ARM will have to break through this first rock, then more bigger rocks behind.
Even ARM can get their CPU ready by next year, there is a long lead time needed by server OEM to certify new product. Coupling with the new architect, new software, and everything, it is really a long road.
I hope it is not too little too late for the ARM camp to redesign.
"Though Intel has not confirmed it, coupled with the fact that Centerton uses a new socket (FCBGA1283), we believe that Centerton is a new Atom design rather than just being a server-branded version of Cedarview."
How different are you talking? Different atom core entirely? Or just a different die with slightly different stuff on it?
Different in that it's not just a Cedarview rebrand. It's a unique die, similar to how Intel's 2C IVB parts are different from their 4C IVB parts, or 4C SNB-E is different from 4C SNB.
Basically you're looking at a Saltwell-based SoC with a more capable IMC (for ECC) along with a PCIe controller supporting more lanes.
And no, it does not have on-die graphics (at least not any that are listed in any official document).
The easiest way to check would be to get die sizes and compare them. Intel has a history of fusing off such features as well as offering the same die in different package with such variety.
It's a SoC, so it contains everything to run the thing. CPU, GPU, memory controller, bus controller, ... So if Intel adds ECC support, 4 more PCIe lanes etc. they have to integrate it in the SoC. But this means more pins are required. So they have to change to a different Socket with more pins. Thus, it's a bit far fetched to say that Centerton is, because of the socket change, a newly designed Atom. The CPU core is most probably plain old Cedarview, whereas some parts in the SoC got replaced with server specific things which require more pins but consume less power. Else they would have switched to 22nm already, which they didn't, so it's probably Cedarview with the consumer parts replaced with server parts.
Seems like AMD could just use their Atom equivalent which is better anyway.
I'm confused by Atom (or ARM) for servers though...what are they good for? I mean is Atom/ARM actually more efficient than Intel (or AMD)'s mainstream CPU lines? Yeah those use more power-but they're also vastly more powerful. Atom has comparably terrible single threaded performance too, which of course can matter in any environment depending.
I've just not been understanding what the point of ARM and now Atom for this use IS. Is it just kind of an inane "of these CPUs use less power so we are saving money!" thing or is there actually something to it?
For me, the biggest advantage I see is VT-x support on the Atom CPU. That will allow me to run VMWare ESXi or something similar on the base hardware of an Atom system and they run some OS on-top of that with having minimal impact on performance.
At least for home, I use an Atom based system as my low-power routers and for other general purpose linux systems. Being able to virtualize it makes it easier to manage.
I could see server makers building a 1U rackmount system with 10 Atom CPUs in them, and then that system could run 10 VMs under it. This would give each VM a dedicated CPU, without having to pay the price for the power of Xeon CPUs... Assuming it would be possible to put a lot of Atom CPUs in a single system.
$54 per CPU for the cheapest is $540 and 60-80 Watts for 20 cores/40 threads. A suck ass GPU, so no GPU workload. More sockets on a mobo mean a more expensive motherboard.
It may still make sense to roll with a single Xeon because of the IPC performance. We need system power numbers and a little more details like memory DIMM size (like laptop SODIMM or full sized DIMMs).
I really think it make more sense for low compute need, dispersed systems. i would snatch one up for a small NAS or home proxy server. UPS needs are lower with this vs. a Xeon.
What would kick ass is to get these on a PCIe add-on card. And make a desktop blade server. A poor man's blade system for small business.
I would have rather saw an AMD bobcat variant with ECC.
As Haswell promises to drop power draws to the 15Watt range as soon as next year, I don't know if there's is a low term point to atom/arm for the server realm. But there is a density option.
A 6W atom cpu + msata SSD + single NIC system on a board would allow for a blade chassis of incredible density without putting out the heat of a fusion reactor. One use case would be offering colo/web services to people who don't want to share. A single rack could have hundreds to close to perhaps a thousand of these and still be at only 10KW.
But I have a hard time seeing a use case for singular units that would sell significantly. Certainly there would be a few good spots where saving another 5-10 watts is helpful, but enough to make it worth everyone's time to develop/support?
These are server parts. Basically the "same" Atom that goes into mobile devices maybe have 2-4W, but has 8W for servers. The same would go for the Haswell part of Xeon. It would be a lot more power hungry.
Intel has has 17W dual core IVB Xeon; the same TDP as their coolest mobile IVB parts at the time it was launched. They also have a 45W quad core Xeon; the same TDP as all but one of hteir mobile quadcore chips. At the moment they only offer a single Xeon at each power point; but if the demand is there nothing stops them from offering more choices in the future. The dual core Xeon is actually faster than any of the 17W mobile parts available. Granted there's probably binning involved in; but as long as they don't dominate demand Intel can skim off the top of production to badge as Xeon to absorb the higher TDP of ECC at all but the lowest base levels.
On the one hand, Intel doesn't sell Xeons at $54... If a Xeon is the right solution for your task, buy a Xeon --- but expect to pay for it. Just like, if a POWER7 is the right solution for your task, buy it --- but you will pay for it.
The point of Atom (or server ARM) is not to replace Xeon, it is to be a cheaper alternative for those (frequent) cases where a Xeon is way more CPU than is needed, as I described above.
The REAL issue you all should be focussing on is not the quality of these CPUs (as measured by things like CPI), it is the quality of the memory system.
At the memory controller level, who is doing the better job of predicting which pages should be opened or closed?
Who is using advanced academic (not yet commercial) techniques like Idle Rank Prediction (the idea here is when you have cache write back, you don't write immediately, you store the write in a buffer and use a prediction mechanism to predict when the RAM is likely to be idle, so your write does not block a read).
There are even fancier things you can do like much more sophisticated rank/bank mapping. Don't just extract bits from the address. If you have, say, 32 banks available, don't just extract bits from the address for chip-routing, use a divide-by-31. This gives you 31 virtual banks which will have their addresses effectively scrambled over 32 physical banks, so that long stride sequential accesses don't keep hitting the same bank (just like Cray was doing back in the 1970s).
Even at the basic level, does ARM have something equivalent to Intel's SMI/SMB? (This splits the memory controller into a part on the main CPU, doing the real thinking, connected via a custom very high speed bus to a baby memory controller sitting right next to the actual DIMMs. The point of this is that the physical path from the DIMMs to the baby memory controller is made as short as possible, which means the DIMM bus can be run at higher speeds. The long run from the DIMMs to the CPU goes over the custom bus, which has separate read and write lanes and so can be way faster.) And if not, what is their plan here? One solution (IMHO) would be for ARM to get heavily into packaging, and to encourage that their server chips are all sold with DRAM stacked PoP on the chip.
The point is, the days of decoupled CPU and memory are gone --- you cannot drive a decoupled bus fast enough. ARM needs a server solution to this fact which plays to ARM's strengths, not a mindless copying of the way Intel and AMD and IBM handle this (which reflect there rather different markets and strengths).
"I'm confused by Atom (or ARM) for servers though...what are they good for?"
SOME computers do a substantial amount of computation. You want those computers to have powerful execute cores.
But OTHER computers mainly route data from disks to memory to the network. Computers like this have little need for a powerful execute core, since most of their time is spent waiting on memory. They are best served by having a sophisticated memory controller (which can predict accurately which DRAM pages to open or close) hooked up to a number of independent threads of execution, each of which can just pause while it waits on RAM.
Atom and ARM in servers target this latter market. IBM does so as well, through a very interesting design. Hooked up a massively powerful memory system (way beyond what we are talking here) a POWER7 core consists of two sub cores, each of which can run two threads. For compute-intensive tasks, a core runs one thread which uses the resources of both subcores (so it has two integer units, two load-store units, two FPUs available to it). For memory intensive tasks, each subcore runs two threads, so you have four threads per core, and you're back in the world I described above --- each thread spends most of its time waiting, and the memory system is 100% busy serving up data to each waiting thread.
You will NEVER understand the design of modern CPUs until you realize that computation is cheap, memory access is expensive, so everything is twisted around making your memory work as hard as possible; and if that means your logic is idle 90% of the time, well, that's life.
So this is basically meant for low load servers that don't have to do much real work ever and instead of letting a powerful Xeon server waste power while they idle along under very low load you use a low power atom server at higher load levels?
It does make sense as long as the load never changes or increases much then an atom will be fine but a xeon server would be able to scale far better across all sorts of different workloads if there any possible future changes. I guess I can see a place at the very low end or specialized projects where things won't change much.
I'd really like the list of the "light" server apps by name where an atom can beat a xeon as the hp graphic implies with it's performance per watt as things scale out but I have to wonder how the per dollar aspect comes into play when you start scaling up with multiple servers and what loads/app scale out across multiple atom servers vs xeon servers .
"It does make sense as long as the load never changes or increases much then an atom will be fine but a xeon server would be able to scale far better across all sorts of different workloads if there any possible future changes"
(a) Don't throw around words like "scaling" without thinking what they mean. If your scaling dimensions are along the axes of disk or network, the CPU doesn't matter much. If your workload is dominated by memory performance, then faster CPU doesn't matter much.
(b) The BIG boys (Apple, Google, MS, Facebook, etc) aren't buying servers on the idea that "well, let's get something fast and we'll figure out ways to use it". They have very definite roles that are played by different machines --- some are computer servers, some are cache servers, some are network servers, some are IO servers. These ARM and Atom CPus are targeted at all these roles except compute server.
Why only 8GB if they already did the work for 64-bit support? Support for ECC registered 2x16GB would make it compelling for MemCached, NoSQL, and other distributed memory-capacity intensive workloads.
The memory controller likely does not support registered memory which would limit DIMM size to 8 GB. Being single channel that limits the memory controller to two DIMM slots. That does raise the question of why it does not support two 8 GB unbuffered ECC DIMMs for a total of 16 GB.
No support for registered memory. Yup, 16GB is still better than 8GB. I know one of the things holding back ARM in the server market is the 4GB limit. This should easily be able to handle 2x8GB then and makes for a more compelling product at minimal cost.
$54 for S1220, $64 for S1260... the low power part, S1240, will likely be a hair more expensive, say $66, knowing Intel's pricing schemes.
Comparing that to existing Atom boards, and Atom server boards, I would expect a Supermicro Atom S1200 board to run from $149-$249, depending on features, when it hits retail.
My question is this: will the S1200 series support multiple sockets per motherboard? Most ARM concepts I've heard of are for 32+ ARM processors per server. Nothing I have seen so far says whether the S1200 Atoms can do SMP.
Prices will be higher for final products, I think. I believe current Atom pricing includes a bundled chipset while these may not. That means you may have to factor in the additional chipset cost.
These probably only have one QPI so single socket. By the way, that's why AMD bought SeaMicro.... for their "fabric" that connects the individual CPUs.
The 32+ ARM processors per server probabaly communicate over Ethernet?
The main IO for this chip is likely the eight PCI-E lanes. QPI's main advantage are coherent, low latency links for usage in multi socket configurations. These Atom chips are strictly single socket.
I do think that there will be boards with several of these chips on them, but each chip will be seen as an independent node. Communication will either be done via a PCI-e to some hub or via embedded Ethernet.
ARM needs those silly 32 processor-in-a-box setups to reach competitive performance per rack unit. I wouldn't expect Intel to do something similar with Atom, considering that Xeons are better in every way once a certain threshold of power is reached.
There is also a 17w rated 2core Intel Xeon E3-1220L V2 nominal 2.30GHz (3.5GHz max). You just need to scale down the I/O, chipset and the reduce power consumption of other mainboard components. A low power CPU in a server board could still idle with >100W at the wall. It would be interesting if they could ad 5w to the Atom based cpu's to allow increasing the clock speed or additional CPU resources (increase IPC). Take the P3 and core2 etc. ideas and create a 2 or 3 decode, simple branch predictor, low power cache, 2 ALU, 1 slower FP/SSE/AVX multiply/div, 1 slower FP/SSE/AVX add/sub. Reduce the number of instructions in the pipeline and fast recovery from stall. Aim for a core <0.5 the complexity (&size) of Ivy Bridge at the same nm fab. Chipset & MB from notebook computer form factor and you have something to fit between Atom and i7 or just kill Atom in terms of perf/watt.
"Given the relatively low pricing of these processors, it will be in Intel’s interests to ensure that they are complementary to their existing Xeon processors and not significantly competitive."
This kind of says it all about Intel's own internal problem. They won't try to make Atoms as best as they can possibly make them, to compete with ARM chips, and eventually, they'll get to the point where they have to compete with ARM so much on performance, that it would endanger their Core chip market. And Intel will have a HUGE dilemma about what to do next then. Keep Atom wimpy? Or cannibalize the Core chips?
In the meantime, ARM has no such conflict of interest, and they'll be as aggressive as they can in the server market, and try to cannibalize the Core chip market as soon as possible. Intel can't, and more importantly, WON'T do that to their own chips, that cost 10x more for the same die area, and give them more than 10x the profit.
Yes, Only if ARM can do that. And by the way, the ARM server chip is no where cheaper than the Intel S1200. The Calxeda ECX-1000 powered HP redstone server cost 1.2 million and have 1600, 4 chips node, which put the per-socket cost is 187.5$.
You're placing a lot of faith in ARM designs being able to hit Xeon-level performance without a Xeon-level TDP. Not sure that's going to happen anytime soon.
You're also ignoring the effect of virtualization. If my only choice is a big, expensive Xeon chip which is going to be under-utilized, there is a strong incentive to run something like VMware on it and run a bunch of VMs. That can add considerably to the cost, and Intel sees none of that money. On the other hand, if instead of one Xeon plus a hypervisor, I can buy 5 Atoms, and run the same workloads on the bare metal (and not even have to change my OS since it's all x86), Intel sees a larger chunk of revenue that otherwise would have gone to VMware or some other company, even if the margins per chip are lower.
This is mostly a rehash of Intel slideware, after all (and would it really hurt the writer of this piece to spell out the use of every acronym the first time it's mentioned in an article? Basic tech writing principle. I would never hire Ryan for my tech pubs team). That said, it is good to know Intel is on the beat in this area, because I am highly skeptical of ARM's ability to do any serious damage here. I bought some Intel shares recently and plan on buying more due to the info presented in this article.
Thanks for the article. Useful information: "Calxeda is rightfully annoyed by these claims and put up a blog about it. Marvell seem to simply not understand the press, so we have no idea what they think, but you can be sure that they are not pleased with the terminology. " Classic Marvell corporate communication!
This said the Centerton holds it own against its current 32 bit ARM server competitors. And Avoton more than holds its own. Calxeda is aimed at a subset of many threaded work loads and should do okay in its targeted market. The rapidly growing micro server TAM can support several product skews.
First of all it's a foolish idea to have micro servers and an even more foolish idea for arm to try to get in. Intel should stick to Xeon and arm should stick to low power computing.
Atom servers are great for NAS with services, not just "dumb NAS", (home and SOHO) or micro servers and colo for small business. I run 3x Atom servers. We are a small Management Academy with just 3 employees and 10 contractors. One Atom runs the webserver. One runs the fileserver. One runs on an offsite location and mirrors the fileserver for local access and also to provide backup functionality.
Building and maintaining this setup is childsplay. And cheap.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
35 Comments
Back to Article
nofumble62 - Wednesday, December 12, 2012 - link
ARM will have to break through this first rock, then more bigger rocks behind.Even ARM can get their CPU ready by next year, there is a long lead time needed by server OEM to certify new product. Coupling with the new architect, new software, and everything, it is really a long road.
I hope it is not too little too late for the ARM camp to redesign.
extide - Wednesday, December 12, 2012 - link
"Though Intel has not confirmed it, coupled with the fact that Centerton uses a new socket (FCBGA1283), we believe that Centerton is a new Atom design rather than just being a server-branded version of Cedarview."How different are you talking? Different atom core entirely? Or just a different die with slightly different stuff on it?
Do these chips have on-die graphics?
Ryan Smith - Wednesday, December 12, 2012 - link
Different in that it's not just a Cedarview rebrand. It's a unique die, similar to how Intel's 2C IVB parts are different from their 4C IVB parts, or 4C SNB-E is different from 4C SNB.Basically you're looking at a Saltwell-based SoC with a more capable IMC (for ECC) along with a PCIe controller supporting more lanes.
And no, it does not have on-die graphics (at least not any that are listed in any official document).
Kevin G - Wednesday, December 12, 2012 - link
The easiest way to check would be to get die sizes and compare them. Intel has a history of fusing off such features as well as offering the same die in different package with such variety.UpSpin - Wednesday, December 12, 2012 - link
It's a SoC, so it contains everything to run the thing. CPU, GPU, memory controller, bus controller, ...So if Intel adds ECC support, 4 more PCIe lanes etc. they have to integrate it in the SoC. But this means more pins are required. So they have to change to a different Socket with more pins.
Thus, it's a bit far fetched to say that Centerton is, because of the socket change, a newly designed Atom. The CPU core is most probably plain old Cedarview, whereas some parts in the SoC got replaced with server specific things which require more pins but consume less power.
Else they would have switched to 22nm already, which they didn't, so it's probably Cedarview with the consumer parts replaced with server parts.
MySchizoBuddy - Wednesday, December 12, 2012 - link
AMD should make a great product using this CPU from Intel via it's subsidiary Seamicro.Wolfpup - Wednesday, December 12, 2012 - link
Seems like AMD could just use their Atom equivalent which is better anyway.I'm confused by Atom (or ARM) for servers though...what are they good for? I mean is Atom/ARM actually more efficient than Intel (or AMD)'s mainstream CPU lines? Yeah those use more power-but they're also vastly more powerful. Atom has comparably terrible single threaded performance too, which of course can matter in any environment depending.
I've just not been understanding what the point of ARM and now Atom for this use IS. Is it just kind of an inane "of these CPUs use less power so we are saving money!" thing or is there actually something to it?
extide - Wednesday, December 12, 2012 - link
They should be good for things like serving up static content to lots of users.Kyrra1234 - Wednesday, December 12, 2012 - link
For me, the biggest advantage I see is VT-x support on the Atom CPU. That will allow me to run VMWare ESXi or something similar on the base hardware of an Atom system and they run some OS on-top of that with having minimal impact on performance.At least for home, I use an Atom based system as my low-power routers and for other general purpose linux systems. Being able to virtualize it makes it easier to manage.
I could see server makers building a 1U rackmount system with 10 Atom CPUs in them, and then that system could run 10 VMs under it. This would give each VM a dedicated CPU, without having to pay the price for the power of Xeon CPUs... Assuming it would be possible to put a lot of Atom CPUs in a single system.
Krysto - Thursday, December 13, 2012 - link
Cortex A15 has hardware virtualization, too.eanazag - Friday, December 14, 2012 - link
I was thinking that but...$54 per CPU for the cheapest is $540 and 60-80 Watts for 20 cores/40 threads. A suck ass GPU, so no GPU workload. More sockets on a mobo mean a more expensive motherboard.
It may still make sense to roll with a single Xeon because of the IPC performance. We need system power numbers and a little more details like memory DIMM size (like laptop SODIMM or full sized DIMMs).
I really think it make more sense for low compute need, dispersed systems. i would snatch one up for a small NAS or home proxy server. UPS needs are lower with this vs. a Xeon.
What would kick ass is to get these on a PCIe add-on card. And make a desktop blade server. A poor man's blade system for small business.
I would have rather saw an AMD bobcat variant with ECC.
bsd228 - Wednesday, December 12, 2012 - link
As Haswell promises to drop power draws to the 15Watt range as soon as next year, I don't know if there's is a low term point to atom/arm for the server realm. But there is a density option.A 6W atom cpu + msata SSD + single NIC system on a board would allow for a blade chassis of incredible density without putting out the heat of a fusion reactor. One use case would be offering colo/web services to people who don't want to share. A single rack could have hundreds to close to perhaps a thousand of these and still be at only 10KW.
But I have a hard time seeing a use case for singular units that would sell significantly. Certainly there would be a few good spots where saving another 5-10 watts is helpful, but enough to make it worth everyone's time to develop/support?
Krysto - Thursday, December 13, 2012 - link
These are server parts. Basically the "same" Atom that goes into mobile devices maybe have 2-4W, but has 8W for servers. The same would go for the Haswell part of Xeon. It would be a lot more power hungry.DanNeely - Thursday, December 13, 2012 - link
Intel has has 17W dual core IVB Xeon; the same TDP as their coolest mobile IVB parts at the time it was launched. They also have a 45W quad core Xeon; the same TDP as all but one of hteir mobile quadcore chips. At the moment they only offer a single Xeon at each power point; but if the demand is there nothing stops them from offering more choices in the future. The dual core Xeon is actually faster than any of the 17W mobile parts available. Granted there's probably binning involved in; but as long as they don't dominate demand Intel can skim off the top of production to badge as Xeon to absorb the higher TDP of ECC at all but the lowest base levels.http://ark.intel.com/products/65735
http://ark.intel.com/products/65728
name99 - Saturday, January 5, 2013 - link
On the one hand, Intel doesn't sell Xeons at $54...If a Xeon is the right solution for your task, buy a Xeon --- but expect to pay for it. Just like, if a POWER7 is the right solution for your task, buy it --- but you will pay for it.
The point of Atom (or server ARM) is not to replace Xeon, it is to be a cheaper alternative for those (frequent) cases where a Xeon is way more CPU than is needed, as I described above.
The REAL issue you all should be focussing on is not the quality of these CPUs (as measured by things like CPI), it is the quality of the memory system.
At the memory controller level, who is doing the better job of predicting which pages should be opened or closed?
Who is using advanced academic (not yet commercial) techniques like Idle Rank Prediction (the idea here is when you have cache write back, you don't write immediately, you store the write in a buffer and use a prediction mechanism to predict when the RAM is likely to be idle, so your write does not block a read).
There are even fancier things you can do like much more sophisticated rank/bank mapping. Don't just extract bits from the address. If you have, say, 32 banks available, don't just extract bits from the address for chip-routing, use a divide-by-31. This gives you 31 virtual banks which will have their addresses effectively scrambled over 32 physical banks, so that long stride sequential accesses don't keep hitting the same bank (just like Cray was doing back in the 1970s).
Even at the basic level, does ARM have something equivalent to Intel's SMI/SMB? (This splits the memory controller into a part on the main CPU, doing the real thinking, connected via a custom very high speed bus to a baby memory controller sitting right next to the actual DIMMs. The point of this is that the physical path from the DIMMs to the baby memory controller is made as short as possible, which means the DIMM bus can be run at higher speeds. The long run from the DIMMs to the CPU goes over the custom bus, which has separate read and write lanes and so can be way faster.)
And if not, what is their plan here? One solution (IMHO) would be for ARM to get heavily into packaging, and to encourage that their server chips are all sold with DRAM stacked PoP on the chip.
The point is, the days of decoupled CPU and memory are gone --- you cannot drive a decoupled bus fast enough. ARM needs a server solution to this fact which plays to ARM's strengths, not a mindless copying of the way Intel and AMD and IBM handle this (which reflect there rather different markets and strengths).
name99 - Saturday, January 5, 2013 - link
"I'm confused by Atom (or ARM) for servers though...what are they good for?"SOME computers do a substantial amount of computation. You want those computers to have powerful execute cores.
But OTHER computers mainly route data from disks to memory to the network. Computers like this have little need for a powerful execute core, since most of their time is spent waiting on memory. They are best served by having a sophisticated memory controller (which can predict accurately which DRAM pages to open or close) hooked up to a number of independent threads of execution, each of which can just pause while it waits on RAM.
Atom and ARM in servers target this latter market.
IBM does so as well, through a very interesting design. Hooked up a massively powerful memory system (way beyond what we are talking here) a POWER7 core consists of two sub cores, each of which can run two threads. For compute-intensive tasks, a core runs one thread which uses the resources of both subcores (so it has two integer units, two load-store units, two FPUs available to it). For memory intensive tasks, each subcore runs two threads, so you have four threads per core, and you're back in the world I described above --- each thread spends most of its time waiting, and the memory system is 100% busy serving up data to each waiting thread.
You will NEVER understand the design of modern CPUs until you realize that computation is cheap, memory access is expensive, so everything is twisted around making your memory work as hard as possible; and if that means your logic is idle 90% of the time, well, that's life.
SunLord - Wednesday, December 12, 2012 - link
So this is basically meant for low load servers that don't have to do much real work ever and instead of letting a powerful Xeon server waste power while they idle along under very low load you use a low power atom server at higher load levels?It does make sense as long as the load never changes or increases much then an atom will be fine but a xeon server would be able to scale far better across all sorts of different workloads if there any possible future changes. I guess I can see a place at the very low end or specialized projects where things won't change much.
I'd really like the list of the "light" server apps by name where an atom can beat a xeon as the hp graphic implies with it's performance per watt as things scale out but I have to wonder how the per dollar aspect comes into play when you start scaling up with multiple servers and what loads/app scale out across multiple atom servers vs xeon servers .
name99 - Saturday, January 5, 2013 - link
"It does make sense as long as the load never changes or increases much then an atom will be fine but a xeon server would be able to scale far better across all sorts of different workloads if there any possible future changes"(a) Don't throw around words like "scaling" without thinking what they mean. If your scaling dimensions are along the axes of disk or network, the CPU doesn't matter much.
If your workload is dominated by memory performance, then faster CPU doesn't matter much.
(b) The BIG boys (Apple, Google, MS, Facebook, etc) aren't buying servers on the idea that "well, let's get something fast and we'll figure out ways to use it". They have very definite roles that are played by different machines --- some are computer servers, some are cache servers, some are network servers, some are IO servers. These ARM and Atom CPus are targeted at all these roles except compute server.
DuckieHo - Wednesday, December 12, 2012 - link
Why only 8GB if they already did the work for 64-bit support? Support for ECC registered 2x16GB would make it compelling for MemCached, NoSQL, and other distributed memory-capacity intensive workloads.Kevin G - Wednesday, December 12, 2012 - link
The memory controller likely does not support registered memory which would limit DIMM size to 8 GB. Being single channel that limits the memory controller to two DIMM slots. That does raise the question of why it does not support two 8 GB unbuffered ECC DIMMs for a total of 16 GB.DuckieHo - Wednesday, December 12, 2012 - link
No support for registered memory. Yup, 16GB is still better than 8GB. I know one of the things holding back ARM in the server market is the 4GB limit. This should easily be able to handle 2x8GB then and makes for a more compelling product at minimal cost.Jammrock - Wednesday, December 12, 2012 - link
http://ark.intel.com/products/family/71263$54 for S1220, $64 for S1260... the low power part, S1240, will likely be a hair more expensive, say $66, knowing Intel's pricing schemes.
Comparing that to existing Atom boards, and Atom server boards, I would expect a Supermicro Atom S1200 board to run from $149-$249, depending on features, when it hits retail.
My question is this: will the S1200 series support multiple sockets per motherboard? Most ARM concepts I've heard of are for 32+ ARM processors per server. Nothing I have seen so far says whether the S1200 Atoms can do SMP.
DuckieHo - Wednesday, December 12, 2012 - link
Prices will be higher for final products, I think. I believe current Atom pricing includes a bundled chipset while these may not. That means you may have to factor in the additional chipset cost.These probably only have one QPI so single socket. By the way, that's why AMD bought SeaMicro.... for their "fabric" that connects the individual CPUs.
The 32+ ARM processors per server probabaly communicate over Ethernet?
Kevin G - Wednesday, December 12, 2012 - link
The main IO for this chip is likely the eight PCI-E lanes. QPI's main advantage are coherent, low latency links for usage in multi socket configurations. These Atom chips are strictly single socket.I do think that there will be boards with several of these chips on them, but each chip will be seen as an independent node. Communication will either be done via a PCI-e to some hub or via embedded Ethernet.
madmilk - Wednesday, December 12, 2012 - link
ARM needs those silly 32 processor-in-a-box setups to reach competitive performance per rack unit. I wouldn't expect Intel to do something similar with Atom, considering that Xeons are better in every way once a certain threshold of power is reached.tygrus - Wednesday, December 12, 2012 - link
There is also a 17w rated 2core Intel Xeon E3-1220L V2 nominal 2.30GHz (3.5GHz max). You just need to scale down the I/O, chipset and the reduce power consumption of other mainboard components. A low power CPU in a server board could still idle with >100W at the wall. It would be interesting if they could ad 5w to the Atom based cpu's to allow increasing the clock speed or additional CPU resources (increase IPC). Take the P3 and core2 etc. ideas and create a 2 or 3 decode, simple branch predictor, low power cache, 2 ALU, 1 slower FP/SSE/AVX multiply/div, 1 slower FP/SSE/AVX add/sub. Reduce the number of instructions in the pipeline and fast recovery from stall. Aim for a core <0.5 the complexity (&size) of Ivy Bridge at the same nm fab. Chipset & MB from notebook computer form factor and you have something to fit between Atom and i7 or just kill Atom in terms of perf/watt.Krysto - Thursday, December 13, 2012 - link
"Given the relatively low pricing of these processors, it will be in Intel’s interests to ensure that they are complementary to their existing Xeon processors and not significantly competitive."This kind of says it all about Intel's own internal problem. They won't try to make Atoms as best as they can possibly make them, to compete with ARM chips, and eventually, they'll get to the point where they have to compete with ARM so much on performance, that it would endanger their Core chip market. And Intel will have a HUGE dilemma about what to do next then. Keep Atom wimpy? Or cannibalize the Core chips?
In the meantime, ARM has no such conflict of interest, and they'll be as aggressive as they can in the server market, and try to cannibalize the Core chip market as soon as possible. Intel can't, and more importantly, WON'T do that to their own chips, that cost 10x more for the same die area, and give them more than 10x the profit.
wsw1982 - Thursday, December 13, 2012 - link
Yes, Only if ARM can do that. And by the way, the ARM server chip is no where cheaper than the Intel S1200. The Calxeda ECX-1000 powered HP redstone server cost 1.2 million and have 1600, 4 chips node, which put the per-socket cost is 187.5$.A5 - Thursday, December 13, 2012 - link
You're placing a lot of faith in ARM designs being able to hit Xeon-level performance without a Xeon-level TDP. Not sure that's going to happen anytime soon.Metaluna - Thursday, December 13, 2012 - link
You're also ignoring the effect of virtualization. If my only choice is a big, expensive Xeon chip which is going to be under-utilized, there is a strong incentive to run something like VMware on it and run a bunch of VMs. That can add considerably to the cost, and Intel sees none of that money. On the other hand, if instead of one Xeon plus a hypervisor, I can buy 5 Atoms, and run the same workloads on the bare metal (and not even have to change my OS since it's all x86), Intel sees a larger chunk of revenue that otherwise would have gone to VMware or some other company, even if the margins per chip are lower.Dadofamunky - Friday, December 14, 2012 - link
This is mostly a rehash of Intel slideware, after all (and would it really hurt the writer of this piece to spell out the use of every acronym the first time it's mentioned in an article? Basic tech writing principle. I would never hire Ryan for my tech pubs team). That said, it is good to know Intel is on the beat in this area, because I am highly skeptical of ARM's ability to do any serious damage here. I bought some Intel shares recently and plan on buying more due to the info presented in this article.Krysto - Friday, December 14, 2012 - link
Here's an indepth analysis by semiaccurate about why Centerton can't even keep up with current ARM server chips:http://semiaccurate.com/2012/12/14/intel-and-arm-v...
1008anan - Monday, December 17, 2012 - link
Thanks for the article. Useful information:"Calxeda is rightfully annoyed by these claims and put up a blog about it. Marvell seem to simply not understand the press, so we have no idea what they think, but you can be sure that they are not pleased with the terminology. "
Classic Marvell corporate communication!
This said the Centerton holds it own against its current 32 bit ARM server competitors. And Avoton more than holds its own. Calxeda is aimed at a subset of many threaded work loads and should do okay in its targeted market. The rapidly growing micro server TAM can support several product skews.
DDR4 - Sunday, February 3, 2013 - link
First of all it's a foolish idea to have micro servers and an even more foolish idea for arm to try to get in. Intel should stick to Xeon and arm should stick to low power computing.lemonadesoda - Friday, March 29, 2013 - link
Atom servers are great for NAS with services, not just "dumb NAS", (home and SOHO) or micro servers and colo for small business. I run 3x Atom servers. We are a small Management Academy with just 3 employees and 10 contractors. One Atom runs the webserver. One runs the fileserver. One runs on an offsite location and mirrors the fileserver for local access and also to provide backup functionality.Building and maintaining this setup is childsplay. And cheap.
Horses for courses.