Intel Launches Its First 10-Core Desktop CPU With Broadwell-E 184
Two years since the release of Intel's Haswell-E platform, which popularized 8-core processor to users. On Tuesday, the chipmaker unveiled Broadwell-E family, which consists of an "Extreme Edition" of Core i7 chipset that has 10 cores and 20 threads. (Do note that Intel is intentionally not calling it deca-core.) Intel says the Extreme Edition is designed for games, content creators, and overclockers. From an NDTV report: The 7th generation Intel Core processors are built on the 14nm fabrication process, and are part of the 'semi-Tock' release -- neither in the Intel Tick or Tock cycle. and come with Turbo Boost Max Technology 3.0 for more efficient core allocation for single-threaded processes, giving up to 15 percent better performance compared to the previous Haswell-E generation. All four new Intel Core i7 Enthusiast processors, codenamed Broadwell-E, support 40 PCIe lanes, quad-channel memory, and bear a TDP of 140W. Give Intel $1,723 and the Extreme Edition pack is yours.
A sucker born every minute (Score:2, Funny)
"Intel says the Extreme Edition is designed for games, content creators, and overclockers."
Also known as people too dumb to realize they're paying a thousand percent markup for commodity hardware.
Re: (Score:3)
NUCs are indeed awesome. I've got 3 chugging away in the server closet, having replaced older, slower, bigger, hotter, louder boxes and 1 retrofitted with an extra ethernet port that replaced the consumer grade router. The workload isn't bigger. So the servers can shrink.
I've got one on my desk at work, chugging away on simulations so I don't have to share the data center machines with 10,000 other engineers.
Re: (Score:2)
If you go down this road carelessly, you'll end up with an ultra-wide bus. So consider server back door operation, as this is generally an underutilized port. Make sure you employ Logical Unit Bus Expansion tech in both cases; otherwise bus errors are much more likely to occur. Note that as with all bus expansions, you must arrange for the bus impedance to be low; this requires low resistance, and also, if capacitance is relatively high, you'll find that
Yeah man.... (Score:5, Funny)
but does it go to 11?
Re:Yeah man.... (Score:5, Funny)
Well, it's an Intel chip, so it goes to 10.999739068902037589.
Re: (Score:2)
Only if you pay for the Tufnel Edition Xtreme. Its pricey, but hey, its one more.
Re: (Score:2)
Ocean's Eleven
Original or remake?
deca-core (Score:4, Interesting)
Perhaps somebody could elaborate on this?
Re:deca-core (Score:5, Funny)
deca is one letter away from decay?
Re: (Score:2)
I was thinking the exact same thing. Perhaps this? link [wccftech.com]
That really is a weird comment though.
Re: (Score:2)
deca-core = dick a core
Intel doesn't want to be confused with AMD.
Re:deca-core (Score:5, Informative)
One possibility is that "deca" is the slang for "Deca Durabolin". This anabolic steroid is infamous for causing loss of sexual function as a side effect - which is in a way quite ironically amusing, as one would imagine that most steroid users are trying to make their bodies more attractive for mates of whichever gender they prefer.
The loss of sexual function is called "deca dick".
Still, one would imagine that having your monstrously powerful new CPU named after a steroid isn't the worst thing. This chip certainly is "juiced up", it's the most powerful CPU Intel has ever released.
Re: (Score:3)
it's the most powerful CPU Intel has ever released.
Actually, no. Other chips are faster per core (6700K for one) [arstechnica.com] and Xeons go up to 22 cores [youtube.com].
Re: (Score:2)
It brings the most computing power to a desktop CPU, but yea, as written it is incorrect.
Re: (Score:2)
Maybe it's just time to end that practice. Once we get to 17 cores I don't want to hear "Intel launches it first septendecacore CPU..."
Re: (Score:2)
Oh, it is FAR too late for that...
https://www.amazon.com/Intel-X... [amazon.com]
Heptapentaconta-core!
Deca-core is the correct name, whether Intel calls it that or not. Everyone knows that deca means ten, and the strange stuff I see proposed in this thread (among bodybuilders it is slang for a certain anabolic...) is just not the reason they aren't going with it. They probably found that "ten core" markets better than "deca-core", and they may not even have a reason beyond that.
Re: (Score:2)
Most normal people probably don't know what "deca" means, so Intel wants to call it a 10 core chip to avoid confusion. I think outside of computer and maths enthusiasts quad is about as far as most people go.
Re: (Score:2)
Nah - we know octa because of the terrifying octopuses that plague our seas.
Re: (Score:2)
Time to genetically engineer a decapus. (or be learn to be terrified of squid)
Re: (Score:2)
Up to quad, you say.
Um, surely anyone from the US knows Penta-gon means 5 sides.
Re: (Score:2)
Apple sold their first computer for $666.66
Re: (Score:2)
That's just rounding man, you round, you don't truncate.
Multi-threaded applications (Score:5, Interesting)
Re: (Score:2)
One could even argue that they aren't even that good for gaming. In fact, if I were building for gaming, I'd go for the 6 core version that has the highest of the single-core performance, as that tends to matter a lot more to games, as games as a rule don't have a lot of CPU threads.
However, for what you describe, Intel would steer you toward 'Xeon' server or workstation.
Re: (Score:2)
Yeah, I'm going to save my money until Intel can release a deca-core CPU designed for web browsing enthusiasts.
Re: (Score:2)
I would like to have one for audio processing, but I can't afford it. :-/
Re: (Score:2)
I wish more IDEs had good support for multi-threaded compilation. Atmel Studio lets you use it, but the errors and warnings all get mixed up between modules with no easy way to sort them. The abomination known as MPLAB is even worse... I don't think it even supports it, with Microchip's custom (commercial!) version of GCC.
Re: (Score:2)
Edit the make script to pipe to separate logs.
The autogenerated make isn't multithread aware enough, but there should be a project setting to use a custom make. Of course you will have to futz with each one. Once you get it down to a routine, you will have a really good bug report for the devs.
The OS is doing backflips to mix those errors up like that, I bet fixing it speeds up compiles a tiny bit.
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:3)
Re: (Score:2)
* "Spinning up" threads is a negligible overhead - a few microseconds to start them and then they run for seconds or minutes
* "Scheduling" threads and "swapping" threads in and out of cores is not necessary - that's the whole point of multicore - they all run literally concurrently
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
I originally build these objects to implement a tagging system for my other objects. I have a container in which I can put millions of objects (tested out to 200 million so far) and attach tags to each one. For example: I can create a 'photo' object and store JPEG information within it and then attach all kinds o
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Help us AMD! You're our only hope! (Score:4, Insightful)
With AMD Zen right around the corner (October-ish) I believe Intel is milking their performance monopoly as much as they can with their $1700 CPUs.
The Zen should give us roughly Skylake IPC (Some predict a little better, some predict a little worse.) Being it's AMD, they'll have to undercut Intel's price if they want marketshare. If the arch is good, this will lead to a price war, which should drive down Intel to AMD price levels.
With any luck, high end Zen launch will be a 16-core with Skylake level single thread performance for $999. Sign me up for one of those!
AMD ZEN more pci-e then skylake (Score:5, Interesting)
AMD ZEN more pci-e then skylake. With all the pci-e based storage around Intel's skylake can't even power 2 M2 pci-e cards + 1 video card at full speed.
with amd zen it seems like 2 videos cards at full speed + lot's left over for storage / network / usb / TB 3.0 and more.
Re: (Score:2)
DMI 3.0 is only pci-e 3.0 X4.
Still only have 16+4 3.0 coming off the skylake cpu.
Re: (Score:2)
I like AMD but it's not going to happen. AMD had their chance when Intel didn't do 64bit with the Pentium D, they weren't able to capitalize on that because Intel played dirty and prevented OEM's from using them.
Now that Intel is actually competing even if AMD can produce an equally good processor design Intel's process lead will give Intel a 15% advantage. AMD will never catch up to that. The process advantage is substantial, it allows Intel to have significantly more transistors for the same power usage.
Re: (Score:3)
The process advantage is substantial, it allows Intel to have significantly more transistors for the same power usage.
This doesnt mean AMD gets that fab time (contracts with GloFlo may prevent it), but AMD isnt really Intels competitor. Only in imaginary simpleton land is a cpu design company like AMD a competitor to a semiconductor manufacturer like Intel.
The competitors to Intel are TSMC, GloFlo, Samsung, and Toshiba/Sandisk, At this moment all of these have 14nm, 15nm
Re: (Score:2)
Coils (Score:3)
Do the coils also whine 10 times more, it has been a nightmare with this new Skylake.
Re: (Score:2)
ERROR COIL 1 is on fire!
Too close to "decimate" (Score:2)
Re: (Score:2)
Beaten to death with the bare hands of his fellow soldiers.
Drawing lots would likely produce a better outcome vs. letting the PHB pick (stacked ranking). The Roman legions couldn't have worked with 50% of effort devoted to gaming rank and rate.
Re: (Score:2)
It was military punishment, and meant for soldiers who were mutinous in one form or another en masse. It wasn't universally used. It was felt, not by the soldiers involved, that to whack them all meant you had no soldiers but you could blow some non-mutinous air into the lot if only 1/10 got whacked.
Modern corporations do it all the time but not adhering to the 1/10 rule. Instead it is felt by the MBAs that when bonuses are reduced, this must be because of inattention to their bonuses by the rank and file m
Re: (Score:2)
From what I understand, it was largely a forgotten practice even among Romans. I think it was Marcus Crassus who revived it after his personally funded army lost an initial battle with Spartacus' slave army.
Not sure what a gamer would do with it (Score:3)
Re: (Score:2)
I don't think this is hardware for gamers. But if you can type make -j 20 happily you really will like this CPU.
Re: (Score:2)
Re: (Score:2)
not counting crap like Far Cry 3 where it binds to core 3 for some inexplicable reason
Far Cry 3...core 3...sounds like an easter egg :-P
Devil's Canyon is still the single-thread champ (Score:3)
Re: (Score:2)
All their effort has been on power savings, which for desktop PC's is a pretty darn low priority. I'd much rather they dropped the crappy integrated GPU in the higher end chips and filled that area with a couple more cores. Instead of go beyond 4 cores you have to but the higher end motherboard and pay Extreme pricing.
Who would buy a 6600k or 6700k and not buy a separate GPU?! It boggles my mind. A good 40% of the die area will never get used by the majority of buyers, there is enough room to fit 8 core
Re: (Score:2)
I suppose the integrated GPU is useful when your discrete GPU burns out (it happens). But on a "K" chip I'd prefer the most barest minimest iGPU just-functional-to-get-a-desktop-running, and use the saved space for a core or more cache.
All of my CPUs have 10 cores... (Score:2)
Value Isn't in the Chips, but... (Score:2)
Value Isn't in the Chips, but in the fact that you are less likely to need a less cantankerous dual-CPU motherboard for doing workstation chores such as CAD, Design, Photoshop, Video editing, and moderate raytracing.
Important news: 6 cores is the new 4 (Score:2)
Few average people will be running these LGA2011 boards/processors. The important news is that the mainstream i7 now has 6 cores. It really isn't affecting much else, as workstations have been built with Xeon processors for many years now, which have all had more than 4 cores for quite a few years now.
Only gamers and people with an OC fetish buy "Extreme" processors; everyone else just buys Xeons.
Re: (Score:3)
Re:"Desktop" LOL (Score:5, Funny)
It's obviously not meant for plebians. It's meant for the programmer who makes >$200 an hour, i.e. the time lost to compiling is worth more than this extreme high-end CPU is.
I did a PC refresh project at a Fortune 500 company a few years ago. The initial batch of Dell workstations had six-core processors. But Dell ran out of six-core processors and dropped in eight-core processors. The senior engineers almost broke out into a riot since they grabbed the initial shipment and the junior engineers were getting the eight-core processor workstations, upsetting a delicate pecking order throughout the office.
Re: (Score:3)
Sounds like a poor company culture to me.
More like H1B culture gone wrong. All Indians.
Re: (Score:2)
Don't be a pathetic racist nerd.
The racist nerd is the Indian who has 40 half-empty coffee cups with molds in different states of growth. The guy got fired and a hazmat team had to scrub down all the cubicles surrounding his.
Get outta your basement and your cubicle and meet real people.
Which is it? Basement and cubicle are mutually exclusive. You can have one or the other, but not both.
You don't think Indians are real people? How racist.
When is the last time you had sex?? And the boring motions you have with your lifeless wife everyday does not count as sex.
One, I'm not married. Two, when it comes to worshipping the tech gods, celibacy is highly valued.
Re: (Score:2)
Yeah, about that. Compiling is obviously a process with tremendous interconnected dependencies. While there are multi-core compilers, one would expect mediocre scaling, especially above the "normal" 4 CPUs.
Re: (Score:2)
cd /tmpfs/mybuild
make -j 10
Re: (Score:2)
No shit you can use 10 cores, but how much faster is 10 than 5? Note the word "scaling"?
Re: (Score:2)
The poster claimed that compiling is hard to parallelize. I pointed out that there is, in fact, a lot of opportunity for parallelism in many build processes.
Re: (Score:2)
At least in the C/C++ world most compiles are done on a per-file basis and then linked together at the end. So it's no problem to compile multiple files at once (provided your memory and storage can keep up).
Re: (Score:3)
It's obviously not meant for plebians. It's meant for the programmer who makes >$200 an hour, i.e. the time lost to compiling is worth more than this extreme high-end CPU is.
Oh please [xkcd.com], according to OpenBenchmark you can compile the Linux 4.3 kernel in 62 seconds [openbenchmarking.org] on an Intel Core i7-5960X. Unless you have a developer who just whacks the build button to throw shit at the wall and see what sticks - which is not the kind of person you should be paying >$200/hour - then almost any kind of employee perk or complimentary service would be more effective than 0.1 second off his compile time.
Re:"Desktop" LOL (Score:4, Informative)
The software I support takes about an hour to compile with a 20 way build on an enterprise class server blade farm. Before I optimized and increased the parallelization of the build process it used to take 10+ hours. Not every project compiles and links that quickly.
100 times as long as the kernel, I wonder why (Score:2)
That's interesting that your software takes 100 times as long as the Linux kernel does. My first thought was "I bet that could be reduced by 90%", but you said you already did that, reducing it from 10 hours on 20 build servers. I'm curious why it takes so much longer, what the difference is. Does this project have a lot more tightly coupled dependencies than the kernel does?
Re: (Score:2)
That's interesting that your software takes 100 times as long as the Linux kernel does.
Spoken as someone who has never used Gentoo. The kernel is one of the faster building components. Try some C++ builds.
Re: (Score:2)
It's a C++ project with a large amount of optimization to ensure it fits in the tight memory requirements of an embedded system. It also has to compile a lot of the code multiple times as it targets an embedded system which have dissimilar nodes ( different CPU / memory architectures / devices etc. ).
Re: (Score:2)
Do you think you've reached the realistic limits of how much you could parallelize the build process?
Re: (Score:2)
Theoretically I might be able to improve it, but several of the links involved take ~30 minutes and while I can have several links or other parts of the build at the same time it had to break down linking too much. I did a little though by subdivide the linking into partial links of related code in archive libraries. It reduces the overall optimization but speed up the linking. There still is some build system overhead that I can reduce. One of my largest savings was switching the build system from nested m
Re: (Score:2)
> That's interesting that your software takes 100 times as long as the Linux kernel does.
I don't think so at all. If your standard for production software quality is "the Linux kernel", then you must really look down on almost every software project! Add in that many languages are going to build way slower, and you'll really end up disappointed.
yes, I do (Score:2)
> If your standard for production software quality is "the Linux kernel", then you must really look down on almost every software project!
Yes, I do. I'm with Sturgeon when he said "90% of everything is crap." :)
Re: (Score:3)
Given the price I think that it's meant for a very select audience.
How many compilers support multi-threaded compiling? I can't find any reference that says Visual C++ complies using multiple threads, but of course that doesn't mean that it doesn't (just that I'm not hitting on the right question) but I'm curious if it does? Most references that I found mentioned that compiling a large project is pretty disk intensive and people recommended a SSD and more ram to speed compilation?
Re: (Score:2)
Re: (Score:2)
How many compilers support multi-threaded compiling? I can't find any reference that says Visual C++ complies using multiple threads, but of course that doesn't mean that it doesn't (just that I'm not hitting on the right question) but I'm curious if it does?
At least in the C/C++ world paralellism isn't done in the compiler itself, it's done by running multiple instances of the compiler at once working on different files. gnu make can easilly do this, dunno what things are like in the MS world but I'd be surprised if they don't have a soloution.
Most references that I found mentioned that compiling a large project is pretty disk intensive and people recommended a SSD and more ram to speed compilation?
Yes if you build a buildbox round one of these you will need to make sure the rest of the system is up to supporting the CPU cores. What that entails will depend on what exactly it is you are building. The platform supp
Re: (Score:2)
With respect, I think your approach to the puzzle of how to parallelize things needs more fundamental thought. (I speak from a position of one who has been in this situation!) The compiler doesn't need any "support" for parallelism whatsoever. Any project of non-trivial scale consists of many separate modules. All you do is run "make -j10" instead of "make". WHATEVER compilers make invokes are then run 10 at a time instead of one after another.
Of course I'
Re: (Score:2)
Re: "Desktop" LOL (Score:2)
Re: (Score:2)
Its called "the monopolist determines the prices". AMD is no real competition, and ARM is for mobile devices only. As intel is monopolist for desktop CPUs, they demand what they want.
Re: (Score:2, Interesting)
I believe this announcement is a response to the rumors of AMD's Zen processor, with more instructions per clock than Skylake and with a 95W TDP. Competition is sweet. Bring it on AMD!
Your comment is not new, but still amazes me... (Score:2)
In this day and age, it is still amazing to me that people say of cutting edge technology "Yeah, but whey do we need it? We'll never use it."
We will find a way to use it. We always do. "640K of memory..." and all that. Sheesh.
Re: (Score:2)
I game on a high-end PC and I think typewriters are cool.
Admittedly only if converted into mechanical keyboards. And if I don't have to use it myself.
But it is cool. I still want one of those devices used by the office workers in the film Brazil, they rocked.
But I don't know what a hipster is. I just know that people who claims others are hipsters tend to be total cocks.
Re: (Score:2)
Currently, 8-core for AMD.
Not buying online? Enjoy your 50%+ markup.
Re: (Score:2)
You are essentially destroying the job of the offline seller here. Yes, you also create a job for the parcel delivery, but how long until the drones take over, and the delivery companies pay real shit wages.
Just as using uber destroys the jobs of taxi drivers, or self driving automotives will destroy the jobs of professional uber drivers and truck drivers.
You as customer have choice between new model and the older one. The question of course is what happens with all the workless people this economy creates.
Re: (Score:2)
"You are essentially destroying the job of the offline seller here"
No, they destroyed themselves by failing to follow a basic tenet of business success - diversification.
16 core X 4. Fry's and Microcenter (Score:4, Informative)
AMD has x86 processors with 16 cores. As I recall you can have up to 4 CPUs per motherboard, so 64 cores total. Whether that's appropriate for your "desktop" is your decision. Their APUs have 4 CPU and 8 GPU cores.
Fry's and Microcenter are reasonable choices for brick-and-mortar retailers.
Only if you keep redefining "core", Intel (Score:2)
Each AMD module has two cores. Each core has a control unit (scheduler) 4 execution units (2ALU 2AGU) and an I/O unit to get data in and out. Each core is therefore capable of eight independent single-precision operation or four independent double-precision operations per cycle. With two each of ALU and AGU, arguably that's TWO cores and AMD should call it a 16-core CPU rather than an 8-core.
Several years ago, Intel's CPU cores were what most would call 75% of a core, sharing more parts between "cores" th
But the server / workstation -oriented motherboard (Score:3)
But the server / workstation -oriented motherboards due have dual cpu / lot's of pci-e slots (Some broads have all X16 with slots at 8 or 16 3.0)
Re: (Score:2)
And as always with Intel there is the curious disabling of ECC. With todays memory densities it really is beyond me that the option to put ECC memory on a $1700 CPU is disabled.
Re: (Score:2)
Yeah I'm also really surprised that ECC hasn't become more mainstream. I spent several weeks once chasing down a "compiler / build system" bug that turned out to be the result of a memory bit flip error that had the misfortune to ended up getting cached in the build avoidance system for a fairly static source file. One of the reasons I like server class build farms these days.
Re: (Score:2)
It doesn't matter how "common" they are. All you need is one good source. Just go here [asrockrack.com] and tick the checkbox "Socket 2011-3" or "Socket 1151", depending on which family of up-to-date Xeons you want to use. These are superb for desktop systems and don't suffer from any of the drawbacks you note:
* Boot as fast as any other desktop
* No IPMI[*] or other management features which you "don
Re: (Score:2)