in management (loss of sense of urgency) is unavoidable, its human nature.

Eventually Qualcomm will price itself out of the market, the same thing TI did when it peaked at 80% marketshare. Stagnate, and collapse. RIMM all over again. Nokia all over again. Motorola all over again. There are no shortages of examples.

Apple's time will come, so will Qualcomm. Unavoidable really because the shareholders will insist on management evolving towards adopting a short-term mindset, like HP adopted, to make quarterly numbers and ignore problems that come from under-funding R&D.

Qualcomm is not immune to any of that. For now though, this is their time, so I say nothing against what they are doing now. It is what they do 5 yrs from now that I am speaking to.

 

[Ajay]

Apple's time will come, so will Qualcomm. Unavoidable really because the shareholders will insist on management evolving towards adopting a short-term mindset, like HP adopted, to make quarterly numbers and ignore problems that come from under-funding R&D.

True, though some companies are better at re-inventing themselves (GE, IBM and Intel, etc.), though not without stumbles along the way. I have seen successful R&D teams slowly torn apart because mgmt have put pressure on them to produce more products and less IP - only to kill the product pipeline in the long run (all while giving themselves far greater bonuses). And then they complain about turn over - duh!