Intel Details Upcoming Gulftown Six-Core Processor

MojoKid writes "With the International Solid-State Circuits Conference less than a week away, Intel has released additional details on its upcoming hexa-core desktop CPU, next gen mobile, and dual-core Westmere processors. Much of the dual-core data was revealed last month when Intel unveiled their Clarkdale architecture. However, when Intel set its internal goals for what its calling Westmere 6C, the company aimed to boost both core and cache count by 50 percent without increasing the processor's thermal envelope. Westmere 6C (codename Gulftown) is a native six-core chip. Intel has crammed 1.17 billion transistors into a die that's approximately 240mm sq. The new chip carries 12MB up L3 (up from Nehalem's 8MB) and a TDP of 130W at 3.33GHz. In addition, Intel has built in AES encryption instruction decode support as well as a number of improvements to Gulftown's power consumption, especially in idle sleep states."

Comments

Are most programmes multi-processor?

[SmilingBoy]

Can most programmes really be written to take advantage of so many cores? I am not sure I want to have a 6-core processor, of which 5 spend most of the time idling as I am only running a single-core-aware programme. OK, one more core can be used by the OS to make everything snappy, but the question stands.

Re:Are most programmes multi-processor?

[Anonymous Coward]

MultiTasking

Re:Are most programmes multi-processor?

[Anonymous Coward]

Cores are like girls in 'hot sluts gone wild' scenes - after a certain point you might hardly notice if there's even more of them, but you'd never say "no" to an increase.

Re:Are most programmes multi-processor?

[Anonymous Coward]

When was the last time your computer was only running a single thread/process? (apart from the BIOS initialisation I guess)

Re:Are most programmes multi-processor?

[goldaryn]

Cores are like girls in 'hot sluts gone wild' scenes - after a certain point you might hardly notice if there's even more of them, but you'd never say "no" to an increase.

But my "operating system" can only deal with one hand- I mean core- at once!

Re:Are most programmes multi-processor?

[SmilingBoy]

Wrong question. When was the last time my computer was running a single thread that could use 100% CPU for more than a few milliseconds. Answer: All the time. For example whenever I open Slashdot with Firefox. I rather have less cores at higher speed than more cores.

Re:Are most programmes multi-processor?

[pointbeing]

Can most programmes really be written to take advantage of so many cores?

Yup.

Got a Core i7-920 running at 3.2GHz at home - OS is 64-bit Kubuntu 9.10.

Yesterday I had five two-hour videos I wanted to render to DVD5 format - four were .avi and one was .mp4.

Launched five instances of DeVeDe to render the video and create the DVD file structure and did all five at the same time - then left for work. Took an hour and twelve minutes and the machine didn't melt, explode or let any of the magic smoke out of the box.

Even if an application isn't multithreaded the OS is - so even running a single task a multicore processor will give you a performance boost.

A Core i7 has four cores that'll run two threads each - presents as eight processor cores to the OS. I have no problem using them all ;-)

Re:Are most programmes multi-processor?

[turing_m]

I think I'd prefer two strong oxen. Wait, what were we talking about?

Re:Are most programmes multi-processor?

[TheRaven64]

Most programs can't be written to take full advantage of even one core. Most of the things that you do on a computer will run happily on a 1GHz CPU and still not bring usage over 50% more than occasionally. Most of the things that will tax a modern CPU can be made parallel, so will scale quite well to a number of cores. Even if your processor intensive task isn't using multiple cores, you still benefit a bit from being able to move everything else onto another core. With the recent Intel chips you also have 'Turbo Boost' (horrible name) which underclocks some cores while overclocking others, giving one core a speed boost for that CPU-eating single-threaded app while keeping the power usage and heat generation output. To prevent hotspots on the die, you can move the process around between the cores, giving each a boost for a little while.

Re:Are most programmes multi-processor?

[Anonymous Coward]

Be realistic. Multitasking for the vast majority of your regular user's needs were handled just fine by a single cpu and a pre-emptive multitasking system.

Having a second core was handy for people who like to play world of warcraft in one window and surf web pages in the other (considering how much CPU modern web pages eat for some reason. yay flash?).

Having two more cores beyond that is fairly useless for the vast majority of even power users except for very specific apps that even they are running a very small percentage of the overall time they are using their computers.

If you are routinely running apps that take advantage of all 4 cores in current cpus, such as video encoding or the like, chances are what you really need isn't another 2 cores in your cpu, but a version of your encoder compiled to use CODA or the like on your 192-core GPU.

Re:Are most programmes multi-processor?

[sznupi]

So...you had to wait quite long anyway. Long enough that you decided to do something else in the meantime, because there's no way you would simply wait for the result. Long enough that the speedup offered by Core i7 in comparison to the cheapest Celeron didn't mean much, since in either case the task would be done after you returned.

Re:Are most programmes multi-processor?

[TheThiefMaster]

If you left for work, would it have really made any difference to you if it took five times as long?

Thought it probably wouldn't, as you could have been hard-disk bound quite easily reading five large files while writing five other large files. The seeking alone would be nasty.

Re:Are most programmes multi-processor?

[jimbolauski]

Yes&No, it requires a higher level of skill and understanding to program in parallel. It is even harder to use them efficiently, IO conflicts for instance will make any multi-threaded program run much slower. The most glaring example, Microsoft has yet to solve, simply try to copy two large files separately on a multi-core and then copy the same files together the files copied in series will take much less time because the seeks have been reduced. Some programs will see no benefit Hello World for example.

Re:Are most programmes multi-processor?

[Anonymous Coward]

Only use concurrency when it makes sense. On my system, all audio runs through PulseAudio, which runs in its own process. Input (among other things) is handled by X.org, also running its own process. The scheduler can decide which process runs on which CPU and tries to use all available CPUs (or cores) in the most efficient way. So the operating system is already using concurrent processes itself.

Mobile processors benefit from a low load by entering various idle states which use less power then the active state, which in turn benefits the battery such that it lasts longer on the same charge.

Thus the point is not to try to use all available cores and every available CPU cycle, but rather to use the /least/ cycles possible for any given task. Otherwise there would be no point in adding more cores, because the programs would simply burn more cycles. Although operating systems generally do seem to grow to require more processing power, computers typically run only one operating system at a time. Having multiple cores to work with is meant to benefit the end user by allowing more processes to run simultaneously, such as encoding/decoding an audio and video stream at the same time. Or multiple A/V streams with different camera angles, the possibilities are for the users to explore and there will be plenty.

Re:Are most programmes multi-processor?

[stilldead]

As soon as virtualization is on your mind and not a laptop this makes beautiful sense.

Re:Are most programmes multi-processor?

[Anonymous Coward]

With most browsers becoming multithreaded or multi process, even casual users will potentially gain from this.

Not to mention your machine may have a truckload of background processes going on.

Games etc will definitely benefit from any extra cores you throw at them. Since a fine grained threading library like OpenMP will grab all available threads when sharing out work. Also the Xbox 360 has 3 cores so a lot of games will use at least 3 threads as a minimum (since most games are multiplatform now).

Re:Are most programmes multi-processor?

[stilldead]

Let me rephrase this. As soon as you think virtualization and not just one OS this makes beautiful sense. There, I fixed that myself.

Re:Are most programmes multi-processor?

[pointbeing]

So...you had to wait quite long anyway. Long enough that you decided to do something else in the meantime, because there's no way you would simply wait for the result. Long enough that the speedup offered by Core i7 in comparison to the cheapest Celeron didn't mean much, since in either case the task would be done after you returned.

Most likely the Celeron *wouldn't* have been done when I got home from work - running the same job on my old hyperthreading 2.8GHz machine would have taken about ten hours, just going by past experience. The old P4 took about as long to render video as it did to watch it.

If I *had* been home I'd have reniced the encoding processes so that I could do day-to-day stuff while rendering the video. It'd have taken twice as long but I wouldn't have seen much of a performance hit.

Re:Are most programmes multi-processor?

[pointbeing]

If you left for work, would it have really made any difference to you if it took five times as long?

Well, yeah. I wouldn't have been able to brag about it ;-)

Re:Are most programmes multi-processor?

[physburn]

Most programs are very much not written to take advantage of multi-cores. Even advanced 3D games which might find the extra compute power useful, often can't deal with extra cores. E.g. I had to set the affinity of Borderlands to 1 CPU only to stop it crashing. Multithreaded programming is slowly getting easier as libraries to help it, become available. Java is particularly easy for this, have a look at java.util.concurrent, with i've just started using on the serverside. But most programs are miles behind in the move to being able to work with multiprocessors. Right now 6 cores will have very little to offer the desktop, on the server side however, i'm sure the extra core will have use, but only if the server is particularly loaded with transactions, something with rarely happens.

---

Multithreaded Programming Feed @ Feed Distiller

Re:Are most programmes multi-processor?

[FlyingBishop]

If we get up to 24 cores for consumer hardware I'd wager we can ditch the GPU altogether, which would make cross-platform stuff a lot simpler.

Custom-built GPU decoders/encoders make things really flaky and hard to run in a variety of environments.

6 doesn't sound like quite enough, but from my experience with 4 I don't think the mythical average user would notice the difference between 10 full-powered cores and 1 plus a GPU. Now this may not be cost effective, but if they can eventually offer 10-core CPUs for the price of our current CPUs, it would probably be well worth getting rid of the platform incompatibilities foisted upon us by Microsoft, Nvidia, and ATI.

Re:Are most programmes multi-processor?

[kjart]

Having a second core was handy for people who like to play world of warcraft in one window and surf web pages in the other (considering how much CPU modern web pages eat for some reason. yay flash?).

Having two more cores beyond that is fairly useless for the vast majority of even power users except for very specific apps that even they are running a very small percentage of the overall time they are using their computers.

Not that I particularly disagree with your conclusions overall, but wow can actually be set to run on multiple cores and does get a performance benefit for doing so.

Re:Are most programmes multi-processor?

[war4peace]

Maybe GTA 4 is a very specific app for you.To me, GTA 4 is an important application :)
And that game, my friend, requires a quad-core CPU. Which I don't have at the moment, and even with my Intel E6550 overclocked from 2.66 to 3.6 GHz the game runs at 25-35 FPS at 1680x1050, with both cores at 100%.
I used to have the same approach (multicore is dumb) but now it's a thing of the past.
There are more and more apps and games out there who take advantage of multicore, not to mention that operating systems (such as Windows 7) are better at allocationg sepaarte cores for single-core applications to balance the load.
Therefore, while I don't dismiss the need of having a multicore CPU, in the end it's all about balancing your investment against the benefits. If you need a 6-core CPU and it doesn't cost a kidney and a lung, then get it. Regardless of your decision as a customer, there's always a good thing in this sort of advancements: new stuff pushes down prices for older stuff :)

Re:Are most programmes multi-processor?

[Bandman]

In enterprise computing (the target market for this tech), these processors will be used for one of two things:

Virtualization: You stuff a machine full of these processors, then run 50 guest machines on it. Three machines like that contain more processing power than my entire infrastructure did 3 years ago.

Niche: Stuff a machine full of these processors, then crunch numbers. I've got a pair of 12 core machines with 32GB of RAM apiece that do nothing but monte carlo simulations...and that's not even particularly big for a lot of people.

These multicore processors won't be running web browsers, unless it's because they're hosting a dozens of virtualized desktops that users are working on over the network.

Re:Are most programmes multi-processor?

[sznupi]

Ah, so you don't realize that the cheapest Celeron nowadays is a dualcore 2.5 GHz, essentially a Core 2 Duo with 1 MiB of L2 (irrelevant, encoder will fit and the video is a stream of data) and 800 MHz FSB (irrelevant, mostly limited by the speed of computation, not by sustained transfer of the video stream). It would be done probably in around half of the time you were at work.

If you were at home with a Celeron you could also do day-to-day stuff (yes, it would take even longer - but is that really that important in the case of a rare batch job which in every scenario is too long to be a "smooth" workflow?)

Re:Are most programmes multi-processor?

[Chaos Incarnate]

The Core i5 and i7 can both do this—if you're using fewer cores, it will shut off some and increase the speed on the others (within the same power/thermal envelope). The LGA1156 models are better at it than the LGA1366 ones, though.

Re:Are most programmes multi-processor?

[sourcerror]

But with two cores if one of your programs freezes, you still have a responsive system. (I mean the gui and the task manager.)

Re:Are most programmes multi-processor?

[Anonymous Coward]

Wrong question. When was the last time my computer was running a single thread that could use 100% CPU for more than a few milliseconds. Answer: All the time. For example whenever I open Slashdot with Firefox. I rather have less cores at higher speed than more cores.

Really? So one thread wasn't reading the network traffic, one wasn't parsing the markup, and a third putting things up on the screen? At the same time the page wasn't being saved to your browser cache, while your e-mail program was querying the server for new mail, and cron was checking to see if there were jobs to run this minute? If you're on Windows, all of these activities were probably scanned by anti-virus.

There's a lot going on in a modern system:
$ ps -ef | wc -l
146

Re:Are most programmes multi-processor?

[spiffworks]

For example whenever I open Slashdot with Firefox.

Seriously, why the hell does Firefox struggle so much with slashdot? Chrome seems to do it fine.

Re:Are most programmes multi-processor?

[Big Smirk]

Around here, the programmers never met a thread they didn't like. Add a requirement like - "display dialog box to confirm shutdown" and suddenly the thread count in the application jumps by 4...

Could things be done more efficiently? No, because that would require thinking and thermodynamically it is cheaper just to spawn another thread.

Re:Are most programmes multi-processor?

[asdf7890]

Even if an application isn't multithreaded the OS is - so even running a single task a multicore processor will give you a performance boost.

Not quite. A single single-threaded process will not run faster on a multi-core processor (in fact it might run a little slower if the OS doesn't keep it running on the same core for the whole run, as jumping the task between cores uses L1 and L2 cache less efficiently) because the OS will not know how to try and split it up, but running multiple single-threaded tasks (i.e. your five instances of DeVeDe) will most likely benefit significantly (unless other bottlenecks such as I/O bandwidth/latency kick in).

Re:Are most programmes multi-processor?

[Pieroxy]

Why should it matter if browsers are multi-threaded? Is your browser maxing out the one cpu it's on at the moment? I don't know about the rest of you, but if my browser is holding at 99%+ cpu usage, historically that's told me that something is -wrong-, not that I need more cpu power.

Hiding comments in a slashdot story that has more than 400 comments?

Re:Are most programmes multi-processor?

[Big Smirk]

Real time games are a bad example because in general the trouble with threads is you have to sync them up. The entire program becomes give feedback, gather input, calculate stuff, give feedback. You generally need to make sure the calculate stuff parts starts and stops with some predictability.

Some games seem to run their AI in separate threads. These seems to be a reasonable compromise. So when the game does 'gather input' it asks the AI subsection where it wants to go at that instant.

However, its judging by the stability of games like Fallout3, its unclear if either the programmers know how to deal with threads or the underyling OS is ready for intense real time updates.

Re:Are most programmes multi-processor?

[spinkham]

This is why I bought a 2 core CPU. For the power draw and upfront cost it made the most sense for my workloads.

Occasionally I encode videos for youtube and vimeo, and compile large software packages, but otherwise a 4 core CPU would be drawing power needlessly and unnecessarily expensive for me 99.9% of the time..

Re:Are most programmes multi-processor?

[siride]

This isn't Windows 3.1. If a program freezes, it uses NO CPU, unless it is spinning, in which case its CPU usage is still limited by what the OS is willing to give it. It really has nothing to do with responsiveness, which is a function of the OS scheduler.

Re:Are most programmes multi-processor?

[petermgreen]

Around here, the programmers never met a thread they didn't like. Add a requirement like - "display dialog box to confirm shutdown" and suddenly the thread count in the application jumps by 4...
Lemme guess these programs are also buggy crash prone peices of shit?

having more than one thread doing UI stuff has always struck me as more trouble than it's worth (you need loads of extra locks and a lot of thinking about what does and doesn't constitute a consistent state). Indeed some common gui libraries (swing for example) aren't built to support multiple threads accessing thier components for just this reason.

Re:Are most programmes multi-processor?

[imakemusic]

that game, my friend, requires a quad-core CPU.

Seriously? Is this an official requirement?

If so that would explain why it runs like a bag of shit on my (otherwise fairly good) PC.

Re:Are most programmes multi-processor?

[wwwillem]

Or on a simple Windows XP desktop: open Task Manager and add the "Threads" column. IExplorer 17 threads, System 56, some svchost.exe ranging from 15 to 70 threads. OK, many will be dormant, but I prefer them to sleep on another core. :)

Re:Are most programmes multi-processor?

[edxwelch]

No, but the benchmarks are, in other words this new CPU will be ideal for running benchmarks ;-)

Re:Are most programmes multi-processor?

[Courageous]

You're question is missing something: the Enterprise virtualization market. It's huge. It's inherently parallel. It's basically the main thing that has driven the entire multicore market for the last 3-4 years.

C//

Re:Are most programmes multi-processor?

[Loki_1929]

VMware. /debate

Re:Are most programmes multi-processor?

[ae1294]

I'll take the milfs :) those ladies have experience

And syphilis! enjoy...

Re:Are most programmes multi-processor?

[TheTyrannyOfForcedRe]

I'll take the milfs :) those ladies have experience

They also have skin with massive sun damage, stretch marks, cellulite, and tits that are a mess after two or three births, and that's the ones who've chosen to stay natural. Surgically enhanced MILFs are even worse. Overinflated lips, over-tight facelifts, implants, and botox face...Ugh! None of it ends up looking or feeling anywhere near natural. Were talking Uncanny Valley territory here.

I'll take the 19 year olds, thanks! : ) They can quickly learn to do all the MILF tricks and teaching is fun.

Re:Are most programmes multi-processor?

[LWATCDR]

On server yes. Without a doubt.
On a desktop even if a certain task isn't you are probably running multiple programs in the background that can use the other cores.
Honestly programmers really need to start thinking about how to use these new CPUs.
A good example is where I work our major product does a timed backup of any open document every few minutes.
It can produce a noticeable blip now and then.
In the next version we may just make a copy of the document in RAM and then write the backup in a tread from that copy. On a modern multicore system the end user will never notice it.

Re:Are most programmes multi-processor?

[pointbeing]

Not quite. A single single-threaded process will not run faster on a multi-core processor (in fact it might run a little slower if the OS doesn't keep it running on the same core for the whole run, as jumping the task between cores uses L1 and L2 cache less efficiently) because the OS will not know how to try and split it up, but running multiple single-threaded tasks (i.e. your five instances of DeVeDe) will most likely benefit significantly (unless other bottlenecks such as I/O bandwidth/latency kick in).

My point was that with the kernel and whatever services are running spread out over multiple cores there *should* be more processor resources available per core to run the single-threaded application - at least that's been my experience.

Re:Are most programmes multi-processor?

[pointbeing]

I think what *you* don't realize is that I run this big, bad i7 to compensate for being woefully underendowed.

But seriously, I get your drift on running unattended processes. I buy probably a thousand computers a year and generally buy what people need, not what the latest IT comic book says they need to have ;-)

Re:Are most programmes multi-processor?

[hey]

I read recently that the new Firefox uses a separate core garbage collection.

Re:Are most programmes multi-processor?

[Z80a]

Not really sure, but if i'm not mistaken, everytime you add a extra instruction per cycle, you double the CPU size, while getting another core doubles the total power of the cpu at least in theory.

So, as a intel Core executes like... 4 instructions per cycle, to do one that does 8 you would need 16 times more chip space by that logic, and well, you can get "the same power" by just having two cpus that do 4 instructions per cycle instead.

Re:Are most programmes multi-processor?

[Guspaz]

Examples of things that benefit from more than two cores:

- Modern web browsers such as Chrome
--- Multi-process architecture means that Flash sucks up a CPU all to itself, while various other tabs/domains are in different processes. Javascript-heavy web-apps or the user of other flash like plugins can easily make 3+ cores worthwhile
- Most modern games
--- Games are very CPU intensive. Most modern engines do a very good job of taking advantage of multiple cores. Some games even require 2+ cores in order to get any decent performance; MassEffect 2 (Unreal Engine 3) is unplayable on single-core processors
- Video encoding
--- GPU-accelerated (CUDA, OpenCL, etc) encoding is not yet useful, and isn't likely to be so any time soon. Existing hardware accelerated encoders are extremely limited in flexibility, and are usually of very poor quality (in terms of output).
- Multitasking
--- You scoff at it, but if I've got some demanding application running and I try to do something else at the same time (such as pause a game, alt-tab out of it, try to watch a video), CPU load starts to add up.

Most peoples' needs can be met with the low end dual-core processors with hyperthreading such as the i3 or i5 series, but these days it's not just anybody who can take advantage of a quad core CPU. Pretty much all gamers, for example.

Re:Are most programmes multi-processor?

[Tim C]

Most of the things that you do on a computer will run happily on a 1GHz CPU and still not bring usage over 50% more than occasionally

Speak for yourself.

Re:Are most programmes multi-processor?

[MrHanky]

Unofficial but true. Tom's Hardware has a bunch of benchmarks. The only dual core CPUs that manage 60 FPS are the i5s with hyper-threading, and they are beaten by less expensive Phenom II quad cores.

Re:Are most programmes multi-processor?

[Scarumanga]

Right now, it might just be wiser to go buy a 360 or a ps3....and grab gta4....it would cost the same as buying a new processor, at least then the game would run perfect.

Re:Are most programmes multi-processor?

[Lord Ender]

In my opinion, the advantage of multi-core is not that you can make tasks parallel, but that you can run non-parallel tasks in parallel.

Yesterday I had to do some password-cracking. My fastest system was my quad-core desktop. On a single-core system, the cracking software would have made my desktop unusable. But since this was quad core, I could fire the process off and continue to use my system as usual, with no perceptible slow-down on FireFox, NetBeans, or any other app I was working with at that time.

Re:Are most programmes multi-processor?

[BlackSnake112]

I'll take the 19 year olds, thanks! : ) They can quickly learn to do all the MILF tricks and teaching is fun.

Plus the 19 year olds usually have more energy.

Re:Are most programmes multi-processor?

[u38cg]

Me, I just have BOINC fired up 24/7.

Re:Are most programmes multi-processor?

[networkBoy]

Heck my normal workflow eats a minimum of two cores.

Re:Are most programmes multi-processor?

[Jeremy Erwin]

I heard that Linus Torvalds doesn't even work on the kernel these days. He just downloads the latest source from bittorrent.

Re:Are most programmes multi-processor?

[networkBoy]

I don't see that as MS's fault per se.
I realize there could be some smarts built into the system that would notice the two threads IO blocking each other and put them into serial operations, but really, you would have to make the CP operation more complex and have some sort of -nice kernel flag that each program could register... All in all the complexity is not worth the improvement.
-nB

Re:Are most programmes multi-processor?

[Snufu]

less cores at higher speed

And I want a pony.

The move to multiple cores was driven primarily by Moore's law approaching fundamental physical realities in trying to further shrink conventional CMOS transistors. I'm sure manufacturers would have loved to stay with single cores and bump up clock speed every financial quarter rather than upset the industry's technological architecture.

Solid State Physics 1, Corporate Financial Officers 0.

Re:Are most programmes multi-processor?

[joeyblades]

Actually multi-threading...

Re:Are most programmes multi-processor?

[radish]

Here's a tip for Windows users. Take a look at task manager, and add the "number of threads" column. You'll see that pretty much everything has more than one thread. I'm seeing 12-24 for each IE instance, 16 for Excel, 54 for Outlook, 46 for IDEA, etc. A lot of developers (particularly on slashdot) seem to have a real issue with threading, I think they much be C++ coders, as other more modern languages make multi-threading much easier and more reliable. You still have to be careful, of course, but the benefits can be great.

Re:Are most programmes multi-processor?

[Penguinoflight]

While the 360 and ps3 versions are coded much better they're far from perfect. Cruising down a highway for a long period will get you crashed into an object that hadn't been rendered yet.

Re:Are most programmes multi-processor?

[Overzeetop]

The cool thing about processors now is that they can dynamically allocate the total thermal envelope - and they're getting good at it. So if you only need one core, it runs at (say) 3.5GHz and the others are idle. That might scale up to all 6 processors running at 2.5GHz with a bunch of threads. The key is that the processor will perform better on a single threaded applicatino than it's cover clock speed would indicate. The i7-8xx do this now, and the results are very good in the real world, despite not being able to best the top chips when running full out. It also means that when it's idle, the power dissipation can be throttled way back.

That's bad for folding@home and other distributed apps, as it now costs you electricity to run those apps "in the background." Before, you paid for most of the processor power regardless, so it was "free" operations.

Re:Are most programmes multi-processor?

[joeyblades]

Multi-core and multi-threading is really the only way to get significant performance increases these days. The days of bigger, faster cores are pretty much done, due to a number of technical issues with chip design and manufacture. Some of the key problems being power distribution, heat, and defectivity.

Re:Are most programmes multi-processor?

[grumbel]

Can most programmes really be written to take advantage of so many cores?

No, for most programs that would be quite complicated, however most programs idle around all the time anyway, so that wouldn't be an big issue. Those programs that actually burn CPU (video de/encoding, Photoshop/Gimp filters, etc.) on the other side could for most part easily be parallelized. The issue of course is that basically all program languages we have in use make parallelization hard, so nobody is doing it unless they specifically optimize their stuff for multi-core, which in turn means that most programs in actual use today can't take advantage of multicore.

Re:Are most programmes multi-processor?

[grumbel]

There's a lot going on in a modern system:
$ ps -ef | wc -l
146

Those are just idling process, actual CPU activity is extremely low. It doesn't really make a difference if you have 1 idle threads or 100, you don't need CPU power for those either way and multicore doesn't make idling faster. The processes that actually use CPU power on the other side are seldom broken up into multiple threads, so you quite frequently see your system at 50% usage on a multicore, with the second processor idling along.

Re:Are most programmes multi-processor?

[Scarumanga]

I have never had that happen on my 360 (i cant speak for the ps3) but i install the game to the hard drive. The only game i have ever had that happen on was Saints row and i got took it back to the store after 20 min.

Re:Are most programmes multi-processor?

[tyrione]

OpenCL is not GPU dependent. It's agnostic. OpenCL using Multi-cores and GPGPU cores from Nvidia and Intel to PowerVR and more will be in extensive use this year, not just in Engineering fields.

Re:Are most programmes multi-processor?

[grumbel]

Why should it matter if browsers are multi-threaded?

With Flash, Javascript and dynamic AJAX webpages its easy to have the browser actually burn some serious CPU. Even worse is that it actually can get quite unresponsive while doing so. Having a proper multithreaded browser would mean that plugins and tabs would all run in separate threads and the whole thing could become a lot more responsive.

Re:Are most programmes multi-processor?

[tyrione]

I'll take the 19 year olds, thanks! : ) They can quickly learn to do all the MILF tricks and teaching is fun.

Plus the 19 year olds usually have more energy.

Ever slept with a woman reaching her prime sexually? They aren't 19 years of age.

Re:Are most programmes multi-processor?

[nabsltd]

So...you had to wait quite long anyway. Long enough that you decided to do something else in the meantime, because there's no way you would simply wait for the result

He didn't have to wait, or leave. He could have continued to use the computer and barely noticed that 5 cores were being used.

When I was stress-testing my i7 machine, I used Prime95 (as many people do). Prime95 has a poor interface for this sort of testing in that clicking the close box doesn't stop the app, but rather minimizes it to the tray. This is good for the "real" use of Prime95 (finding large primes), but not so much when you are wanting to check out the way the system responds with and without load.

Anyway, I accidentally left Prime95 running 4 tasks all day while I did other things, and only finally noticed the icon in the tray about 5 hours later. Sometimes, you really do want to run so many programs at the same time that you need 4 cores plus hyperthreading.

Re:Are most programmes multi-processor?

[seventyfive75]

LMAO. My videocard alone has more ram than my xbox 360 and PS3 combined. Maybe we will start comparing F-22's and messerschmitts soon. Perhaps we can look at how the Ferrari California stacks up to the new Honda Fit.

Re:Are most programmes multi-processor?

[TheSunborn]

ps -ef | wc -l
163

ps -ef r | wc -l
3

And of the 3 lines, the first is the headers that ps output, the second is ps itself, so really having more then 1 core does not currently help me.

Re:Are most programmes multi-processor?

[evanbd]

You clearly haven't looked at CPU usage graphs by core when using Firefox. On a complex page, one core will stay at 100% for several seconds, while the other sits underused (10-30%, mostly depending on other stuff, it seems). Those hundreds of other processes aren't doing much.

Re:Are most programmes multi-processor?

[TheSunborn]

Oh crap, r for ps is running, not ready+running. To get usefull stats you need to print process stats and then filter on ready(Ready=processes ready to run, but with no core to run on).

But the man page for ps is rather confusing(Or rather ps itself is, some thing require -, others don't) so I just give up finding the exact syntax.

Re:Are most programmes multi-processor?

[toastar]

Multitasking is usually disk-bound, not CPU-bound. Disks don't handle separate but concurrent loads gracefully. This might change with SSDs. Until then, if you don't have enough CPU load with one application, you're unlikely to saturate the CPU with several applications.

Not when your workstation has >20GB or ram

Re:Are most programmes multi-processor?

[petermgreen]

According to toms hardware a core 2 duo at 3GHz will give over 45FPS on the settings they used. If we assume framerates are linearly proportional to the processor speed then a 1.6GHz C2D would give 24FPS

Given that toms hardware didn't use minimum settings the minimum processor requirements seem to be a reasonable reflection of what is barely adequate to play the game.

Re:Are most programmes multi-processor?

[hlee]

Yes, especially for games, and I don't mean the obvious 3D computations, which in any case is most likely handled by the GPU anyway, but rather parallelism in the game workflow.

Read an article last year on Communications of the ACM (3/2009) entitled "Better Scripts, Better Games" that talked about common design patterns in programming games. I'll focus on the state-effect pattern, which breaks each simulation loop into two phases: effect and update. Effect phase is where each game object is subject to an action and determines its own effect of that action. In the update phase, all the effects are combined, and updates the game state for the next iteration of the loop. Since calculations in the effects do not influence each other, they can be parallelized. The update phase can also be parallelized since it consists only of the aggregation of effects and updates to state variables. Idea is similar to Google's MapReduce.

Re:Are most programmes multi-processor?

[petermgreen]

Ah, so you don't realize that the cheapest Celeron nowadays is a dualcore 2.5 GHz
No it isn't at least it isn't if you shop at newegg. A quick look at newegg shows the cheapest celeron as being a 1.8GHz single core though it is a core 2 based device.

If we ignore the improvements from core 2 to core i7 and assume the time taken to perform his workload is inversely proportional to CPU speed and core count then the celeron would have taken 9 hours 36 minuites, depending on his commute time and working day length it may or may not have finished by the time he got home from work.

Re:Are most programmes multi-processor?

[StikyPad]

GTA 4 requires two extra cores the same way my kid requires two extra wheels on his bicycle.

Re:Are most programmes multi-processor?

[petermgreen]

These multicore processors won't be running web browsers, unless it's because they're hosting a dozens of virtualized desktops that users are working on over the network.
While it looks like there will be a xeon version of this chip for two-socket server/workstation systems it looks like there will also be a single-socket version for the high-end desktop market.

The blunt fact is that neither intel or amd can make the individual cores that much faster so they have pushed multicore into the mainstream. Other than netbooks/nettops I haven't see a new machine with a single core processor in ages and even quad-core processors are seen in fairly low end systems these days. While i'm sure initially these 6-core CPUs will be very expensive I'd expect to see them in midrange desktops in the next few years.

Re:Are most programmes multi-processor?

[DragonWriter]

Can most programmes really be written to take advantage of so many cores?

You don't need most programs to benefit from more cores; you just need a few commonly used ones to.

Or for people to run different programs simultaneously, since separate programs that run at the same time naturally benefit from multiple CPUs without being written especially to do so.

Re:Are most programmes multi-processor?

[default luser]

Can most programmes really be written to take advantage of so many cores? I am not sure I want to have a 6-core processor, of which 5 spend most of the time idling as I am only running a single-core-aware programme. OK, one more core can be used by the OS to make everything snappy, but the question stands.

Intel agrees with you, and that is why their first 32nm release was their mainstream processor codenamed Clarksdale:

2 processing cores
4 threads
4MB cache to make multitasking smooth.

It has all the fancy features of it's bigger bother (including encryption acceleration), but with 1/3 the real cores. You get much better IPC than the Core 2 Duo line, and a great multitasking experience.

The 6-core behemoth to be released later this year is just to make server and workstation owners happy.

Re:Are most programmes multi-processor?

[complete loony]

and still not bring usage over 50% more than occasionally

Let me guess, you have a dual core or hyperthreading processor. Guess what, those programs are stressing a single core. You will only see the CPU usage go over 50% if there are 2 threads running.

Re:Are most programmes multi-processor?

[nanostuff]

http://www.tomshardware.com/charts/2009-desktop-cpu-charts-update-1/GTA-IV-1.0.3,1402.html

One frame more for a quad core processor over the closest dual core.

Not exactly a "requirement"

Re:Are most programmes multi-processor?

[sznupi]

For multitasking you also don't need Core i7, you know...

Prime95 is trivial to not notice even on old, singlecore CPUs. There's hardly any HDD i/o, which has the biggest potential of slowing you down if there's something happening in the background. There's basically just computation.

Re:Are most programmes multi-processor?

[sznupi]

It is in this almost 3rd world country where I am right now (older Celerons are only marginally cheaper and simply not available anymore); I would say that's more representative regarding prices at which those CPUs are bought.

And I assume your estimate was for 1.8 GHz? (what's the point in estimating minutes?) How conveniently you assumed that the performance scales linearly...

Re:Are most programmes multi-processor?

[TheRaven64]

I do have a dual-core processor, but I have per-CPU usage graphs. Right now, they are both sitting at around 10%. In normal use, they stay there. Playing back HD video makes one of them go closer to 50%, playing Flash games makes them go to a lot higher. The only time either gets to 100% is when I am compiling a big project (and parallel make lets me run one compiler on each core trivially) or encoding video (and video encoding is another task that is trivial to parallelise, so uses both cores).

Re:Are most programmes multi-processor?

[war4peace]

Of course, it depends a lot on which settings you use in the game itself. Having a 24" full-HD monitor, I expect a game from late 2008 to run at a smooth framerate at high resolutions. It doesn't. First I thought the bottleneck was my graphic card, so I upgraded from an 8800 GTS to a GTX260. To my dismay, the performance increase was not as high as I expected, so I investigated further. Turns out my CPU limitations choked the framerate, so I turned to a friend and begged for a quad-core loan. At 1680x1050, with the same game settings, the performance jumped from 25-35 FPS to 36-48 FPS.
Now I don't say Tom's Hardware is wrong, all I say is it depends a lot of the in-game settings. There are things in the game that are very CPU-intensive (physics, number of cars, etc). Move the sliders down (in versions 1.0.3.0 or 1.0.4.0) for better performance but less gaming immersion/experience.
Furthermore, I never count avaregare framerate, which might be good. But those performance drops are really pissing me off. If during a certain mission, when things get tense, you are forced to play at 15 FPS, then that's annoying as hell. Who cares that casual cruisin' works smooth? Sure, turn all sliders down and you will have smooth FPS. But stuff will be rendered 50 feet in front of your car and man, that's frustrating as hell.
All I'm saying, in order for my gaming experience to be pleasant, I need a quad-core CPU. Benchmarks... only good for e-peen comparison :)

Re:Are most programmes multi-processor?

[SacredByte]

Your argument seems to be "GTA 4 runs like shit on my dual core CPU, and they say that you really should run it with on a quad, so therefore it will run even better on a 6-core CPU."

No offense, but that's a load of crap. GTA 4 runs like shit because it is for all intents and purposes a completely non-optimized console port. I've seen PLENTY of people with computers hands down above and beyond mine (tri/quad SLI with watercooled and heavily overclocked GTX285/295's (respectively), along with watercooled and heavily overclocked extreme-edition i7's) have difficulty running GTA 4 well.

I have a pretty good computer for these days (Core i7 920 CPU on an EVGA E759 X58 motherboard with an EVGA GTX 275 SC {at least that's how it came flashed}), and I have been able to graphically max out EVERY console port I have ever run at a minimum resolution of 1600 x 1200. -- titles like Call of Duty: World at War (I played it completely maxed out at 1600 x 1200 and I think my FPS was capped at 60 (I think I had it set to not do FPS above the monitor's refresh rate), and Mass Effect 2 (which I run completely maxed out at 1920 x 1440 and it has never gone below 60 FPS in any of my 2-3 playthroughs totaling at least three days ingame so far).

Basically, what I'm saying is that if your argument for more processor cores is based on the fact that badly written/optimized code runs better with more processor power, you need to find a better argument.

Re:Are most programmes multi-processor?

[war4peace]

You are right. In theory. Code should be better optimized, ported games should run faster on PCs und so Weiter. But that's not the point. Knowing where the problem is won't help you run the game/application faster. A better CPU would, though.
You're approaching the discussion from a power user's point of view, which is okay. I am doing the same from a gamer's point of view. I like a game, I bought the PC version, my game experience is not what I would have liked. Ranting and whining about how badly the code is written or how badly it has been ported to PC won't make it run faster. Period. A multicore CPU would.
Now there are lots of ported games out there who require higher than expected PC resources to run better. You could choose to not buy them and not play them, but as far as I'm concerned, this approach won't make me feel better. A multicore CPu would, though :)

Grand Central Dispatch

[TheStonepedo]

Perhaps a jump in number of cores will convince people outside the Apple and FreeBSD camps to port Grand Central Dispatch.
Letting the kernel team handle the hairier parts of multi-threaded design should make it easy for barely-optimized software to use powerful hardware.
Could its Apache license work with the #1 OS family?

Re:Grand Central Dispatch

[Make]

GCD is a nice library with an elegant API. Unfortunately it is not portable, because it requires a compiler which supports a non-standard extension to the C language.

Re:Grand Central Dispatch

[TheRaven64]

Porting libdispatch requires a generic event delivery framework, where the userspace process can wait for a variety of different types of event (signals, I/O, timers). On Darwin, Apple used the kqueue() mechanism that was ported from FreeBSD, so it's quite easy to port the code to FreeBSD (just #ifdef the bits that deal with Mach messages appearing on the queue). Kqueue is also ported to NetBSD and OpenBSD, so porting it to these systems will be easy too.

Solaris and Windows both have completion ports, which provide the same functionality but with different interfaces. Porting to Solaris would require replacing the kqueue stuff with completion port stuff. Porting to Windows would ideally also require replacing the pthread stuff with win32 thread calls. Even Symbian has a nice event delivery framework that could be used, although I'm not sure what the pthread implementation is like in the Symbian POSIX layer.

Linux is the odd system out. All different types of kernel events are delivered to userspace via different mechanisms, so it's really hairy trying to block waiting until the next kernel event. This also makes it harder to write low-power Linux apps, because your app can't spend so long sleeping and so the kernel can't spend so much time with the CPU in standby mode.

If you don't need the event input stuff (which, to be honest, you do; it's really nice), you can use toydispatch, which is a reimplementation that I wrote of the core workqueue model using just portable pthread stuff.

It also adds some pthread extensions for determining the optimal number of threads per workqueue (or workqueues per thread, depending on the number of cores and the load), but these are not required. The FreeBSD 8.0 port doesn't have them; they were added with FreeBSD 8.1.

Re:Grand Central Dispatch

[ronocdh]

If you don't need the event input stuff (which, to be honest, you do; it's really nice), you can use toydispatch, which is a reimplementation that I wrote of the core workqueue model using just portable pthread stuff.

Can you provide a link for this? A Google search for "linux toydispatch" yields 3 hits, one of which is your post above:

Results 1 - 3 of 3 for linux toydispatch. (0.31 seconds)

Re:Grand Central Dispatch

[chaim79]

Actually that isn't the case, I've been keeping an eye on the porting of GCD to other OSs and there are build options for with and without blocks (the non-standard C extension).

As of right now I think the status is that FreeBSD (and other BSDs) can compile GCD with or without block support, Solaris is 90% there (again with and without blocks), and Linux is about 70% there (can compile and parts work, but not all of it).

Re:Grand Central Dispatch

[jo_ham]

Funny enough, it was around the time that really, really multi-cored machines started showing up, so they wrote an open source library to make it easier for people and said "have at it".

If "marketing magic" is writing a webpage on their main site with some nice graphics and a description of how it works, then no wonder Linux is struggling for Desktop acceptance. I would just cal that "making a nice webpage". It's not sorcery. "Marketing? Do we need to compile that?" /tongue in cheek.

Re:Grand Central Dispatch

[mandolin]

Porting libdispatch requires a generic event delivery framework, where the userspace process can wait for a variety of different types of event (signals, I/O, timers). ... Linux is the odd system out. All different types of kernel events are delivered to userspace via different mechanisms, so it's really hairy trying to block waiting until the next kernel event.

I don't understand this. It's true Linux does not have kqueue. (as I recall Linus thought it was "ugly" ... whatever) But in theory (because I haven't actually used them), to achieve the same effect under Linux, you would use timerfd() + signalfd() + (your normal i/o fds like sockets etc.) ... and then epoll_wait()/poll()/select() on all of the fds you were interested in. In this way, one thread could wait for multiple different types of events, including timers and signals.

Would you please point out the flaws w/the above that make it impossible or impractical to achieve the functionality needed by Grand Dispatch? I would be enlightened -- thanks in advance.

Re:Grand Central Dispatch

[TheRaven64]

Subversion repository. Note that it's designed specifically to do stuff in the background for libobjc2. It only implements a tiny subset of the libdispatch functionality, and not as efficiently (one thread per workqueue, for example). It's not intended to replace libdispatch, just to let me use some of the libdispatch APIs in code that has to be portable. The 'toy' in the name is not self-deprecation, it's an accurate assessment.

Oh, and you get better results if you search for 'toydispatch' not 'linux toydispatch' (it's nothing to do with Linux, although it should run there).

Re:Grand Central Dispatch

[tyrione]

GCD is a nice library with an elegant API. Unfortunately it is not portable, because it requires a compiler which supports a non-standard extension to the C language.

LLVM

Re:Grand Central Dispatch

[Hortensia Patel]

Interesting post. I've ignored GCD completely, being under the impression that it was so entangled with Apple's GCC extensions as to be effectively nonportable. Am I utterly wrong? Would it be feasible to use GCD with, f'rinstance, C++0x lambdas instead of Apple's blocks?

Re:Grand Central Dispatch

[TheRaven64]

GCD can be used with functions and void* pointers rather than blocks. All of the dispatch functions have a _f version that takes a function pointer and a void*, rather than a block. Blocks are not just an Apple thing. We were shipping an (MIT licensed) implementation of the runtime with Etoile before Apple shipped theirs and I use them regularly on a variety of platforms. You could probably use them with C++ lambdas if you write a trampoline that dereferences the pointer and then frees them.

240mm square?

[pointbeing]

1.17 billion transistors into a die that's approximately 240mm sq

That's a big chip.

Re:240mm square?

[goldaryn]

1.17 billion transistors into a die that's approximately 240mm sq

That's a big chip.

240 mm sq, that's 15.49mm x 15.49mm

Re:240mm square?

[pointbeing]

240 mm sq, that's 15.49mm x 15.49mm

But not nearly as amusing. ;-)

Re:240mm square?

[IBBoard]

Isn't it 240mm sq = 240mm x 240mm (as in (240mm) squared) and 240 sq mm is 240 x 1mm x 1mm (as in 240 x (square mms))? It's always an awkward one to represent and be clear on.

Re:240mm square?

[realityimpaired]

that'd be how I read it, and I'm pretty sure how the GGP read it, too... 240 mm sq = 240mm squared, which is different from 240 square millimeters.

Re:240mm square?

[vadim_t]

No.

1sq mm is a square with 1mm sides
240sq mm is 240 of them.

The side is sqrt 240 for a square shape.

Re:240mm square?

[mario_grgic]

That's because the actual unit is mm^2, which people pronounce millimeter squared, instead of square millimeter.

So, 240 mm^2 is 15.49 mm * 15.49 mm

Re:240mm square?

[IBBoard]

But the original was "mm sqr" (millimetres square) rather than "mm sqrd" (millimetres squared).

I still thought that "mm^2" was "square millimetres", though, as in "tin of paint covers nine square metres" = "covers nine x 1m x 1m" = "covers 9m^2".

All so confusing, and so very liable to getting things completely wrong by huge orders of magnitude!

Re:240mm square?

[jc79]

No, sq mm means nothing. The proper way to write it is mm.

Shirley the proper way to write sq mm is mm²

Well, it would be if /. supported that particular tag

Re:240mm square?

[Buelldozer]

What? Why?

In every other case I can think of where you need to denote area it's opposite.

240 mm sq is 240 millimeters each side for a total area of 240mm x 2, or 57,600 square millimeters.

If you want to refer directly to the area then the unit descriptor comes AFTER the square designator, like this: 240 square millimeters.

You wouldn't write " 42 feet square" when you meant "42 square feet" would you?

Did I miss something somewhere?

Re:240mm square?

[rrohbeck]

And how much is that in football fields?

Re:240mm square?

[radish]

The correct notation for an area of 240 square millimeters is very hard to type (it's 240mm^2), my guess is the OP just turned "^2" into "sq" (for "squared") leading to the confusion.

Re:240mm square?

[tyrione]

1.17 billion transistors into a die that's approximately 240mm sq

That's a big chip.

That's small. What's big is the pin housing that surrounds it.

Re:240mm square?

[canajin56]

Yeah, you missed something somewhere. You missed order of operations. What's the area of a circle? "Pi r squared", squared only squares r, not the entire thing. 240 square mm is 240 mm^2. Acceleration due to gravity is "9.8 meters per second squared", which means, "9.8 m/s/s", not "96.04 m*m/s/s". Nobody ever says it the way you seem to think is the only correct way, "9.8 meters per square second"

Re:240mm square?

[Buelldozer]

I see what you're saying but have you ever bought flooring?

Walk into the showroom and tell them "I want 144 feet square of carpet." The first thing you're going to get is a very odd look.

It's far more probable that you'll walk in and say "I want 144 square feet of carpet." and then you'll give them the precise x & y dimensions.

Would you say that your house is "2,200 feet square" or "2,200 square feet"?

Again, I'm guessing the latter.

I cannot think of a single instance where I've seen "sq" used behind the unit identifier when discussing AREA.

What I believe YOU are missing is that this isn't an equation. Squared in this instance is not an operand but a statement of measurement, area in this case.

Another mistake I believe you making is the assumption of a rectangular shape with four equidistant sides!

After all, I can reach an "area" of 240mm quite a few different ways, even confined to a rectangular shape. You ASSUME that it's a square with four equi-distance sides but it could also be a rectangle with only two equi-distant sides!

No, I am pretty sure I have it right. This is a statement of area and not an equation. Therefore order of operand does not apply. Additionally you should not assume a shape based on a statement of area.

Re:240mm square?

[nasch]

The correct notation for an area of 240 square millimeters is very hard to type (it's 240mm^2),

Yeah, holding the shift key while you press 6 is just brutal. I think you're right about what happened.

Re:240mm square?

[complete loony]

More likely it's exactly 15mm x 16mm.

They are supposed to ship in March

[alen]

Talked to our HP rep a few weeks ago about them. As soon as HP ships proliant servers with the new CPU's, we're going to buy 4 of them. Just haven't decided if we're going with 36GB RAM or 72GB RAM. 72GB RAM is only $2000 more than 36GB RAM these days.

Re:They are supposed to ship in March

[petermgreen]

I'm guessing that 36GB and 72GB refer to three dimms per channel times 6 channels (three per processor) and 2GB or 4GB modules. IIRC with DDR3 if you put three dimms on a channel you are limited to DDR3-800 speeds.

the best fucking CPU that ever existed

[TubeSteak]

Just so you know, I made this joke almost two years ago:
http://hardware.slashdot.org/comments.pl?sid=465898&cid=22548916

They could have gone to 3 cores, like the competition. That seems like the logical thing to do, but they said "Fuck it, we're going to six". What part of this don't you understand? If two cores is good, and four cores is better, obviously six cores would make them the best fucking CPU that ever existed.

http://www.theonion.com/content/node/33930 [theonion.com]
/I'm just waiting for the day Intel says "this one goes to 11"

It's the CPU joke that will never die.

Re:first

[Anonymous Coward]

first

looks like you need more cores

Intel announces 6 cores, 6 months after AMD.. yawn

[Gr8Apes]

So I skimmed TFA (gasp!) and it appears that Intel is finally following AMDs lead by keeping thermal envelopes constant.

I note that this is still a effectively 2 CPUs with 3 cores each, but that's better than legacy Intel approaches, which would have been 3 sets of dual cores.

It will be interesting to see how independent performance benchmarks play out between the new processors that are coming out.

Obligatory

[Mattskimo]

blah blah Beowulf blah blah

Transistor count

[gr8_phk]

1.17 Billion transistors. Anyone remember the 6502, the 6800, and then the 68000? 68K transistors was a LOT in 1980 and made for a fantastic 32bit architecture. Now we're at 17000 times that count. Sometime you just have to stop for a moment and think just about the numbers.

Re:Transistor count

[Maximum Prophet]

... 68K transistors was a LOT in 1980 and made for a fantastic 32bit architecture....

I'm guessing you're still caffeine deprived, and meant 8 bit architecture.

Anyway, what I want to know is where are the 3.2GHz 6502 and Z80's? You'd think making an existing architecture run like a bat out of hell would be far easier than a new Pentium chip. With less than 1.17 billion transistors, you could put an entire C64 or Apple II on one chip and run all the old software.

Re:Transistor count

[TheThiefMaster]

The Motorola 68000 was/is a 32-bit architecture.

Re:Transistor count

[wtfbill]

No, I bet his caffeine content is fine. The 68K transistors would refer to the 68000 procs from Motorola which were 16- or 32-bit depending on configuration. Some of them could be switched at boot time by holding one of the pins high or low (I forget which...where are those old data sheets I have on those?) Of course the 65xx series and the 6800 series were 8-bit, however, they didn't have close to 68K transistors. But GP is right on, 68K transistors for a 32-bit architecture.

Re:Transistor count

[sznupi]

And yet, latest ARM cores are much closer to that 68k transistors from 1980, while not being nearly that far behind i7 in performance as the relation between numbers of transistors would suggest.

Perhaps ARM found the sweet spot.

Re:Transistor count

[rrohbeck]

I have a feeling that Loderunner would be difficult to play at 3.2 GHz.

Re:Transistor count

[ChrisMaple]

There are some sped-up versions, although nowhere near 3.2 GHz. (For instance, a 25 MHz Z80). There isn't any point to having a 3.2 GHz Z80 if the Z80 architechure is preserved. Z80 instructions required 4 to 17 cycles to execute (IIRC) and multiple memory cycles whenever external data was accessed.

There are many, many things that could be done to improve the Z80 with modern technology and most would break compatibility in some manner. For instance, a great speedup could be made by putting the entire 64k memory space on die, but that would break anything that required memory-mapped external access. Pipelining and wider memory could make all instuctions single cycle, provided the memory access was multiport. A multiply instruction would be nice.

Really, there are better uses of technolgy available than investing $1M or so on a 3.2 GHz Z80.

Nonetheless, I understand the feeling. I'd like to see one, too.

You mention "an entire C64 or Apple II on one chip". The SOC philosophy is a good one, and for a long time chips have been available with many of the I/O functions on-chip. However, I doubt that there's anything that completely duplicates a C64 or Apple II. What could be done is to implement an 8 bit computer, inclding memory, on an FPGA. Then you'd just have to add power supply, level-shifters for RS-232 &c, power supply, disk drive, case... Fun games. Really, an emulator running on a modern computer would be less work.

Re:Transistor count

[Foobar_]

However, I doubt that there's anything that completely duplicates a C64 or Apple II.

That should be "anyone"

http://en.wikipedia.org/wiki/Jeri_Ellsworth

Re:DRM Support

[FlyingBishop]

This is a server processor. They did it for advanced encryption. The only way this would make more powerful DRM is if there were some sort of key embedded in the CPU (and this is not that.)

This is for encryption, which unlike DRM is actually more than security theatre (when used properly.)

Re:DRM Support

[sakdoctor]

What?

AES acceleration will be useful for VPNs, serving SSL websites, VoIP, full disk encryption ... and so on.

To much cores, to little use...

[null8]

Instead of churning out cores they schould tweak the x86 isa to use multiple cores efficently. 1/2-word Atomic compare and swap is not enough, you cannot make atomic lockless doubly linked lists with that. No wonder something as interesting as http://valerieaurora.org/synthesis/SynthesisOS/ is not possible on x86 without major hacks.

Re:To much cores, to little use...

[TheThiefMaster]

1/2-word?
I'm pretty sure that there are instructions for atomic compare and swap of pointer-sized values, at least.

Re:To much cores, to little use...

[master_p]

Isn't the xchg instruction atomic for all sizes (8/16/32/64 bits)?

Re:To much cores, to little use...

[null8]

Sorry I didn't express that more clearly, I meant word as in "machine register size", not as it was used by intel. You can use compare and swap on 32/64 bit values on x86 or on 64/128 bit values on amd64, but still no double word compare and swap where both words are in different memory locations.

Re:To much cores, to little use...

[noidentity]

I've always wondered why x86 hasn't added support for load-link/store-conditional synchronization instructions. Basically these involves an extra hidden address register. Load-link acts like a normal load, except it also sets this hidden register to the address just loaded from. A store-conditional only stores if this hidden register matches the address being stored at. Afterwards, this hidden register is cleared and the condition codes reflect whether the store occurred. A load-link by another core/processor would broadcast over the bus so other processors could update/clear their internal register. I like this scheme because it's extremely simple, and doesn't require any hardware-interlocked instructions as hardware compare-and-swap does.

Re:To much cores, to little use...

[null8]

Sure, but there is no architecture which provides strong load-link/store conditional, so you cannot operate on 2 memory locations because the second will reset the first, so, it's like compare-and swap, only with more limitations.

on-board AES?

[MobyDisk]

Why put AES on-board? I thought AES was relatively fast as encryption algorithms go. Plus, it is inevitable that AES will be replaced at some point, so why include something so specific in a chip now? It will suck to have to implement that in the processor in 20 years when nobody uses AES any longer. This is the whole point of a processor - include generic instructions that are useful for implementing any algorithm.

Re:on-board AES?

[0123456]

Why put AES on-board?

They're not: they're putting extra instructions on-board which help implement AES more efficiently. They may also allow you to implement other algorithms more efficiently, though I haven't looked at them in enough detail to be sure.

I thought AES was relatively fast as encryption algorithms go.

That still doesn't make it fast at an absolute level. Particularly when you're doing full-disk encryption with user account encryption on top and IPSEC on all your network connections.

Re:on-board AES?

[Galactic Dominator]

Yeah AES is relatively fast with the keyword being relatively. Those of us who like to use disk encryption applaud this move since it would great reduce the need for seperate and expensive crypto hardware.

Re:on-board AES?

[MobyDisk]

They're not: they're putting extra instructions on-board which help implement AES more efficiently. They may also allow you to implement other algorithms more efficiently...

Oh good. That makes more sense.

Re:on-board AES?

[Bart Coppens]

Implementing AES directly in the chip makes implementing AES in userspace also less vulnerable to cache-based side-channel attacks against AES. Also note that there has been work to use these AES instructions in implementing (some of the) candidates of SHA-3 (http://crypto.rd.francetelecom.com/sha3/AES/)

Re:on-board AES?

[wirelessbuzzers]

Why put AES on-board?

They're not: they're putting extra instructions on-board which help implement AES more efficiently. They may also allow you to implement other algorithms more efficiently, though I haven't looked at them in enough detail to be sure.

The instructions perform a single round of AES (which has 10-14 rounds depending on key size), either encrypting or decrypting. Certain other algorithms such as Lex, Camellia, Fugue and Grostl use AES S-boxes in their core, and can probably benefit from these instructions. However, they will not achieve nearly so much a speedup as AES.

The AES instructions themselves will approximately double the speed of sequential AES computations. This is very unimpressive; VIA's AES instructions are much faster. They will also make it resistant to cache-timing attacks without losing speed, which is unimpressive because you can already do this on Penryn and Nehalem. The low speed results from the AES instructions having latency 6; if you can use a parallel mode (GCM, OCB, PMAC, or CBC-decrypt, for example) then the performance should be 10-12x the fastest current libraries. Hopefully, this will cause people to stop using CBC mode, but perhaps I'm too optimistic.

Intel also added an instruction called PCLMULQDQ which does polynomial multiplication over F_2. If it's fast (I can't find timing numbers, but hopefully it's something like latency 2 and throughput 1) then it will be very useful for cryptography in general, speeding up certain operations by an order of magnitude or more. This is more exciting to me than the AES stuff, because it might enable faster, simpler elliptic-curve crypto and similarly simpler message authentication codes. Unfortunately, these operations are still slow on other processors, so cryptographers will be hesitant to use them until similar instructions become standard. If the guy you're communicating with has to do 10x the work so that you can do half the work... well, I guess it's still a win if you're the server.

I thought AES was relatively fast as encryption algorithms go.

That still doesn't make it fast at an absolute level. Particularly when you're doing full-disk encryption with user account encryption on top and IPSEC on all your network connections.

AES is fast for a block cipher, but modern stream ciphers such as Salsa20/12, Rabbit, HC and SOSEMANUK are about 3-4x faster. (In other words, they are still faster than AES in a sequential mode on Westmere.) AES is still competitive, though, if you can use OCB mode to encrypt and integrity-protect the data at the same time.

The fastest previous Intel processor with cutting-edge libraries in the most favorable mode could probably encrypt or decrypt 500MB/s/core at 3-3.5GHz. This is fast enough for most purposes, but in real life with ordinary libraries you'd probably get a third of that. So this will significantly improve disk and network encryption if they use a favorable cipher mode.

Cred: I am a cryptographer, and I wrote what is currently the fastest sequential AES library for Penryn and Nehalem processors. But the calculations above are back-of-the-envelope, so don't depend on them.

Re:on-board AES?

[bored]

The fastest previous Intel processor with cutting-edge libraries in the most favorable mode could probably encrypt or decrypt 500MB/s/core at 3-3.5GHz. This is fast enough for most purposes, but in real life with ordinary libraries you'd probably get a third of that

500MB/sec on AES, or something else? If AES, is your library available for use/purchase? Our application needs AES as fast as it can go, and that's pretty darn fast without acceleration hardware.

Re:on-board AES?

[wirelessbuzzers]

The fastest code that I know of for AES in CTR mode is Kasper-Schwabe. It does 8 128-bit encryptions at a time, so it also should be suitable for, say, PMAC if you doctor it. I believe that it does not handle decryption (outside of CTR mode where it's the same as encryption) or other key sizes. Modes other than CTR lose some optimization, and should be ~20% slower. It should be available on Kasper's homepage. It requires SSSE3 and reportedly achieves 6.9 cycles/byte on Nehalem for CTR mode.

My code is available here. On Nehalem, it achieves ~9.4 cycles encrypting, ~11.1 cycles decrypting in essentially any mode. It is suitable for encryption or decryption, and supports all three key sizes (longer keys are slower, of course). A newer (unreleased, experimental) version makes slight performance improvements (maybe down to 9.1 cycles encrypting on Nehalem) and implements an optimization for CTR mode that brings it down to ~7.5 cycles. Email me (mhamburg AT cs DOT stanford DOT edu) if you want to try the experimental version. However, my code fundamentally requires SSSE3, and it performs quite poorly on Conroe.

Also, Dan Bernstein (homepage) has somewhere a fast conventional (not timing-attack resistant, but not requiring any sort of SSE) implementation of AES for several processors, and I've heard Crypto++ is pretty fast too.

I believe that all of the above libraries are public-domain and patent-free.

Out of curiosity, what's your application? Can you just get a VIA Nano or Intel Westmere core and run on that?

Re:on-board AES?

[wirelessbuzzers]

Hopefully, this will cause people to stop using CBC mode, but perhaps I'm too optimistic.

I don't quite get the point. Are there flaws in CBC mode or do you think it's just better to use i.e. OCB or GCM mode because many developers often forget to add proper MACs to their protocols?

I overstated the case. CBC mode isn't terrible, it's just not the best mode out there.

The main problem with CBC mode is that it isn't parallel. This means that most of the new, faster, timing-attack-resistant libraries can't handle it efficiently (the one I wrote is an exception, but it only works on recent Intel procs and it's ~10% slower than Kasper-Schwabe for bulk encryption). It also means that CBC is a factor of 3 slower on Westmere, maybe 1.5 on VIA chips and 1.5 on the PowerPC G4. Due to optimizations, CBC mode is an additional 15-20% slower than CTR mode in most software libraries. It also has a larger attack surface than CTR mode, and unlike CTR mode it requires decryption.

OCB and GCM modes are really nice because of the integrated MAC. Too bad OCB is patented... if you fit the constraints, there really isn't a better mode for AES.

Re:on-board AES?

[wirelessbuzzers]

Intel also added an instruction called PCLMULQDQ which does polynomial multiplication over F_2. If it's fast (I can't find timing numbers, but hopefully it's something like latency 2 and throughput 1) then it will be very useful for cryptography in general...

Latency 15 cycles, throughput 10 cycles. That's a shame. Also, AESENC has thoughput 2 cycles/round, only half as fast as I expected.

Codenamed codename?

[CoffeeDregs]

>Westmere 6C (codename Gulftown)

    Really? I fricking hate codenamed codenames...

Chicken or Egg

[wwwillem]

In the server space we've gone through the same thing. Sun introduced the T1 with 8 cores and 32 threads (now 64 on T2). Lots of software wasn't suitable for this type of horizontal scaling. But over a period of five years, that changed dramatically.

On the desktop, you can expect the same. For now, not many desktop apps will take advantage of the additional cores. But if Intel would have stuck with 1-2 cores, no software will be written to take advantage of multiple cores.

Chicken or egg ....

Re:DRM Support

[Ant P.]

Which is obviously the reason AMD's and VIA's CPUs had the same thing over half a decade ago, right? DRM sure was prolific back then.

Re:Intel announces 6 cores, 6 months after AMD.. y

[TheTyrannyOfForcedRe]

I note that this is still a effectively 2 CPUs with 3 cores each

You need to skim a little slower. It is a single die, native six core chip. From the article:

Westmere 6C (codename Gulftown) is a native six-core chip

Yes, some functional blocks are "split in two" and assigned to three cores. Those blocks are split up because it makes engineering sense to do so, not because the designers were lazy or "slapped together" two non-six core designs.

If you look at a GPU with 10's or 100's of cores you will see that some functions are "split up" with each piece being assigned to (and located close close by!) some subset of cores. That does not make them "effectively X number of CPU's."

Re:DRM Support

[RawsonDR]

VPNs, serving SSL websites, VoIP, full disk encryption ... and so on

Well, yeah, but I think the GP's point is that while some of those things have been in mainstream use for years, the use case for DRM is what made it important enough to be implemented.

Re:Intel announces 6 cores, 6 months after AMD.. y

[TheThiefMaster]

I'm pretty sure it is one die, with communication possible between any cores. It just looks like 2x3 due to the way it is laid out.

42 ...

[skywatcher2501]

... is the number!

Re:DRM Support

[Tim C]

This is a server processor.

Well, the summary says "desktop CPU", while the URL of TFA says "mobile"...

Who cares...

[the linux geek]

This comes the same month as the release of 16-core processors by IBM and Oracle, and a 12-core from AMD. This isn't that impressive.

Re:Who cares...

[Microlith]

Sure, but neither the Oracle or IBM chips will be available for less than several grand, and never in consumer level equipment (I can't exactly order one off Newegg.) And there's no telling how long it will be until the AMD chip trickles down from Opteron class to Phenom class, while it will probably be short order for the Core i9 to appear in stores.

I suspect that AMD will drop the 6-core version as an X6 pretty soon, but it will likely be outperformed (possibly significantly) by the Gulftown.

Re:DRM Support

[rliden]

In addition, Intel has built in AES encryption instruction decode support

You can bet the sole reason for including this was to support DRM-protected content.

It may very well be useful for DRM, but I would venture that there are several reasons to include further AES instructions in hardware; one of the chief being full disk encryption performance.

AMD responds...

[DarthVain]

with an AMD X3 Core 2 Duo or AMD X3X2...

Though I wonder why we are going to 6 rather than 8. Core 2 Quad Duo's? Head 'esplodes....

I just can't wait till the Quad Quads... or something spiffy, like Quad Squared. 16 is probably a ways off from the consumer market anyway.

Re:AMD responds...

[Yvan256]

The nice thing about running Doom on a Core 2 Quad is that all weapons do four times as much damage, all the time.

Re:Most programs are MULTITHREADED, thus ready

[metrix007]

Jesus Christ APK, you're useless.

multi cores

[itzdandy]

Having 6+ cores is a complex issue for the desktop. In the 'olden' days it way easy to do the math that per Mhz, the less cores the faster ( a pair of 1Ghz chips is slower than a 2Ghz CPU if all all other things are equal)

But now, each CPU can have a dedicated memory bus. That means that 2 cores can be FASTER than a single core because 2 cores can have twice the memory bandwidth. Its not necessarily about clock cycles but bandwidth.

More cores also can improve the desktop experience. because flash is currently stuck on a single CPU, your system cant get totally rocked by flash. If you have a really intense process, you could set the CPU affinity to 5 of the 6 cores to make sure you had one left over. The process scheduler should try to do this on its own but its not always successfull.

I run some VM hosts on dual quad cores and processor affinity for virtual machines is very important. On one box I have CPUs that are 2 dual cores on a single CPU and then 2 sockets. A VM on CPU1&2 (numerically 1-8, not 0-7) is ideal. A VM on 2&3 is slow and a VM on 4&5 is rediculously slow because you go from a signle chip, to two chips, then to a core on seperate sockets. This is an extreme case but it is true on the desktop as much as it is anywhere else.

Any high transaction count server that runs multiple processes or threads will make very good use of 6 cores as these process tend to not jump between cores ( or shouldnt if the programmers are concerned about performance)

24 isn't enough

[toastar]

24 x86 cores just doesn't compare to 1 Fermi with 512 striped down vector processors

Re:24 isn't enough

[wooferhound]

640 cores should be enough for anybody

Taming CPU Hogs

[nuckfuts]

CPU hogging programs can be a real pain, especially on a multi-user system. I know of businesses that run keyboard polling applications on a Terminal Server, for example. Each instance easily pin a core at 100%.

If you ever run into a problem like this (on Windows), check out ThreadMaster

How do I get one of these in my phone!

[richardkelleher]

Or a game boy or something like that. Of course you would have to wear asbestos gloves to play with the thing... :)

Graphics core

[wirelessbuzzers]

It's a shame they didn't make a mid-level version with no graphics core, a la Core i7 860. As a crypto/security guy, I'd like to try out PCLMULQDQ, the AES instructions and maybe the IOMMU. But if I'm going to get a fancy new computer, I might as well put a decent graphics card in it, at which point their on-die graphics card is simply a waste of space, power, money and latency. And no, I'm not dropping $1k for a 6-core Gulftown.

Re:Most programs are MULTITHREADED, thus ready

[petermgreen]

Well... take a peek @ your Taskmgr.exe (with the processes tab open & the threads column selected + viewable)...

You'll probably see pretty much as I do here, and, as I have for YEARS now no less: That most of the processes running on your system now have 2-N threads running PER PROCESS already
No they have 2-N threads per process, afaict there is no indication in task manager how many of them are running/runable.

Most threads spend most of their time blocked waiting for some kind of event, while occasionally doing some quick action in response and then blocking to wait for an event again. The number of tasks doing serious processing work is usually in the range 0-1 unless you either start multiple intensive tasks at once or you have software that is specifically designed to take advantage of multiple processors/cores.

I'd rather have a better scheduler

[Chemisor]

I'm running 2.6.32 with the fabulous new scheduler and all those "low latency" settings painstakingly compiled in. Nevertheless, I'm currently making some DVD isos and while mkisofs is running, firefox has to think for five or six seconds before loading every page. Somehow I doubt that having two extra cores is going to help me here...

Re:Intel announces 6 cores, 6 months after AMD.. y

[petermgreen]

"which would have been 3 sets of dual cores."
The have been previous intel hex-core processors which were three dual-core dies in one package.

This is one die though it does seem to have two seperate L3 caches. feeding off a central memory controller and queue (with the two QPI interfaces on the other end of the L3 caches).

Re:Intel announces 6 cores, 6 months after AMD.. y

[Gr8Apes]

It was announced in Oct. It has a single shared L3 cache. Anon Intel fanboy? Whatever for? Embarrassed?

Re:You need to read up on 3 things

[petermgreen]

I think you misunderstand me, most threads aren't blocked because they need access to data. They are blocking because they are waiting for something to happen (the user doing something, a network packet coming in, a timer tick, a soundcard asking for more audio data etc). Just look in task manager and see all those 0% entries in the CPU column.

Desktops are rarely doing more than one intensive job at a time and even more rarely doing more than two. This means that unless your apps can break an individual job into multiple threads (and as I said above the number of threads shown in task manager is no indication of whether an app will do this) or you run a lot of background batch tasks adding cores (particularly beyond the first two) has little benefit.

castle

[johan85]

watch castle

Re:Intel announces 6 cores, 6 months after AMD.. y

[SacredByte]

I'm not the world's leading expert on processor design, but from what I can tell from the schematics and die pictures I've seen of the Westmere 6-Core CPU's, it only *looks* like they stuck 2 3-core processors on the same die. Everything I have read about Nehalem (and now this current die-shrink) has said how modular and scalable it has been designed to be; Intel saw the limitations of their early dual-core designs in that they *weren't* particularly modular or scalable, and I can't imagine that they would repeat that mistake.

That said, from what I can tell, they laid it out like they did (so that it looks like three cores on one side and three on the other) in order to minimize the maximum distance(s) between certain key components (they stuck the queue and most of the uncore in the middle). It is my assumption that although there appears to be two L3 caches on the die, that any core can directly access any of the cache. I base this mostly on my understanding of how the QPI system works on a DP platform -- one link is for communicating with the Northbridge (Current DP NB is tylersburg 5520 while SP boards usually use X58) while the other is for direct communication with the other processor --- I would assume that if processor 0 can get data directly from the cache on processor 1, that any core on processor 1 can access any data on it's internal shared cache.

Re:Intel announces 6 cores, 6 months after AMD.. y

[Gr8Apes]

It will be interesting to see if this is borne out in the benchmarks. I'm very curious regarding these details. It'd be very nice if the #1 CPU maker was on the same playing field as the current tech of the day. We would all benefit from that situation.