Slashdot Mirror
Qualcomm Debuts 10nm Server Chip To Attack Intel Server Stronghold (tomshardware.com)
An anonymous reader quotes a report from Tom's Hardware: Qualcomm and its Qualcomm Datacenter Technologies subsidiary announced today that the company has already begun sampling its first 10nm server processor. The Centriq 2400 is the second generation of Qualcomm server SOCs, but it is the first in its new family of 10nm FinFET processors. The Centriq 2400 features up to 48 custom Qualcomm ARMv8-compliant Falkor cores and comes a little over a year after Qualcomm began developing its first-generation Centriq processors. Qualcomm's introduction of a 10nm server chip while Intel is still refining its 14nm process appears to be a clear shot across Intel's bow--due not only to the smaller process, but also its sudden lead in core count. Intel's latest 14nm E7 Broadwell processors top out at 24 cores. Qualcomm isn't releasing more information, such as clock speeds or performance specifications, which would help to quantify the benefit of its increased core count. The server market commands the highest margins, which is certainly attractive for the mobile-centric Qualcomm, which found its success in the relatively low-margin smartphone segment. However, Intel has a commanding lead in the data center with more than a 99% share of the world's server sockets, and penetrating the segment requires considerable time, investment, and ecosystem development. Qualcomm unveiled at least a small portion of its development efforts by demonstrating Apache Spark and Hadoop on Linux and Java running on the Centriq 2400 processor. The company also notes that Falkor is SBSA compliant, which means that it is compatible with any software that runs on an ARMv8-compliant server platform.
It takes a LOT of cache and very clever data paths to keep 48 cores fed with data. Intel cores typically have 2.5MB of local level 3 cache for each core and multiple ring busses so cores can access the whole cache and not waste precious off-chip bandwidth trying to read from main memory. If this is a special purpose chip for executing deep learning algorithms that's one thing, but for a general purpose server where tasks are uncorrelated, it ain't easy to prevent stalls while cores wait for data.
You're entirely right that the memory subsystem is 90% of the battle for most server workloads once you exceed ten cores.
For integer workloads with unreasonable parallelism and unreasonable cache locality (that Intel's AVX doesn't already handle almost ideally), I'm sure this design will smoke Intel on the thermal management envelope, a nice niche to gain Qualcomm some traction in the server mix, but hardly a shot heard around the world.
And Qualcomm better be good, because Intel will soon respond with Omni-Path Knights Hill—perhaps also larded with HBM—that could probably take on the same workload between power sprints (less power efficiency in the CPU itself—which isn't always the main power draw—and probably more flexible as part of a tidy one-vendor-rules-them-all server mix).
I'm all for vendor diversity, but let's not get ahead of ourselves thinking that 10 nm levels the playing field, sucking down the data aquifer through a double-wide handful of drinking straws.
Yes, core count matters, but size matters even more when it comes to the hose.
—
Looky looky, the bow moveth:
Intel announcements for AI: Nervana 100x faster than GPU, Knights Crest & Mill 4x faster, SKL mid-17
Kx Streaming Analytics Crunches 1.2 Billion NYC Taxi Data Points using Intel Xeon Phi
It would be interesting since AMD cancelled their ARM efforts in the server space.
Really we should use AMD's market success as some sort of indicator? What about their 80x86 failures? Is this the market?
How's that? AMD had failures in the 80x86 market? Well, depends on what you call a failure.. If we just look at the CPU market......
With a few exceptions, I find the AMD x86 processor family a pretty good value for the money. (Your mileage may vary) Yea they tend to run hotter and take more power than the Intel offerings, but they perform well enough in most environments to be viable. They may not do as well in the mobile and Server spaces (due to power consumption being higher at the same processing capacity) but they do manage to soak up part of that market too. In a desktop, AMD is more than adequate and cheaper than similar performing Intel options. Don't get me started in the advantages of AMD over Intel in the overclocking world with their unlocked clock multipliers...
AMD staying afloat in the face of Intel's market share is a pretty amazing feat. It hasn't been easy being the distant second while keeping up the pressure on the #1 player but AMD has kept going for decades. I expect AMD to continue to be the distant second competitor, but being second doesn't mean you are a failure...
Then there is the whole GPU market.... AMD may be less of a player here and I don't recommend their GPU offerings, but that doesn't mean they are a failure here either.
"File to fit, pound to insert, paint to match" - Aircraft Maintenance 101
My understanding is most server farms are connected to dedicated nuclear power plants anyway, so power consumption isn't an issue. Heat dissipation? Yeah, that might be an issue.
With recent news that Google is shooting for 100% renewable energy for its datacentres (and many others will follow suit), I'm not quite so sure that's true any more.
None of the features will be as small as 10nm in size, just like none of Intels 14nm chips have features even close to as small as 14nm.
The current trend in labeling since transistors started on their vertical adventures is to extrapolate an "equivalent" feature size based on overall transistor density. These TSMC-made chips have about a 30% higher density than their "14nm" chips, just as Intels "10nm" chips will have about a 30% higher density than their "14nm" chips.
"His name was James Damore."
It's called an "example". There are millions of servers that do almost nothing but run a bunch of Apache threads, many that do nothing but smtp, many that do nothing but nosql lookups, etc. It's very common, especially for companies with thousands of servers, to have servers dedicated to a single task.
But you really can't now since most components of the server that used to be discrete chips are embedded into the CPU. This is the case now with most systems. CPUs in consumer electronics have embedded graphics chips, audio, etc..
The server is not much different these days. You no longer have discrete memory controllers (you do sometimes), or discrete north and south bridges to handle things like expansion cards, network connectivity, etc. Now, a lot of that resides in the CPU. So the CPU determines what sort of (and how much) memory a server can accept, how many PCI lanes are available, and the number of sockets. It doesn't help either that the socket type is determined by the CPU manufacturer.
Intel in the previous century needed AMD for Intel's own survival for several reasons:
In the 80s and 90s when Intel was considered a small player in computation, many contracts called for a second supplier of CPUs in case Intel failed or failed to deliver. AMD was that company, which is why it was a near-perfect clone of Intel chips until the 386. AMD kept its license to make x86-compatible, independently-developed chips for a couple of reasons, which evolved over time.
Later, when Intel's dominance in the home computer market made it a natural monopoly, Intel used AMD's existence to argue against US-Justice Department litigation.
Even later, AMD's better technical decisions, IMO, gave it a performance lead at the same time Intel made a serious tech blunder with the Pentium-4. AMD became a better processor than an Intel. So Intel mobilized their hugeness and designed chips which outperformed AMD both in performance and efficiency, in the Core series.
AMD became a player in the graphics chip side through acquisition. Intel tried to develop GPUs but proved to be inept at it. Now, Intel is contemplating using AMD GPUs integrated into their desktop offerings. http://www.pcworld.com/article...
https://www.extremetech.com/co...
http://www.nasdaq.com/article/...
Intel's relationship with AMD is existential.
Well, at least performance wasn't thoracic.
Bruce Perens.
Anti-trust.
Bruce Perens.
Absolutely, if your WordPress blog needs about 1/4 the resources of a server, a virtual machine is a good way to do that. I offer that for our smallest customers. (We call it "Half Server", two cores and 8GB dedicated to each customer.)
If you need a cluster of 4, 40, or 400 nodes in your cluster of Squid proxies, the virtualization works the other way around - a true cluster is a rack row of machines that look and act like one. Each node, each piece of hardware, is an interchangeable and disposable part of of the whole. There's no reason to run a hypervisor on the nodes, the whole row, the whole cluster, is a virtual service.
Or use a GPU (http://shader.kaist.edu/packetshader/)
Shai Schticks:"You don't make peace with friends, you make peace with enemies"
Nobody learned.
Look at any standard library or application framework and you will not find any cache oblivious algorithms.
Linked lists are just traditionally implemented linked lists. Hash tables are just traditionally implemented hash tables. Trees are just traditionally implemented trees. Even sorting will be a ham-fisted quicksort.
pretty much only assembly language programmers give a shit, mainly because they are the only ones that understand the issues. Any exceptions you find are the exceptions that prove the rule.
"His name was James Damore."
I don't know much about foundries but I remember TSMC had some problems getting to this node as does everybody. What I do know is that fabs are all TSMC does. Intel is a bigger beast that does fabbing, software, motorboards, chip design, etc.
It is this, but I don't think its in the way you think.
Intels problem is that it cannot sell FAB time because they are vertically integrated. Intel builds a FAB and runs its next gen chips off of it for a few years, then they are stuck looking for something to do with the FAB when it is no longer current-gen. The problem is this specifically. Intel is competing with just about every FAB on the planet in this older-gen market (unlike with their desktop chips) so therefore margins are thin even on much older FAB's that are good enough to satisfy the bulk of the markets needs for all these secondary sub-products (drive controllers, etc...)
There are 3 kinds of semiconductor fabricators:
1) Vertically integrated like Intel. Only they can use their FABs.
2) Integrated device manufacturer like Samsung. They can sell FAB time to other companies so long as there isnt a conflict of interest.
3) Fair-play like TSMC. They only sell FAB time.
TSMC's revenue is now approaching Intel's, and unlike Intel they can keep all their FABs busy making money, so the outlook for Intel is grim without a serious restructuring, which they are doing (see recent massive layoffs, and bullshit marketing about their new "cloud strategy")
I've posted more than once about this on slashdot, and each time I end with the same recommendation: Sell your Intel stock.
"His name was James Damore."
Basically Intel and by extension x86 won in a large part by exploiting a FAB advantage. That FAB advantage is over, and the chip architectures that managed to survive have an opportunity to come back from life support. So the likes of Power, Sparc, MIPS and ARM now have a chance to compete on a level technological playing field with x86.
Coupled with the increasing use of open source which also negates the value of the x86 instruction set lock in then interesting times indeed.
AMD had a unique market opportunity to build up a good manufacturing base w/ quality fabs for their CPUs, but didn't. Intel gave top priority to their fabs, and are the standard
AMD spun off their fabs for precisely this reason. Building fabs is insanely expensive and the only way to do is to amortise the cost over a lot of chips. Even at its peak, Intel was producing 4-5 times as many CPUs as AMD and had a load of lower-end products (e.g. network interfaces) that they'd start using the fabs for once they were a generation old. There was absolutely no way for AMD to compete head to head with Intel in fab technology, because they couldn't get the economies of scale.
This does; however, highlight just how bad Intel is at CPU design. AMD has been able to achieve rough parity for decades (and been ahead a couple of times, with the original Athlons and Opterons) in spite of always being at least one process generation behind in fabrication technology.
I am TheRaven on Soylent News
I searched on Google. Found this in under two seconds. Took me more than that to write this reply.
http://www.theverge.com/2016/9...