AMD Launches ARM Based Opteron A1100 SOCs For Datacenter – With Up To 8 A-57 Cores and DDR4 Support

[SurvivorNVL]

Then why did they even bother?

They owe it to their investors.  They have to try.  Even if they are fighting a losing battle, they have to try and gain even small portions.  It may be grasping at life, but if it can keep them surviving in to 2018 - then it is good.

[Dabombinable]

They owe it to their investors.  They have to try.  Even if they are fighting a losing battle, they have to try and gain even small portions.  It may be grasping at life, but if it can keep them surviving in to 2018 - then it is good.

Their ARM server chip won't have when its got no actual advantages....

[zMeul]

Then why did they even bother?

probably fooled some OEM by promising them then moon and the sun

[SurvivorNVL]

Their ARM server chip won't have when its got no actual advantages....

Yeah... in this case it's wasted money in all likelihood.  AMD are.. I don't even know.  They blew that deal with the investment firm a few months ago, and lost out on a huge injection of cash, and I just--I've sort of lost most of my hope for them since then.  I am sure they've got a lot of good stuff in the works aside from this, but.. well, at least we can be rest-assured if we own Radeon cards - Intel will likely snatch up RTG if it all implodes, and they'd be covered under Intel.  I don't know.  Hopefully they know what they are doing.

[patrickjp93]

This is meant for the low power server market, x86 can't beat ARM in power consumption.

Tell that to Avoton and Xeon D. They've taken up the majority of dense microserver orders already. Intel closed the gap for datacenter needs.

[Prysin]

Tell that to Avoton and Xeon D. They've taken up the majority of dense microserver orders already. Intel closed the gap for datacenter needs.

could you show us Intels order books?

 

since you claim you have seen/have access to it that is...

[Doobeedoo]

Good to see them keeping momentum. Will be very interesting year.

[patrickjp93]

could you show us Intels order books?

since you claim you have seen/have access to it that is...

No need, just look at Samsung's withdrawal from their server market inclinations, Qualcomm's added delay to their 24-core ARM chip, and Cavium's uncharacteristic silence for the past few months despite their usual boasting of Thunder-X. Xeon D is proving to be an incredible obstacle where performance is concerned, and Intel has the better multicore scaling, so Avoton is playing very nicely against the more specialized low-power parts.

Even Huawei is giving up their 64-core MIPS chip for servers, shortly after Xeon D was unveiled.

[Prysin]

No need, just look at Samsung's withdrawal from their server market inclinations, Qualcomm's added delay to their 24-core ARM chip, and Cavium's uncharacteristic silence for the past few months despite their usual boasting of Thunder-X. Xeon D is proving to be an incredible obstacle where performance is concerned, and Intel has the better multicore scaling, so Avoton is playing very nicely against the more specialized low-power parts.

Even Huawei is giving up their 64-core MIPS chip for servers, shortly after Xeon D was unveiled.

all of this is speculation based upon insinuations.

let me start with qualcomms ARM: it wasnt 64bit, it was AFAIK RISK based, or with similar instructions. So even though server software space usually is built around the hardware, if the hardware is too far away from the software, then the hardware must be reconsidered to maintain low costs.

MIPS has the same issue. the limitations of the instruction set will set a limitation to the implementation of software.

Due to x86 and x64 having access to wastly higher amounts of memory and some other features then ARM based offerings do, this too comes into play.

I will agree that Xeon D is a strong contender, and that AMD is by far too late to this market. They will not gain any traction at this point, because a equally priced or higher performing offer is already established in the market.

[Master Disaster]

It's not like they could use 14nm. Sheesh. Some people make comments without understanding why it's 28nm.

 

Zmuel just hates AMD and always shit posts in every AMD topic.

 

Anyway, so does this mean we might be moving towards a future where datacenters are using ARM and RISC based CPUs?

[TechGod]

just ignore him

 

Apple was able to get the A9 (a custom ARMv8-A chip) in the Iphone 6S and 6S+, 14nm FinFET from Samsung and 16nm from TSMC and AMD can't!?

No! YOU'RE the moron that should be ignored. 

[WynLore]

if memory serves, AMD has been working on this for far longer than it took them to tape out the Arctic Islands GPUs and Zen CPUs

---

http://www.anandtech.com/show/9956/the-silver-lining-of-the-late-amd-opteron-a1100-arrival

What the fuck has AMD been doing these past few years? I don't understand them. No, really, I don't understand what they have been doing. They've been working on that longer than Arctic Islands and Zen, yet there's that...

[Dabombinable]

No! YOU'RE the moron that should be ignored. 

Every other manufacturer is getting close to or already has their CPU at 14nm. Because of that they can get higher efficiency and performance than AMD. There is no reason to choose AMD's ARM SOC over Xeon D-which I might add were seen in June last year: http://www.anandtech.com/show/9185/intel-xeon-d-review-performance-per-watt-server-soc-champion

 

Face it-AMD has done too little too late to actually compete again. And before you go off the rails, this is a Xeon D: http://ark.intel.com/products/87039/Intel-Xeon-Processor-D-1540-12M-Cache-2_00-GHz , and now compare it to the new ARM based Opteron: http://www.anandtech.com/show/9956/the-silver-lining-of-the-late-amd-opteron-a1100-arrival

 

The Opteron was trailing in every single way before it even launched. Also read the articles so you actually get the full context for once.

[TechGod]

Every other manufacturer is getting close to or already has their CPU at 14nm. Because of that they can get higher efficiency and performance than AMD. There is no reason to choose AMD's ARM SOC over Xeon D-which I might add were seen in June last year: http://www.anandtech.com/show/9185/intel-xeon-d-review-performance-per-watt-server-soc-champion

 

Face it-AMD has done too little too late to actually compete again. And before you go off the rails, this is a Xeon D: http://ark.intel.com/products/87039/Intel-Xeon-Processor-D-1540-12M-Cache-2_00-GHz , and now compare it to the new ARM based Opteron: http://www.anandtech.com/show/9956/the-silver-lining-of-the-late-amd-opteron-a1100-arrival

 

The Opteron was trailing in every single way before it even launched. Also read the articles so you actually get the full context for once.

You talk about how AMD is behind and doesn't have 14nm but you conveniently "forget" Zen. It should be obvious this product is a stop gap.

[Dabombinable]

You talk about how AMD is behind and doesn't have 14nm but you conveniently "forget" Zen. It should be obvious this product is a stop gap.

Zen isn't a SOC like the 14nm x86 Xeon D and 14nm ARM based CPU made by other manufacturers. Zen is also something that you don't forget about as its been 5 years since AMD actually had a CPU architecture that made sense. Also, a stop-gap that was in develpement for years? A stop gap is overclocking the FX 8350 to 5GHz and selling it as a new CPU for up to $999, or rebadging Hawaii to Grenada and saying its new.

[HalGameGuru]

I wonder if we're trying to shoehorn this SoC into an Intel mould and discounting it out of hand?

 

Based on some of the things made mention of, i.e. possible lowered clock/TDP, small package size, "convergence" of different workloads, and focus on datacenter, might this be used in untraditional ways to leverage a different use case from Intel? Or a similar use case in a different way? Imagine several chips populating a single module, at full load each chip running at 15 watts and 1.2 or 1.5 Ghz or whatever that break even point is, and under differing loads running only a handful of cores at full 32/35 watt full clock, or switching whole cores or SoCs off under differing workloads?

 

x86 is putting ARM on notice but that's not to say uses for ARM, or this chip in particular, couldn't be made. At 28nm and built out how they say, imagine populating a datacenter with a higher number of cores at a lower price and competitive power usage, for services that aren't what we traditionally think of with these products, something other than social media, scientific calculation, or netflix streaming. I'm not saying that's how it is going to happen, but we are making huge industry wide discounts to this product because we are presuming it is a direct competitor to the industry leader. Intel may make the best freight train or cargo jet, but this may well be a semi or barge. It's nominally doing the same job but in a different use case or under different set of criteria.

 

I do worry it may be a zeppelin in the age of jets when such dirigible platforms may not have found their new niche.

[Coaxialgamer]

yay , "low power" servers !