News AMD Q3 Earnings Results

[Hitman928]

EPS is $0.75, beat by $0.14
Revenue is $4.31B, beat by $200M
Q4 revenue projection is $4.5B, above $4.25B estimates.
Q4 GM projection is 49.5%

For the full year 2021, AMD now expects revenue to grow approximately 65 percent driven by growth across all businesses, up from prior guidance of 60 percent growth. AMD expects non-GAAP gross margin to be approximately 48 percent for the full year 2021.

Advanced Micro Devices EPS beats by $0.07, beats on revenue

Advanced Micro Devices (AMD): Q3 Non-GAAP EPS of $0.73 beats by $0.07; GAAP EPS of $0.75 beats by $0.14.Revenue of $4.31B (+53.9% Y/Y) beats by $200M.Press ReleaseQ3 non-GAAP gross...

seekingalpha.com

More to come.

 

[Hitman928]

Why would consoles need the latest node? They didn't even get the latest ZEN gen when it was already out.

If intel needs customers for their new fab business, even if it is just to advertise with them, then they might just do it(agree to a low price) for that alone.
Also intel could do a 2C/6c CPU and increase gaming performance due to much much higher single thread, that would be relatively cheap for intel and even on 14nm it would be good enough since the e cores would use much less power than full cores, if the Xe could keep up it would be a good package.

You think Intel could put together a solution for next gen consoles using their 14 nm process that would some how outperform their current gen consoles on TSMC 7? Never going to happen.

Why, because they did supply the CPU in as large a quantity as MS asked for?
Maybe some engineers are still sour about it because they planned everything for AMD but why would MS as an company care?
Also everybody seems to think that wintel is a thing and that would be a, by far, bigger bond and then there is xbox as an service because that's where MS seems to be steering towards, everything running on windows on mostly intel systems anyway.

Reports were that Intel was both unwilling to make requested changes to the IP to tailor it for MS's needs as well as work with MS to reduce costs over time. Basically the attitude from Intel was you'll take what we give you and like it or find another supplier. So MS found another supplier who was much more willing to work in partnership with them in the design and production of the required chips, first in IBM and then in AMD.

 

[sandorski]

Intel would have to make one hell of a chip to take the Console Market, unless MS/Sony have some kind of issue with AMD. I could see one of them going Intel if it is competitive and they want to differentiate themselves, but the AMD platform has been solid and they are familiar with it.

 

[maddie]

Intel would have to make one hell of a chip to take the Console Market, unless MS/Sony have some kind of issue with AMD. I could see one of them going Intel if it is competitive and they want to differentiate themselves, but the AMD platform has been solid and they are familiar with it.

Unstated, but they also share at least part of the development costs with each other at a deep level, then they start to differentiate.

 

[lobz]

Yes in deed, if AMD asks for more money while intel comes out with Xe, big.bigger, with opening fab to 3rd party and with having a much better supply it might look very different but not necessarily in favor of AMD.
Intel might even swallow all the R&D money to develop a console SOC just to be a dick to AMD.
As long as they keep performance per core/thread and GPU performance the same there would be zero need to change anything else other than maybe some driver support.

in a parallel universe all these things could sound perfectly feasible

 

[gdansk]

Nvidia is currently more positioned to compete in the console market. They could offer customized variants of their automotive line products, for example. Intel has nothing yet competitive. I'm not sure it's worth competing in that market for Intel unless they cannot find other customers for their products.

 

[sandorski]

Nvidia is currently more positioned to compete in the console market. They could offer customized variants of their automotive line products, for example. Intel has nothing yet competitive. I'm not sure it's worth competing in that market for Intel unless they cannot find other customers for their products.

Nvidia's problem is they are difficult to deal with.

 

[TheELF]

You think Intel could put together a solution for next gen consoles using their 14 nm process that would some how outperform their current gen consoles on TSMC 7? Never going to happen.

Well if the starting argument was for AMD to wait until the next gen to renegotiate their deal then forget about it, that's going to be a lot of years until that happens.

 

[BorisTheBlade82]

So if that entire division was $1.9 billion, then all the consoles were at most $800 million.
Using up all those scarce TSMC 7nm wafers for so little revenue is crazy.

I suspect AMD are keen to hide how little the console make them, but is there enough information there for us to try and deduce some figures?

Something like taking that $800 million and dividing it by the combined numbers for PS5 and both the Xbox'es for the quarter?

Well, we could do some arguably extreme calculation in order to assess the impact on turnover and ASP:
The console-SoCs have an area of around 300mm2. So with that area you could produce 3,x Zen3 CCDs not even taking into account the much better yield due to decreased size. What will a console SoC be priced? IMHO not much more than 200$ when looking at the other components and final prices for the consoles. But each of the CCD could go into an EPYC where demand seems much higher than supply. And as such they could earn 500$ per die. So basically a loss of 1300$ per console SoC.
I guess, that even AMD did not know some 5 years ago, how much demand and how tight a supply would exist today.

 

[Doug S]

Why?

Dividends and stock buybacks are equivalent in every way except tax consequences, which currently favor buybacks for many investors. It's one thing to complain a company should have spent money for growth instead of returning it to the investors, but complaining about the method they did the latter is just incomprehensible to me.

Not true, there is one other very important difference. If you hold stock options buybacks make your options more valuable, while dividends do nothing for you. Or if you don't believe held cash is discounted in the trading price of a stock, buybacks are neutral while dividends make your options worth less - because money is leaving the company to shareholders but not to you.

Now ask yourself who is likely to be holding a lot of stock options, and who is making the buyback vs dividend decisions?

 

[Hitman928]

Well if the starting argument was for AMD to wait until the next gen to renegotiate their deal then forget about it, that's going to be a lot of years until that happens.

Of course it would have to be next gen. There's no way the console makers would switch suppliers and architectures mid-gen. That would be unheard of and would create a whole host of issues. Even if they wanted to, there's nothing Intel could offer today on 14 nm that could compete with what they already have from AMD.

 

[Saylick]

Nvidia's problem is they are difficult to deal with.

That, and they like margins. Like, a lot. So much so that they'd really never make a console SoC unless it could be made super cheap (i.e. on a deprecated process) and sold fore more, OR the vendor could somehow package it into a premium console and charge consumers premium prices. Unfortunately, consoles are traditionally a price-sensitive product and success depends on having a broad install base so that software sales can be made. Hence, the only console products Nvidia have provided are for the Switch, which was not even a custom SoC and was on an old process.

 

[moinmoin]

So if that entire division was $1.9 billion, then all the consoles were at most $800 million.
Using up all those scarce TSMC 7nm wafers for so little revenue is crazy.

I suspect AMD are keen to hide how little the console make them, but is there enough information there for us to try and deduce some figures?

Something like taking that $800 million and dividing it by the combined numbers for PS5 and both the Xbox'es for the quarter?

AMD never was in the position of using those wafers for its own products. It's best to consider wafers used for console chips to be ordered by Sony and Microsoft using AMD as a middleman.

Also AMD has to hide the console numbers somehow because otherwise Sony and Microsoft could deduce each others' numbers from them which is neither in the interest of these two (internal business decisions) nor AMD (independent contract negotiations with these two).

In pretty much every earnings report, analysts have poked AMD about their low and slowly increasing margin rate, but like you said they are trying to raise it by increasing the average sale price across the board and to increase the mix of sales towards products with naturally higher margins, e.g. enterprise and server.

High margins are what made Intel incapable of expanding beyond its core competence, eventually shuttering any effort beyond it. It also caused Intel to milk 14nm for all it was worth and spend as little as possible for the floundering foundry efforts. Now the margin sinks like a stone and Intel is in a hurry to right the ship.

I feel AMD is aware of the pitfalls high margins bring and purposefully tries to keep them and the growth thereof in a controllable range.

It's definitely lower than Intel and Nvidia, and it's largely due to them having to historically offer products with better perf per dollar and/or lower margins in order to take market share. (...) Intel appears to be raising their prices for Alderlake to match AMD, and there's rumors that the next generation of GPUs from all three vendors will be even more expensive. Due to every player offering more expensive products, I think it's fair to say that the days of low margin are going to be behind AMD for the foreseeable future.

I don't know about GPUs, but MSRPs for AMD CPUs didn't change perf per dollar wise, AMD still offered better perf per dollar, and Intel hasn't adapted to that yet. We will see if Alderlake changes that.

Nvidia is currently more positioned to compete in the console market.

Nvidia currently competes excellently as part of Nintendo's popular Switch family of hybrid consoles.

 

[Joe NYC]

Going into the earnings, I was really wondering if AMD was going to revel if all the projects are on track, looks like everything is on track.

Whew, what a relief!

While I have no idea what "on track" means, and we don't have any roadmap, but good to know that things are on track.

 

[eek2121]

Nvidia is currently more positioned to compete in the console market. They could offer customized variants of their automotive line products, for example. Intel has nothing yet competitive. I'm not sure it's worth competing in that market for Intel unless they cannot find other customers for their products.

Nvidia's problem is they are difficult to deal with.

That, and they like margins. Like, a lot. So much so that they'd really never make a console SoC unless it could be made super cheap (i.e. on a deprecated process) and sold fore more, OR the vendor could somehow package it into a premium console and charge consumers premium prices. Unfortunately, consoles are traditionally a price-sensitive product and success depends on having a broad install base so that software sales can be made. Hence, the only console products Nvidia have provided are for the Switch, which was not even a custom SoC and was on an old process.

As stated above, NVIDIA is already in the console market. They make the SoC in the Switch.

NVIDIA does like margins, however. They probably welcomed business from Nintendo to add volume to a low volume sector.

 

[gdansk]

As stated above, NVIDIA is already in the console market. They make the SoC in the Switch.

Surely they are all aware. I was considering companies who would compete to take existing customers from AMD. And I would consider the Switch evidence that Nvidia is more likely to compete than Intel.

 

[DrMrLordX]

Intel would have to make one hell of a chip to take the Console Market, unless MS/Sony have some kind of issue with AMD. I could see one of them going Intel if it is competitive and they want to differentiate themselves, but the AMD platform has been solid and they are familiar with it.

Intel won't enter the console market due to the low margins involved. Also, MS and Sony effectively buy the wafers using designs provided to them by AMD, which is something that wouldn't (necessarily) work with Intel unless they fully spin off their fabs into a separate company. It was perfect for AMD's semi-custom unit.

While I have no idea what "on track" means, and we don't have any roadmap, but good to know that things are on track.

It means AMD will wait as late as possible to tell anyone anything about upcoming products.

As stated above, NVIDIA is already in the console market. They make the SoC in the Switch.

Correction. Nintendo uses a slightly-modified version of the Tegra X1, a design from early 2015. NV is discontinuing production of X1 soon:

Nintendo Says the OLED Switch Uses Same Tegra X1 Processor, RAM

Nvidia is ending production of the Switch's Tegra X1 Mariko processor this year.

www.pcmag.com

It's hard to say that NV is "in the console market" when the only company that will use their components is still using an off-the-shelf chip that's about to leave production. Nintendo could have switched to newer Tegra variants ages ago, but chose no to do so.

 

[moinmoin]

It's hard to say that NV is "in the console market" when the only company that will use their components is still using an off-the-shelf chip that's about to leave production.

A piece by Nvidia is in every single Switch. Nvidia earns something with every single Switch. What is your definition of "in the console market"?

 

[DrMrLordX]

A piece by Nvidia is in every single Switch. Nvidia earns something with every single Switch. What is your definition of "in the console market"?

If they were building custom or semi-custom hardware for their clients, and/or if their clients were buying their latest designs (or at least the most-recent designs that NV would produce for a console customer; e.g. one should not expect an RTX3090 in an NV-based console).

Nintendo wouldn't even touch the X2.

 

[Asterox]

After the original Zen /Ryzen launch, i remember a lot of people have said that AMD stock will never jump more than $30.

https://finance.yahoo.com/quote/AMD/

 

[moinmoin]

If they were building custom or semi-custom hardware for their clients, and/or if their clients were buying their latest designs (or at least the most-recent designs that NV would produce for a console customer; e.g. one should not expect an RTX3090 in an NV-based console).

Nintendo wouldn't even touch the X2.

Because X2 with those custom Denver cores is a deadend even Nvidia likely wants to avoid at this point. Nintendo on the other hand is all in into Nvidia's ecosystem, using the whole development chain (which is a huge improvement over what they had before). That alone makes it unlikely that Nintendo is moving away from Nvidia anytime soon. Considering the success Switch is so far, and the fact that Nvidia is essentially removing X1 from the open market, I expect Nvidia to work on a suitable successor chip that's then used only in a Switch follow on.

Also note how Nvidia never put its high end graphics in a console. It had its chance with PS3, and that arguably worked out even worse than off-the-shelves X1.

 

[DrMrLordX]

Because X2 with those custom Denver cores is a deadend even Nvidia likely wants to avoid at this point.

Carmel was also an option.

If Nintendo decides to finally update their SoC and goes with an NV product, then we can at least acknowledge that NV has a lasting footprint in the console market. But I'm highly suspicious of Nintendo's stance on continuing use of NV products when they deliberately stuck with a soon-to-be-EoLed SoC instead of moving on to something like Carmel immediately.

It had its chance with PS3, and that arguably worked out even worse than off-the-shelves X1.

PS3 was never supposed to need a dGPU, but Cell proved inadequate to the task of compute AND graphics. NV was the backup plan. It's not surprising that that relationship was short-term.

 

[TheELF]

Of course it would have to be next gen. There's no way the console makers would switch suppliers and architectures mid-gen. That would be unheard of and would create a whole host of issues. Even if they wanted to, there's nothing Intel could offer today on 14 nm that could compete with what they already have from AMD.

What architecture would they switch? From x86 to x86 or from dx to dx (whatever graphics API sony uses)
The consoles are running an OS since several gens now, it's all about the APIs.

Even if they wanted to, there's nothing Intel could offer today on 14 nm that could compete with what they already have from AMD.

i would need to see numbers for that and so would sony and ms, consoles don't run cinebench or avx512, I have no idea how well rocket lake would perform in games at that power envelope, do you?

 

[TheELF]

After the original Zen /Ryzen launch, i remember a lot of people have said that AMD stock will never jump more than $30.

https://finance.yahoo.com/quote/AMD/

Well sane people also said that gamestop stocks will never rise and that they are going out of business...
The stock market often doesn't make any sense at all.
Intel stock was around $30-40 at that time and intel is an established company for being very stable, it doesn't make any sense for an incredibly more financially unstable company having higher stock prices...unless it's because of people gambling, it's the same reason that crypto currency has a high price, everybody hopes that it will go higher but it's an incredibly dangerous bet to make.

 

[NTMBK]

PS3 was never supposed to need a dGPU, but Cell proved inadequate to the task of compute AND graphics. NV was the backup plan. It's not surprising that that relationship was short-term.

Not according to one of the architects of the Cell. In "Race For A New Game Machine", David Shippy says that Sony had an internal team working on a graphics processor for the PS3. That project failed, and Nvidia were brought in... but Cell wasn't even meant to be the only processor.

 

[Tuna-Fish]

Not according to one of the architects of the Cell. In "Race For A New Game Machine", David Shippy says that Sony had an internal team working on a graphics processor for the PS3. That project failed, and Nvidia were brought in... but Cell wasn't even meant to be the only processor.

Wasn't it kind of yes and no? IIRC the plan was for the Cell to do the early part of the pipeline, and then have a really simple yet beefy "GPU" that just did the rasterization and pixel shading? This way, the part of the pipeline that required high-precision compute would be in the Cell, and then the gpu part would just do fixed-function and short integer pixel shader math. This would let it be both super cheap and super fast. And the way it failed was that they found that while, yes, a Cell could in principle do vertex shading and the necessary setup steps, there was no way in hell they could move that data out of the cell and into the GPU fast enough to be competitive with GPUs that did the whole shebang.