AMD Q1 2015 Earnings - 23 cents a share loss, to exit dense server (SeaMicro)

[Qianglong]

Lets hope the PS4 and XBONE do well in China which is one of the few consistent revenue source for AMD, albeit not that much.

 

[AtenRa]

Well, AMD has to prove everybody now that it can deliver. Of course, you can look at it both ways, pessimistic or optimistic way. That's fine by me

What nobody have realized so far, is that AMD has managed to have the same Throughput per single Core/thread with a fewer execution units/resources on the Bulldozer architecture (that was not mend for high single thread performance) against Star/Thuban in Phenom II CPUs.

Now, since ZEN is a new design with different architecture and from what it seams it will utilize a SMT design, Single thread Throughput will also have a tremendous increase against the Bulldozer based Kaveri/Carriso. Not only that, but Perf/watt will increase exponentially due to the new design and the 14nm FF node.

The new architecture will be effective both in server/high-end Desktop but it will also be very competitive in Mobile (Laptops) with an added bonus of a new highly efficiently new GPU architecture. The combination of a new CPU core with new GPU architecture and the 14nm FF node (not to mention HBM) has all the right things to make the 2016 AMD APUs perhaps the best products for the Mobile (Laptops) segment(10-35W TDP).

 

[Phynaz]

Lets hope the PS4 and XBONE do well in China which is one of the few consistent revenue source for AMD, albeit not that much.

How did the consoles work out for IBM?

They didn't save IBM's chip business and they won't save AMD's.

 

[Phynaz]

What nobody have realized so far, is that AMD has managed to have the same Throughput per single Core/thread with a fewer execution units/resources on the Bulldozer architecture (that was not mend for high single thread performance) against Star/Thuban in Phenom II CPUs.

Now, since ZEN is a new design with different architecture and from what it seams it will utilize a SMT design, Single thread Throughput will also have a tremendous increase against the Bulldozer based Kaveri/Carriso. Not only that, but Perf/watt will increase exponentially due to the new design and the 14nm FF node.

The new architecture will be effective both in server/high-end Desktop but it will also be very competitive in Mobile (Laptops) with an added bonus of a new highly efficiently new GPU architecture. The combination of a new CPU core with new GPU architecture and the 14nm FF node (not to mention HBM) has all the right things to make the 2016 AMD APUs perhaps the best products for the Mobile (Laptops) segment(10-35W TDP).

You sure state a lot of absolutes. Are these facts or your opinion?

 

[AtenRa]

We already know how the current products will do. Just look the guidance provided by AMD.

Lisa Su - President and CEO
On the positive side, we did see progress in several of our strategic initiatives. Mobile APU ASPs and revenue increased from the year-ago period, highlighted by increases in commercial client APU shipments and revenue from the fourth quarter, setting a record for commercial client processor sales. Our focused commercial client strategy is gaining momentum and our investments are driving awareness and generating pull with commercial and government buyers.

As i have explained before, the big decrease in Revenue was from the Desktop sales (Both AMD and Intel) in Q1 2015. Laptop volume and Revenue were fine for both.
Carriso will have a big influence in Q2 and Q3 2015 as OEMs are preparing for Back to School and Win 10.

Lisa Su - President and CEO
Looking forward in the year, given the ongoing macroeconomic and currency uncertainties, it is hard to predict when the PC environment will normalize. We expect that the overall PC market will remain a challenge as our OEM customers and channel partners focus on carrying lean inventories based on the uncertain market conditions. However, we are preparing for a better second half of the year, with Windows 10 and our new product offerings as catalysts for the business.
Regardless of market conditions, we are taking important steps to improve our CG business by completing our channel inventory rebalancing, introducing strong new APU and graphics products, and continuing to grow commercial client sales.

Devinder Kumar - SVP and CFO

Now turning to the outlook. For the second quarter of 2015, AMD expects revenue to decrease 3% sequentially plus or minus 3%.

 

[AtenRa]

You sure state a lot of absolutes. Are these facts or your opinion?

Those about Kaveri and essentially Carriso are facts made from personal measurements. You can see here how much better the 45W TDP Kaveri (28nm Planar) is against 100W TDP Star (32nm SOI)

For ZEN, its coming from what we know so far. New SMT architecture that means wider single Core/thread in order to accommodate the SMT second thread and a new GPU architecture since AMD have stated they will have a new GPU arch (for the APUs as well) every two years. Last one was GCN 1.1/1.x in Kaveri/Carisso.

 

[ShintaiDK]

As i have explained before, the big decrease in Revenue was from the Desktop sales (Both AMD and Intel) in Q1 2015. Laptop volume and Revenue were fine for both.
Carriso will have a big influence in Q2 and Q3 2015 as OEMs are preparing for Back to School and Win 10.

AMD CPU and GPU division went from 861M$ revenue to 532M$ revenue YoY. Intels went from 11184M$ to 11099M$ YoY.

And you still think its positive?

How much of those 532M$ are GPU and CPU? Considering AMD is slaughtered in the GPU segment as well. We can say maybe 350-400M$ revenue is from the CPU part. That leaves AMD with a best case scenario of 3.6% of the revenue. They are the new VIA. Irrelvant and on an ever decline.

Remember Q2 is a growth quarter compared to Q1. And AMD expects a 3% decline from Q1.

 

[Ajay]

For ZEN, its coming from what we know so far. New SMT architecture that means wider single Core/thread in order to accommodate the SMT second thread and a new GPU architecture since AMD have stated they will have a new GPU arch (for the APUs as well) every two years. Last one was GCN 1.1/1.x in Kaveri/Carisso.

That's not what an SMT architecture 'means'. If AMD has dropped CMT and gone with a more traditional 'big-core' architecture - then ST performance should increase. SMT uarch's like Intel's HT improve core utilization by allowing another thread to run whenever the current active thread stalls (usually to fetch data for a register). Intel's HT has become less effective as Intel has improved ILP.

I'd like to know more about Zen, but I don't know if AMD has made any concrete statements about the changes yet (and I wouldn't expect them to for a 2016 product).

 

[MiddleOfTheRoad]

How did the consoles work out for IBM?

They didn't save IBM's chip business and they won't save AMD's.

IBM was a massively larger company. If AMD can survive this downturn and downsize to an nimble tech firm -- they could make it work.

I think the future for AMD looks a lot more like Qaulcomm and a lot less like Intel. I'm not sure if they have enough cash to pull it off.

 

[AtenRa]

That's not what an SMT architecture 'means'. If AMD has dropped CMT and gone with a more traditional 'big-core' architecture - then ST performance should increase. SMT uarch's like Intel's HT improve core utilization by allowing another thread to run whenever the current active thread stalls (usually to fetch data for a register). Intel's HT has become less effective as Intel has improved ILP.

I'd like to know more about Zen, but I don't know if AMD has made any concrete statements about the changes yet (and I wouldn't expect them to for a 2016 product).

SMT works better with wider cores because its able to use the resources left from the first thread. It also benefit when you have a stall of the first thread but that is not the only reason SMT was created. By making the core wider you increase single thread throughput but not every thread can utilize the entire execution units/resources of the core. By having a second thread (SMT) you use all the core resources left from the first thread. Thus SMT increases the utilization and efficiency of your Core throughput.

edit: Also to add that SMT in Haswell is more effective than before because Haswell core is wider than the rest of the Core arch like Ivy and Sandy.

 

[Phynaz]

I guess Zen will save AMD. Just like Steamroller did. Or Piledriver before that, or Bulldozer before that. Just like Phenom did.

Wait, I thought Excavator was going to save AMD...I'm confused now.

 

[AtenRa]

I guess Zen will save AMD. Just like Steamroller did. Or Piledriver before that, or Bulldozer before that. Just like Phenom did.

Wait, I thought Excavator was going to save AMD...I'm confused now.

Steamroller and Excavator are the best archs AMD have created so far. The big difference with ZEN is it will also have an exceptional and very efficient 14nm FF node something the Bulldozer arch CPUs never had.

Nobody expecting single thread performance to be same as Skylake but if we judge from Kaveri/Carriso perf/watt on 28nm Planar then ZEN at 14nm FF will be miles better even if they would just port Excavator on that node.

 

[mrmt]

I think the future for AMD looks a lot more like Qualcomm and a lot less like Intel. I'm not sure if they have enough cash to pull it off.

Qualcomm is able to develop an entire SoC for themselves and is an IP powerhouse. AMD is... something far distant from it.

But overall I agree with you, AMD should be able to downsize and become a very nimble company, just like VIA.

 

[Erenhardt]

Anyone dreaming of 3-4Ghz SB IPC class chips are in for a huge dissapointment.

You mean 10-11GHz SB IPC? cause that is about equal:thumbsdown:

 

[pablo87]

I called that Seamicro acquisition a stupid decision from day 1 (either here or at Never right).

On a positive note, Lisa failed fast and cleaned up the mess.

 

[CHADBOGA]

I called that Seamicro acquisition a stupid decision from day 1 (either here or at Never right).

On a positive note, Lisa failed fast and cleaned up the mess.

Was it Andrew Feldman, who AMD acquired along with SeaMicro, who shot his mouth off at the time of the AMD acquistion, saying that Intel had made a terrible mistake in not buying SeaMicro when they had the chance?

Was SeaMicro's IP terribly overrated, or did AMD just run out of money?

 

[mrmt]

I called that Seamicro acquisition a stupid decision from day 1 (either here or at Never right).

On a positive note, Lisa failed fast and cleaned up the mess.

Yes, she changes her mind so fast:

http://seekingalpha.com/article/283...-results-earnings-call-transcript?part=single

(...)

So when we look at ARM and x86, I would say, you know, the majority of the -- the majority of the market will still be x86 for quite some time because of all of the legacy applications that exist. ARM offers a new opportunity in some of the dense server markets, and so we continue to look at that as a growth opportunity where new business will grow. So it's really I would say separate parts of the market. I think where we add value is that we're able to look at that in totality and optimize the server ecosystem for that.

What a difference a quarter can make.

 

[erunion]

Anyone else really worried about AMD's cash on hand?

Its now under a billion, only $906 million. They depleted $134 million during the quarter.

 

[mrmt]

Was it Andrew Feldman, who AMD acquired along with SeaMicro, who shot his mouth off at the time of the AMD acquistion, saying that Intel had made a terrible mistake in not buying SeaMicro when they had the chance?

Was SeaMicro's IP terribly overrated, or did AMD just run out of money?

The business was written down and closed down, meaning that it was a total failure. If AMD had ran out of money they would be selling the business to someone else, and not swallowing the financial loss and burying the IP on their patent portfolio.

As I said before, AMD management pursuing a market trend is a very good indicator that this trend is wrong. It takes a lot of talent to make such a string of bad choices like AMD does, Seamicro was just one more.

 

[Fjodor2001]

Anyone else really worried about AMD's cash on hand?

Its now under a billion, only $906 million. They depleted $134 million during the quarter.

Yeah that's worrisome. But I guess they can bring in more capital through a new stock issue.

Also, put it into context: $1B is what Intel loses on mobile alone per quarter (!).

 

[pablo87]

Was it Andrew Feldman, who AMD acquired along with SeaMicro, who shot his mouth off at the time of the AMD acquistion, saying that Intel had made a terrible mistake in not buying SeaMicro when they had the chance?

Was SeaMicro's IP terribly overrated, or did AMD just run out of money?

No idea about the IP, my point at the time was that AMD couldn't afford it.

 

[mrmt]

Yeah that's worrisome. But I guess they can bring in more capital through a new stock issue.

Also, put it into context: $1B is what Intel loses on mobile alone per quarter (!).

Do you mind explaining how an operating loss in a segment of one company can share the same context of a negative cash flow for the entire business of another company?

Anyone else really worried about AMD's cash on hand?

Its now under a billion, only $906 million. They depleted $134 million during the quarter.

They aren't bleeding too much cash and they have a revolver credit line of around 500MM to draw upon, so cash isn't really the main issue today, they have plenty of time to caught on VIA in terms of downsizing.

 

[Arachnotronic]

Yeah that's worrisome. But I guess they can bring in more capital through a new stock issue.

Also, put it into context: $1B is what Intel loses on mobile alone per quarter (!).

Much of that loss is due to the fact that Intel allocated costs of development of processor cores, modem, process technology, etc. that's shared across the portfolio to a division that generated basically zilch for revenue.

If Intel pulled the plug on mobile, it would save far, far less than that per quarter.

I highly suggest that if you are interested in understanding Intel's financials that you listen to the presentation that Intel CFO Stacy Smith gave at the company's investor meeting last November.

 

[pablo87]

Anyone else really worried about AMD's cash on hand?

Its now under a billion, only $906 million. They depleted $134 million during the quarter.

The problem is not so much the cash balance but the burn rate.

By my calculation, they burned thru close to $200M in cash in Q1 - $181M of cash/debt/marketable, $205M if you include receivable and payables.

They also agreed to give GF $250M EVERY quarter for the rest of the year.

 

[AtenRa]

They also agreed to give GF $250M EVERY quarter for the rest of the year.

Not correct, they made a commitment to buy 1B worth of wafers from GF in 2015. They have bought 161M worth of wafers in Q1 2015.

Devinder Kumar - SVP and CFO
We entered into a Fifth Amendment to our wafer supply agreement with Global Foundries. Under the terms of the agreement, we expect to purchase approximately $1 billion of wafers in 2015 on a take-or-pay basis. In the first quarter, we spent $161 million on wafer purchases with Global Foundries.