SED - Samsung M&A AMD research news

Page 3 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

mrmt

Diamond Member
Aug 18, 2012
3,974
0
76
IP doesn't necessarily mean products...
Doesn't change what I said. I think AMD would prefer to license or sell their IP to someone else last quarter than to slash their R&D budget another 15%, or that they would prefer to have better margins in the semi-custom sales by funding its own R&D instead of relying on customer's money and having to sell their chips with crappy margins.

There's an awful lack of cash at AMD and I really doubt that if they could fetch a reasonable price for their IP they would have done so. Given the 0 interest in Nvidia IP, which is of better quality and the sore state of AMD CPU development, I really think it is lack of interest.
 

podspi

Golden Member
Jan 11, 2011
1,982
102
106
Doesn't change what I said. I think AMD would prefer to license or sell their IP to someone else last quarter than to slash their R&D budget another 15%, or that they would prefer to have better margins in the semi-custom sales by funding its own R&D instead of relying on customer's money and having to sell their chips with crappy margins.

There's an awful lack of cash at AMD and I really doubt that if they could fetch a reasonable price for their IP they would have done so. Given the 0 interest in Nvidia IP, which is of better quality and the sore state of AMD CPU development, I really think it is lack of interest.

Aren't they licensing IP?


Doubt they'd sell the IP, unless they planned to unwind and liquidate. Without their IP they don't really have much of a business left. Personally I doubt SS will purchase AMD. The ROI would be terrible, and if they really wanted the GPU IP they could just license it.
 

podspi

Golden Member
Jan 11, 2011
1,982
102
106
They may have found a licensing partner there, but 1 customer doesn't much of the big picture, which shows most of the market finding AMD IP unattractive, or at least not attractive enough, to pursue licensing deals.

You wouldn't be saying that if the one partner was Intel :cool:

What other major SoC players are there who might be interested in AMD's IP? I can only think of Samsung. QC has their own (formerly AMD's) IP, going to go out on a limb and say nVidia isn't going to be interested, Intel and Apple seem pretty happy with Imagination, though Intel will probably transition to its own IP as well.

Also keep in mind Sony and Microsoft can be considered licensing partners as well. I'd say they are doing OK. Certainly not a replacement for actually successful products, but they aren't like nVidia where there aren't any takers at all (which probably says more about nVidia and how they negotiate than anything).
 

ShintaiDK

Lifer
Apr 22, 2012
20,378
146
106
Intel licenses nVidia IP. And does so for 1.5B$ over 6 years. And its such a small deal that nobody made a fuss out of it.
 

podspi

Golden Member
Jan 11, 2011
1,982
102
106
Intel licenses nVidia IP. And does so for 1.5B$ over 6 years. And its such a small deal that nobody made a fuss out of it.

You and I both know that is not the same type of deal. If we are counting deals like that, Intel also is licensing AMD's IP (and AMD is licensing Intel's).

The (afaik rumored) MediaTek deal is not a cross-patent deal, its slapping GCN into MediaTek SoCs. Comparing the two deals is intellectually dishonest and you know it.

Edit to make this clearer:

http://arstechnica.com/business/201...l-look-for-nvidia-gpu-on-intel-processor-die/

And their correction:

Correction: NVIDIA wrote in to tell us that our original headline was not accurate. An NVIDIA spokesperson said, "Licensing a technology is different than incorporating an entire processor. The settlement provides Intel with access to our IP and patents, such as Sandy Bridge which already uses NVIDIA technology. The license enables Intel to extend that model for the next 6 years."

Also, I deleted the following text from the article: "On the Intel side, NVIDIA CEO Jen-Hsuan confirmed that Intel could use the licensing agreement to produce a Sandy Bridge successor with an on-die GPU based on NVIDIA technology." It looks like NVIDIA's stance is that there's already NVIDIA IP in the Sandy Bridge IGP, because Sandy Bridge's GPU infringes on NVIDIA patents. This wrinkle wasn't at all clear from the announcement or the call—at least, it wasn't clear to me.

Emphasis added.
 
Last edited:

IlllI

Diamond Member
Feb 12, 2002
4,929
11
81
seems like the past few years amd has just been trying to keep their head above water. I would rather seem the company bought, than to just have it continue the slow decline into nothingness.
 

ShintaiDK

Lifer
Apr 22, 2012
20,378
146
106
In terms of Samsung, they would also have to radically increase their R&D budget.

bulletin20150224Fig01.png
 

podspi

Golden Member
Jan 11, 2011
1,982
102
106
Is that really ALL of Samsung's R&D, or just their R&D for fabrication?

A lot of Intel's R&D goes into fabs, but a lot of it also goes into IC design. I find it hard to believe Samsung manages to do so much with so little, even if they are using off-the-shelf ARM cores.
 

geoxile

Senior member
Sep 23, 2014
327
25
91
Is that really ALL of Samsung's R&D, or just their R&D for fabrication?

A lot of Intel's R&D goes into fabs, but a lot of it also goes into IC design. I find it hard to believe Samsung manages to do so much with so little, even if they are using off-the-shelf ARM cores.

Well now they're working on their own custom microarchitecture that implements ARMv8 ISA, so clearly they're doing something right.

IIRC they poached AMD's bobcat team who put together that uarch on a shoestring budget, so maybe they're hard at work for Samsung.
 

Idontcare

Elite Member
Oct 10, 1999
21,110
64
91
Is that really ALL of Samsung's R&D, or just their R&D for fabrication?

A lot of Intel's R&D goes into fabs, but a lot of it also goes into IC design. I find it hard to believe Samsung manages to do so much with so little, even if they are using off-the-shelf ARM cores.

That's not the direction from which you should approach those data.

Instead, approach it by starting first with TSMC's R&D expenditure as that is the a true reflection of what it costs to develop the leading edge process node technology.

TSMC spent ~$1.6B in 2013 and ~$1.9B in 2014. The 2013 number is basically going to reflect nearly 100% expenses sunk into 16FF+ development (20nm was 98% done in 2013, ramp to manufacturing was all that was left), the $300M bump in 2014 was 10nm (TSMC doubled their R&D headcount and duplicated R&D teams for 10nm).

At any rate, basically the $1.9B number from TSMC is going to be more or less the same "process tech and fab tech" expenditure contained within Samsung's R&D expense as well as Intel's. Intel's you could argue might be a bit higher just because they do a tad more path-finding research but at most you are talking about a 10% cost-adder (not a 2x multiplier).

So if it costs TSMC (and Samsung and Intel, provided their R&D efficiencies are on par with those of TSMC) ~$1.9B to polish off 16FF+ development while going fullsteam at 10nm and the early years of 7nm, then that leaves Samsung with a cool $1B ($2.9B - $1.9B = $1B) of "excess" R&D expenditures.

Now Samsung also has to spend R&D on developing dram and nand flash process node technologies, which is probably about right on the money as to where the remaining $1B goes when you compare Toshiba's and Micron's R&D budgets.

Meanwhile Nvidia, Mediatek, Broadcom, and Qualcomm show you just how silly expensive it is to design and develop the chips themselves that are fabbed on $2B process nodes.

Then you can turn your attention to Intel and gain an appreciation for how much (or little) one can reasonably assume their $11.5B R&D budget goes towards process tech development...to be sure it is more than TSMC, but do you think it is more than Samsung where their $2.9B is spread across the development of leading edge CMOS (14nm), Dram and Nand flash whereas Intel's is mostly just CMOS facing plus a bit of longer range path-finding?

That leaves a very healthy budget for design in Intel's R&D budget. But it really makes your head spin when you look at Qualcomm's monster R&D budget compared to the product lineup coming from it.
 

pablo87

Senior member
Nov 5, 2012
374
0
0
This makes sense. What Samsung would be after is GPU/ATI for standalone GPUs, workstation, ARM based APUs, convert gaming business to ARM, Smartphones of course. What they bring is deep pockets for R&D, Leading edge process. AMD the public entity, the x86 licensee, they don't need. Just cross license 3 way (except x86) between SED, AMD, Global Foundries - probably to SED's net advantage.

+ gpu external biz
+ gpu fills fabs
+ gpu for APU ex chromebooks
+ gpu for smartphone
+ block others fm same
+ license revenues
+ tighten GPU/memory integration

minus - a few Trillion won

Winner - Samsung

Losers - everybody else.
 
Jan 6, 2015
25
0
66
This makes sense. What Samsung would be after is GPU/ATI for standalone GPUs, workstation, ARM based APUs, convert gaming business to ARM, Smartphones of course. What they bring is deep pockets for R&D, Leading edge process. AMD the public entity, the x86 licensee, they don't need. Just cross license 3 way (except x86) between SED, AMD, Global Foundries - probably to SED's net advantage.

+ gpu external biz
+ gpu fills fabs
+ gpu for APU ex chromebooks
+ gpu for smartphone
+ block others fm same
+ license revenues
+ tighten GPU/memory integration

minus - a few Trillion won

Winner - Samsung

Losers - everybody else.

Samsung like other companies designing processing chips are aware that it is the GPU and its compute capabilities that are rapidly increasing relative to other functions.

They can hire engineers from AMD like Apple did, but they still do not have the critical IP and the brightest people in the industry like JKeller, RKoduri and others accept the challenge to lay the foundation for the next wave in computing.

Samsung also left Intel behind when they became top partner in HSA foundation (AMD ARM Qualcomm Mediatek TI) and also when they decided to hedge HMC and proceed with manufacture of AMD Hynix HBM.

If Samsung does proceed with AMD M/A and start a 14nm chip price war with CPU/GPUs for mobile/desktops, then that will be great for consumers, but most likely not so good for competitors.
 
Jan 6, 2015
25
0
66
LOL Coz HSA is just rocking the world. D:
Why do you think Intel is in a rush to buy Altera?

Because of creation of HSA concepts and foundation and software like OpenCL is making FPGA a threat to Intel dominance in server market.

FPGA leaders Xilinx and Altera have adopted OpenCL and it allowed Microsoft to have faster development of their Catapult FPGA for servers that outperform Intel.

AMD was leader in FPGA coprocessor concept with Torrenza platform and OpenFPGA. AMD evolved Torrenza into HSA Foundation with major companies in tech industry like Samsung.

What make Samsung more prominent in HSA is that they have very important board of directors in representation:

Jay (Jeongwook) Kim – Corporate Vice President, Samsung Advanced Institute of Technology, Samsung Electronics

He is the leader of Reconfigurable Processor Group at SAIT (Samsung Advanced Institute of Technology), Samsung’s corporate research center.

Michael Shebanow

Michael Shebanow joined Samsung Research America in 2012 where he is Vice President of the Advanced Processor Lab (APL) in San Jose, CA. At APL, he is managing a multi-site team developing GPU IP in co-development with teams in Suwon, South Korea and Bangalore, India.
http://www.hsafoundation.com/board-of-directors/
 

Idontcare

Elite Member
Oct 10, 1999
21,110
64
91
Why do you think Intel is in a rush to buy Altera?

Do you have any link to support the claim that Intel is in any such rush?

AMD was leader in FPGA coprocessor concept with Torrenza platform and OpenFPGA.

Uh....did you really just write that?

AMD is the leader in FPGA coprocessor concept? Because they have something on paper that you find more valuable than all of AMD's consumers and shareholders combined?

Do you know what is means to be a leader? You kinda have to be in the lead, akin to Altera if you need a reference.

What make Samsung more prominent in HSA is that they have very important board of directors in representation

Having important board of directors only matters if you need unsavory behind-the-scenes deals to be made so as to secure your own IP in front of the IP of others.

Surely you aren't so new to the industry as to believe a board of directors is what makes the difference to the success or failure of your IP? Customers do that. Not smokey filled backrooms.
 
Jan 6, 2015
25
0
66
Do you have any link to support the claim that Intel is in any such rush?



Uh....did you really just write that?

AMD is the leader in FPGA coprocessor concept? Because they have something on paper that you find more valuable than all of AMD's consumers and shareholders combined?

Do you know what is means to be a leader? You kinda have to be in the lead, akin to Altera if you need a reference.

Celoxica confirmed benefits at AMD Torrenza Initiative Seminar.

Note the date and circumstances of AMD in industry. AMD during the early and mid 2000s was starting to outperform and outdeliver Intel in desktop and servers CPUs.

The Torrenza initiative was yet another development that added another front that would jeopardize Intel as a viable competitor and company.

These developments would cause Intel to do the unthinkable and actually pay companies not to do business with AMD. The payment would mount into the billions and the actual amount could be around 10 billion $US or more.

http://www.celoxica.com/about-us/

http://www.businesswire.com/news/ho...Accelerated-Computing-Vision-AMD#.VRvtARop6NE
 

mrmt

Diamond Member
Aug 18, 2012
3,974
0
76
Celoxica confirmed benefits at AMD Torrenza Initiative Seminar.

Note the date and circumstances of AMD in industry. AMD during the early and mid 2000s was starting to outperform and outdeliver Intel in desktop and servers CPUs.

How much money Torrenza made for AMD or someone else? 0? Then it's worthless, no matter Celoxica and others say it's good.