AMD Announces The AMD Opteron A1100 Series SoC

H

HardOCP News

[H] News
Joined
Dec 31, 1969
Messages
0
AMD marks a major step toward delivering choice and innovation in the datacenter with the launch of the AMD Opteron™ A1100 System-on-Chip (SoC), formerly codenamed “Seattle.” Jointly with its software and hardware partners, AMD is accelerating time-to-deployment of ARM®-based systems and driving forward ecosystem support for ARM in the datacenter.

The AMD Opteron A1100 SoC represents a key milestone for establishing ARM in the datacenter as well. “The AMD Opteron A1100 processor brings a new choice in scalability across network infrastructure and datacenters,” said Lakshmi Mandyam, director of server systems and ecosystems, ARM. “AMD brings recognized expertise in the server and embedded markets, making them an ideal partner to deliver a 64-bit ARM processor with the impressive balance of performance and power-efficiency to address an increasingly diverse set of workloads.”

The AMD Opteron A1100 Series SoC is the first 64-bit ARM Cortex-A57-based platform from AMD. Utilizing ARM Cortex®-A57 processors with high-speed network and storage connectivity and outstanding energy efficiency, the AMD Opteron A1100 Series SoC delivers a balanced total cost of ownership for storage, Web and networking workloads. AMD Opteron A1100 Series SoC specifications:

  • Up to eight ARM Cortex-A57 cores with 4MB shared Level 2 and 8MB of shared Level 3 cache
  • 2x 64-bit DDR3/DDR4 channels supporting up to 1866 MHz with ECC
  • 2x 10Gb Ethernet network connectivity
  • 8-lane PCI-Express® Gen 3
  • 14 SATA-3 ports
 
Seems like an interesting SoC. 8 A57's with some serious cache should do pretty well. Anyone know what process this is on? I think it's on an older one, like 28nm, but not sure.
 
Anxious to see MBs with this.
Click to expand...
These would be nice for zfs based fileservers. I can see lots of NAS devices using this as the platform.

I would like to see this with some additional LSI SAS2 ports.
 
meh i am waiting an unlocked intel e5 2699 v3/v4/v5 unlocked 32 core 64 thread competitor
 
So, if AMD is making arm chips. Do we talk about them here or in the non-x86 section?
 
Darakian said:
So, if AMD is making arm chips. Do we talk about them here or in the non-x86 section?
Click to expand...

I think the AMD Processors section is most appropriate, as that forum is more of a catch-all for all the other ARM chips.
 
ars technica said:
There are three models in total. Two are 8-core parts, both with a 32W TDP, running at 2 or 1.7GHz. The third is a 4-core, 1.7GHz part, with a 25W TDP.

AMD has been vague about both pricing and performance. The company says that the top-end part will cost around $150, with the others coming in below. This is quite a bit cheaper than Intel's Xeon D processors, which pack Broadwell-class processor cores with a bunch of I/O connectivity. These parts start at $199 and go up as high as $675. However, AMD concedes that the Xeon Ds are considerably faster. The company says that Intel's 2013-era Atom-based C2000 series systems-on-chips are a better comparison for these new models.

Price-wise, this comparison makes more sense—these processors cost between $43 and $171—but the performance implications are significant. AMD has refused to offer benchmark scores, but for it to be pricing and pitching its new processors against ones that are now two-and-a-half years old suggests that the A1100 range isn't going to blow anyone's socks off.
Click to expand...
Ouch, same old same old.

For anyone thinking ARM is ripe n ready for the datacenter, take a trip to ARM's own home page and browse around and download some stuff. Why, yes it is slow as molasses for such a relatively low traffic site. It's also the flagship example of how pathetic ARM server performance is, even in something simple like cloud/web server.
 
Yeah, I always expected this to be a wash from the moment it was announced. Compared to custom silicon, A57 is nothing special - just a rushed 64-bit port of the A15.

If there was actually any real market for ARM in servers with A57s, everyone would be doing it. Just ask Rockchip or MediaTek how many months it would take them to bang-out a bog-standard 8-core A57 chip (2 clusters, each with 2MB L2), and attach some I/O to it, and a custom L3 cache.

And that's why this is a bad move for AMD: because even if they beat out some market share, they'll lose it tomorrow to the ultra-cheap Chinese! Nvidia learned that the hard way :D
 
Last edited:
Digital Viper-X- said:
I'd like to see some benchmarks, the 8x A9 based servers, were not that far behind the ATOM offering.
Click to expand...

According to anandtech the cores in seattle are ~80% the speed of a comparable atom core. I think the IO is better on the AMD chips which is maybe a selling point.
 
Digital Viper-X- said:
I'd like to see some benchmarks, the 8x A9 based servers, were not that far behind the ATOM offering.
Click to expand...
That was against the older in-order Atom core, which was discontinued a while ago. Performance between Silvermont core (out of order execution) Atoms and Cortex-A15 was closer when comparing 4c/8c per node A15 against multiple 2c/4c per node Atoms.

Silvermont/Rangeley in the C2000 series is about to be replaced this quarter with a new Airmont/Denverton chip, which should get a nice bump in performance per node both in I/O and memory speed increases.

Back to the point, Clover Trail (also Silvermont based) Atom was 50% to 200% faster than Cortex-A9 chips core for core across benchmarks in the 2014 ExtremeTech test, hilariously with the largest margin in "server" benchmarks.

Darakian said:
According to anandtech the cores in seattle are ~80% the speed of a comparable atom core. I think the IO is better on the AMD chips which is maybe a selling point.
Click to expand...
Yes, but server Atom chips based on Airmont/Denverton are being released this quarter, which negate the I/O advantage (10Gb, plus x16 PCIe), and also increase the efficiency by moving to a smaller node (2.5 year old Atom C2000 is a 22nm chip) in addition to other updates (DDR4-2400, 128GB max memory, rumored up to 16 cores per SoC, etc).

The Opteron A1100 was announced almost exactly 2 years ago, and the specs were underwhelming back then. Going back over history, the reason x86 took over the server market is because it had tangible benefits (lower cost, faster performance), so I'm not sure why AMD took this death match to release. Larger companies pulled back on server ARM chips because there's still no market for them and no inherent advantages sufficient for the market: ARM performance is poor and power efficiency/performance isn't even a saving grace.
 
Last edited:
pxc said:
That was against the older in-order Atom core, which was discontinued a while ago. Performance between Silvermont core (out of order execution) Atoms and Cortex-A15 was closer when comparing 4c/8c per node A15 against multiple 2c/4c per node Atoms.

Silvermont/Rangeley in the C2000 series is about to be replaced this quarter with a new Airmont/Denverton chip, which should get a nice bump in performance per node both in I/O and memory speed increases.

Back to the point, Clover Trail (also Silvermont based) Atom was 50% to 200% faster than Cortex-A9 chips core for core across benchmarks in the 2014 ExtremeTech test, hilariously with the largest margin in "server" benchmarks.

Yes, but server Atom chips based on Airmont/Denverton are being released this quarter, which negate the I/O advantage (10Gb, plus x16 PCIe), and also increase the efficiency by moving to a smaller node (2.5 year old Atom C2000 is a 22nm chip) in addition to other updates (DDR4-2400, 128GB max memory, rumored up to 16 cores per SoC, etc).
Click to expand...

http://www.anandtech.com/show/8357/exploring-the-low-end-and-micro-server-platforms/5

I think this one is silvermont-ish time frame, but that is also against an arm v8 series chip. While it doesn't beat the atom, it is pretty close to it all throughout the benchmarks. Though the power efficiency is terrible on it.
 
It's still AMD chasing intel on cpu performance. lol

But that said, a small pc that runs android or chromeos or linux or whatever with 8 arm cores, 14 sata ports and 2x10gb ethernet for potentially <$150 (assuming you use a cheaper chip) is still pretty interesting.
 
Digital Viper-X- said:
http://www.anandtech.com/show/8357/exploring-the-low-end-and-micro-server-platforms/5

I think this one is silvermont-ish time frame, but that is also against an arm v8 series chip. While it doesn't beat the atom, it is pretty close to it all throughout the benchmarks. Though the power efficiency is terrible on it.
Click to expand...
I saw that one before. The X-Gene 1 SoC is a custom chip with integer performance in the (stock) A57 range, a bit sad given that a stock A57 has no L3 cache while X-Gene 1 has 8MB L3.

Isn't the 32W A1100 model a 2GHz chip? If so, it's harder to make comparisons here since an A57 based A1100 with L3 cache running 20% slower (2.4GHz X-Gene vs 2GHz A1100) is also a factor. I wouldn't count on it being much faster, if at all, than that X-Gene SoC.
 
But as long as we're talking the appliance/small server market (non-x86), AMD has MORE competitors than just ARM and Intel!

There's Freescale, and their competitive QorIQ AMP series:

http://www.nxp.com/products/microco...080-multicore-communications-processors:T4240

The T4160 is 8 cores at under 30w with 2x10Gg Ethernet ports, exactly the same as Seattle. It's also been available since 2012.

The QorIQ has three integer pipers per-core (should be competitive with the A57). The dual threaded nature should result in higher I/O throughput per-core than the A57.

The 4080 is probably closer to the same performance as this. And the cheapest price I can find there is $349 in quantity.

So AMD has the price advantage. But will that be enough?

And Broadcom has leveraged MIPS all over the place, so there's THAT architecture to compete with as well!

https://www.broadcom.com/products/enterprise-and-network-processors/processors/xlp400-series
 
Last edited:
I don't think AMD went into this to be the competitor but to give them more semi-custom business. Building chips for other manufacturers of whatever tech needs them is a big business solution.
 
JustReason said:
I don't think AMD went into this to be the competitor but to give them more semi-custom business. Building chips for other manufacturers of whatever tech needs them is a big business solution.
Click to expand...

I think they went into this because ARM server market is still very new and there's a lot of room to grow in it.
 
So, the A1100 is really more of a 'development platform' if you ask me. They tossed this SoC out the door to build support for an ARM ecosystem. Note the stuff about ACPI, etc. Obviously this thing is just using A57's with a bunch of cache, so the performance will not be amazing. Hopefully AMD can get the K12/Custom ARM core out the door in a reasonable amount of time, and by then there will be an ecosystem for those SoC's to move into. So the A1100 chips are a long term play from AMD here. I doubt they are expecting to sell very many of them, but they at least have a product on the market and showed up to the fight. That's what these chips are about.

In the meantime, they look like great platforms for NAS's and stuff like that.
 
pxc said:
*snip*
Click to expand...

Indeed. It's not a terribly exciting chip unless 64 bit arm means something to you. AMD took this 'death march' because they wanted to launch with RedHat support out the door. One item in the anandtech review noted that other arm server chips shipped with custom kernels while the AMD has mainstream kernel support (a big deal to linux people).
 
Darakian said:
Indeed. It's not a terribly exciting chip unless 64 bit arm means something to you. AMD took this 'death march' because they wanted to launch with RedHat support out the door. One item in the anandtech review noted that other arm server chips shipped with custom kernels while the AMD has mainstream kernel support (a big deal to linux people).
Click to expand...

Yeah see this, and the stuff I posted in my post above yours, are the reasons why they released this chip. They wanted to pave the way to having ARM chips be 1st class citizens in the software support side just like x86 stuff is.
 
You must log in or register to reply here.
Back
Top