AMD 4x4 Quad Father, Quad Core CPU Details Emerge

JiminyDigits writes "AMD recently revealed a few more details of their upcoming quad-core platform architecture called 4X4. With CPU bundles affectionately dubbed 'Quad Father,' AMD is taking advantage of the inherent benefits of their HyperTransport interconnect technology to directly connect a pair of dual Athlon 64 desktop chips together with system memory. Details here show a dual socket motherboard that support a whopping 12 SATA connections, four X16 PCI Express slots (x16,x8,x16,x8 configuration) and few other bells and whistles. Supposedly Quad Father kits will come with matched CPUs from 2.6GHz up to 3GHz."

Fuck it, we're going to five cores.

[WilliamSChips]

Hey, it had to be said.

Just like the SX/DX line

[HiggsBison]

I fully expect Intel to make a 2-1/2 core CPU called the Dual-Core-3 and a 3 core called the Dual-Core-4.

Re:Fuck it, we're going to five cores.

[lordofthechia]

Ahh all the ludicrous amounts of cores, when will it end?

"Sir it's AMD.... they've gone plaid"

It won't.

[Millenniumman]

Eventually, we're going to move to processors that dynamically create MicroCores (TM), as they function. MicroCores will exist in another dimension such that they can endlessly multiply without taking up any space.

These systems will allow Windows Panorama (codename: Holstein) to run, although not with the new SuperTransparentyandFlashyandGooeyWoohoo interface, of course.

Of cores...

[jd]

...it will have to end eventually. As with any system, it is limited to the ability to communicate. In this case, Intel can't communicate with their chip designers and AMD won't communicate with anyone who isn't stinking rich. (I've yet to get them to do so much as reply to an e-mail, answer the phone or even return a call.)

Re:

[buswolley]

All your cores belong to us.

Re:

[BeerCat]

Yeah, and our cores go all the way up to 11

Re:Fuck it, we're going to five cores.

[mochan_s]

It's 4x4 so next step is 8x8 and ultimately to 64x64.

Maybe the number of cores will be the new Ghz?

Re:Fuck it, we're going to five cores.

[scum-e-bag]

Perhaps not. It depends on the layout. The inner cores may have trouble with cooling when you scale out the size.

Re: 5, common. How about 80.

[theexcogitator]

Didn't you see the article on intel. http://techfreep.com/intel-80-cores-by-2011.htm [techfreep.com] . Talk about an expensive chip.

Re:

[WilliamSChips]

What's sad is that Intel's probably going to do this. Well, actually, they fucked it and went all the way to eighty cores. No precision floating-point though, I think that's an oxymoron for Intel ;)

Re:Fuck it, we're going to five cores.

[buswolley]

My Quatro shaves just fine thank you anyway.

Responding to my post with a serious one suggests:

[WilliamSChips]

  _|= <---Joke

   o
  -|- <---You
  / \

(the joke looks like a chair because it was originally a Steve Ballmer joke)

Vista

[DaMouse404]

And this just about meets the minimum specs for Vista..
-DaMouse

Re:

[msobkow]

Or a full Linux install with OpenOffice, Mozilla applications, dev tools, utilities, etc.

Sad to say, XP vs. Linux isn't much of a performance competition any more. With a slow enough old box, you'll find they both take forever to boot... ;)

What worries me with Vista is the memory expense of full-application rendering regardless of surfaces displayed, as well as the application expense of always rendering a full screen of widgets instead of skipping over clipped/obscured regions.

The graphics hardware

Re:

[644bd346996]

It only takes a few hours with a fast computer to build a uClibc based gentoo system in a chroot and transfer it to a pentium class machine. The resulting system can surf the web faster than any version of windows, even with Firefox. (OK, so FF was using glibc, so what?) Not even Opera on Win 3.11 was more responsive. The boot time was also faster than Win9x. OO.o and Writely were a bit slower than Office 97, but they looked nicer and had better interfaces.

Sure, running a full redhat distro on an old box li

Re:Vista

[Jinky Williams]

"build a uClibc based gentoo system in a chroot."

When I read this, I snickered uncontrollably.

That is all.

4 * x16 == x16+x8+x16+x8?

[spectral]

Ok, I know I don't understand PCI Express, but isn't that 2*x16, and 2*x8? Yeah, it's 48 PCI-Express lanes, according to the page.. but saying that there's 4 x16 ports is a bit confusing, is it not?

Re:4 * x16 == x16+x8+x16+x8?

[FuturePastNow]

I think they're going by the size of the slot rather than the number of PCIe lanes it has. An x8 slot can support graphics cards fine, if it has the x16 physical connector.

Re:4 * x16 == x16+x8+x16+x8?

[masklinn]

There are something like 3 parts to PCIe-speak on motherboards:

• The number of lanes, which depends of the motherboard chips. That's the total PCIe bandwidth your motherbord can handle
• The physical size of the PCIe slots. That tells you what you can fit in the slots. For example, graphic cards use x16 slots, but can hum along perfectly with only 8, 4, 2 or even 1 lane (albeit with a much reduced bandwidth to work with).
• The number of lanes in every slot, which gives you the bandwidth per slot: all PCIe devices must support x1, but they can use up to x32

What they're saying here is that you're getting 2 x16 and 2 x8 lanes slots, but all the slots have a physical x16 size, which means that you can plug pretty much anything in it, including 4 PCIe graphic cards at once (since graphic cards require physical x16).

I'm not sure I've been perfectly clear though, anyway it's fairly clear when you talk about slot size versus number of lanes.

Re:

[spectral]

That makes sense. I wish that some other terminology had been used for distinguishing between physical size and lane count, but oh well. Thanks for the info!

Re:

[Fweeky]

Note this is the reason for those swappable "Single Graphics Card/SLI Mode" PCB's on many SLI motherboards; when the chipset can only allocate 16 lanes to graphics, one mode sets all 16x to one slot and the other splits it into 2 8x slots. It's nothing new, and though there are motherboards these days which can drive 2 full 16x slots, I don't think they differentiate themselves much in performance from 8x just yet.

Re:

[masklinn]

Yep, but regular SLI works well with x8 slots, and I had to switch to SLI setup (2*x8) on my computer because after watercooling the motherboard's chipset I realized that I couldn't put my graphic card into the top x16 slot anymore (me = toopid, next time i'll check that)

Re:

[Chris Snook]

With PCIe, you can have x16 slots that don't have 16 lanes committed to them, and this can be configured in an appropriately featureful BIOS. So you could plug in 4 x16 cards, and reconfigure the bandwidth to them without opening the case, but you'd have a maximum of two cards running at x16 speed. Apple already has something like this going on in their Mac Pro workstations.

Wow

[Anonymous Coward]

my text editor will just fly. I can't wait to spend shitloads of cash on this.

Re:

[eh2o]

Well yes... if that text editor happens to be vi.

Forced Overkill

[FuturePastNow]

"2.6GHz up to 3.0GHz"

Which means it will cost $1000-$2000 just for CPUs and motherboard. AMD's and Intel's quad cores will cost a grand also, which limits all of this to people with more money than sense.

If they're going to allow dual processors, why not let people use the $150 2.0GHz dual cores? Then the whole thing will come in under $500 and have much wider appeal.

Re:Forced Overkill

[joe 155]

Well, it seems expensive now, but I remember when a DVD-R drive was over £500; early adopters expect to pay quite a bit for bleeding edge stuff. In a couple of years these will start to show up in regular computer shops for much more reasonable prices.

Also, $1000 doesn't seem that expensive, spending about $2500 on a computer (which you probably wouldn't need to upgrade for about 5 years) wouldn't be that crazy, would it? It seems cheaper than spending $1000 every year and a half (which might be an average upgrade cycle)

Re:

[Abreu]

Oh yeah, I remember paying for a 16x CD-RW drive more than it costs today to get a DVD-RW/+RW drive

Re:

[modecx]

Heck, I remember paying 900 US bucks on a CD writer way back when 2x cd recording was blazing fast, and 8x reading was just becoming available... Oh, and discs could scarcely be found for less than $10 each at that time.

I considered it some of the best $900 I ever spent, and I still do. No regrets. In fact, it's still humming along in my Indigo2, which I pulled out of the scrap bin some years later.

$1000 bucks for a system loaded with quad processors won't scare many people off. $1000 for a motherboard m

Re:

[Dunbal]

I remember paying $200 at a computer show for a 2x CD-ROM drive.

      I remember paying $350 for a blazing fast, 2400 baud Hayes modem!

Re:

[raduf]

You guys are crazy, right? I just "upgraded" my computer (meaning I kept the sound board and monitor) for less then $500 and there isn't anything except bleeding edge games that I can't do with it. What else do you spend your money for?! Oh, and that's 500 per two years at least, maybe with $100 per year for bells and whistles.

Forced Overkill is Right!

[Memnos]

I'm just glad that my Dad wasn't a 4x4 Quad Father, or my Mom would have died during conception.

Re:Forced Overkill

[ocbwilg]

Which means it will cost $1000-$2000 just for CPUs and motherboard. AMD's and Intel's quad cores will cost a grand also, which limits all of this to people with more money than sense. If they're going to allow dual processors, why not let people use the $150 2.0GHz dual cores? Then the whole thing will come in under $500 and have much wider appeal.

The target price is under $1000 for the CPUs and (presumably) board. That really doesn't price it out the range of people who were previously buying Athlon FX and Intel EE CPUs. Keep in mind that this is a high-end enthusiast-class platform, rather than the future of AMD's mainstream computing. If you just want dual CPU dual cores, you can buy an Opteron 200-series workstation for less probably. You won't get 4 PCI-E x16 slots and 12 SATA ports, but who needs that anyways? Or, you could just wait until 3Q of 07 and get a native quad core CPU.

Would it be great if they made it cheaper so that everyone could have one? Absolutely. But then they would be cannibalizing the sales of their other higher-end CPUs (why buy a $700 FX-series when you can spend $300 on low end X2 CPUs and get more performance?).

Re:

[WuphonsReach]

You won't get 4 PCI-E x16 slots and 12 SATA ports, but who needs that anyways? Or, you could just wait until 3Q of 07 and get a native quad core CPU.

Those of us who want to drop in PCIe RAID cards and dual/quad port PCIe NIC cards? (Both of which are usually only available in PCIe x4 sizes.) Plus for less expensive servers, 12 SATA ports could allow the use of Software RAID without having to use up a PCIe slot for a SATA card.

When you get into NIC bonding, it's not unusual to want 4-8 gigabit NICs in

Re:Forced Overkill

[Lord_Dweomer]

Which means it will cost $1000-$2000 just for CPUs and motherboard. AMD's and Intel's quad cores will cost a grand also, which limits all of this to people with more money than sense.

This is called an "early-adopter price". You see, there ARE people with a lot of money...and contrary to your statement, they may, and probably do have plenty of sense, they just have more disposable income than you. They buy these when they first come out, and a year or two down the line when they are buying the next hottest toy on the market, companies will be forced to drop the prices on this bad boy so that the rest of us can afford it.

Don't bitch about the price of this just because you're jealous you can't afford it. Just realize that that is how the market works.

Think outside the Desktop Market...

[RotateLeftByte]

Couldn't this sort of beast be aimed at the Server Market? I have an application that would eat up this sort of config.
Curently we use a Dual Xeon or a Quad Xeon and these get maxed out at times.

Think outside of the Desktop Beige Box.

After a while, the technology will filter down to desktops but the server end is where people will pay top dollar/yen/euro/rouble for a system that really performs.

Re:Think outside the Desktop Market...

[Brian Stretch]

Couldn't this sort of beast be aimed at the Server Market? I have an application that would eat up this sort of config. Curently we use a Dual Xeon or a Quad Xeon and these get maxed out at times.

4x4 uses low-latency unbuffered RAM while servers use ECC RAM. More importantly, you can already buy dual CPU Opteron motherboards and chips. They've been capturing LOTS of market share from the Xeon, especially at the quad chip (8 core) level where the Xeon's obsolete FSB architecture falls down. Some vendors even have 8 CPU (16 core) boxes. And then there's Cray's Opteron-based supercomputers...

4x4 is basically an Opteron 2xx-series platform adapted for the desktop enthusiast market.

Quad 2.0GHz? No problem

[Wesley Felter]

4x4 is just a different name for Opterons in Opteron motherboards. Feel free to buy a 4x4 mobo and drop in some Opteron 2212s. However, since the 2212 is almost $400 and the FX-70 is only $500, you're probably better off buying what AMD wants you to buy.

Re:

[medelliadegray]

Wake up. This isnt forced overkill.

Bleeding Edge Technology is rarely purchased by Joe Public, even the vast majority is not even adopted by gamers. Big businesses generally absorb the early adoptors cost of this kind of technology.

A perfect example of a use for this is a vmware environment. Their licenses are based off the number of PHYSICAL cores you have, and each license for their Infrastructure 3 (Standard and Enterprise versions) are well above the cited $1,000 to $2,000 you mention.

Re:

[perlchild]

Because they hard to reengineer them for quad core(esp. memory bus wise, but also bios, etc..), otherwise, they would have launched quand WHEN they launched the 2.0Ghz core. Back when it was 2000$ a pop.

p.s. You don't launch an expensive technology based on cheap commodity parts, that's like subsidizing their waiting with your R&D dollars... You don't want them to wait, and the whole point of it being more expensive at launch, is that the target audience is those early adopter folx. They pay the R

Re:

[Chandon Seldon]

The cheaper processors don't have the additional inter-socket hypertransport link that's needed for a dual socket setup. They *could* create low end Socket F processors for 4x4, but it's more effecitve for them to use their production capacity for higher margin stuff. Remember that these 4x4 FX processors are basically identical to their current-generation Opterons - a market that they can't really afford to either give cheaper processors or reduce production for.

These cores go up to eleven..

[Channard]

At AMD HQ

AMD PR Rep: The chips have four cores. Look, right across the board, four, four, four and...
Tech Columnist: Oh, I see. And most chips go up to two?
AMD PR Rep:: Exactly.
Tech Columnist: Does that mean it's more powerful? Is it more powerful?
AMD PR Rep:: Well, it's two more powerful , isn't it? It's not two. You see, most blokes, you know, will be playing games with two. You're on two here, all the way up, all the way up, all the way up, you're on two on your PC. Where can you go from there? Where?
Tech Columnist: I don't know.
AMD PR Rep:: Nowhere. Exactly. What we do is, if we need that extra push over the cliff, you know what we do?
Tech Columnist: Put it up to four.
AMD PR Rep:: Eleven. Exactly. Two better.
Tech Columnist: Why don't you just have two and make them a little more powerful?
AMD PR Rep:: [pause] These have four cores.

Re:

[phaxkolumbo]

We need to be able to moderate to (Score:6, Funny) for comments like these...

4x4 eh?

[lordofthechia]

They trying to say that all 4 cores get traction or something?

That aside the dual x16 PCI express Mobo looks sweet. I can finally have my triple headed, neigh, quad head display! Note that a quad cpu quad display setup might be useful for MMO gold farmers... they could have one machine running 4 bots unencumbered and have the ability to monitor all 4 at the same time...

Re:

[Anonymous Coward]

They trying to say that all 4 cores get traction or something?

Absolutely. AMD full-time all-core processing provides outstanding traction with almost any air- or water cooled system. It constantly monitors processing conditions, sensing any loss of traction and automatically transfers processes from the cores that slip to the cores that grip. And cores that grip are especially nice if you're into Doom3 or any other game that demands a lot from a processor. Like grid computing, where AMD is a consistent

Re:

[Brian Stretch]

They trying to say that all 4 cores get traction or something?

Actually, that's not a bad analogy. Each AMD CPU has its own memory controller and bank of memory so there's lots of memory bandwidth to go around, whereas an Intel dual CPU config has both processors accessing memory through an obsolete FSB architecture. Accordingly, an Intel dual CPU machine will be spinning its wheels in situations where an AMD 4x4 has memory bandwidth to spare.

Re:

[mrchaotica]

That aside the dual x16 PCI express Mobo looks sweet. I can finally have my triple headed, neigh, quad head display!

Actually, you could have an 8-head display, because you can put graphics cards (albeit not super-high-end ones) in the two x8 slots also.

= 4 Acentral Processing Units

[Doc Ruby]

With two CPU chips with 2 cores each, shouldn't that be called "2X2"?

Hey, with 2 microprocessors, can they still be called "Central Processing Units", when each is "offcenter" to the other?

Re:= 4 Acentral Processing Units

[ocbwilg]

With two CPU chips with 2 cores each, shouldn't that be called "2X2"?

It was explained awhile back, but 4x4 isn't directly related to the core count. Otherwise, why wouldn't a dual CPU workstation class system with dual core CPUs be considered 4x4?

4x4 actually is in reference to 4 CPU cores and 4 video cards, at least that is the way that it was explained to me.

Re:

[syzler]

Otherwise, why wouldn't a dual CPU workstation class system with dual core CPUs be considered 4x4?

Actually, I think that was the point of the grandparent's post. The name 4x4 to those unfamiliar with the term in the context of motherboards is misleading. I thought the name referred to a quad core chip in a quad chip configuration. The grandparent's question and a few of the other comments I read implies I am not the only one to make this mistake. To the uninitiated a dual CPU workstation with dual core

Re:

[Kjella]

4x4 actually is in reference to 4 CPU cores and 4 video cards, at least that is the way that it was explained to me.

I tbink 4x4 is suffering from the Moore's law syndrome - it applies to whatever you want it to apply to which happens to fit the concept (quads, exponential growth).

Re:

[Chandon Seldon]

I'm pretty sure they're referring to four cores and *four graphics cards*. The "x" isn't multiplication, it's more like a funny looking comma.

Re:

[Doc Ruby]

I never thought that the "4x4" designation for "all wheel drive" cars made any sense, either. It's 1x4, engine to wheels, as opposed to the standard 1x2+2 (1 engine driving both front or both rear wheels). 4x4 makes sense when each wheel has an independent engine, which is pretty rare. Though apparently the AMD designation really refers to "4 cores with 4 GPUs", which is not even the config (GPUs) they're marketing under this buzzword.

But then, they're inserting two "father packages" into a single "motherbo

Re:= 4 Acentral Processing Units

[Dun Malg]

I never thought that the "4x4" designation for "all wheel drive" cars made any sense, either.

I helps once you understand that the designation isn't limited to "cars", but applies to ALL wheeled vehicles. The format is (total number of wheels) x (number of driven wheels). For example, the US Army's M-939A2 [fas.org] 5 ton truck is a 6x6-- 6 wheels, all driven-- and the M1074 PLS [fas.org] is a 10x10! Civilian trucking, by comparison, will usually make do with 10x8 on the tractor unit, being more concerned with weight capacity than offroad ability.

Re:

[Doc Ruby]

That's really interesting (and Insightful :), thanks. Maybe you could update the 4x4::Terminology [wikipedia.org] Wikipedia page section. Because its description makes sense in the larger scope now that you clarified, but didn't discuss enough examples to understand the meaning of the two symbols in the combined "WxD" designation.

Re:

[Doc Ruby]

What is "animal style"? Is that better than "the works"?

Would Paris Hilton please demonstrate an "animal style with the works, 4x4 down on the floor" to me?

Re:

[Jeremi]

What is "animal style"? Is that better than "the works"?

"Animal style" means you have to hide in tall grass, sneak up on the cow, tear its throat out, then use your teeth to pull steaming hunks of raw meat off of its still-warm carcass. It is indeed better than 'the works', if you're into that sort of thing.

Re:

[Doc Ruby]

Oooh, now I really want to demonstrate that to Paris Hilton, rather than she on me.

So where's the quad core cpu?

[timeOday]

After reading the article, I didn't see anything about a quad core CPU. Quad Father simply seems to be a dual cpu board with dual-core CPUs in it. That has been possible all along, no?

"Enthusiast Megatasking" is a lousy catchphrase

[timeOday]

AMD is pushing [hothardware.com] multitasking, a model of parallel processing that will never do desktop users much good beyond a small handful of processors. (Yes I know you currently have 57 processes running, and no that does not mean you'd benefit from 57 processors). If AMD presents these silly examples like being able to play two instances of a video game simultaneously, nobody will see any value. Instead, AMD (and for that matter Intel) should be doing all they can to promote fine-grained parallelism so individual applications can easily harness multicore chips without a huge extra developer burden. All too often I am sitting waiting for a job and my CPU utilization is only 50% because the app can't use both cores. (Come on, where's dual-core gzip?) You can say it isn't the chipmakers' problem, but if it prevents me from needing their products, it is their problem.

It's an OS problem

[joe_n_bloe]

Multitasking on *nux has worked fine since the 70s. Threading has been evolving on *nux since the 1980s and there is no shortage of threading support in that world.

The problem is with Windows and its tireless efforts to fill memory with dirty pages that get flushed at the most inconvenient times. Lots of CPU-intensive Windows applications support multithreading. It's not as if multiple CPUs are a new thing in desktop PCs. The old thing is the crappy NT scheduler and the OS's bizarrely dysfunctional memory m

Re:

[Mad Merlin]

Indeed, consumer hardware in general is held back by Windows and it's countless deficiencies. With memory for example, you basically can't use more than 2G of RAM with consumer level hardware because a) Windows still has miserable 64-bit support and b) Windows scales very poorly with more RAM anyways. So even those of us that aren't directly crippled by Windows, still have to put up with underdeveloped hardware.

Re:

[Mad Merlin]

(Come on, where's dual-core gzip?)

Gzip is sufficiently fast that I suspect in most cases it's more limited by your hard drive speed than your CPU speed. There is however, parallel bzip2 [compression.ca], which most certainly does benefit from parallelism.

Re:

[timeOday]

I back up from one drive to another, so can push up to 50 MB/s or so into gzip. And yet if I use the --fast option, what you say is often correct.

In fact I was going to go on a second diatribe in my previous post about how my new dual-cpu computer seems more disk-bound than ever, but I don't have any good suggestions on how to fix that.

Re:

[SillyNickName4me]

In fact I was going to go on a second diatribe in my previous post about how my new dual-cpu computer seems more disk-bound than ever, but I don't have any good suggestions on how to fix that.

Back when supercomputers were a lot less powerfull then todays desktop pcs, someone made the comment that supercomputers are machines that turn compute bound problems into io bound problems. It should be little surprise that this applies to many a modern computer now..

Re:

[timeOday]

Thanks. I've looked into it previously, but bzip2 is so much slower than gzip, it swamped the gain from parallelism. (Of course the files do come out a little smaller). (Also it doesn't yet support pipe input, which is a real problem for backing up an entire disk).

from "Memory Space"

[Anonymous Coward]

Come on, where's dual-core gzip?

Peter Gibbons: What would you do if you had two cores?
Lawrence: I'll tell you what I'd do, man: two gzips at the same time, man.
Peter Gibbons: That's it? If you had two cores, you'd do two gzips at the same time?
Lawrence: Damn straight. I always wanted to do that, man. And I think if I had two cores I could hook that up, too; 'cause processes dig CPUs with cores.
Peter Gibbons: Well, not all processes.
Lawrence: Well, the type of processes that'd double up on a PC like this do.
Peter Gibbons: Good point.
Lawrence: Well, what about you now? what would you do?
Peter Gibbons: Besides two gzips at the same time?
Lawrence: Well, yeah.
Peter Gibbons: Nothing.
Lawrence: Nothing, huh?
Peter Gibbons: I would idle... I would sit on my ass all day... I would do nothing.
Lawrence: Well, you don't need two cores to do nothing, man. Take a look at my cousin: he's got a 386, don't do shit.

Re:

[RhettLivingston]

The problem is that we keep trying to come up with programs to automagically twist programs written in a single tasking paradigm into a multitasking paradigm. Unless we could completely represent the program mathematically and apply a mathematical transformation, that will never be efficient. It would be easier to create a new paradigm to begin with. But we're blinded by our current focus on data. A focus on process oriented programming paradigms that utilize microprocesses to perform small bits of work

Re:

[timeOday]

That sounds somewhat like the blackboard [stsci.edu] architecture, have you looked into it? I agree that's much closer to how nature works. Imagine if all the cells in your body had to take turns :)

Re:

[Dunbal]

these silly examples like being able to play two instances of a video game simultaneously, nobody will see any value

      Yeah, instead they should present real situations, like playing oblivion/F.E.A.R./Doom while ripping and burning DVD's and seeding torrents...ooops

Re:

[adrianmonk]

Come on, where's dual-core gzip?

Well, I don't know of a dual-core gzip, but there is an SMP bzip2 [compression.ca]. Is that good enough? Their tests seem to indicate that the speedup is very close to linear up to around 30 processors.

Re:"Enthusiast Megatasking" is a lousy catchphrase

[Jeremi]

AMD is pushing multitasking, a model of parallel processing that will never do desktop users much good beyond a small handful of processors.

I'll bet you a beer that in 10 or 15 years you'll look back on the above statement and admit that you were completely wrong. You may be right that current apps, and even current types of apps, will receive limited benefit from dozens of processors, but what you're missing is that massive parallelism will enable new types of application that are barely imagined now.

Perhaps you remember the famous (apocyphal?) quote, "640K of RAM ought to be enough for anybody". I suspect that the author of that statement thought that because all apps at the time ran in 80x40 monochrome text mode, and what text-mode app could possibly need so much RAM? He didn't forsee the migration to GUI-based apps that was made practical by the availability of large amounts of RAM.

Re:

[timeOday]

Gzip, like all bit-serial encoders, is an inherently serial operation. You can't parallelize it without either attempting to speculate (which may or may not help, and in the case of gzip is probably too fine-grained to give you good performance on modern machines) or running two gzips on two halfs of your files, which will not get the same compression ratio.

This is the problem with uninspired formalism. You throw up your hands and say "impossible," when a small (and very reasonable) compromise would yei

Re:

[karlm]

This is the problem with uninspired formalism. You throw up your hands and say "impossible," when a small (and very reasonable) compromise would yeild very useful results. Here's what the pbzip2 manpage says about that issue:

There is something to be said about being overly formal. There is also something to be said about comparing apples to oranges. The GP made a statement about bit-serial compressors, and the parent cited information from a batch compression algorithm.

bzip2 is based around the Burro

My upgrade path...

[Josiah_Bradley]

I upgraded from Socket A to Socket AM2 this summer with 4x4 in mind, but now they say it's only being supported on socket 1207. I bought a nice 150$ 3800X2 planning on saving up and getting another one with this new 4x4 I have been hearing about for a while. They keep saying things are future proof, yet they go and change the socket type and then make it so you can only buy the top-end cpus for it to work. Where is the AMD of socket 939 when they had everything from the low-end to the high end totally covered. 4x4 just looks like they are taking their server/workstation tactics and trying to apply it to gamers.

Re:

[mako1138]

But X2s don't even have the extra HyperTransport links necessary to support multisocket processing. AMD charges a premium for the parts that do (Opteron 2xx, 4xx, 8xx).

And well, yeah, 4x4 is obviously aimed at people with lots of money.

Re:

[Chandon Seldon]

AMD never claimed that they would be using socket 939 for 4x4, or that any Athlon X2 processor would work. In fact, if you thought about it a little bit you'd realize that that isn't even possible... socket 939 doesn't have enough hypertransport links.

From an upgradability perspective, Socket 939 is pretty solid. The fact that they needed a new socket for their riduculus enthusiast platform isn't something that you should really be upset about.

Re:

[mrchaotica]

He didn't say 939, he said AM2.

two dual-cores?

[DreadSpoon]

I thought AMD was bragging about how their qaud-core CPUs were going to be "native," unlike Intel's which were going to just be two dual-core CPUs on one die? Or is this 4x4 platform not meant to be their real quad-core solutions, just an interim "hack" until the quad-cores come out in 2007?

Re:

[karlm]

I thought AMD was bragging about how their qaud-core CPUs were going to be "native," unlike Intel's which were going to just be two dual-core CPUs on one die? Or is this 4x4 platform not meant to be their real quad-core solutions, just an interim "hack" until the quad-cores come out in 2007?

Intel's current quad-core solution is 2 dual-core dies in a single package, not "two dual-core CPUs on one die".

In this context, my definition of "a hack" is engineering short-term solutions that don't have much ben

Re:

[Dunbal]

2 quad-core cpus in a 4x4 system

Let me be the first on slashdot to ask, "but will it run Vienna?"

AMD 4x4 - The off roading CPU

[suggsjc]

Whenever I see 4x4 I think of a truck. So, is this going to be able to "off road"?

Re:

[Awedaura]

It is not something you just dump something on. It's not a big truck. It's a series of tubes.

Re:

[Wesley Felter]

4x4 is the Hummer of computers: designed as a military-industrial complex work machine, used by posers with more money than sense to show off.

I'm finally cashing in!

[qodfathr]

As I've been the real 'quad father' since 1991 (that's the prefered pronounciation of 'qodfathr'), I'm expecting a big payday for such blatant copyright infringment!

Not as good as intels quad core?

[JustNiz]

AMD's quad soultion is two dual-core cpus, qhile Intel's is 4 cores in a single package.

TFA seems to suggest that somehow AMD' hypertransport system gives it an edge over Intel's solution, however any external bus (i.e. hypertransport) is going to be slower than package-internal interconnects.

Re:

[KitesWorld]

They're talking about access to system memory for independant applications.

Basically, if you farm out four tasks to a 2xDual intel setup, the memory bandwidth available doesn't scale. IE, you can add more dies, but at the cost of reducing the memory bandwidth available to each of those dies (to/from system).

With AMD's setup, adding a new die also adds a new memory controller (they're on the die, remember?), which in turn increases the amount of memory bandwidth available (to/from system).

It's already bieng

Re:

[Wesley Felter]

TFA seems to suggest that somehow AMD' hypertransport system gives it an edge over Intel's solution, however any external bus (i.e. hypertransport) is going to be slower than package-internal interconnects.

That might be true, but Intel has no package-internal interconnect. Kentsfield has a single FSB that connects the two processor dice and the northbridge; there is no on-package fast-path between the dice. 4x4 has twice as many memory channels and several HT links in addition.

Kentsfield may still end up fa

the obvious reaction

[illegalcortex]

*drool*

*pant*pant*pant*

*gasp*

*faint*

Having just seen Murderball...

[dr_dank]

QuadFather takes on a different meaning altogether.

Quad Father? Charming.

[BitwizeGHC]

I'm bracing myself for the gaming PCs based on this CPU setup, and sold under the name "Quad Damage"...

Re:Quad Father

[SigILL]

Was somebody at AMD a babylon 5 fan?

Nah, it's a father to stick in your mother board.

Re:

[Tim C]

Was somebody at AMD a babylon 5 fan?

I don't get it... "Quad Father" ~= "God Father"...

Re:So there's...

[Anonymous Coward]

As I was running SPECint
I met a man with 4 computers
Each computer had 4 CPUs
Each CPU had 4 cores
Each core had 4 pipelines
Pipelines, cores, CPUs, computers
How many were running SPECint?

(Answer: one, me. This guy was trying to boot Vista.)

Re:

[HiredMan]

No, you just suck.

[rimshot]

I'll be here all week. Tip your waitress....

=tkk

Re:

[Poltras]

what? you're saying you actually read it? *shiver*

Re:

[Chandon Seldon]

No... it's not just you. In fact, this comment has appeared something like five times before you said it.

And, beyond that, the comment is idiotic because adding more cores continues to be useful for quite a while - it's not a useless marketting thing like the seventh razorblade.

Re:

[Chandon Seldon]

Current games support two cores pretty well. Games currently being written are generally written to support either "up to 4 cores", "up to 8 cores", or "a whole shitload of cores".

Since core count is going up, multi-core systems is where it's at for future high performance applications.

Re:

[Ant P.]

A lot of games already work better with dual cores. Quake 2 or 3 had a SMP version. UT2004 does all its audio processing in a separate thread (and since it's using OpenAL there's a lot of processing going on).