AMD Q414 results

[krumme]

As a comment on APU's, I have not been impressed by AMD's efforts. They always seem to be bandwidth limited. Its probably heavily influenced by OEM's who just want to hit price points (they're all still bundling Winchester HDD's). Hopefully Carrizo is better but how many times have we heard this before.

Plus when you go to buy, the shelves are replete with various generations. And it seems they don't actually destroy obsolete SKU's but use these written down parts to get design wins - I know they did this in the past. Sheer folly. Its a big problem - most of the Q4 conference call revolved around inventory.

Also, I did a bit more research and Kumar told Hans Mosemann of Raymond James at their analyst conference with respect to Global Foundries per the transcript: "its just under $1.2B. We're on track to go ahead and take that from Global Foundries". But on the Q4 earnings call 6 weeks later we have this gem.

"But from a purchases standpoint, the way it works in WSA is the wafers get delivered, and like I said, against the 1.2 that we expected when we closed 2014 WSA, we purchased $1 billion worth of wafers. As we ended 2014, we made a mutual decision to reduce the supply and purchases based on market conditions, and the 2014 WSA is complete."

Where did those $200M worth of wafers go???

More from Raymond James conference: "And of course,The relationship with GlobalFoundries both at the GlobalFoundries level and with our partners in Abu Dhabi is the best in the history of the relationship. You know I’ve been involved in this right from the inception of when GlobalFoundries was formed in 2009, but the relationship is just great at this point. From an execution standpoint, GlobalFoundries execution is significantly better today than it was even a year ago, and as you know they have brought in new management team there and I think that’s doing really well for us from a Foundry partner standpoint. But we do our business with them. Everything is playing out even for 2015. We are on track in 2013 and 2014. We’ve been very fortunate with our partner in GlobalFoundries."

I understand now why somebody was buying June puts.

I have to understand the last sentence right as i am not into this stock option market. Please explain?

 

[mrmt]

It's clear you've made up your mind long ago regardless, there is no point discussing it further.

I'd be glad if you could share metrics to judge a successful M&A.

 

[mrmt]

I have to understand the last sentence right as i am not into this stock option market. Please explain?

GLF forgave 200MM in take-or-pay charges from the wsa 2014 quota. Why they do that 6 weeks after AMD said they were on track to reach the 1.2 billion quota is up to speculation.

 

[cbn]

Why would AMD benefit from APU's to take chunks of the discrete graphics market?

Well, NV aren't going to be able to do x86 based APU's and AMD do still seem to be ahead of Intel in terms of integrated graphics. The latter might not last of course.

But if the graphical performance of iGPUs gets high enough that people buy them on that basis they'll have a fighting chance in that market. Like with the console contracts really.

Could really do with a more power efficient CPU tech of course. Can they get there in one piece?

Trouble is those APUs have problems that go beyond just bandwidth.

Assuming AMD is able to fix the bandwidth problem with HBM, there are also these factors to consider (for desktop):

1. CPU throttling under igpu load with the current APU (Kaveri). Apparently there are software fixes for this, but I am not sure if they work with the stock coolers.

2. Cost of integration: Making a die 47% bigger (or larger if we count Kaveri's special DDR3 PHY) via 512 stream processor iGPU doesn't add 47% cost. It adds more cost than that for various reasons including "higher chance of defect as chip size increases", edge of wafer loss etc.

3. Balance of CPU to GPU: AMD's dual module CPUs are not considered particularly strong (at least compared to Intel) , so there is only so much iGPU AMD can add. This especially if the CPU is also throttling (lowering performance even further) due to sharing additional thermal load from the iGPU.

So IMO, APU competing against cpu and gpu as separate chips (on desktop) is pretty tough. And this has been born out in numerous cost comparisons in these forums even using AMD's own chips. (eg, Athlon x 4 + R7 dGPU vs. A10 APU, etc.)

This not mention people just like to choose and swap in their own video cards. And having the silicon integrated takes away that personal choice.

With that mentioned, I feel better about large iGPU APU on mobile. Certainly for a gaming laptop where the form factor is space constrained integration makes a lot more sense. Hopefully AMD's cpu efficiency is good on Carrizo, otherwise they will have problems with customers choosing either Intel GT3/GT4 or Intel + dGPU for mobile as well.

 

[AnandThenMan]

AMD is up ~10% right now, proves you simply cannot ever rely solely on financial performance when investing. Yesterday people sold off in a panic, now people are buying when low.

Smart.

 

[Blitzvogel]

AMD just losing more and more money means less money for product development which means continued dwindling revenues. I do not want to see the graphics division dragged down with the failures of AMD's executives.

I think the x86 license isn't transferable to anyone that buys AMD, but assuming it was, who would make a good candidate for buying AMD?

Apple could easily profit well off their graphics and ARM tech. With enough injected cash flow, AMD could make x86 or ARM APUs that are actually worth a damn. Only other problem is Apple would just keep all the tech for themselves, except maybe graphics.

Qualcomm. I doubt they have any real interest in the desktop chip market, but their resources could turn what used to be AMD into an x86 powerhouse again. Qualcomm could have a stake in the ARM server wars and perhaps take on Intel in x86 servers.

Samsung. AFAIK only second to Intel in terms of revenue and IC manufacturing tech. However, I would doubt that Samsung would forgo using Intel processors in their products unless they could get AMD up to par, which is possible.

 

[turtile]

AMD may be dead if they never purchased ATi.

1. AMD would have kept their fabs longer which which would have bled them dry.
2. Intel would have continued to offer integrated graphics (just a little later) while AMD would have to license another chip on board - making Intel cheaper to produce.
3. Still would be behind Intel in server chips since their own Fabs would be far behind
4. Would have never been in all consoles

So while they did pay a lot for ATi, it was required. Why would investors in ATi not hold onto a profitable company? They want a premium to sell. I doubt S3 had the know how and IP of ATi.

 

[Blitzvogel]

On the flip side turtile, AMD could've been in a better position to provide a better response to Core 2 with more money for product development.

 

[turtile]

On the flip side turtile, AMD could've been in a better position to provide a better response to Core 2 with more money for product development.

That's the thing. While it's true that AMD could have competed with Core 2 Duo, it wouldn't have been sustainable.

Management was the biggest problem at AMD.

They took way to long to integrate the two companies after the buy out which wasted tons of money.

Rory Read concentrated too much on low-end products with small margins at the beginning. I think Lisa Su will do a much better job of selling products at larger margins. Rory was too used to the consumer PC market which is race to the bottom on margins. You can't do that with a microprocessor company which has tons of R&D expenses.

AMD's huge mistake with ATi was not pushing graphics into HPC from day one.

 

[Blitzvogel]

They also took way too long with Fusion. While Bobcat made them alot of money, I think they should've focused on Llano and focused on developing a worthwhile software ecosystem.

 

[cytg111]

In the end, can AMD look back and say that they reached the acquisition objectives and the financial results were in line with projections that supported the 5 billion value? I guess the answer is no, and if so, AMD overpaid and the acquisition was a failure.

That fails logic 101 by my standards. Business is a game of incomplete information, like poker, politics or war, there is no single move or calculated set of moves that will ensure or guarantee your victory, you can only take one optimized move at a time and in that context hope you are not optimizing towards a local minima. Your concept of failure and thus success requires timetravel. Good luck with that.

 

[Enigmoid]

1. CPU throttling under igpu load with the current APU (Kaveri). Apparently there are software fixes for this, but I am not sure if they work with the stock coolers.

Throttling is relatively minor and likely so that the chips could fit in the power target.

 

[Blitzvogel]

Ok so what are the best moves AMD can make now? Seems that they should be looking for buyers right now.

 

[antihelten]

Samsung. AFAIK only second to Intel in terms of revenue and IC manufacturing tech. However, I would doubt that Samsung would forgo using Intel processors in their products unless they could get AMD up to par, which is possible.

Samsung doesn't actually use any intel chips in any of their current products as far as I can tell (they have announced that they will be dropping their laptop business).

Plus they are currently sitting on underutilized fabs (due to Apple switching a portion of their orders to TSMC) which share IP with GF (possibly making porting of AMD designs easier).

They don't currently have their own custom designed ARM core, putting them largely at the mercy of either Qualcomm or stock ARM (with stock ARM apparently winning the next gen due to overheating issues with Qualcomms SOCs). AMD is working on a custom ARM design (K12).

As far as the x86 license goes, I doubt Samsung would care all that much about losing out on the Bulldozer family, but the Jaguar family certainly has some potential if Samsung wanted to chase after the windows 8.1/10 tablet/2-in-1 market, so that would arguably be to biggest stumbling block in that regard. The 2-in-1 market looks like one that Samsung might be interested in, but also one where they might be satisfied with android solutions (stuff like Samsung Galaxy Tab Pro 12.2).

 

[turtile]

Ok so what are the best moves AMD can make now? Seems that they should be looking for buyers right now.

I think it all depends on how Zen and ARM turn out. They are targeting the high end on graphics. The rumors that Carrizo will not be socketed makes sense. I doubt anyone wants any left over motherboards for such a short lifespan (right before Zen, ARM).

Kaveri failed because it used to much power and required fast memory. Carrizo changes that so it should have more design wins if it performs well.

 

[cbn]

Throttling is relatively minor and likely so that the chips could fit in the power target.

When furmark was used in this review, the cpu cores on 65 watt A8-7600 dropped down to 2.4 Ghz during Prime 95.

Granted, that is Furmark....but that is a lot of throttling.

The A10 Kaveri APUs have also been known to have a considerable amount of throttling as well.

 

[Erenhardt]

When furmark was used in this review, the cpu cores on 65 watt A8-7600 dropped down to 2.4 Ghz during Prime 95.

Granted, that is Furmark....but that is a lot of throttling.

The A10 Kaveri APUs have also been known to have a considerable amount of throttling as well.

Isn't that usual behaviour in graphics, where drivers detect torture tests and throttle on purpose.

Is noticed quite a difference in kaveri throttling between OCCT cpu stress test + furmark and gaming.

 

[Qwertilot]

Sure the APU's aren't perfect, especially not now

They need a CPU that can run as a decently performing quadcore in a 30-40w power envelope. Then you put a big net TDP in, get 60+w for the GPU portion and can fit a lot of iGPU performance in.

Also mobile chips of course. It won't make them overwhelming products or anything, but it gives them a logical, sensible niche which they're rather struggling for right now.

Of course, were they super healthy, big R&D budgets etc they'd very likely have transferred at least one of those console memory bandwidth boosting ideas over to some desktop chip with a big iGPU just to show what's possible. Oh well.

 

[ctsoth]

I find it interesting that the general trend in thought is that AMD, as far as market value goes, is not helped by its x86 portfolio.

Now, it is obvious that AMD's current desktop and server CPUs are not raking in the cash, and I am not of the opinion that Zen/whatever are going to be the next best thing, but it is quite likely in my opinion that if R&D was gifted with a generous budget expansion that AMD's x86 designs could become very competitive.

In reply to a post from much earlier in the thread, if AMD really is valuable, why hasn't anyone purchase it? Well, personally, I would wait for the company's market value to hit rock bottom before purchasing it. You're still getting the same IP, licenses etc, why overpay?

 

[Vesku]

I find it interesting that the general trend in thought is that AMD, as far as market value goes, is not helped by its x86 portfolio.

Now, it is obvious that AMD's current desktop and server CPUs are not raking in the cash, and I am not of the opinion that Zen/whatever are going to be the next best thing, but it is quite likely in my opinion that if R&D was gifted with a generous budget expansion that AMD's x86 designs could become very competitive.

In reply to a post from much earlier in the thread, if AMD really is valuable, why hasn't anyone purchase it? Well, personally, I would wait for the company's market value to hit rock bottom before purchasing it. You're still getting the same IP, licenses etc, why overpay?

Because the x86 license deal is tied to AMD as it is now and most likely won't transfer from a buyout or bankruptcy. You'll also be stuck with the WSA they have with GF with less designs to actually utilize those wafers. Main thing of value to buy from AMD is their graphics IP and engineers and selling that off would kill the company.

AMD may be dead if they never purchased ATi.

1. AMD would have kept their fabs longer which which would have bled them dry.
2. Intel would have continued to offer integrated graphics (just a little later) while AMD would have to license another chip on board - making Intel cheaper to produce.
3. Still would be behind Intel in server chips since their own Fabs would be far behind
4. Would have never been in all consoles

So while they did pay a lot for ATi, it was required. Why would investors in ATi not hold onto a profitable company? They want a premium to sell. I doubt S3 had the know how and IP of ATi.

Well the serious alternative to paying a lot for ATI was merging with Nvidia but Ruiz didn't like that Jen-Hsun Huang wanted to be the CEO of the merged company. Intel dodged a bullet with that, Nvidia with an unrestricted x86 license. Ruiz and Meyer really messed up AMD. Still shaking my head that an engineer like Meyer bought into Intel's P4 vision, probably with a dash of IBM Power envy, and used it as the basis for Bulldozer, i.e. GHz sells.

Apparantly, NVIDIA's boss Jen-Hsun Huang insisted on going on to become the CEO of the proposed AMD-NVIDIA combine, an idea that didn't fly too well with AMD's Hector Ruiz.

http://www.techpowerup.com/161109/amd-talked-to-nvidia-before-acquiring-ati-report.html

 

[frozentundra123456]

They also took way too long with Fusion. While Bobcat made them alot of money, I think they should've focused on Llano and focused on developing a worthwhile software ecosystem.

I agree about taking too long. In fact even after spending 5 billion for ATI, intel was the first to come out with an integrated cpu/gpu.

At the time I was eagerly looking forward to "Fusion", but could not believe that it was the same old cpu with just a gpu stuck on after it finally came out. And the clockspeeds for the cpu were really slow.

 

[AnandThenMan]

Well the serious alternative to paying a lot for ATI was merging with Nvidia but Ruiz didn't like that Jen-Hsun Huang wanted to be the CEO of the merged company.

That's why Ruiz is probably on a beach somewhere and Jensen is a successful CEO.

 

[CHADBOGA]

How often do you go there? You seem to know a lot of their forums (I never go there).

Pretty much everyday.

I get a perverse joy at reading so many delusional posts that appear on that site.

Totally forgot about AMDZone, how do you know it's not active anymore?

Because I check it out frequently as well. Not as frequently as I once did, because the posting level has diminished so much.

 

[cbn]

I find it interesting that the general trend in thought is that AMD, as far as market value goes, is not helped by its x86 portfolio.

The x86 part is relatively OK, it's the other non-x86 parts on the chip(s) that are the problem with AMD's x86 portfolio IMHO:

1. Excessive die size allocation for cache on AM3+
2. Excessive die size allocation for iGPU and DDR3 PHY on Kaveri.

If AMD could get these least productive non-x86 parts of their chips under control I think we would see a much higher uptake of their processors as well as a better profit to AMD (due to a reduction in die size).

P.S. Just eyeballing the Kaveri die, I think I could almost fit four Steamroller modules in the space iGPU occupies (for a total of six modules or 12 cpu cores in the existing 245mm2 die space). Of course, other parts of the die might also need to be increased, but it just boggles my mind how much of Kaveri isn't even x86 CPU. (With that mentioned just imagine how cheap a hexcore with a small iGPU would be to make relative to Kaveri, not to mention how much more desireable it would be for desktop users)

 

[Idontcare]

If AMD could get these least productive non-x86 parts of their chips under control I think we would see a much higher uptake of their processors as well as a better profit to AMD (due to a reduction in die size).

With AMD's "take or pay" requirements at GloFo, a reduction in die size won't actually reduce AMD's expenses. It will just mean they need fewer wafers to make the chips they can sell, and will have to beg/plead/etc with GF to give them another break on reducing the WSA.

The only time AMD needs to worry about the diesize of their chips is when they are produced at TSMC, where they pay for every wafer with no minimums. Or in the unlikely event that they start to sell enough product that they begin to exceed their WSA minimums at GloFo. Only then do they need to reign in their die-sizes from a financials perspective.