Intel's Quad Core CPU Reviewed

Gr8Apes is one of many to let us know that Tom's Hardware Guide has posted a review of Intel's new Kentsfield quad core processor. From the article: "Even expert opinions are deeply divided, ranging from 'more cores are absolutely necessary' to 'why do I need something more than my five-year-old PC system?' Although the Core 2 quad-core processors are not expected to hit retail channels before October, Tom's Hardware Guide had the opportunity to examine several Core 2 Quadro models in the test labs. We would like to make it clear that these samples were not provided by Intel."

Comments

Same old dilemma, new format.

[sdaemon]

Some applications will make use of it, some won't. More cores is pretty much the same as more CPUs.

Re:Same old dilemma, new format.

[fm6]

Gosh darn, you ended the flame war before it had a chance to start. Shame on you!

Re:Same old dilemma, new format.

[sdaemon]

My apologies. The jury will disregard my previous comment. Proceed with flames...

Re:Same old dilemma, new format.

[bangenge]

so does this mean that the parent is gonna be modded flamebait?

Re:Same old dilemma, new format.

[Kjella]

Some applications will make use of it, some won't. More cores is pretty much the same as more CPUs.

And it was quite obvious which, parallelizable processes like media encoding saw 80% improvement. Some of the rendering tests saw very impressive improvement as well, but that's not really new and that's why those people have been paying for SMP setups in the past too. The gaming improvements were at the moment none, unlike the Core Duo which spanked the PIVs. They're really stretching at straws saying "it's a must for HDTV editing" - indeed, but how many are editing/compressing HDTV on their media PC? HDV home video is still rare as hell, HD-DVD / Blu-Ray can't be recompressed because of DRM. So maybe the ones with MythTV and OTA HDTV (or the few HDTV channels in Europe), but that's pretty much it. For what it's worth, I think the recommendation is to use hardware compression (or direct stream rips) anyway, and I imagine it won't take that long to get hardware cards which'll put out HDTV in one of the better HD-DVD/Blu-Ray formats, either VC-1 or H.264. At that point, who needs it? Maybe if they get the power consumption down to 2x Core 2 Duo (was that confusing or what) it'll be a nice way to pack them even tighter in servers. Beyond that... what?

Re:Same old dilemma, new format.

[SharpFang]

One core dedicated to enemy AI.
One core for physics calculations
One core for the game itself.
One core for OS, daemons and to bring them all and in the darkness bind them

Sure nothing for any -current- game.

Re:Same old dilemma, new format.

[Schemat1c]

Hell is rare?

From what I've heard it's more like well-done.

Re:Same old dilemma, new format.

[ChrisA90278]

One core dedicated to enemy AI.
One core for physics calculations

That is NOT the way you design multi threaded aplications. No, what you do is to identify varius proceses that you need to do. Perhaps one of them is to read user inputs, read inputs comming in from the network (Internet). another to move the "camera" and then one process for each object to be simulated. So you might have 40 or 100 threads running. Then you would depend on the operating system to allocate physical "cores" to software threads. With this type if design performance can scale as you add cores. ASo in 5 to 10 years when we all are using 32 or 64 core machines it really will run 8 times faster How many threads is an enginerring decision. For example if you have four cars running on a raod do you use a thread for each car or one thread for all cars, another for all people. Thread per object is a radical design but thread per type of object is less radical

Re:Same old dilemma, new format.

[LWATCDR]

"As I recall, the 6510 in the C-64 didn't even have hardware multiplication!"
that is correct and also no floating point. But it had zero wait-state memory and was actually pretty fast for it's day :)
Also only 2 registers and a single accumulator. Oh and those where all 8 bit.
You kids multiply? It is nothing but a repetitive add anyway.

It's the bandwidth stupid!

[siewsk]

It's the bandwidth stupid! Does not matter how fast the CPU is if it is bandwidth limited.

Re:It's the bandwidth stupid!

[osho_gg]

You are only saying this because the current state of the art is bandwith limited. It is much too trivialized generalization to say it "it is bandwith stupid!". Computer architecture and preformance is all about compromises, bottlenecks and parameter tuning. Not so far in future, there will be more bandwidth and then the limiting factor could be the speed/size of the memory. Or, it could be the power envelop of the entire system, Or, it could be back to the raw performance of CPUs. Or, it could be limiting opportunities for parallelization in most common workload for a typical office/home user. A little further out in future, there will be integrated GPUs, maybe even memory etc. and then it could be dinosaur hard-drive technology that may be limiting factors of overall system performance.

At any point of time in the history of computer performance, to say that, "it is stupid *anything*!" is much too simplistic point of view IMHO.

Osho

Re:It's the bandwidth stupid!

[ZachPruckowski]

From the looks of things, a Kentsfield on a 1066MHz FSB did similarly to a Kentsfield on a 1333MHz FSB, suggesting that bandwidth isn't the key issue. In fact, they scored the same on several tests (both were at the 2.67GHz standard speed). That surprised me, but I'm not complaining. Now, it may be that those tests never utilized all four cores, but it still scored head and shoulders above a E6700, suggesting that at least 3 cores were used.

Re:It's the bandwidth stupid!

[convolvatron]

of course you are right to a large degree. global memory bandwidth
is the cause of the day and given coherency, its not trivial to
architect around. the parent may have been a little terse, but
as you point out, overall throughput doesn't go up if all the
cores are too starved to issue.

however the memory latency picture isn't changing very much, and the
most compelling method to hide it for general purpose machines is through
thread parallelism (ignore vectors for a moment, its kind of a special
case of the same thing).

this is what makes the multicore picture interesting. assuming a workload
that can exploit it, you can really turn the scale knob pretty far up.
unfortunately the whole affair pivots on being able to get past the
crappy heavyweight thread-and-lock model the software people have grown
so fond of. and the software community isn't particularily light on its
feet.

Re:It's the bandwidth stupid!

[adam31]

Not so far in future, there will be more bandwidth and then the limiting factor could be the speed/size of the memory. Or, it could be the power envelop of the entire system, Or, it could be back to the raw performance of CPUs.

Ah, but the future is now. Cell has already addressed these issues: 25.6 GB/s main memory bandwidth, 256 kb of L1 cache per core, OoO sacrificed to minimize heat, maximal raw performance of CPUs in FP, integer, FP, load/store, FP, and main memory transfer (DMA engine) without any silliness like 8 GP registers (128). Even when multiple Cells are hooked together, it's over a 35 GB/s IOIF port.

Also for onboard multitasking, you forgot about being latency-bound by atomic operations, which is something that would be really bad with separated L2 caches. This issue is also elegantly handled by cell by having only a single bus-snooped L2.

It must be frustrating for the hardware guys. You address all the bottlenecks in a pretty uniform way, and they still criticize: "But... uh, the software guys need a refresher course in hypertasking..."

Re:It's the bandwidth stupid!

[uglyduckling]

OoO sacrificed to minimize heat

Surely it's not necessary to sacrifice openoffice.org, can't it just be tuned a bit to keep its processor useage down? Maybe it will be the bottleneck for a few more years to come, but eventually it will make better use of system resources, I'm sure. Or are you just a Microsoft Office fanboy?

Experts?

[Angst Badger]

Even expert opinions are deeply divided, ranging from 'more cores are absolutely necessary' to 'why do I need something more than my five-year-old PC system?

These are obviously experts who have never heard of servers.

I'm perfectly content with my 1.2GHz single-core single-processor laptop, but I'd sure as shit like to have more muscle in the database cluster I'm responsible for maintaining at work. Whether these chips are a good solution remains to be seen, but that's a separate question.

Re:Experts?

[Aadain2001]

Both AMD and Intel know they can't focus soley on the server or desktop market, but on both. While these chips are great for servers (what high powered chip isn't??), they will also be targeted at desktop machines (and may laptops in the future). With the performance of single core CPUs reaching their limit (thank you leakage current and high temps), multicore and multiproc systems will be the future of computing. Yes you can quote me on that.

Re:Experts?

[steveha]

The companies who are really serious about servers are particularly interested in CPU power compared to heat dissipation -- thermal density. This new Intel CPU is high performance with high heat--more of a gamer chip. At least so far it is; it's a very early sample and Intel hasn't had time to tune the power management features.

Intel's latest chips are fabbed at 65nm, while AMD is still only shipping chips fabbed at 90nm. This should give Intel a serious edge in the performance/heat ratio, but AMD's chips are so much more energy efficient that they are still competitive. (The current best performance/heat is the AMD Athlon64 X2 3800+ ADD chip.) When AMD finally ships 65nm Opterons, those ought to be really great for dense server installations.

It's telling that even Dell is planning to ship servers with AMD chips. They announced a 4-core server; two dual-core Opterons. It wouldn't surprise me if they will be 65nm Opterons when they finally are released.

The article says that Intel is going to transition from 65nm to 45nm sometime in 2007, and to 34nm sometime in 2009. They beat AMD to 65nm big-time. They may well be at 34nm before AMD can make it to 45nm! Just imagine some sort of server chip with 16 cores... or more likely, 8 cores and a whole bunch of cache.

But we shouldn't count those chickens before they hatch. Right now Intel is at 65nm and AMD will be there soon.

steveha

Re:Experts?

[julesh]

I've heard a lot of experts suggest that scaling outwards (i.e. adding more nodes to the cluster) is a better solution than improving the performance of individual nodes. They look at google as a model of how to build a high-performance database application.

I'm not convinced, but that's one point of view that's often expressed.

Props for Intel for being early

[elh_inny]

Some nice example where more processing power (even in parralel) is nice is virtualisation, whether at home or on servers. Running multiple OSes in parallel will saturate all your processing power nicely.
What's more quad-core surely gives more processing power per watt and per cubic meter which is a very important factor for big folks like Google or whereever hosting space is expensive.
Even John Carmack who used to be very much against multi-cores for gamins recently elaborated much on this area in his keynote. Practically any modern (lets call it nextgen :D) gaming platform is now multi-core.

So I'd say overall it's nice that Intel is pushing this so fast, if developers start to realize that multi-cores are hitting mainstream, they will have to take that into account and by the time Intel and AMD launch 8-cores, there should be more software to take advantage of it.

but the real question is

[Rooked_One]

how many FPS can I get in quakeworld? With the +1000 FPS it would give, i'm sure I would be able to bunnyhop all the way across the 2fort5 outdoors area!

Re:but the real question is

[RuBLed]

it could be possible (maybe someone would tweak it in the future), but I doubt your existing crt/lcd would support that :)

What would improve is the quality of the games both in physics and in AI.

Games are going parallel

[savuporo]

Coincidentally, Gamasutra just two nice feature articles on rearchitecting the game engine flow to better parallelize the tasks so that multi-core can be taken advantage of, utilizing OpenMP
"Multithreaded Game Engine Architectures "
http://gamasutra.com/features/20060906/monkkonen_0 1.shtml
"Multi-Threaded Terrain Smoothing"
http://gamasutra.com/features/20060531/gruen_02.sh tml

Re:A BIG question???

[Nom du Keyboard]

Did I miss something here... I thought it was called Kentsfarm?

Kent's Farm is where Superboy grew up before he became Superman. It was a rights issue.

260 Watts!

[dbIII]

260 Watts! We've been looking at the wrong factors - it appears that global warming is related to Moore's law.

Re:One other thing

[Jeremi]

Core Duo, Core Solo. What are we going to call this one?

Based on Intel's recent naming conventions, I think they'll call it the "Core 2 Duo Duo", so as to generate as much confusion as possible :^P

If not Intel...

[abshnasko]

"We would like to make it clear that these samples were not provided by Intel"

Then who were they provided by, exactly?

Re:If not Intel...

[AcidLacedPenguiN]

*tinfoil hat*
 
AMD.

more power is often bad

[r00t]

Typically this requires more fans. Fans are moving parts, thus likely to fail. Your super-fast computer could crash because of a 13-cent Chinese fan. Dead computers only go fast when you drop them out the window.

Fans are noisy. This causes other people to accelerate your computer at 9.8 m/s/s. They can sneak up on you because you're going deaf from the noise.

Duo 2 Sexo?

[BikeRacer]

TFA spends a little time describing that Intel doesn't have enough package area to use this iteration of the Core 2 Duo to make a 4 die, 8 core part. So, my question is: Ignoring likely heat and bandwith issues, is there a SMP architectural reason they can't put 3 dies in one package?

Re:Duo 2 Sexo?

[james b]

I've wondered in the past why multi-core/multi-processor systems usually seem to have a power-of-two number of cores. This quote is interesting:

Besides, it's very rare for users to need an odd number of processors (in any of the parallel codes I've seen at least). Most parallel problems are able to work in parallel by decomposing some sort of domain (be it physical, a mathematical matrix, etc.), and this decomposition usually happens in more than one dimension (this generally is done to optimize computation vs. communication). So generally worst case are prime numbers of nodes and the best cases are powers of two.

So perhaps it's a convention borne of parallel computing algorithm design? But there could be a more fundamental SMP architecture reason - anyone know?

Spare Capacity Usage.....

[IHC Navistar]

Hmmmmm.....

Maybe now I will have the ability to put all of my excess computing power into figuring out why women are the way they are instead of searching for extraterrestrial intelligence...............oh wait.....

-----

Sig Sauer

Re:One other thing

[evilviper]

Based on Intel's recent naming conventions, I think they'll call it the "Core 2 Duo Duo", so as to generate as much confusion as possible :^P

No, that's much to straight forward for Intel.

I expect something along the lines of: "Core 2 Tre Quad Pentium 405".

And AMD's AM3 5235+ 3.1G X4 Thunderon is faster and cheaper, anyhow.

With four cores...

[Bazman]

There's an ad for PC World or Currys or something where the salesguy explains to some students that having an Intel (bing-bong-bing-bong) dual core processor means it can do two things at once - like sending an email and downloading music!

Re:With four cores...

[Rodyland]

Talk about dumbing it down. Still, maybe they should put it in terms the average real computer user would understand.

How about, you can illegally download a DVD whilst watching another illegally downloaded DVD.

Or, more likely, you can play your pirate copy of Civ4 while compressing your CD collection so you can share it illegally over the internet.

Or, even more likely, you can watch illegally-downloaded pr0n on your monitor whilst watching more illegally-downloaded pr0n on your TV through your video-out.

I reckon with 4 cores hyper-threaded you could do all of the above at the same time. *swoon*

from intel's point of view

[Anonymous Coward]

it should be spelled the quad core due owe, they think it is their due for you to owe them a lot of money for something you really don't need.

The last couple of years I am finding it extremely hard justifying upgrades. I can surf fast, web pages render fast, can already watch a nice dvd movie or listen to fairly good sound, etc, and other multitasking things-all with a barely past 1 ghz cpu and half a gig of old slow RAM and a 45 dollar vid card. If I need some "upgrade", well, I still have two empty RAM slots, and that's cheap and probably a lot more cost effective in real world use than having to upgrade to some new chip that takes a brand new mobo as well. I am trying to see where having some webpage open 5 milliseconds faster or something like that is any sort of huge advantage that I should pay hundreds of dollars for. I's not like I would kick if it was only 20 bucks or something, but....not seeing it, computers hit a "good enough" level a few years ago now it seems, and I have heard that from any number of people in meatspace as well.

  I like a variety of gadgets, not just computers, and I'd say for the bulk of humanity out there, what they have right now (I am just generally speaking now) for computing power is actually filling the bill quite well. I realise that companies have to keep selling to stay in business, but this arms race is also creating a lot of planetary waste and over production of honestly unneeded stuff when it comes to computers. And cellphones are even more stupid is it REALLY necessary to get a new one and throw the old one away every few months? I would rather see this silicon (and peoples extra money they would have blown on an unneeded new computer) go to stuff like solar panels for the next few years ahead, that industry needs it a lot more and will help to drop prices there and help society as well.. I like consuming electricity and what it does, but I would like to step up to the plate and *produce* a little of it as well, especially with something that will come with a 30 year warranty like most PV panels do now.

Anyway, just a thought on this subject, I think it would be better if they told these chip places to just go into idle from a flat out race for a spell, give people a chance to really *use* what they already have, do some more R&D, and skip to every third generation or so to mass produce and sell.

Re:from intel's point of view

[aywwts4]

Damn, forgot the breaks, If you want to actualy Read my post, this might be more legible.

Vista will hopefully solve that problem, and 'progress' can continue unabated. ;)

But you have excellent points, I would have modded you up but I would rather respond.

Processor makers need to really work on energy efficiency of all their desktops, these speeds were achieved through sheer increases in heat and power consumption, and its really flatly unacceptable (My current desktop heats up my office to a toasty 89 degrees in the middle of winter if I close the door)

Truly computers have become terribly wasteful, Other than environmental impact (less lead and harmful metals) I doubt the $399 computer of today contributes less waste than the $2000 dollar computer of 1990, The only change is my 386 lasted for 7 years, and the 399 dollar computer, isn't very upgradable, comes crippled from the start, and gets thrown out when infested with spyware, because the price point makes it a disposable commodity.

The power saving features built into computers now are great, but its only so good if the processor is just short of having a nuclear powered core.

Not only that, but with your idea hopefully less money can be put into stopgap number increases down dead end roads and make processor manufacturers build their architectures with a little more forethought. Maybe if every 3-5 years there was a responsible and substantial leap in computing power people would upgrade in regular phases, without wasting so much energy, and without so many confusing acronyms and model numbers. Each generation would have a nice leap in power, and a new OS to match, each computer manufacturer would keep models to a minimum, and just slide the prices down during that generation until it hits everyone's sweet spot. Upgrades during the generation can focus on power consumption instead of computing power, programmers can aim and test for certain generations of computers, and system requirements will be simplified greatly, Less confusion, less waste, morestandardized component generations. "This software runs on generation 4 and 5 computing components"

Of course gaming is to blame for this constant demand, So I figure nothing will ever change until we hit that mythical peak in which everyone says "Good enough" and the race to real time photorealism is over.

(And yes, I too am very happy on my 6 year old system, I just wish I could give it a ram upgrade, The only problem being... I mistakenly thought RD-Ram was the future, Check the prices on a gig of that and have a laugh at my expense, In conclusion, sorry for the rambling, Slashdot is my outlet for insomnia)

Re:from intel's point of view

[joto]

Processor makers need to really work on energy efficiency of all their desktops, these speeds were achieved through sheer increases in heat and power consumption, and its really flatly unacceptable

Didn't you pay attention in class? All the processor makers have started doing this. For the last one or two years, the mantra has been "computing per watt", pretty universally, no matter which company you were from. And by the way, if you compare your 386 and a modern computer, I'll bet the modern computer gives more "computing per watt", no matter how you decide to calculate, so all improvements can't have been "achieved through sheer increases in heat and power consumption".

Maybe if every 3-5 years there was a responsible and substantial leap in computing power people would upgrade in regular phases [SNIP] Of course gaming is to blame for this constant demand

Ha ha ha. Hi hi hi. Ho ho ho. Hilarious! So you think if it wasn't for the gaming industry, we would somehow magically have every manufacturer on the planet agreeing to wait releasing their improved products untill someone (who?) said it was time for a new "generation"?

Apart from the fact that such a thing has never happened in other industries (or would you care to give a counterexample?), that it is uncompatible with the idea of a free market, and that it is bad for consumers as well as producers, please explain why you think that would happen AND why you think that would be a good thing.

So I figure nothing will ever change until we hit that mythical peak in which everyone says "Good enough" and the race to real time photorealism is over.

Yeah, because when graphics are suddenly "good enough", nobody needs computers for other purposes. People don't need plain old office computers, workstations, databases, web-servers, B2B-applications, industrial control software, or anything else. Because the only thing driving the computing industry forward is gaming. And when photorealism in gaming has been achieved, all computing innovation will stop. Right!

And besides, there will never be a "good enough" for graphics. In my view, audio has been "good enough" since long before the invention of the CD. That doesn't mean that there aren't people still working on creating better and/or cheaper audio components.

Re:from intel's point of view

[keesh]

Speaking as a software developer, I need the extra cores. Not for customers to run the software. For me, to compile things.

My previous development box was an Athlon XP 1700+. It did a full compile and test run of my current project in about sixteen minutes. I've just been upgraded to a Core Duo 2, which does it in three (make has a parallelise option, so it can use both cores). Give me a box that's twice as fast (which twice as many cores is, for compiling) and the compiler will finally be able to keep up with the coding, which means no time wasted sitting around.

Is it worth the money? For some people, yes. For others, clearly not. I couldn't justify getting a super expensive IBM pSeries box as a development system, but a cheapo x86 desktop doesn't even show up on the budget...

Re:from intel's point of view

[keesh]

Not the compiler. Compilers are complicated enough that making them threaded is just a recipe for disaster. The build system is what should be doing the work -- in a project with lots of files, a decent build system (like, say, make) will be able to use all cores quite happily. Heck, make -j can use all 32 CPUs on our big release box without any difficulties.

If your build system doesn't support parallelisation, you should look into switching build systems. Developer time is not cheap.

Re:from intel's point of view

[jtwronski]

I am trying to see where having some webpage open 5 milliseconds faster or something like that is any sort of huge advantage that I should pay hundreds of dollars for.

If all you do is surf the web and watch movies, well then you'll be fine for some time. I edit movies quite often at work, and at 720x480x2000kb/s, I'm screaming for as much horsepower as I can get my hands on. I'm currently on a P2.4 with 1Gb of memory, and i'd like at least 4 times as much power as that. Cost be damnned. It'll pay for itself in saved time even before it becomes obsolete.

Re:Dear "Expert"

[Rodyland]

Until such time as MS/Oracle/VendorOfChoice decides to (re-)institute per-core (or even better - per-virtual-core) licencing...

You act like being raped by MS/Oracle/VendorOfChoice isn't a priviledge and an honour.

Re:Dear "Expert"

[Vendekkai]

But people like me wanting to roll out a database cluster, getting a cpu with four cores could save me $75,000 per CPU running M$ SQL Server. Oh the dream of running 16 cores on 4 CPU's

Till Oracle and Microsoft revise their licensing terms to take into account multiple cores, that is.

Or do you think they're going to sacrifice all that potential revenue?

V.

Bring on the more complicated architectures!

[Rodyland]

As a sofware developer, I can't help but think the move to multiple cores is a good thing. In my mind, anything that makes software development MORE complex can only improve my employability.

Re:One other thing

[PlusFiveTroll]

Four core and seven years ago chipzilla brought forth on this on this die, a new processor, conceived in vanity, and dedicated to the proposition that all CPUs shall require a nuclear power plant to function.

Guess they should have called the platform Gettysburg.

FSB

[Dersaidin]

They should have made the FSB 4 Mhz faster. That'd just be too cool. Fight back AMD! Don't take that shit!

Re:FSB

[Dersaidin]

Its just that
1333 + 4 = 1337

:/

One thing they didn't benchmark on it....

[gsasha]

Is Linux Kernel compilation. It should rock there, that's an inherently parallelizable task.

As a programmer, I want one. No, I want two :)

Re:One other thing

[PlasticArmyMan]

It'll render graphics and give you a close shave at the same time!

... Next page....

[KillzoneNET]

After

... Next page ...

getting up

... Next page "Interested in..." ...

to the first

... Next page "No please..." ...

page of

... Next page "Buy our offer..." ...

the actual

... Next page "We are desperate..." ...

charts,

... Next page "Buy now! Limited time offer!" ...

I lost what

... Next page "SPECIAL OFFER!" ...

the art

... Next page "Brought to you by Fixed Random Ads Inc" ...

icle was about...

Re:One other thing

[tygerstripes]

"How can we make our CPUs more attractive without actually improving their performance?"
"Well, they've had twin-core - let's give them FOUR cores! And an extra one on the back!"
"It's... it's brilliant!"

Yes, Intel are the new Gillette. ;-)

Re:Us coders are delaying the Singularity!

[joto]

Why is it so hard to get developers to write decent multi-threaded code? It's not that hard,

Let me put it this way. If all the developers in the world were as smart as you think you are, it would not be that hard. As it is, however, coming up with scalable, manageable, efficient ways of writing multi-threaded code, in a way that is future-proof, as opposed to simply optimized for todays generation of hardware, is hard. Very hard. Not as in research-subject hard, but as in continuing-research-for decades-has-still-not-brought-us-much-closer-to-a- solution hard!

, and using threads properly can almost always improve performance and/or responsiveness on single proc/core machines to boot.

Let me rephrase part of the above sentence: "using threads properly can...". Did you notice which word I emphasized? Can you guess why?

Any idiot can use threads. The difficulty is to find the right granularity of threads (which is related to what kind of hardware you've got), which tasks are parallelizable, which parallelizable tasks should (or should not) get parallelized because of communication overhead and other factors (which is also related to what kind of hardware you've got), and so on.

It is also important to note, that few existing programs are designed from scratch today. In fact, almost all existing programs were designed in the past! In the past they didn't have affordable multicore (or multi-CPU) computers. And thus, those old designs didn't take that into account.

Re:One other thing

[SharpFang]

Unfortunately, it's not MACH, it's ix86.

Re:Us coders are delaying the Singularity!

[ardor]

Why is it so hard to get developers to write decent multi-threaded code? It's not that hard, and using threads properly can almost always improve performance and/or responsiveness on single proc/core machines to boot.

Because it IS harder. It introduces new pitfalls (deadlocks, livelocks, race conditions), debugging is harder (gdb with multithreaded programs.. brrr), old paradigma have to be thrown overboard (and new ones introduced, such as task- or stream-based processing). Also, threads NEVER improve performance on a single-core machine. They do help with responsiveness, however. If you want performance boosts, use a multicore machine.

Re:One other thing

[Hast]

I hear AMD is going with 5 cores.

http://www.theonion.com/content/node/33930

Exactly

[joss]

I grew up programming transputer clusters cos I figured Moore's law wqould have to slow down sometime and then we would have to move to multiprocessor systems. Efficiently using more than a couple of cores is *not* easy.. and it opens up a whole realm of interesting algorithmic work where basic problems with established solutions suddenly become open again.

Its about fucking time...

Re:Us coders are delaying the Singularity!

[Kjella]

Also, threads NEVER improve performance on a single-core machine.

Unless you call blocking functions (like IO), where one thread'll block and the other(s) will keep going just fine. But yes, if you're 100% CPU bound then making it 2x50% won't help at all.

Re:One other thing

[CrackedButter]

My friend i have just one word for you "Thundercougarfalconbird".

Re:Us coders are delaying the Singularity!

[ardor]

Here we see the problems with measuring performance and distinguishing it from responsiveness - sometimes it is very hard to do. The blocking IO almost always harms responsiveness, not performance. A file manager scanning thousands of files for thumbnails generation has responsiveness problems if this is done without multithreading.

In fact, I would see responsiveness as the bigger problem nowadays. It affects the user directly (for example, the file manager not reacting to any kind of input while scanning a large directory) and is ignored too often.

Re:Us coders are delaying the Singularity!

[joss]

Are you being deliberately stupid or what ?

Some things are easy to parallelise, a lot of things arent. Processing an image.. fine.. give each processor a chunk.. but wait.. you get edge artifacts since each pixel needs neighbour pixel information.. that has to be shared.. at what point does it take *longer* using multithreads.. doing a large matrix inversion that takes hours.. ok, parallelise it.. but wait.. standard algorithms that you can see in numerical recipes in C assume single thread.. u need a completely new algorithm, and u need a completely new mechanism for stabilisation bla bla bla.. and this is before u even get into resource contention etc etc.. These are not compiler issues, they are design issues, some of which are at least as hard to solve as writing the entire single thread program. Until computers are smarter than humans, they are not going to get solved automatically.

How long will it be ...

[LaughingCoder]

before Microsoft starts charging more for multi-core installations? Seriously, if quad core means fewer boxes in the rack, it means fewer licenses.

Re:How long will it be ...

[LordEd]

What software is asking for extra licenses? Microsoft has explicitly said they are going on a per-processor model:

On October 19, 2004, Microsoft announced that its server software that is currently licensed on a per-processor model will continue to be licensed on a per-processor, and not on a per-core, model. This policy will allow customers to recognize more performance and power from Microsoft software on a multicore processor system without incurring additional software licensing fees.

Microsoft's Multicore licensing policy

[LordEd]

Google says...

On October 19, 2004, Microsoft announced that its server software that is currently licensed on a per-processor model will continue to be licensed on a per-processor, and not on a per-core, model. This policy will allow customers to recognize more performance and power from Microsoft software on a multicore processor system without incurring additional software licensing fees.

Early?

[Tmack]

Nah, they are still playing catch-up. SUN has already been selling their Niagara (Ultra Sparc T1), which has 6-8 cores each, with an additional 4 threads per core, and the ability to have 4 in a box (8*4*4 = 128 threads!) blows Intel out of the water. Yes, I am perfectly aware that this chip is a SERVER cpu, but you mention servers and virtualization (vmware, parallels, etc), which Sun basically designed this chip to do. As a sysadmin, this cpu also get my attention over Intel for using very little power, something Intel isn't exactly known for. Less power == less heat generally, and with the number of cores the T1 has, we can run multiple VMware instances on one box instead of several 1u servers or blades. The drawbacks to the T1 are its relative lower speed (1.2Ghz, tho it is a Sparc based CPU), and a single shared FPU per cpu, but for general purpose servers, the T1 is it. For our heavyweight servers, the ones that need the most CPU horsepower regardless of heat/power consumption, we still pick AMD opteron based Suns (Sun Fire x4600). The memory bus to the Cores alone far surpasses Intel when it comes to throughput, which directly translates to faster operation with bigger datasets, not to mention that you can get an 8xdual core 3Ghz Opteron setup...

AMD beat Intel to the dual core CPU, and will be releasing a 4core opteron soon as well. IBM's Cell cpus are also multi-core, though in a different arrangement, and will be used in the ps3 and IBM's next megasupercomputer.... Multi-core itself isnt too special, it just takes multi-cpu and puts them in the same physical chip to save space/heat/power. The advantages for programmers/gamers/etc (neglecting the environmentals like heat/power) are basically the same as if each core were a different physical chip. Intel got behind, and is now rushing to one-up AMD. The same old CPU war rages on

tm

Re:Us coders are delaying the Singularity!

[Sebastopol]

The problem with threads:

http://www.computer.org/portal/site/computer/menui tem.5d61c1d591162e4b0ef1bd108bcd45f3/index.jsp?&pN ame=computer_level1_article&TheCat=1005&path=compu ter/homepage/0506&file=cover.xml&xsl=article.xsl

Basically, even the simplest tasks require significant armor plating to run correctly.

OTOH: Multi-PROCESS programming is far simpler than multi-THREADED programming.