One thing I think IBM has failed to grasp about POWER architecture is how to get people to use it.
Lots of people in tech talk about being interested in trying it, it has a number of interesting characteristics, but the barrier for entry is fairly high.
It's possible to get a POWER8 based server on Softlayer, IBM's cloud product, but not as a VM. You can get one as a bare metal server, but you can't get one for an hourly fee. You have to pay for a full month, which starts out at around $1000.
There are very few individuals that would be willing to make such a commitment, but so many that would be willing to spend a few tens of dollars on spinning up a VM for a few hours to see if it provides value.
If you want people to get excited about it, or interested in using it, you really need to make it easy for people to test it on a small scale.
There are no entry-level POWER machines, SPARC machines. There is no entry-level IBM i or Z. There isn't even a software emulator for i (and even for Z, where there is a very nice emulator, you can't legally run anything remotely current on non-IBM hardware).
You can get a cheap POWER or SPARC on Ebay, with a sharp price drop when the equipment is EOL'ed, but that's no way to entice people to build new stuff on a platform.
I love my PA-RISC, my RS/6000's and my SPARCstations, but, neat as they are, they aren't anything that would make me consider building for their modern descendants. And I won't pay more than a decent, proven, certainly useful Xeon workstation worth of money for a cool, but exotic-that-may-not-run-my-software-well, POWER9 performance equivalent.
If IBM wants me to pay Xeon Platinum prices for their production gear, they'd better allow me to pay Core i prices for development gear, or else I'll just keep deploying on Xeon, which is good enough.
Typed on a very comfortable and reasonably priced Core i7 laptop.
Seconding what you just said, and another thing is that the cost barrier to entry to developing x86-64 software on a very current Linux platform is almost zero. If you are a curious 15 year old who wants to learn Linux stuff you can get an older Core 2 Duo desktop PC for free, or almost free ($20) from a number of recycling places and install the latest debian-testing AMD64 on it. Put it on your home LAN, ssh into it, and start doing whatever you want.
I know of companies that regularly throw away second generation core i5 dual and quad core systems with 8GB of RAM.
The same software you develop on your $20 headless desktop PC can be easily moved over onto a $15/month KVM virtual machine somewhere in hosting. Or onto a used $250 1U server that you can pay to colocate somewhere for $65-100/month with bandwidth. All x86-64 platform.
The massive, massive economies of scale for x86-64 platform stuff are going to be very hard to get people to move away from unless there is an amazingly compelling reason. The ability to cannibalize random recycled computers to make one working computer out of "free" parts is a big thing for the developing world.
A very important point is that you mostly don't need to buy a new computer. Chances are a lot of teens already own perfectly good x86 computers that can do a lot of things they want and be the machines they learn to develop software on.
IBM might be having trouble letting go of their old model. If you want their amazing hardware they’ll do everything for you. Which means on day one you have none of the expertise needed to reverse that decision.
If you come from a world where lock-in is normal then you may well build something that locks everybody out.
You aren't their target market share. They are focusing solely on the HPC Market at this point. Their lowest in shape has 16 cores. Contrast that with a xeon at 6 cores. I think you were hoping that they have a desktop processor that you could play around with the actual architecture on. That would be nice, but they likely don't want to put all their eggs in too many baskets right now.
At the same time, you need mindshare at the lower end to drive demand for the high end. I previously developed for POWER by getting an Apple Mac Mini. Very affordable, and while not a full POWER instruction set was sufficient to test and validate my code. Today, there is no equivalent, and I'm about to dispose of the Mini. I couldn't justify use of POWER today even though we have a datacentre stuffed full of machines for HPC, and could likely make use of it. We have zero experience with it, and no easy way to get any experience. Being able to buy a cheap low-end system for experimentation is important. If I wanted to do this for ARM, I could buy half a dozen cheap boards online this morning and have them delivered by Monday. But POWER is out of reach.
Except for special US DoD things (like supercomputers that simulate nuclear weapons) for Sandia Labs and the like, the math also doesn't work out for HPC...
Let's say that a 44RU rack of 1U, narrow (half of 17.5" width), dual motherboard, single socket systems costs $N if it's built with Xeon or EPYC CPUs. As an example here the Supermicro barebones which are two long, narrow motherboards in 1RU with single sockets and front to rear wind tunnel airflow. Same rack of systems costs $N multiplied by 2.25 if built with power9, but only performs 1.25 times faster.
the cheapest power8 server I have ever seen was probably 4x more expensive then the equivalent dual socket xeon. That info is a bit stale. But there is also a complete lack of good information online about actual dollar figures for power8 systems (a lot of "contact us for more info!"). The platform's pricing is aimed at the sort of CTO people who don't care about the cost of Oracle licensing. Not the sort of people who care about $/MIPS for HPC applications or cloud scale hypervisor platforms.
I would very much like to be proven wrong so if you have access to some specific "this hardware spec costs this much money and you can buy it now" info, that i can compare to barebones supermicros, please do share it.
You're right in that the sticker price is much, much higher. I'm surprised you were even available to find one. IBM tends to like to keep things secret until you reach out to them, for better or for worse. I would like to see some public power 9 benchmarks to compare to xeons instead of just marketing speak as well.
But I will say that your comment about $/MIPS isn't accurate. There are many people, including Oakridge national labs, who are using it because of the higher performance and v100 integration. Intel is obviously not too keen on letting Nvidia succeed.
somehow, compared to the vast amounts of money they spend on other things, I think pricing is way down the list of things that are important to organizations like LLNL, Argonne, Sandia, Idaho National Labs, PNNL, etc. Basically everything that uses HPC for DoD top secret projects.
I sit on the POWER Customer Advisory Board, what you have raised has certainly been something I've told IBM multiple times over the last few meetings I've had with them.
All of IBM DNA is structured to sell big contracts and deal with a relative small number of customers with very deep pockets.
Even when they try some new trend ( aka cloud computing for the masses ) the economics aren't that good because the way they are structured, it's a culture thing.
It's like Intel developing ARM CPUs, they have the means, but they can't do it successfully.
One of the things that Nvidia quite successfully did (and Intel to some extent) is a ridiculously good academic outreach program. Making it easy (and cheap) for students and researchers to use you hardware leads to long-term buy-in.
From IBM's history, I'd assume this is chalked up to IBM institutional inertia (enterprise customers make big deployments, nothing else matters) than willfully doing it for some other profit-oriented purpose?
We (Oregon State University Open Source Lab) have been working with IBM for the last few years to help with this problem. We host a set of P8 (and soon P9) machines for them so FOSS projects can work on porting issues. You're welcome to sign up if you're interested [1].
Not any worse than the past decade of arm server "promise"
Bootstrapping an alt arch is hard, it really needs a loss leader OR a significant performance delta (which they now have with accelerator workloads). Google deploying it in prod is a major milestone and will help with volume and confidence.
Supermicro has a P8 and P9 line. It is price competitive with x86. If you are interested, send their sales a note.
Come on, you can buy a load of ARM 64 bit development boards starting from just a few tens of dollars. There is no problem at all getting access to ARM. And from there you can scale up to Cavium 2 or Amberwing which have Xeon-like performance (better than Xeon in some ways). If there's a problem with ARM I'd say it's still the lack of standardization, and also very few mid-range servers.
POWER is nothing like this. There are no development boards at all, for anything less than $thousands, and the real servers have great performance but sky-high prices.
You can't actually buy TX2. I don't know about market availability of Amberwing either (if you know please let me know). There is an abundance of arm stuff in the embedded space where it is king, and it's true you can pick them up for next to nothing but somewhat orthogonal to servers (flattened device trees vs ACPI/UEFI etc)
You can web buy an S821LC straight from IBM for $5k for the past couple years. The AC922 is GA. Supermicro will sell you a P8 right now and P9 in May for nominal prices and the CPUs are cheaper than Skylake by a very wide margin.
Not directly related to your comment but I get the feeling a lot of people complaining about price are navel gazing and have no idea how much a production server costs and how the costs break down. Right now storage is generally 50+% of the cost. DRAM is a very high fixed cost at the moment as well. Intel flatted the quad socket SKUs into the "Scalable Series" so Skylake represents a big price increase for a lot of builders. To help re-calibrate people, IBM doesn't have $6-12k CPUs in the dual socket config but Intel does at bins people would want for common workloads.
> Not directly related to your comment but I get the feeling a lot of people complaining about price are navel gazing and have no idea how much a production server costs and how the costs break down.
They're not asking for production servers.
$5k for a workstation, for a minority architecture, is basically a fancy way of saying "no" to the army of tinkerers you need to widen the base.
I have my own Cavium rep. Show me a real CPU, board and worthwhile sheet metal (i.e 10+ nvme) I can buy in volume because you might know something he doesn't and I'd be happy to eval it.
"At STH we are working with Gigabyte and Cavium and will share more about the ThunderX2 architecture as we are given the go-ahead. We have heard the next production run is in the Q2 2018 timeframe."
This has been my experience with cavium, bait and switch slideware and when you do finally get a sample it has so far been undesirable (octeon, TX1)
They are still useful though. If you want to build an 7-node cluster of 2GB nodes on a single Mini ITX motherboard, it's even cheap. And requires no messy cables.
That's neat. I had heard of the PINE64 board but never the SOPINE. Curious, have you implemented any projects on this? Seems like it might be neat for testing a small distributed system.
Those SBCs are still stuck with using very outdated SoCs usually with a quadcore A53. Even a 7 node cluster of SOPINEs is going to be slower than an iPhone X.
But it is worse. Over 12 million Ras Pis have been sold, I see them just lying around on developer's desks all the time.
I don't think I've ever even seen a bootable Power system.
That pretty much tells the story. A single, last man standing vendor like IBM with a business model of making money on Mainframe/“Midrange”/Old-School Unix boxes will never, ever be Intel. POWER is cool, but who cares.
It’s more productive to think about how to harness lots of cheap ARM cores if intel isn’t doing it for you.
My employer has a few. It’s fast and looks impressive. But when you cut away the bullshit, it mostly exists because the sales guy presented a story where the cost of a new POWER box is a better deal than maintenance on the old one.
If you really dig into it, there’s no scenario other than a license play where a transition to Intel isn’t more cost effective. Even in those scenarios, you can usually engineer a solution (Oracle, etc) where you deliver a better ROI on commodity hardware or cloud hardware.
The local UUG I was a part of had one available through the IBM innovation center. You were able to schedule time on it to experiment. Even with that close access it seemed like a PITA.
I will concede one point: it was not easy to purchase. I prefer to host my personal site on Power hardware; I started with AIX in the 3.2.5 days and I ran Floodgap on an Apple Network Server 500 for the better part of 14 years. I wanted to get a POWER7 to replace it in 2010, I budgeted $15k for it, and IBM wouldn't take my money. I couldn't find _any_ IBM VAR who would do an end-user sale because I wasn't going to buy the service contract.
Eventually I found a reseller who was more than happy to take $10K of my budget for a decent 2-year-old POWER6. It had a backplane burp a couple years ago but otherwise has been pretty damn spiffy.
I'll concede IBM has to do a lot more to get these systems into people's hands to achieve a critical mass and it certainly wouldn't hurt to make them cheaper up to a point, but the systems are out there, and you can get them (and find them).
As a postscript, in my current job (a large local government agency) I was in the CIO's office one day and the regional IBM salesdroid dropped by. Just to needle him I told him this story and he gave me his card and told him to call him with any parts requests, any time. I still buy from the reseller, though. They've earned my personal business.
Even in your example, the barrier or entry is still so high — I can buy a very nice new x86 machine for $10k, instead of a 2 year old POWER system. I am interested in the POWER architecture, but not buy $10k worth of old hardware interested.
IBM used power as a sort of step-down from Mainframe when big enterprises moved from cobol to java. That came with all of the bullshit like leasing CPU/hours and getting more software revenue using their "value unit" model.
The other thing with power is that it has been a scale up vs. scale out product. That makes sense when you want to optimize your Oracle/SAP licensing or something similar. It doesn't make sense for modern use cases where you're using open-source or other solutions where growing infrastructure into your use case makes more economic sense.
”The Power Cloud that enables developers offers no-charge remote access to IBM hardware, including IBM POWER8, IBM POWER7+ and IBM POWER7 processor-based servers on the Linux, IBM AIX and IBM i operating systems.”
An interesting discussion to be sure. As far as cloud based access go, off the top of my head both Oregon State and Nimbix give access to P8/P9 instances for developers, free or cheap.
One of our members, Raptor do a nice high performance dev workstation, pretty sure they do everything from a mobo+CPU combo to complete systems.
I think lower end than that might well come, we certainly see it as something we'd like to facilitate through the OpenPOWER Foundation.
I absolutely agree. It baffles me that after TWELVE years from the launch of S3 and EC2, IBM still hasn't understood what made them popular. (hint: the pay-as-you-go model).
> One thing I think IBM has failed to grasp about POWER architecture is how to get people to use it.
I don't think IBM cares about people using it, especially people who can't afford $1000 dollars. They are only interested in high margin corp/gov pork barrel deals and need something that they can plausibly claim is superior in order to charge 10X for it.
They definitely aren't charging 10X for it, it's very much in line and competitive at the Xeon level in terms of pricing, and they shine for certain workloads, especially with the new interconnects (CAPI, NVLink). POWER8 was harder to find, but the performance characteristics were good (in my experiments). And now the software is better (more solid ports) and cost is down. I'm eagerly awaiting the chips to hit this year and trying to find access to some.
Plus IBM licenses the design these days (through OpenPOWER), so several of these players are building their own chips, boards, etc. Someone could inevitably enter the low-end market but lower-end devices have thinner margins and a lot more competitors.
Even right now you can get real chips to go on a board on pre-order, just under $400 (TALOS II preorders) -- the mobo is the pricier part, but part of that is likely due to the BOM choices on that piece from Raptor Engineering. A smaller form factor motherboard (maybe with 1 socket) could land in the sub $2000 range for a whole mobo+cpu -- which is certainly competitive with similar HEDT/workstation prices[1]...
[1] I just dropped $1,500 on a 1950X threadripper and associated mobo earlier this year, so this price range is certainly alive and kicking, IMHO.
Threadripper also only has 8MB of L3 cache per CCX, and you can't hold more than 8MB of data. (true, there are 4x CCX per chip, but you can't "combine" the datasets. You're effectively limited to 8MB per 4-cores. To communicate cross-CCX requires an expensive ping to main-memory) Its a great architecture, but cross-CCX communication is relatively slow and definitely is a concern for some work problems.
Power9 has 120MB of L3 cache that's shared between all cores. Which means you can ACTUALLY have a full 120MB-sized problem set and share all that information between cores.
In short: Threadripper is great for sure, but its not necessarily a fair comparison. In an apples-to-apples comparison (ie: Monero Mining), it seems like a Power9 server is 3x better than Threadripper (Power9 gets ~3000 hash/sec, while Threadripper is roughly 1000 hash/sec).
> Using the xmr-stak-power PPC64LE-focused Monero miner, they are seeing great performance with it running on dual pre-production 16-core POWER9 processors. There's a hash rate of 2945H/s while this POWER9 system is pulling 350 Watts DC power.
And mind you: Monero / Cryptonight only uses 2MB of L3 per core. So that's practically Threadripper's ideal problem. Imagine if you actually had a dataset that was larger than the 8MB per CCX that Threadripper is limited to.
Not to hate on Threadripper at all. Its cheap and high performance. I'm seriously considering a Threadripper system myself. But these Power9 specs are incredible, and I'd definitely like to test one if I could afford one.
That's a dual socket system with 16 cores per socket. Dual socket Epyc 7351 (also 16 core per socket) is faster and more power efficient than that. 3200H/s at 332W.
Each Power9 is consisting of either 2-super slices or 4-super slices, depending on which Power9 you get. (Corresponding to 4x SMT or 8x SMT respectively).
Threadripper has 4x integer pipelines, 4x floating point pipelines, and 2x load/store units (called AGUs by AMD) and supports 2-threads (aka: 2x SMT).
So Threadripper definitely is "broad", but Power9 is "broader". The 8x SMT Power9 can perform 8x loads / stores per cycle per core, while AMD's Threadripper can only perform 2x loads/stores per cycle per core.
The last presentation on Power at Hotchips left me with the impression that "Core" in IBM terms is 9/10ths marketing speak and whatever Power might have that their engineers might think of as "cores" probably has little to do with what their marketing people are telling their customers and software vendors who do their licensing "per core".
Could you expand a bit on this? Are you saying they are inflating their core count or the reverse? Everything I've looked at indicates significantly more powerful cores.
Sorry but I just noticed this reply. I don't want to claim that IBM is "inflating" their core count, nor the inverse. Rather that what passes for what we might think of as a "core" based on the ways software vendors describe things isn't directly transferable to what IBM has actually built.
IBM has extremely flexible hardware which, using IBM's hypervisor, can be configured on the fly to present itself in a number of different ways. For example: few powerful compute units, many medium compute units, or very many less capable compute units.
So right up front the idea (which a lot of people might assume without really thinking about) that a given chip has a fixed number of "cores" doesn't really hold true. More to the point the next question that presents itself "what makes up for a core and what determines how many a given processor has" is best answered with "it depends" (at least as far as IBM Power goes).
One of the main reasons this matters is that all sorts of businesses use the idea of "cores" as a fixed entity as part of their pricing structures... and now it's all muddied.
They "inflated" it from 12 to 24 but recently it looks like they deflated it back to 12 cores. However you count it, a full Power9 chip is competitive with a 24-core Xeon or 32-core Epyc.
A full POWER9 chip should be compared with a 48-core Xeon.
In regards to cores, the distinction is between the PowerVM and Linux (OpenPOWER/PowerNV) ecosystem variants. Both are made to process 96 threads, but the difference is in whether the thread processing units are grouped eight to a core (SMT8) or four to a core (SMT4).
The PowerVM version, some say for licensing reasons, gets the SMT8 cores. Either way, you get 96 threads:
I will believe it when I can pull out my visa card, go to some online stores (such as those that resell Supermicro, Tyan, MSI, Quanta, etc), buy a Power9 capable ATX motherboard from one of the top-ten taiwanese motherboard manufacturers for $350, put a $300 CPU in it, $300 to $400 of RAM and build a 1U server with it. Until that happens it will not get into the hands of enough developers/sysadmin/devops/engineers to matter.
Note that the other DIMMs aren't completely arbitrary, they are buffered and require Centaur controllers to allow for much greater capacity and bandwidth (at increased latency). But for scale out workloads the on chip DDR controllers are great for both lower cost and lower latency.
I'm really excited to get my Talos finally in my hands (was one of the early orders). Good to see Tim Pearson and the demo system. I'm very hopeful the wait will be over soon.
That's great! Thank you already in advance! I am playing with the idea to get a TalosII computer. But so far I didn't find much experience reports from users and on top, on the https://raptorcs.com/ they are only on pre-order.
No, it was a workstation powerful enough to be a reasonable alternative to x86 that wasn't x86. I like Power, it's the architecture I'm most familiar with (other than the 6502 8-), and I think Intel needs a big kick right in the effective monopoly.
Sure, I could have bought one of the AmigaOne systems, which would have been somewhat cheaper, but those are basically embedded systems in ATX cases. The 12-year-old Quad G5 I'm typing this on would mop the floor with one of those. This isn't a slam on the Amiga community, who tried hard to bring in systems at not-completely-eyewatering prices, but this ends up producing boutique systems running underwhelming chips in unimaginative designs because of the small numbers and the rigid price points. More to the point, I don't have a strong Amiga history, so the AmigaOS side of those machines would be mostly wasted on me.
If there's going to be an architectural alternative in the grunt ballpark with x86, and I don't think ARM has gotten there yet, right now it's going to be Power ISA. And everyone in this thread complains about the cost. Understandable, but if no one steps up and buys one, no one will make any more of them and they will never achieve the necessary economies of scale. Fortunately I'm a PPC bigot^Wzealot with more money than sense. So I'll take the plunge. :)
The fact the firmware is auditable, no management engine/PSP crap, I get full schematics, etc., is just a bonus in my view.
I purchased one for the openness/privacy and for an alternative to x86. Currently my stack is entirely x86, though I have taken the time to purchase hardware that is freeable/mostly free (ASUS KGPE-D16/8, x200, and librem 13).
It seems I have close to no experience with POWER (other than older Macs), so, I am a bit concerned about what I will and won't be able to do on it.
When I receive mine I hope to be able to use it for most things, but as a fall back it should be able to be used for some self hosting.
Any particular place POWER guys hang out? I haven't been able to find a community that talks about day to day off x86 (albeit I haven't done much looking).
That would be nice, but the problem with these systems historically is that they're ridiculously expensive compared to x64 servers for the end user. As a result end users and SMBs never get their hands on them and nobody develops anything for them or supports them.
Server cost is dominated by storage at the moment. DRAM is also a large static cost with little negotiation room. The cost of a P9 and motherboard is a small subset of a system build and below the price of Skylake for me.
which is why they're not being sold directly to enterprises but to Cloud Providers. I'm not a hardware engineer so not familiar with the fine print but what I've read so far seems to suggest that Power systems are more efficient at running Database-centric workloads and the power hypervisor is an order of magnitude more efficient than intel ones http://www.computerweekly.com/opinion/Intel-x86-and-IBM-POWE....
Disclosure: I work on IBM cloud, but only tangentially with the Power systems.
As for the future, I can't say much publicly (le sigh) but let me just say that personally I think future of power is bright, especially with cloud providers.
The article addresses this at a couple of points, do you know factors which would counter its argument, or are you just pointing to the same history the article pointed to?
e.g.:
"Power 9 should do better than its predecessors given its costs, bandwidth, and ease of porting. Power 9 is IBM’s first to use standard DIMMs, opening a door to other standard components that are, overall, cutting system costs by 20% to 50% compared to the Power 8, said IBM’s partners."
IBM will never understand this but few, if any, architectures have made it into the datacenter that were not first affordable for experimenters and enthusiasts.
The burden of proof is on those claiming the pricing does not present a barrier to entry to the ecosystem.
I am curious about developing for Power. Please link me to an entry-level system I can buy on the web.
Well there are some counter examples, network switches, fiber channel, and Infiniband come to mind. And of course IBM "invented" the data center by defining environmental requirements for their machines in the 50's and 60's.
Setting that aside though, its true that easy experimental access can allow a technology to "sneak in" to a market that previously ignored it, but that isn't really the case here. (nor was it the case for SPARC or Itanium) These systems are being built for folks who are going to deploy a lot of them and they already know their cost of ownership numbers for existing x64 boxes.
This is an excellent way to cater to the legacy app market, but not a very good way to ride the next wave.
The next wave is probably running on commodity x86 hardware running Linux and using gamer-grade GPUs. They'll probably be first deployed to production on x86 virtual hardware in a cloud provider.
If you want companies to adopt your technology, make their developers adopt it first. That's how most of the current enterprise DevOps tools happened.
No one has accused of IBM of 'riding the next wave', sometimes I wonder if they even start paddling as the wave swells :-).
At least at Google the driving factor is total cost of ownership, full stop. When you need as much compute as they do to deliver on what they deliver, saving a few percent on the TCO flows right to the bottom line.
That said I don't see them going full on with this technology if its QP$S[1] isn't competitive with x64.
Buying the perfect hardware for your workload works really well when you have a couple million boxes to run it. POWER9 has the insanely fast NVLink to connect to specialized silicon Google has enough volume to design and build.
Where it can make a substantial dent in the x86 server is precisely this HPC niche. A 10% faster $300 box is worth $30 more, hardly enough to warrant extra work, but a 10% faster $10 million dollar machine is worth a million more.
Sun/SPARC for example, benefited greatly in Dot-Com v1.0 because so many universities had great prices via edu discounts. The sysadmins and other Unix users all cut their teeth on Sun boxes; so when they got jobs they took along their familiarity with Sun, which resulted in a lot of sales.
And as a developer you could get a basic v100 (or was it v120?) 1U 'server' from Sun really cheaply. Sure it was a single CPU with a PATA disk and no hotswap or redundant components, but it was a cheap SPARC you could test your stuff on.
Why does it matter what enthusiasts use? There are piles of Enterprise software written in Java that can be shifted to any architecture you care to name on a moment's notice. Lots of Java shops would bite your hand off if offered a 20% price cut for instances in exchange for switching architectures.
A. Many of the people who today specify datacenter hardware came into that career as enthusiasts (aka "early adopters") and are more comfortable recommending what they have experience with.
B. Enthusiasts bring the consumer volume that drives down production cost. This is a necessary driver of Moore's law.
It's worth noting that even entry-level machines were quite expensive compared to midrange PCs and were more or less in the same price range as high-end machines.
If, however, IBM can get me a 32GB single-core/8-thread POWER9 machine for the same price Lenovo can give me a performance-equivalent 4-core/4-tread Xeon, I'll be tempted. Worst case scenario, I'll still run my x86 workloads on my trusty Lenovo and use the POWER9 as a nice X terminal.
This is only good. Everyone is arguing here about stuff that really does not matter if this move the needle forward.
x64 by AMD pushed Intel to do better. Things became faster / cheaper. AMD fell behind and Intel rested and priced for high end went up. AMD released Ryzen and all of the sudden we see movement from Intel again. ARM on the low end pushed Intel to look at their lower powered chips and try to do better. IBM pushing P9 to win at the top at a reason price point pushed on AMD and Intel. Competition is only good here and brings value to all of us.
The architecture was something I was very excited about for low latency workloads; it seemed that only thing stopping IBM from overtaking x86 in this space was the actual availability of hardware. POWER9 has been "in the works" for, what, two years now [1]? Google have been running POWER9 internally for at least 12 months [2], and IBM seem to have been caught up in the AI hype train by pivoting the POWER9 to some kind of "AI processing system" [3].
Had the P9 been released before Skylake, it might have been successful. By the time it is GA, it will be competing with x86 chips two generations higher than were available at the original P9 announcements and SPECInt benchmarks.
This has been made even more frustrating by the fact that Oracle effectively killed off SPARC during this time. Perhaps Fujitsu will continue to run with it, but I think it's fair to say that development will stagnate and support will dwindle as it is increasingly obscure and niche architecture.
I can see China running with ARM and POWER, ditching Intel entirely.
Google, on the other hand would need a reason to switch. Somewhere there's a spreadsheet with a number for energy cost savings that would make it worth it for Google to switch architectures. I've heard it's as low as 10%.
> I can see China running with ARM and POWER, ditching Intel entirely.
China has Zhaoxin, which develops domestic x86-compatible cores that have received massive state funding, despite little commercial success. They way it looks, I think that China has decided to stick with x86, with Zhaoxin as an "escape valve" for the case where chip supply from Intel and AMD gets threatened.
If I were China I'd be cultivating a friendly chip fab for my high assurance platforms. There's no way China trusts Intel or AMD chips for classified processing.
MIPS was the thing in China... until it faded away. Then Power8 was the thing but it never shipped. Now we swear ARM is really the thing. Really for real this time.
If you read the article, they Demo'd Zaius (POWER9 based server) and apparently have a flag which compiles all (or most? Not clear) their apps to target POWER arch.
I think the formula is simple — if the amount of idle developer time waiting for development gear + the cost of development gear + the cost of porting your software to POWER > the potential cost saving of POWER, don’t touch it.
If POWER unlocks something fundamentally new and historically impossible on X86, like ARM unlocked cell phones, and such, maybe that’s another reason to try it.
At the end of the day, nobody got fired for choosing x86. Its well understood, it’s cheap, development tools are accessible, and it works.
From what I’ve read, POWER was affected by spectre/ meltdown too. However, Intels initial response to play down it’s significance can’t have gone well with large enterprise customers. In addition to the performance boosts the article refers to, I wonder how much of an impact the recent kerfuffle over Intels response had on POWER demand.
IBM already released patches for this, at least for POWER7+. Interestingly the Meltdown issue seems "only" to affect the L1 cache, so L2 does not need to be evicted. Raptor says all Talos systems will be patched prior to shipping.
Patching POWER for Spectre/Meltdown was a pain but we are done with it.
You have to apply patches to the LPAR/OS, VIOS and firmware. IBM has acknowledged that there will be a performance hit but have not provided any quantified numbers that I have seen so far.
"All POWER8 and POWER9 results in this table reflect performance with firmware and Operating System updates to
mitigate Common Vulnerabilities and Exposures issue numbers ... known as Spectre and Meltdown"
How did you verify if the patched system is not affected anymore? I'm really curious what means the users have to verify the effectiveness of such patches.
For well-optimised floating point calculations, this is generally true (even taking into account the SMT capabilities of Power8). Intel also has 8-wide (and more) SIMD in terms of intrinsics (AVX/AVX512), whilst Power is still limited to 4-wide with AltiVec, although with Power9 VSX should allow wider (I think).
Where Power8/9 really shines is memory bandwidth - it's orders of magnitude faster than leading edge Intel stuff currently. But if you're not memory constrained, Intel machines can still have a bit of an edge.
Where Power8/9 really shines is memory bandwidth - it's orders of magnitude faster than leading edge Intel stuff currently. But if you're not memory constrained, Intel machines can still have a bit of an edge.
That makes Power sound very appealing. I've never met a real life scientific simulation that wasn't often constrained by memory bandwidth. I'm sure they exist, but they aren't all that plentiful. I'd love to try running the code that matters to me on Power. (For one thing, I don't even know if it would pass basic regression tests; numerical code is touchy. Are gcc/gfortran reliably adequate on Power like they are on x64?) Unfortunately, as everyone else is commenting, entry level x64 is dirt cheap and entry level Power isn't.
gcc is perfectly cromulent on Power (speaking from experience with my own PowerPC and POWER6 systems, and when I had time on one of Raptor's POWER8 systems). The speed differential between that and IBM xlc is pretty small IME nowadays, too.
I'm more hesitant to say the same about clang/llvm but it's been awhile and I'm probably not current.
It's true that raw FP performance isn't the be-all and end-all, but things aren't necessarily memory-bound either. HPC codes at scale are often dominated by communication.
I'm surprised if numerical issues are that much of a worry. x86 has now caught up in doing FMA, in particular. I don't know what the build systems actually are, but most of Fedora (and Debian?) is available for POWER, and will typically have at least minimal tests in the package. It's not as if POWER is new for HPC; the Daresbury HPC-X machine was quite high in the top500 15 years or so ago, as an example I was physically close to.
Yes, but even when throttled, it can often be faster than full-clock less wide instructions, if your workload allows for it. So it's sometimes worth it and there is a benefit.
It's power throttled by design. I don't think it's heat throttled. I'm sure Intel could push for higher power or optimize power to allow full bandwidth performance, perhaps there isn't a strong demand for sustained, full-width vector execution, yet.
"Processors have two modes of thermal protection, throttling and automatic shutdown. When a core exceeds the set throttle temperature, it will start to reduce power to bring the temperature back below that point."
"Orders of magnitude" is implausible, and certainly not supported by published STREAM results for POWER8, even excluding HBM in Knights Landing, for instance. I'd love to see some actual performance figures for POWER9, though, or even facts about its SIMD, in particular.
Lots of people in tech talk about being interested in trying it, it has a number of interesting characteristics, but the barrier for entry is fairly high.
It's possible to get a POWER8 based server on Softlayer, IBM's cloud product, but not as a VM. You can get one as a bare metal server, but you can't get one for an hourly fee. You have to pay for a full month, which starts out at around $1000.
There are very few individuals that would be willing to make such a commitment, but so many that would be willing to spend a few tens of dollars on spinning up a VM for a few hours to see if it provides value.
If you want people to get excited about it, or interested in using it, you really need to make it easy for people to test it on a small scale.