08-15-2017, 04:36 PM
Alright, I wanted to wait to reply to this thread until the reviews were live and we all had a chance to get a good range of data points to digest, I read every post daily, even though I don't often reply 
This has a couple of different elements to it, and both of them are rather important. One is the rate at which either division needs to advance, and one is a straight balance sheet issue.
If I were to tell someone today that I had an i5 3570K in my primary gaming rig, nobody would bat an eye, it wouldn't be given a second thought. That chip was mid range in 2012. Mid range in 2012 on the other end? GTX 660. You all, I would hope, would be busting my balls about running that dinosaur in anything with gaming rig in its duties in 2017. The performance disparity in most games between a 3570k and a top of the line current $1K Intel CPU(or AMD) is going to be less then 100% almost all of the time and the majority of the time it will be *significantly* less. The difference between a 660 and a 1080Ti? Can we just throw out 400% and agree that is conservative?
From an engineering perspective the CPU team was "chasing" a snail strolling up hill and spent years and billions of dollars to not quite do it. YoY CPU improvements have been well below 20% for a very long time now. The GPU team, on the other hand, is facing competition moving so fast that they have managed to swipe ~$1Billion in data center sales for the first half of this year with a fucking graphics card. MIT has named nVidia the smartest company on the planet(links have been giving me issues when posting, you can Google it, no joke).
So from an engineering perspective on one side you have a division using a couple billion to chase down that snail, and almost managed to do it. On the other side, you have a division that in the last five years went from neck and neck with the competition to being a year late, significantly more expensive to build and slower by a staggering margin(this is the engineering side- Vega 64 goes toe to toe with Titan, 56 with 1080Ti- that is engineering reality) while consuming a borderline insane amount of additional power.
To make matters worse, you are staring down a *major* architectural upgrade which will hit before you have production fully ramped likely relegating your halo product to somewhere in the 4th or 5th tier from the competition, and likely using about 20% of the power.
That is the first side to look at, and that is where they look best.
Now on the business side we are simply taking the realities of the financial aspects into account.
The Vega 56 costs more to produce then the TitanXP. That isn't spin, that isn't me trying to bash AMD, that is simply a point of fact. It is larger, uses far more expensive memory, uses considerably more power requiring more power circuitry and a better cooler.
The Vega 64 in any realistic sense is DoA, only drooling fanboys are going to touch it, and the Vega 56 is only somewhat viable because of the miners impact on 1070 pricing(MSRP is $350 for the 1070). Given that this product is over a year late, is more expensive to produce then the Titan and can only reasonably be sold for around $400, what is a realistic estimate on margins? I would say we could be very generous and call it 30%, that being very generous. Next we have some market numbers to look at-
http://www.anandtech.com/show/10613/disc...ins-on-top
The two elements we want to focus on are the total shipments and and the segment breakdown. Less then 10% of the market is in the enthusiast segment, probably closer to 5% but I'm trying to err on AMD's side here, so we'll call it 10%. Based on their sales, we'll round up for their benefit again, and we get 300K units per quarter. We are giving them 30% margins which works out to gross margins of roughly $45 Million per quarter.
To put that number into perspective nVidia's quarterly R&D budget was $85 million........... in 2005. They have now pushed over $400 million a quarter and implied that would quickly grow to over $500 million a quarter.
Now wait a second here Ben, you are leaving out that crazy console money, fair point, let's be very generous once again and give AMD a whole lot of loving on that end.
We are going to call it a $25 *margin* on every console sold for this entire generation and we will give all of the credit to the GPU side, sound good?
Right now we are sitting at 90 million consoles shipped for the two AMD powered platforms over the course of four years give or take. I'm going to do AMD a solid and I'm going to say they will move 200 million consoles over the course of seven years, sound generous enough? That gives them a whopping $5 billion in straight margins, broken down to a quarterly basis it puts us at about $179Million per- if we add that with Vega margins we end up at about half of what nVidia is using for R&D right now. Let's be clear, nVidia was making considerably less then that for margins on their contracts and both MS and Sony turned their nose up to nV this generation because they were too greedy, but like I said, I'm going all in on trying to prop up the AMD side of the equation here.
Now we still have the other 90% of the discrete market to look at, which is true, so let's take that 2.7 million units and we'll give them another 30% margin on an ASP of $250, $202Million more per quarter coming in for straight margins.
Anyone got any problems with my numbers? Anyone thinking I am being unfair to AMD in any way? I'm bending over backwards to give them every benefit I can, and these are the numbers I'm hitting.
In this dream land pie in the sky scenario- AMD can equal nV's current quarterly R&D budget and make *$0* profit. It actually turns negative if I get down to the final figures- they would be losing a few million a quarter with that rub one out for AMD fan boy dreamland scenario I spelled out.
So now we must put the two different factors together- the engineering one and the financial one. The engineering side is way behind nVidia, and normally that would indicate a need to increase R&D to a level that would allow you to catch up to your competition. If nVidia was a reasonably competent company, a reasonable assumption would be that matching their R&D would allow for you to close the gap, but this is "The Smartest Company in the World"- it stands as a reasonable stance to take that outspending them to close the gap would be logical.
That's where we run into the next problem. Intel, the 800lb gorilla of the computing world for decades, has outspent nVidia by orders of magnitude and is losing ground to them at a staggering rate. Intel had their pick of all the top engineering talent for many years, now nVidia has that luxury while AMD would likely be, at best, the fifth option on most enthusiasts lists of places to work for(Qualcomm, Apple, Samsung, Sony, MS etc).
If you are the CEO sitting back, going over the financials and combining that with the engineering reality that you part is roughly 40% slower *and* consumes 30% more power *and* is more expensive to produce- is that a path you want to try and rectify given the financial realities?
All of this is looking at it from AMD's perspective. nVidia is moving into so many different markets, their margin for error simply keeps growing- they have leveraged GPUs into the fastest growing commodity in the market, and have done so while continually setting new high marks for performance/watt and revenue generated per $ of R&D(by diversifying the segments effectively covered by their technology).
If you are sitting their as CEO of AMD today, do you really look at all these numbers and green light any new projects?
AMD got their modern shot at relevance due to the brilliant design team it acquired- ArtX. Those were the people responsible for the legendary R300- the first time we saw a company go toe to toe with current nVidia tech and have a compelling argument as to why there were flat out superior. This was a group of former SGi engineers, they designed the N64 rasterizer and had monopolized Nintendo set top designs since. So with that said, and with nVidia opting out of consoles altogether, what happened with this generation of Nintendo?
They went with a years old ultra mobile off the shelf chip from nVidia. This while requiring their markedly higher margins. Can AMD risk the other two heading in a similar direction? Vega vs Volta- which is where the majority of this generation will be fought, could realistically be seeing AMD's 350 watt parts struggling against nV's 100 watt offerings. As a console manufacturer, margins or not- where do you draw the line? Additional power usage requires better cooling, bigger power supply, lowers reliability- I mean, how much is worth it? What makes matters worse, nV already has laptop chips that can go toe to toe with Vega 64- and they are markedly cheaper to produce.
Honest analysis- we are staring down a situation(Volta vs Vega) where nVidia's mid tier laptop parts are likely to perform roughly equal to clearly superior to AMD's 350watt desktop parts. So we could be looking at a situation where nVidia is demanding 60% margins versus 30% for AMD and nV could end up being *markedly* cheaper.
Again, if you are the CEO- honestly assessing the competitive landscape and the realistic limitations of your GPU design team, are you going to spend that $400 million to try and go toe to toe with nVidia and, best case scenario- only lose tens of millons a year?
Quote:A. It's how they make a living. They're not going to just say, "The jig is up, we're beat. Time to fish.". They've lingered around in the CPU market for decades, rarely competing.
This has a couple of different elements to it, and both of them are rather important. One is the rate at which either division needs to advance, and one is a straight balance sheet issue.
If I were to tell someone today that I had an i5 3570K in my primary gaming rig, nobody would bat an eye, it wouldn't be given a second thought. That chip was mid range in 2012. Mid range in 2012 on the other end? GTX 660. You all, I would hope, would be busting my balls about running that dinosaur in anything with gaming rig in its duties in 2017. The performance disparity in most games between a 3570k and a top of the line current $1K Intel CPU(or AMD) is going to be less then 100% almost all of the time and the majority of the time it will be *significantly* less. The difference between a 660 and a 1080Ti? Can we just throw out 400% and agree that is conservative?
From an engineering perspective the CPU team was "chasing" a snail strolling up hill and spent years and billions of dollars to not quite do it. YoY CPU improvements have been well below 20% for a very long time now. The GPU team, on the other hand, is facing competition moving so fast that they have managed to swipe ~$1Billion in data center sales for the first half of this year with a fucking graphics card. MIT has named nVidia the smartest company on the planet(links have been giving me issues when posting, you can Google it, no joke).
So from an engineering perspective on one side you have a division using a couple billion to chase down that snail, and almost managed to do it. On the other side, you have a division that in the last five years went from neck and neck with the competition to being a year late, significantly more expensive to build and slower by a staggering margin(this is the engineering side- Vega 64 goes toe to toe with Titan, 56 with 1080Ti- that is engineering reality) while consuming a borderline insane amount of additional power.
To make matters worse, you are staring down a *major* architectural upgrade which will hit before you have production fully ramped likely relegating your halo product to somewhere in the 4th or 5th tier from the competition, and likely using about 20% of the power.
That is the first side to look at, and that is where they look best.
Now on the business side we are simply taking the realities of the financial aspects into account.
The Vega 56 costs more to produce then the TitanXP. That isn't spin, that isn't me trying to bash AMD, that is simply a point of fact. It is larger, uses far more expensive memory, uses considerably more power requiring more power circuitry and a better cooler.
The Vega 64 in any realistic sense is DoA, only drooling fanboys are going to touch it, and the Vega 56 is only somewhat viable because of the miners impact on 1070 pricing(MSRP is $350 for the 1070). Given that this product is over a year late, is more expensive to produce then the Titan and can only reasonably be sold for around $400, what is a realistic estimate on margins? I would say we could be very generous and call it 30%, that being very generous. Next we have some market numbers to look at-
http://www.anandtech.com/show/10613/disc...ins-on-top
The two elements we want to focus on are the total shipments and and the segment breakdown. Less then 10% of the market is in the enthusiast segment, probably closer to 5% but I'm trying to err on AMD's side here, so we'll call it 10%. Based on their sales, we'll round up for their benefit again, and we get 300K units per quarter. We are giving them 30% margins which works out to gross margins of roughly $45 Million per quarter.
To put that number into perspective nVidia's quarterly R&D budget was $85 million........... in 2005. They have now pushed over $400 million a quarter and implied that would quickly grow to over $500 million a quarter.
Now wait a second here Ben, you are leaving out that crazy console money, fair point, let's be very generous once again and give AMD a whole lot of loving on that end.
We are going to call it a $25 *margin* on every console sold for this entire generation and we will give all of the credit to the GPU side, sound good?
Right now we are sitting at 90 million consoles shipped for the two AMD powered platforms over the course of four years give or take. I'm going to do AMD a solid and I'm going to say they will move 200 million consoles over the course of seven years, sound generous enough? That gives them a whopping $5 billion in straight margins, broken down to a quarterly basis it puts us at about $179Million per- if we add that with Vega margins we end up at about half of what nVidia is using for R&D right now. Let's be clear, nVidia was making considerably less then that for margins on their contracts and both MS and Sony turned their nose up to nV this generation because they were too greedy, but like I said, I'm going all in on trying to prop up the AMD side of the equation here.
Now we still have the other 90% of the discrete market to look at, which is true, so let's take that 2.7 million units and we'll give them another 30% margin on an ASP of $250, $202Million more per quarter coming in for straight margins.
Anyone got any problems with my numbers? Anyone thinking I am being unfair to AMD in any way? I'm bending over backwards to give them every benefit I can, and these are the numbers I'm hitting.
In this dream land pie in the sky scenario- AMD can equal nV's current quarterly R&D budget and make *$0* profit. It actually turns negative if I get down to the final figures- they would be losing a few million a quarter with that rub one out for AMD fan boy dreamland scenario I spelled out.
So now we must put the two different factors together- the engineering one and the financial one. The engineering side is way behind nVidia, and normally that would indicate a need to increase R&D to a level that would allow you to catch up to your competition. If nVidia was a reasonably competent company, a reasonable assumption would be that matching their R&D would allow for you to close the gap, but this is "The Smartest Company in the World"- it stands as a reasonable stance to take that outspending them to close the gap would be logical.
That's where we run into the next problem. Intel, the 800lb gorilla of the computing world for decades, has outspent nVidia by orders of magnitude and is losing ground to them at a staggering rate. Intel had their pick of all the top engineering talent for many years, now nVidia has that luxury while AMD would likely be, at best, the fifth option on most enthusiasts lists of places to work for(Qualcomm, Apple, Samsung, Sony, MS etc).
If you are the CEO sitting back, going over the financials and combining that with the engineering reality that you part is roughly 40% slower *and* consumes 30% more power *and* is more expensive to produce- is that a path you want to try and rectify given the financial realities?
All of this is looking at it from AMD's perspective. nVidia is moving into so many different markets, their margin for error simply keeps growing- they have leveraged GPUs into the fastest growing commodity in the market, and have done so while continually setting new high marks for performance/watt and revenue generated per $ of R&D(by diversifying the segments effectively covered by their technology).
If you are sitting their as CEO of AMD today, do you really look at all these numbers and green light any new projects?
AMD got their modern shot at relevance due to the brilliant design team it acquired- ArtX. Those were the people responsible for the legendary R300- the first time we saw a company go toe to toe with current nVidia tech and have a compelling argument as to why there were flat out superior. This was a group of former SGi engineers, they designed the N64 rasterizer and had monopolized Nintendo set top designs since. So with that said, and with nVidia opting out of consoles altogether, what happened with this generation of Nintendo?
They went with a years old ultra mobile off the shelf chip from nVidia. This while requiring their markedly higher margins. Can AMD risk the other two heading in a similar direction? Vega vs Volta- which is where the majority of this generation will be fought, could realistically be seeing AMD's 350 watt parts struggling against nV's 100 watt offerings. As a console manufacturer, margins or not- where do you draw the line? Additional power usage requires better cooling, bigger power supply, lowers reliability- I mean, how much is worth it? What makes matters worse, nV already has laptop chips that can go toe to toe with Vega 64- and they are markedly cheaper to produce.
Honest analysis- we are staring down a situation(Volta vs Vega) where nVidia's mid tier laptop parts are likely to perform roughly equal to clearly superior to AMD's 350watt desktop parts. So we could be looking at a situation where nVidia is demanding 60% margins versus 30% for AMD and nV could end up being *markedly* cheaper.
Again, if you are the CEO- honestly assessing the competitive landscape and the realistic limitations of your GPU design team, are you going to spend that $400 million to try and go toe to toe with nVidia and, best case scenario- only lose tens of millons a year?