Linked by Thom Holwerda on Tue 16th May 2006 22:43 UTC
AMD AMD confirmed details of its "Next Generation Processor Technology" today, but it's really business as usual for the company. As AMD heads to four-core country, the company will continue to improve the bandwidth of its processor package, tweak memory and rely on help from partners to compete with an upcoming line of revamped chips from Intel.
Order by: Score:
traderjb
Member since:
2006-05-16

Dual-core, quad-core, in a couple of years, we'll be talking about 10-core units. I just have two questions, dear friends, has software caught up with these new technological developments, taking advantage of these multiple cores? And are we finally going see a drop in price in this things? AMD used to be the "cheap/affordable" choice in CPUs. I walked into a computer store and saw how AMD's chips were more expensive than Intel's dual-core chips!

Reply Score: 1

happycamper Member since:
2006-01-01

has software caught up with these new technological developments, taking advantage of these multiple cores?


i can be wrong if so plese somebody out there do correct me. But think inorder to use the dual-quad core CPU. the OS needs to have a multithreaded filesytem and a SMP kernel that should do. AMD used to be the cheapest,but any more.

Edited 2006-05-16 23:35

Reply Score: 2

cerbie Member since:
2006-01-02

First, they quit the cheap stuff with the Athlon, which have regularly been up there with their Pentium counterparts, except the few times they couldn't keep up (Barton until Clawhammer). I'd love to see dual-core Semprons, though.

Yes, we will have 10+ core CPUs. Not in desktops in a couple years, but surely in servers around 5-7 years.

Software has not caught up [for general use]. But it will. Games and encoding software (including compression!) are catching up right now.

BTW, those four cores are for a server chip, where software can use its power as soon as it is released.

Reply Score: 1

Duffman Member since:
2005-11-23

"Yes, we will have 10+ core CPUs. Not in desktops in a couple years, but surely in servers around 5-7 years. "

In 5-7 years ????

Sun already have an 8 cores CPUs, I don't think it will take 5 years for them to get more than 10 cores.

Reply Score: 1

cerbie Member since:
2006-01-02

Yes, in 5-7 years. In 2-3, probably 8--but they could just make faster 4-core CPUs, too.

Those Sun CPUs are not general purpose CPUs, like an Opteron or Xeon. AMD does not appear to be making anything remotely like them, nor Intel. They're looking at general use CPUs that can handle anything you throw at them 'pretty well', not multithreading monsters like Niagara, or narrow number-crunchers like the Cell.

Multicore CPUs that we will have in desktops need to handle single-threaded apps as well as single-core CPUs (benefits coming in running many instances, and multitasking, when dealing with apps that do much multithreading). Most servers will use the same chips, or at least variants of them.

Reply Score: 1

Phillip.Fayers Member since:
2005-12-14

Software has not caught up [for general use]. But it will. Games and encoding software (including compression!) are catching up right now.

Some software may make good use of multicore CPUs, most won't. Games and a few other performance sensitive codes may start to make use of multicores but not a lot else will.

Parallel computing has been around for years but its still quite hard to turn a serial code into a parallel one, or to design an application which will work happily across a varying number of CPU cores. It'll take vendors years to rewrite their applications to make decent use of the available power so, for those of you who fancy yourselves as software authors, you have a nice window of opportunity to develop parallel apps from scratch.

Reply Score: 1

netpython Member since:
2005-07-06

I just have two questions, dear friends, has software caught up with these new technological developments, taking advantage of these multiple cores?

Are you kidding?
It's the software becoming bigger and bigger and slower at the same time that makes faster hardware a perceived necessity.On purpose or not.

Reply Score: 1

CharAznable Member since:
2005-07-06

More cores does not necessarily equal faster.

Reply Score: 1

...
by suryad on Tue 16th May 2006 23:41 UTC
suryad
Member since:
2005-07-09

I am not sure but there is already a good amount of software that is able to handle dual cores. I think XP is able to handle dual cores most of the time quite well without you havign to resort to setting affinity and so on. I am not sure but I read somewhere about Vista's kernel enhancements and it is supposed to have a better task scheduler. But yes you are still right that software besides the OS needs to be multithreaded to take advantage of all those cores. I am sure that number of software is going to increase drastically in the next few years.

Reply Score: 1

v AMD markup madness
by stormloss on Wed 17th May 2006 00:18 UTC
RE: AMD markup madness
by abraxas on Wed 17th May 2006 02:47 UTC in reply to "AMD markup madness"
abraxas Member since:
2005-07-07

AMD have lost it. Seperon what a crippled barton/thunder bird with ss2, pfft I'll have the new Celeron.

There is two versions of the sempron. There is one based on the AthlonXP and there is another one based on the AMD64 socket 754. It's a much better deal than a celeron.

http://www.amd.com/us-en/Processors/ProductInformation/0,,30_118_11...

Edited 2006-05-17 02:48

Reply Score: 1

v RE[2]: AMD markup madness
by stormloss on Wed 17th May 2006 03:20 UTC in reply to "RE: AMD markup madness"
RE[3]: AMD markup madness
by abraxas on Wed 17th May 2006 13:15 UTC in reply to "RE[2]: AMD markup madness"
abraxas Member since:
2005-07-07

Of course the informatation from Amd.com would never be wrong lol, I smell fan boys not realists.

So you're claiming that AMD is lying in their specifications? Do you have any evidence to back that up?

btw the socket 754 Seperon is still only sse2, the a-socket is worse using sse.

BTW the SSE3 instruction set is not the end all be all of performance. I would rather have an overall faster, better thermally designed, higher memory bandwith cpu than a piece of junk with SSE3.

It's funny that even without SSE2 on the AthlonXP, it still wiped the floor with Pentium4s and then when SSE3 came out and AMD64s didn't have support for it they still wiped the floor with Pentiums.

Reply Score: 2

v RE[4]: AMD markup madness
by stormloss on Wed 17th May 2006 14:24 UTC in reply to "RE[3]: AMD markup madness"
RE[5]: AMD markup madness
by abraxas on Wed 17th May 2006 14:31 UTC in reply to "RE[3]: AMD markup madness"
abraxas Member since:
2005-07-07

Semprons are cripped barton's, BTW i'm tying this on a AMD XP 2800. I've stumble into a den of blind fan boy.

Ok so you think AMD is lying about having a sempron based on the socket 754? I don't understand what you are getting at. I apologize if you are a non-native speaker but your english is horrible.

I guess you just haven't been up on CPU tech in the past few years because it is well known FACT that AMD processors are more powerful than Intel and they have better bang for the buck. I will admit that recently AMD has raised prices in comparison to Intel but that has nothing to do with the discussion because a 754 Sempron will still wipe the floor with a celeron anyday, just like it did when it was 754 AMD64 vs Pentium4.

Reply Score: 1

v RE[6]: AMD markup madness
by stormloss on Wed 17th May 2006 16:13 UTC in reply to "RE[5]: AMD markup madness"
RE[7]: AMD markup madness
by embleau on Wed 17th May 2006 16:29 UTC in reply to "RE[6]: AMD markup madness"
embleau Member since:
2005-12-05

"BTW Embleau try typing it ALL in caps, it's not loud enough! "

umm I'm sorry... but I do believe I cut and pasted from a website.. Blame them for the caps, not me. Notice that the rest is not in caps.

I'm sorry to hear that you guys get shafted on the price down under though.. ouch

Reply Score: 1

RE[7]: AMD markup madness
by abraxas on Thu 18th May 2006 01:22 UTC in reply to "RE[5]: AMD markup madness"
abraxas Member since:
2005-07-07

Abraxas your still a biased fan girl/boy (let not exclude the fairer sex, which I'm assuming Abraxas is)

In Australia (yes we speak the Queens English there)


Well, considering your assumptions about me I'm going to assume that you are around 14 years old because your english is terrible and you arguments are even worse.

Intel Celeron 331 (2.67GHz LGA775 EM64T) CPU - AUD$82.50

*AMD Sempron 3000+ 64bit (S754) - AUD$124.30


Look harder. Those prices are outrageous. Not only that but that Sempron will wipe the floor with that Celeron. It's not exactly and even comparison.

BTW Abraxas old girl, being condescending about my English is one thing ...
How about try practicing what YOU teach.
It's not "sempron" it's "Sempron" names have capital letters. omg


Forgeting a to capitalize a letter is not exactly the same as completely mispelling the name, and what's worse is that you cannot even form a full and correct sentence.

Reply Score: 2

RE[3]: AMD markup madness
by cerbie on Wed 17th May 2006 22:27 UTC in reply to "RE[2]: AMD markup madness"
cerbie Member since:
2006-01-02
RE[2]: AMD markup madness
by Trollstoi on Wed 17th May 2006 10:06 UTC in reply to "RE: AMD markup madness"
Trollstoi Member since:
2005-11-11

I wouldn't say that both are a better deal... the 462 Sempron (which I believe is not produced anymore) is in fact just a crippled Barton as the guy said. Now the 754 Sempron I have to agree is a very cool choice because of the on chip memory controller (or however is its name). If I were to assmble a new PC today, I would still buy AMD, 754 Sempron or any Athlon64, but that can change instantly if Intel starts offering better deals. I don't belong to AMD church :-)

Reply Score: 1

Underdog
by CrLf on Wed 17th May 2006 01:00 UTC
CrLf
Member since:
2006-01-03

One can tell how much AMD is being successful (in the marketplace) by the number of posts claiming Intel chips to be the best thing since sliced bread.

I mean, centrinos beat AMD's offerings on the laptop market hands down, but on the (performance) desktop and server markets the Athlon64/Opteron still come on top. And both are selling well.

AMD is still behind Intel, but Intel is starting to look like the underdog, and people love the underdog...

Reply Score: 1

Why multicore
by jamesd on Wed 17th May 2006 02:17 UTC
jamesd
Member since:
2006-01-17

Well people say that current apps don't need multicore processors. I don't know what they are using on there desktop. Ok I don't care how they slice it I hope my, word processor never turns turns into a multi-threaded monster that makes full use of a 10 core CPU. One core for text processing, one for spelling/grammar checking and a background process to print/cache my output should I be printing is the max that I ever hope to see my Word Proccessor or spreadsheet using.

On the other hand, I run more than one application on my desktop. I have a browser that thanks to java, ajax, and flash. and other extensions, it does almost constant animation and updates using up multiple threads, not to mention multiple tabs those alone make them multicore friendly. Even if you discount those saying they those are a single core. there are other applets can be take full advantage of cores such as google desktop that is constantly searching files I create. The OS that is defragmenting my harddisk. The virus checker and spyware checkers that are using up still more cores.

Those are just the typical desktop, when we move the equation to Linux the OS it self implements multiple threads to do its work. X and services are seperate threads. Along with the possibility of multiple users and possibly multiple OSes(Xen) or vitualization makes further use of multiple cores.

Should the user make use of java apps like netbeans or eclipse they allready make use of muliple cores.

Of course the typical gentoo user could always dedicate one or more cores to upgrading to the latest bells and whistles.

I'm sure that NTFS is multithreaded enough to use as many cores as you can throw at it, as is vista.

Even the latest benchmarks of Linux on the 8 core/32 thread niagara based T2000 shows that its capable of scaling to multiple cores with ease.

Solaris and ZFS goes with out mentioning its multithreaded ability since out of the box it scales to 144 core/cpus. And ZFS is even more scalible.

Reply Score: 3

RE: Why multicore
by JacobMunoz on Wed 17th May 2006 04:46 UTC in reply to "Why multicore"
JacobMunoz Member since:
2006-03-17

I know BeOS/Zeta is supposedly 'dead', but that desktop OS supports up to 16 CPUs - and makes FULL use of them. And a tastey treat is the ability to toggle individual CPUs on-and-off while running. It may not be 144 cores, but on a desktop (like you said) one hopes not to NEED 10 or more cores.

Nice to see what some 1995 technology has to offer...

Reply Score: 2

RE[2]: Why multicore
by evert on Wed 17th May 2006 05:53 UTC in reply to "RE: Why multicore"
evert Member since:
2005-07-06

> By JacobMunoz (1.69) on 2006-05-17 04:46:30 UTC in reply to ""
> I know BeOS/Zeta is supposedly 'dead', but that desktop OS supports up to 16 CPUs

Yes, it was wonderful technology. If Haiku can reproduce that, it would be great.

One of the advantages of BeOS was that the API agressively promoted multitreading - all apps written using the BeOS API were automagically multitreaded.

Reply Score: 1

Multicore
by chicobaud on Wed 17th May 2006 03:07 UTC
chicobaud
Member since:
2005-08-14

Threaded software (and 64 bits) will be a must in the next 2 years; of course a word processor doesn't benefit much.
But games and database software will; other popular software like html editors (code cleaning) and graphics (photoshop, flash) will benefit very much also.

I am an AMD fan since my k2 cpu notebook (1999). Never looked back at Intel.

Reply Score: 1

Multicore
by chicobaud on Wed 17th May 2006 03:10 UTC
chicobaud
Member since:
2005-08-14

...bugger

Edited 2006-05-17 03:11

Reply Score: 0

I love Athlons
by JacobMunoz on Wed 17th May 2006 04:25 UTC
JacobMunoz
Member since:
2006-03-17

I've been an AMD customer for a long time.

In 1990, my parents bought a 286 - which I dicovered a decade later while dismantling the machine was an AMD.

In 96, I had a K-something AMD (don't remember the model) for a little while - it got destroyed in a basement flood. ;)

In 2001, a friend of mine gave me his old Athlon 766 Mhz (which I still use to this very day).

In 2005, I got hired by 'the Man' and now work with an awe-inspiring 128 node dual-socket (256 CPU) behemoth Opteron cluster.

I admire their products - especially since Intel was supposed to put them out of business so long ago. They are a terrific set of engineers that come up with some amazing chips. The Opteron is a thing of beauty!

Reply Score: 1

kaiwai
Member since:
2005-07-06

Apart from possibly OpenBSD, name an operating system today which isn't mult-threaded and SMP capable.

Reply Score: 1

elektrik Member since:
2006-04-18

"Apart from possibly OpenBSD, name an operating system today which isn't mult-threaded and SMP capable."

In a phrase: Windows XP Home-at least for SMP (unless you count hyperthreading, but even that is a different discussion). It is (or should be?) common knowledge that the desktop/workstation versions of NT at most use only *two* processors (Indeed, XP Home only uses one). I think a valid point should be made that it isn't a matter of SMP, in the case of Windows XP; Rather it should be a matter of whether an O.S. will be able to take advantage of the multicore CPUs inside the chip. In the case of XP, if I buy the (future) Gargantuon (or whatever it will be called =])by AMD with 4 cores, XP Professional will only be able to control two, and XP Home only one....

Reply Score: 1

Sean746 Member since:
2006-05-18

XP Home will support multiple processors as long as they're on the same cpu die. I do not know if this extends to more than 2 processors on a die however.

Reply Score: 1

RE: Why multicore
by kaiwai on Wed 17th May 2006 05:41 UTC
kaiwai
Member since:
2005-07-06

On the story of Xorg and the likes; there was a move made many years ago to make a multithreaded X server, but it never actually got off the ground - the justification was that the complexity far out weight the so-called 'performance boost' it might possibly get.

With the eventual mainstream adoption of multicore processors, is this the time to revist the idea of a multithreaded X server?

Reply Score: 1

RE[2]: Why multicore
by spikeb on Wed 17th May 2006 16:22 UTC in reply to "RE: Why multicore"
spikeb Member since:
2006-01-18

yes, but the conclusion might end up beign the same

Reply Score: 1

Use of Multiple Cores
by zlynx on Wed 17th May 2006 14:26 UTC
zlynx
Member since:
2005-07-20

Software is already using it. Quake 4 and the nVidia graphics drivers both have dual-core support. Photoshop filters support it. Software build environments support it.

Any serious game in the future will support multi-cores, because the engines have to, to get the most out of the new game consoles.

Reply Score: 1

No longer Cheaper??
by embleau on Wed 17th May 2006 15:03 UTC
embleau
Member since:
2005-12-05

What do you mean no longer cheaper?

www.mwave.com is the source....

INTEL PENTIUM 4 630 3.0GHZ EM64T W/2MB CACHE 800MHZ LGA775-PIN RETAIL BOXED W/COOLING FAN (3 YEARS WARRANTY)

$172

AMD ATHLON 64 3000+ (ADA3000BPBOX) W/512KB CACHE 90NM (VENICE) 64-BIT SOCKET 939 RETAIL BOXED W/COOLING FAN (3 YEARS WARRANTY)

$111


PLUS the MoBo and RAM required for the Intel Proc is more expensive.

Even with Intel's more Cache memory, this AMD Proc Still wipes it up in benchmarks.... All because of the integrated memory controller on the die and Effective 2Ghz Hypertransport (FSB basically) ;)

My Home rig has AMDInstead and IntelInTrash

Edited 2006-05-17 15:05

Reply Score: 1

Not all are fanboys, I'm a fangirl ;P
by WereCatf on Thu 18th May 2006 02:32 UTC
WereCatf
Member since:
2006-02-15

I just came to think that I didn't notice anyone saying even single-threaded apps will gain something from these multi-cores. Sure, running a single single-threaded app on a 8-core 2ghz CPU would be as fast as running it on a 1-core 2ghz CPU, but one must remember there are constantly other task running on the system, too. The OS itself, for example, could be running on one core, the app on the other. Atleast under Linux, you don't need to do anything to run multiple single-threaded apps evenly across all the cores, and as such, all of them will have a bit more CPU-time for themselves. But yes, multi-threaded apps will have the biggest performance improvement with multi-core.

And to take a stand on the Celeron D versus Sempron:
I'd go for a Sempron any day. Why? Well, HyperTransport (similar to FSB) is lightning fast. Integrated memory controller. 64-bits, which Celeron D apparently isn't. And, I quote stormloss: "Celeron D is sse3 and in my part of world cheaper than the Seperon." Well, does this "Seperon" support 64-bits, 3DNow! or 3DNowExt? I don't think so. Besides, having support for one extension doesn't make it faster than one with support for other extensions. The price itself isn't a good reason, either. I just checked, the Sempron is about 5 euros more expensive ;)

Reply Score: 2

RE[3]: Why multicore
by kaiwai on Thu 18th May 2006 06:20 UTC
kaiwai
Member since:
2005-07-06

No quite, the issue was that it was on an SMP system, and SMP systems were not that common, and if they were, they would be only dual, hence, there would be no real performance boost.

Fastward to today, there are now quad core systems shipping, and not too far in the future we'll have dual processor, quad core (8 cores in total) shipping - the number of processors and the volume of them would justify the extra work given the must improved throughput today, given the hardware, vs. what people were up against 15-20 years ago.

Reply Score: 1