My understanding of the motivation behind NV's Fermi and future product developments

[RussianSensation]

"Nvidia continues to face challenges in the graphics market from players like Intel and AMD, which are increasingly providing integrated graphics chips to PC makers that reduce the need for separate (discrete) graphics cards." https://www.trefis.com/company?from=widget:forecast&ovd_urlid=132075#/NVDA

Question (I asked myself): What is the bread and butter of NV business model?
- The most honest response that came to my mind was: Discrete and integrated graphics of course....and I was wrong.

Analyzing the factors that investors looked at in determining the stock price related to NV forecast (i.e., future) revenue streams (taking into consideration the resultant profits) and cash position reveals the following:

1. Professional Graphics - 31.5%
2. Discrete Graphics - 18.7%
3. Mobile and Game Console Computing Chips - 10.8%
4. Integrated Graphics - 9.8%
5. DRAM Graphics memory - 3.0%
6. Net cash - 26.3%
https://www.trefis.com/company?article=22029#

Takeway #1: Therefore, for NV as a firm, professional graphics with its superior profit margins is just as important as discrete + integrated graphics markets combined.

Another question: Ok, so now that I know that professional graphics is crucial for NV, how does this relate to my graphics card?
- Actually it does, because NV's focus can't just be on gaming graphics. "There is a risk, however, that Nvidia’s leading [discrete graphics] share will be eroded as integrated graphics cards become more powerful and if spend on high end graphics cards, like those made by Nvidia, is curtailed due to broader economic concerns."

- Thus, not only does NV have to create a graphics card that's competitive in gaming, but to preserve 31% of stock value, it has to be good in professional graphics. Wait, not just good, it has to be better than good.

- The average of forecasts for Nvidia Pro Graphics Cards Market Share created by Trefis members in the last week indicated a projected decrease from 85.6% in 2010 to 81.5%.

Takeaway #2: Therefore, for NV to be successful as a business, it has to maintain its professional graphics market share in order to sustain the superior margins and yet deliver performance to clients that are willing to pay for that performance. Failing to deliver on this front is a direct negative impact on NV profit margins, revenue and bottom line.

One more question: So how is NV able to maintain its lead in professional graphics?

- Simple, NV designed Fermi architecture for the workstation space.

- Its Quadro 5000 is one of the first models to use the company's GF100 graphics processor. Basically, Quadro 5000 is more a less a twin of the GTX465 352 SP core with added ram and slower clock speeds. On the other hand, its primary competition is ATI's flagship FirePro V8800. Nothing less than 1600 SPs Cypress chip.

Takeaway #3: In summary, it’s clear which card comes out on top: Nvidia’s Quadro 5000 is superior to AMD’s FirePro V8800 in almost every benchmark, usually by a clear margin.
http://www.tomshardware.com/reviews/quadro-5000-firepro-v8800-workstation-graphics,2701-8.html

"Nvidia certainly appears to be the premium choice, both where performance and cost are concerned. Its faster card is naturally more expensive." As a result, NV is able to sell GTX465 core chips that can outperform a 5870 chip and sell them for more $$$ in the professional market - this is the bread and butter of NV business model.

Conclusion: This is why Fermi was truly designed with professional graphics in mind since it's the only space where NV dominates, achieves greatest profit margins and has little competition.

This also explains why Fermi and probably every future NV card to follow will be a compromise between gaming performance and the desire to exceed customer expectations for professional graphics demands. This creates additional pressure on making a card that performs well in more than 1 environment.

My own takeway is that it probably would be better for NV to design 2 separate chips - 1 for professional graphics and 1 for games. However, due to R&D costs and manufacturing constraints, this is probably not possible given the size of the company. By designing Fermi as an "all-purpose" GPU, NV is thus able to amortize/share its R&D costs across at least 3 product lines including Consumer Graphics, Tesla and Quadro lines. In addition, this business strategy minimizes risks. If one of these product lines suffers a decline in revenue as a result of market forces, chances are that the other 2 product lines will remain healthy until a recovery can occur. By NV not putting all of its eggs in one basket (i.e., consumer products only), it is able to withstand changing market conditions to a greater degree.


============================================================================================
Update (September 8, 2010):
Nvidia continues to solidify positions in the professional graphics segment.

This segment also continues to grow. 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 of workstations worldwide in Q2, resulting in sequential growth of 9.6% and a year-over-year increase of 32%.

 

[ModestGamer]

You restating the obvious doesn't make it any less or more real. several of us have been telling you this for days.

Yet now you need ot read it somwehere else.

expect sales points 2-5 to disappear to nvidia leaving them with 1 and 6

 

[bunnyfubbles]

I have to admit, I'm getting into video editing with Adobe Premiere CS5 which supports CUDA with the GTX470 and am heavily considering swapping my 5850 for it.

Although being able to get a GTX470 for $250 after discounts and a rebate isn't hurting when I could probably sell my custom cooled 5850 for as much 😛

 

[RussianSensation]

You restating the obvious doesn't make it any less or more real. several of us have been telling you this for days.

A lot of people put out information on the internet without concrete evidence or facts to backup their opinions. I enjoy information which has some support. Also, which is obvious to you, is less obvious to me. I wasn't aware that Professional Graphics market is almost 2x the size of the market of Discrete Graphics for NV as a company.

This means that for a long time NV has focused its efforts on making videocards that perform well in both environments. Since ATI is not concerned about professional graphics, I was under the view that as a market, it was rather small. Again, I was wrong since I assumed the professional market is smaller than the discrete market.

Since NV has 86% of professional graphics card market, while ATI commands 51% of discrete graphics vs. 49% for NV (http://www.thinq.co.uk/2010/7/29/amd-overtakes-nvidias-gpu-market-share/), and given that only 18.7% of NV is discrete graphics, this implies that:

The professional graphics market for profits is almost as large as the the discrete graphics market of both ATI and NV combined:
(i.e. -> 31.5% / 86% = 36.6 Professional Market Pro-rated to 100% size
18.7% / 49% = 38.2 Discrete Graphics Market Pro-rated to 100% size)

If this was obvious to you, then it's news to me.

 

[railven]

You restating the obvious doesn't make it any less or more real. several of us have been telling you this for days.

Yet now you need ot read it somwehere else.

expect sales points 2-5 to disappear to nvidia leaving them with 1 and 6

If I were to believe your opinion, does that mean they can't make one focal professional build and gimp it to meet the gaming desires?

Ala GF100 to GF104.

Sell GF100 only in the high margin professional realm
Sell GF104 only in the low margin mid-range discrete sector.

Could preserves point 1 and 2.

 

[ModestGamer]

If I were to believe your opinion, does that mean they can't make one focal professional build and gimp it to meet the gaming desires?

Ala GF100 to GF104.

Sell GF100 only in the high margin professional realm
Sell GF104 only in the low margin mid-range discrete sector.

Could preserves point 1 and 2.

The comming APU's from AMD and intel pretty much kills off discrete for them below the middle mid range $100 and under cards. The highest volume sellers. The also loose the integrated graphics chipset sales.

They saw this comming years ago when AMD bought ATI and they knew exactly what it meant then and shifted bussiness strategy.

So it is what it is.

 

[ModestGamer]

A lot of people put out information on the internet without concrete evidence or facts to backup their opinions. I enjoy information which has some support. Also, which is obvious to you, is less obvious to me. I wasn't aware that Professional Graphics market is almost 2x the size of the market of Discrete Graphics for NV as a company.

This means that for a long time NV has focused its efforts on making videocards that perform well in both environments. Since ATI is not concerned about professional graphics, I was under the view that as a market, it was rather small. Again, I was wrong. The professional market is larger than the discrete market for both ATI and NV combined. If this was obvious to you, then it's news to me.

Its becuase you have a reactive mind not a proactive mind.

The question should always be.

what motivates my opponent and how to use that to my advantage.

AMD will move into more rendering cards once they finish killing nvidia out of this market. Just not right now.

 

[Skurge]

Correction, the V8800 is a single cypress chip.

Its drivers suck too as its barely faster than a V8700 that has half of everything.

 

[ModestGamer]

Correction, the V8800 is a single cypress chip.

Its drivers suck too as its barely faster than a V8700 that has half of everything.

I wouldn;t expect that disparity to remian once AMD gets all the other porfolio stuff launched. If you really look at what AMD is doing right now. Its a massive upheaval for the company.

I bet middle of next year thing will have settled down for design and they will get into a good rythm and we will see some really big stride in drivers etc.

My beef still is and has been that the GPU manufacturers are having to optimize drivers in the first place. That job should fall to windows and the API for the game.

Bad code requires mutating drivers to get it working.

 

[railven]

The comming APU's from AMD and intel pretty much kills off discrete for them below the middle mid range $100 and under cards. The highest volume sellers. The also loose the integrated graphics chipset sales.

They saw this comming years ago when AMD bought ATI and they knew exactly what it meant then and shifted bussiness strategy.

So it is what it is.

Sorry to not believe in propaganda/marketing to the level of which it is peddled at me.

The only GPU+CPU product I've seen (and tinkered with) was a Clarksdale and it wasn't doing jack to make me tell a customer "don't bother with a dedicated GPU, this will play anything you want."

I'll wait to see what AMD/Intel roll out before I start to say the mid-range market is gone.

 

[RussianSensation]

Correction, the V8800 is a single cypress chip.

Its drivers suck too as its barely faster than a V8700 that has half of everything.

Thanks Skurge. Noted.

Ya, I don't understand how ATI can't get their drivers together. Why leave NV with 80%+ market share in a market with massive profit margins?

 

[Lonyo]

Quadro and Tesla:

Mid to high 50's in gross margins eventually.
Highly differentiated business a big part of business. Quadro and Tesla and Tegra, where software is a large part of it, growing margins will continue to reflect that. In those markets gross margins are much better [than in consumer graphics cards]. 2/3rd of profits from these three businesses, even though only a small part of total revenues.
Over time as those businesses grow, margins should improve.
Mixed with mainstream Geforce business which will "continue to be competitive" margins in the 50s should be obtained.
Quadro ahead of where they expected to be today. Pent up demand for CS5 due to recession will help Quadro business grow.
Tesla record revenue. Fermi "unlocked new demand" in supercomputing, energy and finance. Long term success through adoption by OEMs (e.g. Supermicro)
Tesla for servers is going to ramp up substantially. First Tesla designed standard for OEM servers (didn't happen with previous Tesla products).



There was very little mentioned about mainstream and low end Fermi based
Obviously Tesla and Quadro (where Fermi is designed to make up shortfalls) are considered very important as they give NV the bulk of their money, and they even expect margins to go into the 50's, to which one of the guys asking questions had to ask if the NV guy had really said 50's because it seemed like a crazy goal. (High 50's type area is where Intel typically sit and they manufacture their own products).

The margins on the workstation and professional stuff, and HPC products are absolutely massive compared to the consumer space. That's where NV already makes most of their money.
I (and probably others) have been saying for a long time that Fermi is a money maker by aiming at these markets and not designed for discrete graphics cards.
The base architecture is designed to let them enter/improve their position such markets, and they don't even need to be wildly successful in areas like HPC, just lay the groundwork for the future.

 

[bryanW1995]

1. professional market isn't that big from a sales %, but it is from a PROFIT % for nvidia, and it's expected to have an even larger future impact (hence the prof is higher than discreet plus mobile combined).

2. net cash is nearly as large a stock price determinant as professional graphics. net cash would be enormously higher if they built consumer cards that people actually wanted to buy (other than gtx 460), so the consumer market needs more love than it got when they originally designed gtx 4xx series. hopefully they aren't making that mistake this time around.

 

[Madcatatlas]

Sorry to not believe in propaganda/marketing to the level of which it is peddled at me.

The only GPU+CPU product I've seen (and tinkered with) was a Clarksdale and it wasn't doing jack to make me tell a customer "don't bother with a dedicated GPU, this will play anything you want."

I'll wait to see what AMD/Intel roll out before I start to say the mid-range market is gone.

Did he say mid-range?

low-end is definetly out if we are to belive Anandtech and others.

 

[akugami]

Thanks Skurge. Noted.

Ya, I don't understand how ATI can't get their drivers together. Why leave NV with 80%+ market share in a market with massive profit margins?

I think it's a case where they are spread too thin. ATI has mostly cleaned up the quality of their drivers in the consumer games market. Enough to say nVidia and ATI are roughly equal. But it seems they still have a lot of work to do in the professional space.

And while I saw nVidia's focus seem to shift to the professional space more than towards gamers I didn't realize that they derive so much revenue from pro level cards. This explains a lot.

 

[ModestGamer]

Sorry to not believe in propaganda/marketing to the level of which it is peddled at me.

The only GPU+CPU product I've seen (and tinkered with) was a Clarksdale and it wasn't doing jack to make me tell a customer "don't bother with a dedicated GPU, this will play anything you want."

I'll wait to see what AMD/Intel roll out before I start to say the mid-range market is gone.

How are the largest group of computer buyers ?

Home Office
Home PC
Office applications.

They are most likely 85% of the market IIRC. low end Enthuaist gamers might buy a $100 card rarely. They make up 8% of the market the other 5@ are the hardcore FPS gamers who have to have the latest and greatest "excluding enterprise CPU stuff"

As you can see the APU products work fine for 85% of applications based on that alone. A market where nvidia has done well with integrated. Well Intel and AMD just sucked that market dry with cheaper MB's and lower parts counts and better themral for notebooks, desktops, and netbooks and laptops.

Your not seeing the forest for the trees.

People say glenn Beck is a popular guy. This is simply not true. He has 2 million viewers.

dancing with the stars has 70 million.

do the math.

 

[railven]

Did he say mid-range?

low-end is definetly out if we are to belive Anandtech and others.

Actually he said:

The comming APU's from AMD and intel pretty much kills off discrete for them below the middle mid range $100 and under cards. The highest volume sellers.

I bought my HD 4850 for $180 on release, if I recall the MSRP was $200, at that point in time I would call the 4850 mid-range.

5770 and lower to me is entry. And from what I've seen I positive a Clarksdale can't rival a GTX 460 1GB/HD 5830 (Mid-range middle if had to pick specifically). And I'll wait to see what AMD/Intel rolls out to see if it can beat a GTX 460 1GB/HD 5830.

 

[HurleyBird]

The big question is how much of Nvidia's dominance in professional graphics is due drivers, and how much (if any) of it is due to architecture.

 

[railven]

How are the largest group of computer buyers ?

Home Office
Home PC
Office applications.

They are most likely 85% of the market IIRC. low end Enthuaist gamers might buy a $100 card rarely. They make up 8% of the market the other 5@ are the hardcore FPS gamers who have to have the latest and greatest "excluding enterprise CPU stuff"

As you can see the APU products work fine for 85% of applications based on that alone. A market where nvidia has done well with integrated. Well Intel and AMD just sucked that market dry with cheaper MB's and lower parts counts and better themral for notebooks, desktops, and netbooks and laptops.

Your not seeing the forest for the trees.

People say glenn Beck is a popular guy. This is simply not true. He has 2 million viewers.

dancing with the stars has 70 million.

do the math.

And that would completely remove the mid-range? The people who are buying in the $200+ range?

I think we both have different opinions on what mid-range is.

I consider the 57xx and lower entry level. Above that point is mid range.

 

[ModestGamer]

And that would completely remove the mid-range? The people who are buying in the $200+ range?

I think we both have different opinions on what mid-range is.

I consider the 57xx and lower entry level. Above that point is mid range.

2million people agree with you and 70 million don't. where are you on this scale. Most people buy pcs under $500.

 

[Madcatatlas]

Actually he said:



I bought my HD 4850 for $180 on release, if I recall the MSRP was $200, at that point in time I would call the 4850 mid-range.

5770 and lower to me is entry. And from what I've seen I positive a Clarksdale can't rival a GTX 460 1GB/HD 5830 (Mid-range middle if had to pick specifically). And I'll wait to see what AMD/Intel rolls out to see if it can beat a GTX 460 1GB/HD 5830.

Didnt know you could get 460, 5830 or even the 5770 for under 100 bucks.

He means the cards under 100 bucks. And i would also consider his statement to mean a comparishon between the new APUs and the new discreet GPUs. Not age old Radeon9700, which in your book would be highend.

 

[Genx87]

How are the largest group of computer buyers ?

Home Office
Home PC
Office applications.

They are most likely 85% of the market IIRC. low end Enthuaist gamers might buy a $100 card rarely. They make up 8% of the market the other 5@ are the hardcore FPS gamers who have to have the latest and greatest "excluding enterprise CPU stuff"

As you can see the APU products work fine for 85% of applications based on that alone. A market where nvidia has done well with integrated. Well Intel and AMD just sucked that market dry with cheaper MB's and lower parts counts and better themral for notebooks, desktops, and netbooks and laptops.

Your not seeing the forest for the trees.

People say glenn Beck is a popular guy. This is simply not true. He has 2 million viewers.

dancing with the stars has 70 million.

do the math.

Why do you think Intel has such a large % of the market??? Are you expecting integrated level performance to gobble up more of the market than it already does?

 

[railven]

2million people agree with you and 70 million don't. where are you on this scale. Most people buy pcs under $500.

What does that have to do with anything? I thought the high volume products made something mainstream, but it doesn't change its performance category.

By all means, list out the price bracket's you'd use for the common discriptors such as low-end, entry, mid-range, high-end, and enthusiast.

I remember in a thread earlier someone said "GTX 480 isn't high end" and the crowd laughed at him because they clearly understod it was. So if GTX 480/5870 is High End, and HD 5970 is Enthusiast, what is HD 5830/GTX 460 768MB? It definetely isn't low end/entry. I'd peg it at low mid-range, and like I said before going off Clarksdale, I wouldn't recommend the CPU+GPU package for someone wanting to game at 1680x1050 resolutions (the more common LCD resolution.)

Didnt know you could get 460, 5830 or even the 5770 for under 100 bucks.

He means the cards under 100 bucks. And i would also consider his statement to mean a comparishon between the new APUs and the new discreet GPUs. Not age old Radeon9700, which in your book would be highend.

Show me where I said they did?

Unless you are of the same mentality as him that Mid-Range excludes the GTX 460 and HD 5830?

 

[ModestGamer]

Why do you think Intel has such a large % of the market??? Are you expecting integrated level performance to gobble up more of them market than it already does?

lets revisit this anology. what percentage of pc users ever buy a graphics card ?

its a prebuilt desktop and enterprise productivity war on the horizon. If you can't see it. your just being ignorant.

what percentage of computers get a upgraded graphics card vrs what came stock ?

This is the market AMD and Intel are going to do battle in next and nvidia has been shown the door.

Get over it.

 

[Martimus]

-snip-
My own takeway is that it probably would be better for NV to design 2 separate chips - 1 for professional graphics and 1 for games. However, due to R&D costs and manufacturing constraints, this is probably not possible given the size of the company.

This was a good post that I would not expect in the video forum. It was a logical progression of points, and portrayed your opinion well with properly referenced backup data. I am used to seeing that kind of thing in the CPU forum but not here.

However, the responses to your post reminded me why I don't visit the GPU forum very often.

It is news to me that the Professional market is as large as the Gaming Market as well. After thinking about it, I realized something though; Part of the reason for that is that nVidia is able to name the price for their components since they have a near monopoly on the market (81.5% if your quoted number is correct). Since there aren't many viable options, you can get close to the maximum price that people are willing to pay for the product they are selling (because of a lack of competitive options). So even though the volume is much lower than discrete consumer level graphics cards, nVidia is able to make up for it by getting the maximum gross profit the market would bear; just because they drive the market.

Of course this is just an opinion, based on my market experience. The problem with high profit margin items, is that they are desirable to sell, which means that someone may make a push to get a larger piece of that market. AMD has this ability, but I am wary about their history of lackluster support. Intel, in my opinion, is the bigger threat for this market. After a few generations, Larrabee may actually be competitive with Quadro in the future, along with acceptable support, which would drive down the profit margins on the products.