News AMD 2Q20 Report

Hitman928

Diamond Member
Apr 15, 2012
5,177
7,629
136
EPS 13 cents, beat by 1 cent.
Revenue 1.93 billion, beat by 70 million.

developing. . .
 

Hitman928

Diamond Member
Apr 15, 2012
5,177
7,629
136
“We delivered strong second quarter results, led by record notebook and server processor sales as Ryzen and EPYC revenue more than doubled from a year ago,” said Dr. Lisa Su, AMD president and CEO. “Despite some macroeconomic uncertainty, we are raising our full-year revenue outlook as we enter our next phase of growth driven by the acceleration of our business in multiple markets.”

Computing and Graphics segment revenue was $1.37 billion, up 45 percent year-over-year and down 5 percent quarter-over-quarter. Revenue was higher year-over-year driven by strong Ryzen processor sales. The quarter-over-quarter decline was due to lower graphics processor sales.

Enterprise, Embedded and Semi-Custom segment revenue was $565 million, down 4 percent year-over-year and up 62 percent quarter-over-quarter. Revenue was lower year-over-year due to lower semi-custom product sales largely offset by higher EPYC processor sales. The quarter-over-quarter increase was driven by higher EPYC processor and semi-custom product sales.

For the third quarter of 2020, AMD expects revenue to be approximately $2.55 billion, plus or minus $100 million, an increase of approximately 42 percent year-over-year and 32 percent sequentially. The year-over-year and sequential increases are expected to be primarily driven by Ryzen and EPYC processor sales and next generation semi-custom products. AMD expects non-GAAP gross margin to be approximately 44 percent in the third quarter of 2020. Gross margin is expected to increase year-over-year primarily driven by Ryzen and EPYC processor sales.

AMD now expects 2020 revenue to grow by approximately 32 percent compared to 2019 driven by strength in PC, gaming and data center products. Non-GAAP gross margin is expected to be approximately 45 percent.

 
Last edited:

DisEnchantment

Golden Member
Mar 3, 2017
1,590
5,722
136
Some highlights
  • CDNA this quarter.
  • 7nm is tight.
  • Console will ramp mainly in Q3 and slower towards the end of H2.
  • Zen3/RDNA2 will ramp in Q4.
  • Rome will continue through the next year.
  • RDNA2 will be top to bottom refresh of the stack.
  • 1.5 Billion USD in inventory to support launches in next quarters.
 

maddie

Diamond Member
Jul 18, 2010
4,723
4,628
136
Some highlights
  • CDNA this quarter.
  • 7nm is tight.
  • Console will ramp mainly in Q3 and slower towards the end of H2.
  • Zen3/RDNA2 will ramp in Q4.
  • Rome will continue through the next year.
  • RDNA2 will be top to bottom refresh of the stack.
  • 1.5 Billion USD in inventory to support launches in next quarters.
So, was the rumor of client Zen3 on 5nm in 2021 still a fantasy? What do you think?
 

DisEnchantment

Golden Member
Mar 3, 2017
1,590
5,722
136
So, was the rumor of client Zen3 on 5nm in 2021 still a fantasy? What do you think?
It is very unlikely. Maybe a refresh in H2 21?
But their roadmap is packed. Don't know if they can squeeze 5nm Zen3 refresh in there.
Bigger issue now seems to be capacity, they are supply constrained. Lisa said they are working to increase capacity.
Like I have alluded before, there is some truth to the rumor about PS5 launch numbers and Lisa mentioned they are working to meet increased demand.
Basically there is little capacity left in Q3.
But since we are almost in August we should be hearing about Zen3/RDNA2 launch in 5-6 weeks and available 4-5 weeks thereafter. It is not far off.
(I ordered my 3900X in early July and got it late Sept last year... so don't know how it is going to be this year)
 
Last edited:

Saylick

Diamond Member
Sep 10, 2012
3,084
6,184
136
Listening the call, seems AMD is VERY capacity limited
Everyone wants a piece of TSMC's 7nm node, including Intel, but at the same time, I don't fault Dr. Su and AMD for ordering a conservative number of wafers from TSMC. Better to forecast a conservative number of wafers needed to fulfill demand than to risk over-ordering and ending up with a surplus/inventory stockpile that they end up writing off. It does end up putting a choker on their growth if demand exceeds their expectations but better to be in that predicament than to overestimate demand in my opinion.
 

amd6502

Senior member
Apr 21, 2017
971
360
136
Listening the call, seems AMD is VERY capacity limited

They have all the 12nm they want. They should focus on chiplets which are very efficient use of 7nm die are. Picasso for value oriented. And Pinnacles can still grab some mainstream DIY market if priced right.

Desktop oriented chiplet based APU would be good with 12nm Vega + IO hub. But who knows they may have had little idea that 7nm would be tight ~1.5 years ago when that ship sailed.
 
  • Like
Reactions: Olikan

amd6502

Senior member
Apr 21, 2017
971
360
136
Is there any chance GloFo could upgrade some of their fabs and license 7nm?

I'm also curious, does anybody know what sort of new products they are fabbing in big volumes?
 

NostaSeronx

Diamond Member
Sep 18, 2011
3,683
1,218
136
Is there any chance GloFo could upgrade some of their fabs and license 7nm?
By themselves no, so far the only group that is willing to invest is KIC/Samsung/Hynix.
I'm also curious, does anybody know what sort of new products they are fabbing in big volumes?
Do to the drainage of customers from 14LPP/14HP/12LP/12LP+, GloFo is re-establishing their 28BLK -> 22FDX -> 12FDX roadmap.

Skywater FDSOI nodes -> Samsung FDSOI Node -> 22FDX -> Samsung FDSOI Shrink -> 12FDX -> CEA-Leti FDSOI nodes.

HVM at 22FDX aren't big products yet however there are; Rockchip, Synaptics, etc
 
Last edited:
  • Like
Reactions: Tlh97 and amd6502

ksec

Senior member
Mar 5, 2010
420
117
116
Everyone wants a piece of TSMC's 7nm node, including Intel, but at the same time, I don't fault Dr. Su and AMD for ordering a conservative number of wafers from TSMC. Better to forecast a conservative number of wafers needed to fulfill demand than to risk over-ordering and ending up with a surplus/inventory stockpile that they end up writing off. It does end up putting a choker on their growth if demand exceeds their expectations but better to be in that predicament than to overestimate demand in my opinion.

This, it is not TSMC is capacity limited, it is AMD being conservatives with estimate.

And whether that is AMD's fault is a matter of one's opinion.

But Seriously $500M in Datacenter Segment ( Assuming Zero Consoles Sales ) is not good enough in my opinion.

On one hand they are a far too conservatives, they have reached their low target of 10% Enterprise Market Share, on the other hand, AMD has a very tight budget and margin of error. Hopefully their next cycle is much better with Zen 3.
 
  • Like
Reactions: spursindonesia

ksec

Senior member
Mar 5, 2010
420
117
116
Is there any chance GloFo could upgrade some of their fabs and license 7nm?

I'm also curious, does anybody know what sort of new products they are fabbing in big volumes?

Their New York Fabs could, theoretically from Samsung. But you then run into the problem of getting ASML's EUV TwinScan which is behind their shipping schedule as well due to pandemic and are extremely overbooked.

Not to mention without a valid path and sustainable business case upgrading the New York Fab is a waste of money.
 

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
@ksec

I would be surprised if anyone let GF back into the foundry game. Would be funny if it happened twice though. Samsung already bailed out GF once with 14LPP.

Otherwise:

I'm a little concerned about AMD coming out and stating that they'll continue Rome deliveries. I know hardware qualification is a lengthy process, but pushing customers onto Milan should be top priority. Very interested in what CDNA brings to the table. I fear us plebs may not get much data on it though.
 
  • Like
Reactions: Tlh97

Kedas

Senior member
Dec 6, 2018
355
339
136
Zen4 5nm end next year. (2021) :astonished:
We haven't even seen Zen 3 yet this year.

I assume Zen4 will be an improved Zen3 version with focus on 7nm to 5nm.

I just hope they don't call the Zen3 APU next year the 5000 series because that would mean AM4 and AM5 in the same series line.
Now it's confusing but at least it's still always for the DDR4 AM4 motherboard.
They are smart, make it 4700G-Z3 or 4705G or something (pick a crazy number just for one year), so we can all happily jump to AM5, DDR5 with the 5000 series on 5nm.
 
  • Like
Reactions: Tlh97 and amd6502

ksec

Senior member
Mar 5, 2010
420
117
116
@ksec

I would be surprised if anyone let GF back into the foundry game. Would be funny if it happened twice though. Samsung already bailed out GF once with 14LPP.

Otherwise:

I'm a little concerned about AMD coming out and stating that they'll continue Rome deliveries. I know hardware qualification is a lengthy process, but pushing customers onto Milan should be top priority. Very interested in what CDNA brings to the table. I fear us plebs may not get much data on it though.

Yes. And Samsung already has problem filling up its Fab. Helping GF doesn't really help them in anyway.

I am, concern ( or not understand ) about Rome as well. But I am waiting for more Milan information before making any sort of judgement or opinion. I absolutely agree on pushing ALL to Milan should be top priority.

Isn't CDNA just a rebranded Vega? ( Or at least that was what I thought ). Is it something entirely new?
 

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
Isn't CDNA just a rebranded Vega? ( Or at least that was what I thought ). Is it something entirely new?

CDNA is the next evolution of Vega. While RDNA/RDNA2 is more aimed at consumer graphics with some compute capability, CDNA is aimed squarely at compute. It is not even clear if AMD intends for CDNA-based accelerator cards to have any display or rendering capabilities. Which means RDNA2 would be aimed at GPU render boxes/farms while CDNA would be used for workloads currently foisted onto GPGPU clusters. But that is just conjecture. At the present, nobody knows exactly what CDNA will be like, or if CDNA variants will find their way into APUs like Cezanne. Since Cezanne is apparently NOT RDNA-based.
 

Atari2600

Golden Member
Nov 22, 2016
1,409
1,655
136
I'm a little concerned about AMD coming out and stating that they'll continue Rome deliveries. I know hardware qualification is a lengthy process, but pushing customers onto Milan should be top priority. Very interested in what CDNA brings to the table. I fear us plebs may not get much data on it though.

I wouldn't be concerned at all about that - indeed I'd be somewhat encouraged that they see legacy demand as significant enough to entail continued production.

Don't forget, price is a function of supply & demand - not necessarily outright performance.

If businesses have partially rolled out Rome, see great performance in it, and want to fully roll it out - and either skip Milan or upgrade later - then AMD can still significantly charge them for it.

Consider a business that has just switched to AMD after a decade (or more) of Xeon. They want to buy more Rome to mirror what they have and cannot obtain it - how do you think that would make them think of AMD's supply reliability in future?


You CANNOT push customers onto using (and therefore requalifiying) a new mission critical CPU. You'll only end up with them going back to Intel for dependable, predictable supply. Enticement is the name of the game, a carrot and stick approach does not work, only the carrot.
 

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
You CANNOT push customers onto using (and therefore requalifiying) a new mission critical CPU. You'll only end up with them going back to Intel for dependable, predictable supply. Enticement is the name of the game, a carrot and stick approach does not work, only the carrot.

Think you kinda got it backwards. Selling more Rome makes it easier for people to stay with Intel. If there really is a situation where customers are being sticks-in-the-mud and saying "we will buy Rome, not Milan!" then it means AMD has done a poor job of getting QS where they need to go. That is what I mean.