Future of graphics business

[T2k]

http://www.xbitlabs.com/news/other/...tion_Market_Begins_to_Stabilize_Analysts.html

Yeah, that's a good one, I just want to see something similar but less than a year old...

 

[v8envy]

FYI Quadros are all based on the same R&D budget discrete graphics sales finance - if discrete graphics profits disappear so will the R&D budget, along with Quadros.

It's more than just R&D. The discrete card volumes are what makes the great profit margins on Quadros and Teslas possible. We know they're the same GPUs, but with different drivers, packaging and memory configurations. So if TSMC order volume goes from tens of thousands of wafers to tens of wafers the volume pricing goes with it. Suddenly it's not a $100 mass market GPU being sold in a $2500 board, it's a $2000 custom, ultra low volume GPU being sold in a $2500 board...

 

[RussianSensation]

some issues...

1. "many decades ago" is obviously silly, since the company is less than 2 decades old
2. Where did you get the "profitability related to this business is barely 10% of profits" from?
3. There's a reason that Lasso never developed...

#1. You are right. I just meant a long time ago in "tech years".

#2. Discussed in this thread here. 9.8% of profits for NV are chipsets.

#3. I am not the CEO of NV, nor do I work for NV. Just relaying information that NV's executives have told us themselves about possible growth opportunities:

"We have two strategies at Nvidia: one is to put graphics everywhere, the other one is to [find more ways to] integrate discrete chips into the box. I think there is definitely a place for [external graphics cards for notebooks], no question. We continue to look at whether this is a GPU [docking stations] or external devices,” said Rene Haas, general manager of the notebook GPU business at Nvidia, in a brief interview with X-bit labs." - February 2010

Originally Posted: http://thetechjournal.com/electroni...rs-for-notebook-by-nvidia.xhtml#ixzz10Hx35eKj

The fact remains a fact: when NV lost its chipset business it was 1/3rd of its entire revenue. Period.
No more chipset business, lost revenue, period.
Stock would tank immediately.

What does Revenue have to do with Cash Flow? (Investors value a company using cash flow as one of the criteria to arrive at the Enterprise Value). If the chipset business produced 1/3 of Revenue for the firm, but was not very profitable (low gross margins), then it makes sense for NV to explore more profitable and growing, rather than shrinking, market opportunities instead.

I am not sure why you want to bring stock performance into the discussion. Still, AMD is in the top most volatile stocks for some time now (this is for last week). On the other hand, NV is in the 12 most attractive stocks trading under $35.

Of course, this is also because ATI is not a standalone company. Regardless, both stocks have suffered significantly with YTD declines from 35-39% as per Google Finance.

Link to back up that 88% claim? I highly doubt it's true when ATI has truckloads of low- and mid-range WS card, usually for less than NV.

"Nonetheless, desktop ATI FirePro graphics accelerators managed to capture 11.4% of professional add-in graphics cards (up 0.9%), but the lion's share of 88.6% still belonged to Nvidia. The latter also managed to gain some share on the market of professional graphics accelerators for mobile computers." - September 2010

This statement confuses me. If they already command an 88% market share, where is the growth going to come from? They either need to figure out a way to make the market bigger, get more profit from each part sold, or fight for the final 12% of the market. None of these options are likely to get much in the way of growth however.

Exactly. It might be a solid revenue source but it won't grow, that's for sure unless he confuses Tesla-sales with Quadro-sales.

You can maintain the same market share and have growth as long as the market/industry in which you are selling is also growing. This is precisely the case with the professional graphics market.

According to estimates of Jon Peddie Research, AMD and Nvidia shipped around 1.3 million of professional graphics accelerators in the second quarter of 2010. The technology and market research firm reports that the industry shipped 795 thousand workstations worldwide in Q2, resulting in sequential growth of 9.6% and a year-over-year increase of 32%.

Totally going down. At the same time AMD was ablo to ship truckloads of mobile DX11 chips so their high-end mobile graphics market share lead also disappeared and now AMD is coming with their intergated CPU+GPU too... future looks pretty dark here, NV has zero chance to beat AMD in this game on the long run, forget Intel's market share...

The notebook market is growing faster than the desktop market. Even if NV maintains similar market share as today, it still is a growing market segment for the firm. Competitors trade market share every so often. This round ATI released great desktop and mobile offerings. This doesn't mean NV is going to stop producing discrete mobile and desktop graphics. These 2 segments are very valuable for the firm. With notebooks continuing to be a growing sector, NV is going to focus on this market more than ever. Remember, they were very late with Fermi mobile because Fermi generation was late itself. Whether or not they will be late every single generation from now on is anyone's guess.

This will include mobile GPUs to target smarphones and tablet pcs too such as Tegra 3 and Tegra 4. I am not sure why you are so anti-NV (NV has provided competition to ATI, which lowered prices on 5770 series and is now going to introduce HD6000 series to remain competitive with NV's next offerings). We need both firms to do well for a healthy competition in the marketplace, no ifs about it.
​

 

[Will Robinson]

and is now going to introduce HD6000 series to remain competitive with NV's next offerings

What next offerings are you referring to?

 

[RussianSensation]

What next offerings are you referring to?

AMD executives think that NV will release cards after Fermi 1, which I presume isn't your view? I suppose there was no need for HD6000 series then, because the last videocard NV will ever make is GTX4xx series....:whiste:

 

[SolMiester]

What next offerings are you referring to?

Does it matter, the fact is they will have another offering!, odds are thought that's its a derivative of Fermi

 

[SolMiester]

<snip>

thumbs up for the post mate!

 

[Viditor]

#1. You are right. I just meant a long time ago in "tech years".

#2. Discussed in this thread here. 9.8% of profits for NV are chipsets.

#3. I am not the CEO of NV, nor do I work for NV. Just relaying information that NV's executives have told us themselves about possible growth opportunities:

#2 I see that it was discussed, but the SEC reports they are listing as their source (10-Q and 10-k) do not actually have those breakouts in them at all...I've gone through the last 5 years worth.

#3 I agree that anything is possible...

 

[RussianSensation]

#2 I see that it was discussed, but the SEC reports they are listing as their source (10-Q and 10-k) do not actually have those breakouts in them at all...I've gone through the last 5 years worth.

The numbers represent future cash flows (i.e. forecasted model, based on historical sales). This is why you won't see them in historical statements.

Viditor, I presume NV's management had foreseen the demise of the chipset business as they knew that AMD's acquisition of ATI would present problems for this business. I am sure they were aware of the fact that AMD would be able to sell systems with onboard ATI graphics (Fusion that was one of the primary reasons for the acquisition). I am sure that they also had an idea that at some point Intel would be able to fit their GPU on-die with the CPU, once the manufacturing technology permitted it. This would leave NV competing in the discrete mobile and desktop gaming space -- the situation we have now.

I think NV's management team had long-ago developed a contingency plan/changing strategy of developing other segments (such as mobile computing space, tablets, increasing foothold in the professional graphics, etc.), at least since the merger of AMD with ATI. Therefore, while the chipset business was important, to suggest that it was crucial is somewhat misleading imo.

 

[Viditor]

Viditor, I presume NV's management had foreseen the demise of the chipset business as they knew that AMD's acquisition of ATI would present problems for this business. I am sure they were aware of the fact that AMD would be able to sell systems with onboard ATI graphics (Fusion that was one of the primary reasons for the acquisition). I am sure that they also had an idea that at some point Intel would be able to fit their GPU on-die with the CPU, once the manufacturing technology permitted it. This would leave NV competing in the discrete mobile and desktop gaming space -- the situation we have now.

I think NV's management team had long-ago developed a contingency plan/changing strategy of developing other segments (such as mobile computing space, tablets, increasing foothold in the professional graphics, etc.), at least since the merger of AMD with ATI. Therefore, while the chipset business was important, to suggest that it was crucial is somewhat misleading imo.

Actually, they didn't expect AMD to survive the merger (JHH said as much) which means that fusion wasn't nearly as imminent as. But that point is moot...the only real death knell for them vis a vis chipsets was the Intel (80% of the market) lawsuit in Feb last year. I suspect that this was when they actually started planning "for realz".

 

[Viditor]

The numbers represent future cash flows (i.e. forecasted model, based on historical sales). This is why you won't see them in historical statements.

Which is another way to say it's guesswork...
I am dubious about chipsets being only 10% of their profits. While I know that the R&D is expensive, IGP represents the VAST majority of graphics in computers today.

 

[cbn]

Nvidia makes some great products and I notice they strongly like to make their technology "proprietary" : CUDA, Physx, TWIMTBP.

I'm not knowledgeable about IT, but does anyone see them going even more in that direction (as a way to grow)? Maybe a proprietary OS (similar to Apple) but using ARM instead of x86? Or are they headed more towards putting that proprietary technology into "Cloud computing"?

 

[Cerb]

Nvidia makes some great products and I notice they strongly like to make their technology "proprietary" : CUDA, Physx, TWIMTBP.

I'm not knowledgeable about IT, but does anyone see them going even more in that direction (as a way to grow)? Maybe a proprietary OS (similar to Apple) but using ARM instead of x86? Or are they headed more towards putting that proprietary technology into "Cloud computing"?

Without Steve Jobs running the show, it will just make them easier to knock down. Intel and AMD are both either in, or going into, every market that nVidia dominates, and in some of those markets, ARM and IBM are players, as well.

If you had to buy twice the hardware right now, but you could move your codebase to another brand's hardware in the future, without a rewrite (even better, if you could compile it straight up, and then incrementally improve performance), why wouldn't you?

FI, let's say develop on Intel's HPC chips, such that the code runs fine on the specialized chip, your desktop with ICC, and your desktop with GCC. Moving to new SIMD, including going away from x86, will not be a trivial matter...but it's going to cost far less than moving away from CUDA will.

OTOH, outside of the hypotheticals, nVidia knows this, and is moving towards directly running C and C++ code on their hardware (if the next gen has as many feature improvements as Fermi, it could very well run its own OS). If they use open libraries for it all, they can keep their hardware attractive to high-margin users for at least awhile, even among those who would be turned off by lock-in, while they figure out some volume mobile money-makers.

 

[maddie]

I posted this today in another thread but it's relevant here. NO ONE responded. Am I talking rubbish?

One big danger AMD has in trying to price products too low, is the demise of Nvidia. They must balance winning without destroying them. The absolute last thing AMD needs now, with fusion starting to arrive , is for Intel to buy Nvidia. AMD would have won the battles but lost the war.

Remember that Intel already has fusion experience. If they had world class graphics, it would take a lot less time to adapt it to their cpu than AMD did, when starting from scratch.

 

[RussianSensation]

Which is another way to say it's guesswork...
I am dubious about chipsets being only 10&#37; of their profits. While I know that the R&D is expensive, IGP represents the VAST majority of graphics in computers today.

IGP is a dead business segment that's on life support. The future is APU not IGP (i.e., integrated graphics inside the CPU). In that future NV has no place (not until they develop an x86 CPU). Therefore, IGP on a chipset is irrelevant to them and has been for at least 2 years. Again, people seem to confuse Revenue with Cash Flow, skipping the Gross Margin altogether. Even if NV can sell 100 IGPs, that's still less $$ to the bottom line than a single high-end Quadro card that can retail at $4000-5000+.

NV's focus has long shifted to improving their foothold in professional graphics, refocusing on notebook graphics in the next 3-5 years and of course Tegra, targeted at smartphones and tablet devices. Why do you suppose they have been pushing CUDA applications so much? They are trying to make GPUs more than devices for games, but general purpose devices which can be used for a variety of applications.

I wouldn't be surprised at all if in 5 years Intel abandoned chipsets altogether. Everything will be on the CPU package (graphics, I/O, sound, etc.).

 

[Dark Shroud]

I wouldn't be surprised at all if in 5 years Intel abandoned chipsets altogether. Everything will be on the CPU package (graphics, I/O, sound, etc.).

That's not going to happen as other parts of the mobo need chips. USB, SATA, & NICs are just the big examples. Intel even has a court order saying they have to provide SATA ports for a set number of years.

And while it might not be completely correct many of us refer to the GPU core as IGPs at the moment.

 

[Viditor]

I posted this today in another thread but it's relevant here. NO ONE responded. Am I talking rubbish?

One big danger AMD has in trying to price products too low, is the demise of Nvidia. They must balance winning without destroying them. The absolute last thing AMD needs now, with fusion starting to arrive , is for Intel to buy Nvidia. AMD would have won the battles but lost the war.

Remember that Intel already has fusion experience. If they had world class graphics, it would take a lot less time to adapt it to their cpu than AMD did, when starting from scratch.

As I've stated previously, it's not legally possible for Intel to buy Nvidia. Neither the SEC nor the DOJ would allow the merger to take place...

 

[maddie]

As I've stated previously, it's not legally possible for Intel to buy Nvidia. Neither the SEC nor the DOJ would allow the merger to take place...

Please point me to such information.

 

[Viditor]

Please point me to such information.

Intel currently owns 50&#37; markestshare of graphics while AMD and Nvidia each hover near the 25% level.
It is illegal for the governement to allow the formation of a clear monopoly through a merger...

 

[Cerb]

That's not going to happen as other parts of the mobo need chips. USB, SATA, & NICs are just the big examples. Intel even has a court order saying they have to provide SATA ports for a set number of years.

And while it might not be completely correct many of us refer to the GPU core as IGPs at the moment.

Why can't all of those go into the CPU, such that they just go out to connectors? There's no need for separate chips. It's certainly cheaper that way, for the time being, but what technical reason must they be separate chips, on separate packages? Even those separate packages are pretty heavily integrates parts.

 

[happy medium]

Intel currently owns 50&#37; markestshare of graphics while AMD and Nvidia each hover near the 25% level.
It is illegal for the governement to allow the formation of a clear monopoly through a merger...

What share of the cpu market does Intel have ? More then 75%?

edit:I'd say more then 81%

http://www.thinq.co.uk/2010/4/30/amd-loses-cpu-market-share-to-intel/

 

[GaiaHunter]

What share of the cpu market does Intel have ? More then 75%?

edit:I'd say more then 81%

http://www.thinq.co.uk/2010/4/30/amd-loses-cpu-market-share-to-intel/

And did they buy anyone to get to 81%?

 

[blastingcap]

And did they buy anyone to get to 81&#37;?

Yeah I think Intel might just try to get some high-ranking NV and AMD defectors to help them out with Larrabee II, if they ever do try another Larrabee. (Never say never.) It's more cost-effective than buying entire companies. If Intel can't get enough key defectors, only then would I think they would buy entire companies.

 

[happy medium]

And did they buy anyone to get to 81%?

2 or 3 more Fermi releases and AMd will have 40% and Nvidia 10% and will be buyable.
Thats what I'm saying.

BUT

That all could change next year, who knows?

 

[GaiaHunter]

It's more cost-effective than buying entire companies. If Intel can't get enough key defectors, only then would I think they would buy entire companies.

I wouldn't bet on that - look at some of the buys that happened this summer (including Intel buying McAfee). Stock markets are a bit crazy with all this inflation created by central banks.

On the other hand, as stated previously, Intel absorbing Nvidia doesn't seem likely due to anti monopoly rules.