There's Still No Sign Of AMD's Low-Cost ARM Development Boards

[Zola]

Well the word in the wilds is that AMD Is ditching all current and future ARM core plans in favor of Zen. Now that would be a plane black cow and by the all means wrong one.
But the right cow is not a wite one nor is that simple. AMD certainly have more expertise in servere design from any one other (excluding the Intel of course) so they should stop chasing the windmills & focus on proven fact that ARM IP's are simply better to start from if they do want to gain on Intel (that by the way is even more stubborn concerning this). Second part is the future of computing (& server workloads there for) doesn't belong to the many general purpose cores, they will be there alright but in more modest count number & only for general purpose intermediate tasks and to alow system to boot up independently. The future belongs to most tasks and intermediate number of tasks as general purpose as possible accelerators in the first place FPGA's. So AMD should be actually also working on merge with some FPGA manufacturer. Intel did well purchasing Altera but naturally AMD can't do something like that & no one actually can purchase AMD without AMD losing its X86 license. What can be done is joint venture. Even Intel is smart enough not to try to push his Atom's on upcoming Altera FPGA products (& that should be telling you something about wite cow). So you see real Zen is somewhere else with opposite Ying to one AMD choose & completely different Yong.

[starshipeleven]

Well the word in the wilds is that AMD Is ditching all current and future ARM core plans in favor of Zen. Now that would be a plane black cow and by the all means wrong one.
But the right cow is not a wite one nor is that simple. AMD certainly have more expertise in servere design from any one other (excluding the Intel of course) so they should stop chasing the windmills & focus on proven fact that ARM IP's are simply better to start from if they do want to gain on Intel (that by the way is even more stubborn concerning this). Second part is the future of computing (& server workloads there for) doesn't belong to the many general purpose cores, they will be there alright but in more modest count number & only for general purpose intermediate tasks and to alow system to boot up independently. The future belongs to most tasks and intermediate number of tasks as general purpose as possible accelerators in the first place FPGA's. So AMD should be actually also working on merge with some FPGA manufacturer. Intel did well purchasing Altera but naturally AMD can't do something like that & no one actually can purchase AMD without AMD losing its X86 license. What can be done is joint venture. Even Intel is smart enough not to try to push his Atom's on upcoming Altera FPGA products (& that should be telling you something about wite cow). So you see real Zen is somewhere else with opposite Ying to one AMD choose & completely different Yong.

I'm going to inform you that your return key seems to be broken.

That said, you're still talking of non-short-term plans.
AMD is focused on getting out of the shit they are drowning in first (if Zen fails as hard as their last desktop FX processors their CPU division is basically disbanded and AMD becomes a GPU-only company), and the most likely chip that sells NOW is x86.

Of course all future plans are on hold.

[LinuxID10T]

Which do you think is the right cow ? Short term ARM server focus or short term x86 followed by longer term ARM ?

IMO, short term x86 and long term x86. I think the whole dense server market was a fad to begin with. I think AMD should have seen this coming, but without real corporate leadership, it was just a ship without a rudder. The company under Lisa Su seems a lot more focused and I think that is what they really need. Once Zen is in the market, I think they will be able to scale down, as opposed to trying to scale ARM up. When it comes down to it, these days, the ISA doesn't really matter from a performance and power standpoint, but software availability does. X86 has software availability, plain and simple.

[Zola]

I'm going to inform you that your return key seems to be broken.

That said, you're still talking of non-short-term plans.
AMD is focused on getting out of the shit they are drowning in first (if Zen fails as hard as their last desktop FX processors their CPU division is basically disbanded and AMD becomes a GPU-only company), and the most likely chip that sells NOW is x86.

Of course all future plans are on hold.

Well Zen whosent a short turn plan it's actually developed for 5 years now (& that is a uper limit for mid range plan's). Short range plan would be license POP IP ARM 28 nm cores paid them with a GPU that you already have on the same lithography trow your branding on it & over it cheap on a market that needs it (like cheap development bord). This all takes 6 months to tape out along with supply chain as you aren't rooting almost anything (supposedly that you have already developed drigers) & that is a very short term plan that do turn the key along with opening the doors in the future for offering more real industrial turn key solutions in the future (as the AMD is still much more trusted company than for instance MTK). & I can tell you that there are many markets and many products for such kind and of products to name a few Smart TV SoC's (MTK currently supplies both Sony & Philips), smart TV boxes & cetera.

[chuckula]

Really? How about doing some reading first sport?

http://www.anandtech.com/show/10353/investigating-cavium-thunderx-48-arm-cores

That's all you got? Those white elephant Cavium products? Yeah, I've been aware of them -- mostly due to their overhyped marketing -- for a very long time. Not impressed with the performance and the bloated core count numbers don't do much for me either. A Xeon D is still a vastly better processor if you are into the microserver market and a higher end Xeon is still the vastly better processor for high end workloads that these Cavium toys can't even run.

[duby229]

So I've been thinking about the position AMD is in. Let me do some pure speculation here. If you don't agree simply keep scrolling.

At least to me it seems pretty obvious Zen is a direct result of the IP license AMD has with Intel. Intel knew years ago that Arm has the potential to compete with Intel's products and they must have known years ago that AMD was getting interested in developing it further. So they decided to gift AMD with a high end x86 architecture, which prompted AMD to contract Jim Keller to sort it out and get it to manufacturability testing. The benefit to Intel is that even if they don't get those sales which AMD gets they are still x86 sales which keeps the platform relevant.

Any sense in this speculation or no?

[starshipeleven]

Well Zen whosent a short turn plan it's actually developed for 5 years now (& that is a uper limit for mid range plan's).

More like 3 years, and it is a total redesign.

Short range plan would be

finishing whatever they begun as changing course now would have wasted the development effort.
That said, ARM space is a very high-competition low-margin market in the consumer segment you talked about.
Big companies don't give a shit about lack of support, they want stuff that works decently now and costs a few bucks. Mediatek usually wins here.

AMD GPUs aren't in high demand in embedded market where at most you need to display 2D content, and I doubt they can make a decent GPU for mobile (decent = powerful AND low-power enough).

[duby229]

More like 3 years, and it is a total redesign.
finishing whatever they begun as changing course now would have wasted the development effort.
That said, ARM space is a very high-competition low-margin market in the consumer segment you talked about.
Big companies don't give a shit about lack of support, they want stuff that works decently now and costs a few bucks. Mediatek usually wins here.

AMD GPUs aren't in high demand in embedded market where at most you need to display 2D content, and I doubt they can make a decent GPU for mobile (decent = powerful AND low-power enough).

You seem to forget that Adreno was once Imageon, a VLIW unified shader architecture very similar to r600 desined by ATi.

[starshipeleven]

At least to me it seems pretty obvious Zen is a direct result of the IP license AMD has with Intel. Intel knew years ago that Arm has the potential to compete with Intel's products and they must have known years ago that AMD was getting interested in developing it further. So they decided to gift AMD with a high end x86 architecture, which prompted AMD to contract Jim Keller to sort it out and get it to manufacturability testing. The benefit to Intel is that even if they don't get those sales which AMD gets they are still x86 sales which keeps the platform relevant.

Any sense in this speculation or no?

Unless you explain more what would be this "high-end x86 architecture" they gifted, and why they would have given this to AMD instead of using it themselves, no.

Also, probably no in any case. Intel does not play cat-mouse game anymore with AMD since a long time ago. Intel is at like 100% in x86 servers, around 90% of laptops and above 85% of desktops.

[starshipeleven]

You seem to forget that Adreno was once Imageon, a VLIW unified shader architecture very similar to r600 desined by ATi.

did r600 use anywhere near the order of magnitude of power that Adreno is allotted? I think not.