SoylentNews is people
SoylentNews
Advanced Micro Devices (AMD) has posted another quarter of disappointing financial results:
The Computing and Graphics segment continues to struggle, although AMD did see stronger sequential growth here with the recent launch of Carrizo. Revenue increased 12% over last quarter, although it is still down 46% year-over-year. This segment had an operating loss of $181 million for the quarter, up from a loss of $147 million last quarter and a loss of $17 million a year ago. Sequentially, the loss is mostly attributed to a write-down of $65 million which AMD is taking on older-generation products. Annually, the decrease is due to lower overall sales. Unlike Intel, AMD processors had a decrease in Average Selling Price (ASP) both sequentially and year-over-year, so there was no help there from the lower sales volume. The GPU ASP was a different story, staying flat sequentially and increasing year-over-year. Recent launches of new AMD graphics cards have helped here.
Alongside the Q3 2015 earnings release, AMD has announced that it is selling an 85% stake in its back-end manufacturing operations. ATMP, "for assembly, test, mark, and pack," is the step in semiconductor manufacturing that takes a finished wafer of chips and cuts them up into individual chips for customer use. AMD retained these operations even after the spin-off of chip fabrication in the form of GlobalFoundries in 2009. Nantong Fujitsu Microelectronics (NFME) will pay AMD $371 million ($320 million after taxes and expenses), and operate a joint venture to produce chips:
[More after the break.]
As for the joint venture itself, this gives NFME the ability to further expand into the market for semiconductor assembly and test services (SATS). With AMD's lower product volumes no doubt making it harder to fully utilize their high-volume ATMP facilities, a joint venture with NFME can bring more work into those facilities by having them work for additional customers beyond AMD. Furthermore NVME also gains the R&D experience that comes with AMD's ATMP operations, which for them is a competitive advantage against other 3rd party SATS providers.
The news comes just days after AMD "Corporate Fellow" Phil Rogers departed for competitor NVIDIA after working at ATI and AMD for 21 years:
As one of AMD's high-ranking technology & engineering corporate fellows, Rogers' held an important position at AMD. For the last several years, Rogers has been responsible for helping to develop the software ecosystem behind AMD's heterogeneous computing products and the Heterogeneous System Architecture. As a result, Rogers has straddled the line as a public figure for AMD; in his position at AMD, Rogers was very active on the software development and evangelism side, frequently presenting the latest HSA tech and announcements for AMD at keynotes and conferences.
[...] Meanwhile of equal interest is where Rogers has landed: AMD's arch-rival NVIDIA. According to his LinkedIn profile Phil Rogers is now NVIDIA's "Chief Software Architect – Compute Server" a position that sounds very similar to what he was doing over at AMD. NVIDIA is not a member of the HSA Foundation, but they are currently gearing up for the launch of the Pascal GPU family, which has some features that overlap well with Phil Rogers' expertise. Pascal's NVLink CPU & GPU interconnect would allow tightly coupled heterogonous computing similar to what AMD has been working on, so for NVIDIA to bring over a heterogeneous compute specialist makes a great deal of sense for the company. And similarly for Rogers, in leaving AMD, NVIDIA is the most logical place for him to go.
(Score: 2) by turgid on Sunday October 18 2015, @08:18PM
I seem to remember reading a couple of years ago or so that the PHBs at AMD had the brilliant idea that they could offshore a lot of CPU design work to China, because they figured cheaper and less experienced people could do the work with "automated layout tools." I believe this lead to hotter-running CPUs and a lack of direction regarding future designs.
Back in the day when HP killed the Alpha [wikipedia.org] in favour of the magnificent itanic, there were all kinds of stories about where the Alpha engineers went. I know for a fact that several went to Sun, but people also said that AMD snapped up a bunch of them, and they were responsible for the Athlon [wikipedia.org] (x86-32) and the Opteron [wikipedia.org] (x86-64) when AMD started kicking intel's behind.
I'm not an AMD fanboy, but I am a loyal customer, and I did work for Sun at the time that the Opteron came out and we put together some boxes for doing Solaris 10 x86-64 ("x64") builds on. I've used Dell servers, Sun servers and workstations, and HP and Dell (x86) workstations. Back in those days a 2.8GHz Pentium IV Xeon was getting its bottom resoundingly spanked by a 1.6GHz Opteron. And the intel boxes didn't scale linearly with multiple CPUs. More than two CPUs in an intel box was a waste of money and power. Opteron and UltraSPARC [wikipedia.org] did scale, though.
Also, in my experience, AMD CPUs/systems feel better when heavily loaded i,e, the multi-tasking seems better. I put that down to a better cache/MMU/TLB/interconnect architecture that the intel stuff. Up until a year ago I was using a lot of intel i7 systems with stupidly fast (expensive) RAM running 64-bit Linux of various flavours, and I wasn't that impressed.
AMD brought out Hypertransport [wikipedia.org] (similar to the Cray/SGI/Sun interconnect) with the Opteron in 2003. I think it was five more years before intel Quickpath [wikipedia.org] came along...
I've been buying AMD CPUs for my own machines since 1999 (before that I used to believe the pro-intel FUD): K6-2/400, K6-2/500, Athlon XP 2000+, Athlon 64 3200(?) 2GHs, Dual core Athlon 64 (2.6GHz?), Phenom II X4 3.0GHz and finally a Phenom II X6 2.7GHz. We've also got a dual-core AMD laptop from about 3 years back. I've always been pleased with them, and they always perform well. Anecdotally, they hold their own against the recent intel stuff for what I do. I use Linux almost exclusively, so everything I run is compiled with gcc (not intel's C compiler or Microsoft Visual whatever Windows uses) and it's fine. I've bought good motherboards that have lasted years, and taken many CPUs.
So I don't believe the benchmark hype as much as many, and I don't care about binaries compiled with other compilers used when doing the official benchmarks. I care about gcc-compiled binaries on x86-64 Linux.
I've heard it said that the AMD Bulldozer [wikipedia.org] ("Faildozer") design is very reminiscent of Alpha 21264 [wikipedia.org], which was a great idea back in the late 90s. I planned to buy one this year, but events overtook me and the money had to be used for other things, but my 6-core 2.7GHz machine is still plenty fast enough. Even despite the negativity surrounding this CPU, it's actually not that bad, and is very good for the price.
The Zen [wikipedia.org] architecture looks really exciting, and I will definitely get one (after the lunatic fringe have found the bugs in the first batch...) 16 real cores on a die looks amazing, especially if I can afford one for home.
My phone has a quad-core ARM in it and I have a Raspberry Pi model B (ARM) running Slackware on my LAN which I use as a compilation target for some home-made software. ARM is definitely still rising, so AMD is backing a winner there.
Finally, my Athlon XP 2000+ (running Slackware) from 2002 is still going strong as a printer server and a 32-bit compile target (I love ssh). It's running a binary-only driver for my laser printer.
I refuse to engage in a battle of wits with an unarmed opponent [wikipedia.org].