Intel Kaby Lake + AMD Radeon product *spin-off*

  • Thread starter Deleted member 13524
  • Start date
If the average person never upgrades anything, then Nvidia consumer side would not make the large revenue and profit it does, especially when it launched Maxwell and importantly with Pascal.
PC Desktop sales are down every year, meaning most sales are coming from the average consumer buying discrete GPUs.
Average consumers obviously don't own desktops. This has been true for many years already. Desktop is a niche. So we can assume that desktop owners don't behave like average consumers.

My experience about desktops (both gamedev workstations and home gaming PCs) is that people only upgrade GPU and HDD/SDD (add another HDD/SSD when space runs out). If you want to change anything else, then you buy a new computer. Sometimes I have seen people upgrade memory, if they have skimped on memory when they bought their computer. Gamedev companies behave exactly like this too. Workstations get bought without GPU and GPUs get replaced eventually, if there's a need for that before it's time to buy a new computer.

There are still some instances where companies buy cheap desktops instead of laptops. Some companies are still old fashioned and don't like the possibility that their employees can take a computer with them outside the office. Most companies nowadays of course feel the exact opposite (extra hours = good). For these kinds of desktops integrated CPU+GPU+memory solutions are preferable. Nobody is ever going to update any computer like this. But the existing Intel GPUs (24 EU GT2 models) are more than fine for this purpose. No need for a fat integrated Radeon. And this is a market that is getting smaller and smaller. Laptops are the workhorses of modern companies. Except gamedev and HPC of course.
 
Average consumers obviously don't own desktops. This has been true for many years already. Desktop is a niche. So we can assume that desktop owners don't behave like average consumers.
.
So by that logic Nvidia is a niche solution in the consumer gaming market?
Their revenue and profit is huge for a niche market (specifically talking about Geforce), think it is still over 50% of all their revenue.
As part of the Q1 2017 revenue announcement it was said:
" Nvidia's gaming business made up the bulk of its revenues with a $1.03bn haul, up 49 per cent on the year-ago quarter."

And the biggest revenue for Nvidia in Geforce is 1050ti and upwards, not sure how $1.03bn a quarter is niche, especially as part of the same market AMD are selling their products as well so the overall figure is rather large.
And gaming business discrete GPUs keeps growing for both.
I think it is wrong to call it niche tbh, especially when one looks where the profit margins lay.
 
So by that logic Nvidia is a niche solution in the consumer gaming market?
Their revenue and profit is huge for a niche market (specifically talking about Geforce), think it is still over 50% of all their revenue.
As part of the Q1 2017 revenue announcement it was said:
" Nvidia's gaming business made up the bulk of its revenues with a $1.03bn haul, up 49 per cent on the year-ago quarter."

And the biggest revenue for Nvidia in Geforce is 1050ti and upwards, not sure how $1.03bn a quarter is niche, especially as part of the same market AMD are selling their products as well so the overall figure is rather large.
And gaming business discrete GPUs keeps growing for both.
I think it is wrong to call it niche tbh, especially when one looks where the profit margins lay.
I was talking about average PC consumer, not gaming market. Intel is the most popular GPU by far in the average PC consumer market. Nvidia is a distant number 2. However gaming desktop GPUs have huge profits and Nvidia is the market leader in this segment. A high end gaming GPU costs roughly twice as much as a cheap laptop. A niche market can be still important business, if your products have 10x+ higher profit margins that GPUs generally found in average consumer PCs.

But this is exactly why the behavior of average PC owner (= mostly laptops nowadays) and a gaming desktop PC owner can't be directly compared. These are entirely different markets (both in size and in profit and commitment and product research done by the buyer).
 
I was talking about average PC consumer, not gaming market. Intel is the most popular GPU by far in the average PC consumer market. Nvidia is a distant number 2. However gaming desktop GPUs have huge profits and Nvidia is the market leader in this segment. A high end gaming GPU costs roughly twice as much as a cheap laptop. A niche market can be still important business, if your products have 10x+ higher profit margins that GPUs generally found in average consumer PCs.
The point is though Nvidia is earning over $1bn a quarter just for their Geforce products, one can add AMD to this and you have a quarterly figure that is getting on to what Qualcom generates if you look more at just their GPU aspect.
High end gaming does not make up all of that $1bn quarterly revenue for Geforce.

However importantly it seems the focus in the discussion here are the very extremes, while Intel is not focusing on entry models.
Intel is focusing on entry performance and prosumer with this product, this is not going in cheap laptops.....

And regarding the context of the market below 1050ti/AMD's equivalent and truly mass market-general.
The margins are incredibly low for such a mass market.
For context here is Qualcomm's earning over the years and covers all aspects (overall revenue would be much lower if looking just at Snapdragon).
A3_Semiconductors_QCOM_chipset-earnings-4Q17.png


This is one reason Intel is trying to target a higher price segment with the EMIB solution 1st.
Some here are talking about entry general, but Intel and AMD has been pretty clear this is targetting above $1k products.
Either this is a challenge to Nvidia at their lower end or it is not, some here have thought so and so do much of the press.
If it is not a challenge to Nvidia, well then it suggests it is more of a challenge to AMD who do target more of the lower end/less margin segment.
 
Last edited:
Here is a figure relating to margins for Intel/AMD/Nvidia, this will be very relevant when considering the semi-custom aspect of this deal and best positioning of a product.

A22_Semiconductors_NVDA-AMD-INTC-profits-margin-3Q18.png
 
Why don't you quote recent quarterly revenue figures for Intel? You might get some meaningful perspective. Hint: $8.9B recently for client computing, i.e. excluding data centre and other "stuff".
 
Why don't you quote recent quarterly revenue figures for Intel? You might get some meaningful perspective. Hint: $8.9B recently for client computing, i.e. excluding data centre and other "stuff".

I am guessing because Nvidia is still pending Q3 release (CC tonight).
 
Also the context is the margins as I said in my post, Intel offsets the margins for their mass lower priced products by higher margin technologies and products even in PC-client computing segment.
The profit margin chart I showed needs to be considered in that the AMD figure would be even lower if it removed the Polaris products, I am not convinced AMD is actually seeing a great benefit from a hard perspective with the EMIB deal when looking at what AMD has to date.
Unlike the positives around consoles that is also low margin but importantly has a strong soft and hard benefit that I touched upon much earlier; consolidates the major consoles under AMD without impacting either of their divisions along with a predictable,stable and long term deal while also being in the best position to negotiate such a solution, with further benefits that it brings ease of development from consoles to PC for games that is expanded with GPUOpen and AMD hardware.
The EMIB deal has the potential to overlap and conflict at some point with the CPU division (and maybe even the lowest discrete GPUs from both manufacturers depending how it evolves) and at least consolidates Intel's position and consumer perception for said market making it much more difficult for AMD to bring in a complete product themselves.
The soft benefit is an increase of GPU client footprint but this is more relevant for AMD if it could also be applied to software development and this requires the solution to be viable for gaming/prosumer rather than more elementary entry models.
But the margin, it would be on the lower scale because AMD would not be in the same position of strength as they were with consoles; gross margin would possibly be realistically anywhere between 14-20% for this deal with Intel, and then look how low operating/net margin would be correlated to that.

If you feel any revenue is better than no revenue (because it is unknown if AMD could ever develop and compete in this segment) then it is academic, but if one feels it potentially limits AMD while not seeming to do much for the bottom line and unknown how the new GPU demograph would influence related services/software then it does raise questions.
But what we do know is this is targetting for now $1k to $1.2k devices (according to AMD), so how mass market is that; the risk IMO is if they expand the solution/tech up and down the consumer ladder it will affect not just Nvidia but also AMD divisions.
 
Last edited:
This is one reason Intel is trying to target a higher price segment with the EMIB solution 1st.
Intel is already dominating the high profit laptop segment with their 15W and below chips. Ultrabook brand was a brilliant move by Intel. High quality aluminum bodies (small form factor & lightweight), SSDs instead of HDDs, high DPI displays (nowadays). Consumers loved this. It was a huge step over 3+ kg plastic laptops. Intel had all the tech for this: low power high perf CPUs, fast enough integrated GPUs to run high DPI displays and even their own SSDs (albeit manufacturers started to use other brands very quickly). They priced their tech accordingly. Laptop manufacturers loved this because now they could charge 1000$+ for a laptop instead of 500$ and people wanted these laptops badly.

IMHO the EMIB solution has too high TDP for ultra thin premium laptops. Intel already added 64 MB of eDRAM to the latest Coffee Lake i7 lineup for this market segment. Intel has an unbeatable high profit margin product there already. Personally I can only think about one target for the EMIB solution: Macbook Pro. Apple values GPU performance. They have been pushing Intel to develop faster integrated GPUs. They are already using Intel + AMD, so integrating them is a perfect move for Apple. Smaller footprint + lower power consumption are both features that Apple would demand. There are of course Windows laptops that are competing against Macbook Pro, but none of them are high volume products that would drive Intel + AMD to work together. However Raja Koduri jumping from AMD to Intel and Intel announcing to produce discrete GPUs (https://www.anandtech.com/show/1201...ete-gpus-hires-raja-koduri-as-chief-architect) tells us that Intel has made big decisions internally. Without knowing anything else about this, it is hard to draw any conclusions right now about Intel's motives.
 
The EMIB deal has the potential to overlap and conflict at some point with the CPU division (and maybe even the lowest discrete GPUs from both manufacturers depending how it evolves) and at least consolidates Intel's position and consumer perception for said market making it much more difficult for AMD to bring in a complete product themselves.
The soft benefit is an increase of GPU client footprint but this is more relevant for AMD if it could also be applied to software development and this requires the solution to be viable for gaming/prosumer rather than more elementary entry models.
But the margin, it would be on the lower scale because AMD would not be in the same position of strength as they were with consoles; gross margin would possibly be realistically anywhere between 14-20% for this deal with Intel, and then look how low operating/net margin would be correlated to that..
Nvidia is currently dominating the laptop GPU market. Nvidia is always paired with Intel CPU. If a customer instead uses Intel EMIB based solution, that customer is using an AMD GPU instead of Nvidia, and AMD gets money. This is one way for AMD to get back to the laptop discrete GPU business.

Of course AMD has their own brand new Raven Ridge 2500U and 2700U products too, but these are targeting lower power segment than the EMIB solutions. The customer base doesn't currently overlap.
 
Nvidia is currently dominating the laptop GPU market. Nvidia is always paired with Intel CPU. If a customer instead uses Intel EMIB based solution, that customer is using an AMD GPU instead of Nvidia, and AMD gets money. This is one way for AMD to get back to the laptop discrete GPU business.

Of course AMD has their own brand new Raven Ridge 2500U and 2700U products too, but these are targeting lower power segment than the EMIB solutions. The customer base doesn't currently overlap.
But it also stops AMD CPU division launching any other product above those because the EMIB deal cements Intel's position in that market all for a low margin revenue for Radeon, which is what I said from the beginning; no-one knows what other products the CPU division had lined up or was thinking of doing.

The margin like I said for the semi-custom contract that then closes the door on any attempt to take a slice of this high margin with their own full CPU solution may not be worth it, especially as Intel is a competitor and this could potentially still evolve slightly up and down the product chain.
The gross margin I think for consoles started at 14% and ended up increasing to 20%; that was from a position of strength. Here Intel holds the cards so real world the gross margin is more likely to remain closer to 14%, then look back at the chart at the operating margin and start to wonder how much actual cash this deal is bringing in while helping Intel.

Or is everyone here now saying AMD does not have the skills/resources cannot create a competitive 35+W mobile APU/possiblywith semi-custom GPU?
Remember much of AMD budget is committed to CPU division, so they are not in the same position as Radeon in terms of R&D/product engineering.
What was the power demand for the new Intel EMIB solution using the Kabylake quad core and also with the semi-custom GPU?

How long did it take Radeon to bring out an enthusiast GPU?
Point being just because the company launches lower performance products 1st does not mean they intend to give up on the more lucrative markets.
But like I said depending upon ones POV either the deal is good or questionable with risks, time will tell either way.
 
Last edited:
But it also stops AMD CPU division launching any other product above those because the EMIB deal cements Intel's position in that market all for a low margin revenue for Radeon, which is what I said from the beginning; no-one knows what other products the CPU division had lined up or was thinking of doing.
That's true. However, how many markets AMD is able to conquer in the next year or two? They already are fighting Xeons with EPYC, i9 with Threadripper, desktop CPUs with Ryzen and 15W-28W laptop CPUs with Ryzen Mobile. Maybe they left gaming/workstation laptops out intentionally to narrow their focus. Maybe Zen2 + Navi combo is going to cater that market later. Meanwhile better to get some steady revenue from Intel licensing deal. Console deals have also been very important for AMD. They wouldn't have survived without them.
 
IMHO the EMIB solution has too high TDP for ultra thin premium laptops. Intel already added 64 MB of eDRAM to the latest Coffee Lake i7 lineup for this market segment. Intel has an unbeatable high profit margin product there already. Personally I can only think about one target for the EMIB solution: Macbook Pro. Apple values GPU performance. They have been pushing Intel to develop faster integrated GPUs. They are already using Intel + AMD, so integrating them is a perfect move for Apple.
My money would be on Imacs and eventually blade products as well. Quad-Core (15W) plus HBM2 (10ish W) plus mid-sized GPU (45-ish W best case) is too much TDP to get rid of in a more business oriented chassis with accetable noise levels. Specialized gaming laptops, their users deal with high noise output, but on a Macbook (Pro), chances are, users want more quiet operation and for that you need cooling material (bulk), not high air flow.
 
My money would be on Imacs and eventually blade products as well. Quad-Core (15W) plus HBM2 (10ish W) plus mid-sized GPU (45-ish W best case) is too much TDP to get rid of in a more business oriented chassis with accetable noise levels. Specialized gaming laptops, their users deal with high noise output, but on a Macbook (Pro), chances are, users want more quiet operation and for that you need cooling material (bulk), not high air flow.
Macbook 2017 with touch bar (15") has 7820HQ, which is a 45W TDP CPU. It also has AMD Radeon Pro 560 discrete GPU. That GPU has 16 CUs. The forthcoming EMIB Radeon GPU will have 24 CUs. I would guess that this EMIB solution fits perfectly to the existing Macbook 2017 TDP range.
 
True, but it was quite loud under load as far as tests go I've read about it, hence my comment. My guess would be that we're seeing very much downclocked version in the mobile space if at all.
 
Back
Top