backtop

Blog: Hardware AMD May Create a New Category of Enthusiasts
Rahul Sood (Blog) - September 8, 2006 3:15 PM
Print E-mail del.icio.us 9 comment(s) - last by Regs.. on Sep 11 at 8:21 PM
As most of you are no doubt aware, AMD announced its 4x4 technology in July. Many questions have come up since the announcement regarding the potential real-world benefits for gamers and enthusiasts

I wrote this article for the most recent edition of CPU Magazine. I was on the panel at the 4x4 launch and I got the general impression that AMD could potentially create a new category of enthusiast with 4x4. They certainly have the architectural advantage and their idea is sound, but there are still some challenges.

For the last few years, AMD has transformed itself from an underdog incumbent to a mainstream alternative with big OEM support. Under the management of Hector Ruiz, AMD was able to take the performance lead over Intel without any warning, which paved the way to where the company finds itself today.

AMD had successfully fought against Intel, and against all odds AMD has gained market share while producing solid numbers. In order to achieve this considerable feat, AMD had to build a better mousetrap, and that’s just what it did—but now it faces a challenge.

In late July, a bloodied and worn Intel finally released a desktop killer. The Core 2 Duo, otherwise known as Conroe, beats AMD’s Athlons in pretty much every major benchmark. Core 2 Duo is by far the best processor on the market right now for enthusiasts, and AMD is busy working to make up the difference.

Intel created a monster with Core 2 Duo by using similar DNA as its Pentium M processors. Yes it is true that Intel has a better platform, but in terms of architecture the company is still behind AMD and Intel’s people know this. By architecture I am referring to the scalability of the company’s current product line. AMD will scale better under multiple processors thanks to its integrated memory controller and HyperTransport, but Intel’s processors are now arguably more efficient as they draw less power and produce less heat.

As most of you are no doubt aware, AMD announced its 4x4 technology in July. Many questions have come up since the announcement regarding the potential real-world benefits for gamers and enthusiasts. With Intel Kentsfield around the corner and AMD 4x4 on the way, the concept of multithreaded gaming is becoming an exciting reality. Of course it’s not all about multithreaded gaming for AMD; there are many more benefits of 4x4 which may... Click here for the rest...

Comments     Threshold


This article is over a month old, voting and posting comments is disabled
But 4x4 Is Far Future Stuff
By TomZ on 9/8/2006 3:40:50 PM , Rating: 2
I think your analysis about scalability for 4+ core machines is probably right for the server market, but not for the desktop market or for games. I don't think that quad-core games are going to come very fast. After all, dual-core in the form of SMP has been available, mainstream, and cheap for several years for desktop, and there is nearly zero market penetration there, and almost no software support either for games or other typical desktop applications.

If the benefits of multithreaded are not there relative to the investment for SMP/dual-core, then I find it hard to believe that quad-core will be, in the short-term, anything but a small niche used by the marketing folks to build expensive machines to run single-threaded games.

Therefore, Intel is right with its timing in terms of that it has a few years to rework its architecture for scalability before this is a serious threat to its desktop market share due to that reason. Of course, servers are another matter.




RE: But 4x4 Is Far Future Stuff
By Targon on 9/11/2006 6:51:58 AM , Rating: 2
Applications are starting to use a multi-threaded design, but it will take time for the industry to make the transition. The vast majority of applications out there could benefit in some way from a multi-threaded design, which many people don't seem to realize.

File I/O, such as saving and loading of files for example. The UI, sound, AI(games), pathfinding, audio. The list goes on and on what could and probably should have it's own thread if you think about it.

And once applications have a multi-threaded design, you don't need to design them around how many cores worth of CPU power a machine has since it scales well across the board.

Now, on the system level, look at Windows XP and Vista. There are so many things going on behind the scenes, from services to background processes that you will want the extra cores just to handle all that junk so it doesn't get in the way of the application you are trying to use. Don't think for a moment that your 3.2GHz P4 or Athlon 64 3200+ won't be slowed down by all the fluff in the OS, because it is.

So, more cores are a positive, though if you spend the time yourself to shut down services you don't use or need you may not see a huge benefit to single-threaded apps.


Nothing new...
By hstewarth on 9/10/2006 11:05:34 AM , Rating: 2
It just a different form of Dual Dual Core systems that are already out there - I assure there are such systems with SLI or crossfire on them.




RE: Nothing new...
By Targon on 9/11/2006 7:13:36 AM , Rating: 2
Crossfire/SLI isn't quite the same though. With graphics, you can scale a LOT further because of the nature of 3D graphics. Once graphics made the transition from working on a single pixel at a time to pixel shaders, vertex shaders, and multiple pixel pipelines, it became obvious that the more of these that you have in your GPU, the better performance would be.

In normal applications though, it's not as simple to split up an application into multiple threads and keep them all working together. That's the real difficulty if you think about it. Making seperate threads really makes sense for many things, but learning how to make these threads, update them, and then kill them when they are no longer used is the hard part.


The Programmers
By barjebus on 9/10/2006 3:37:41 PM , Rating: 2
I think alot of it has to do the with programmers. Programming multi-threaded apps is quite difficult, especially in something as complex as a game. I'm taking a concurrent programming class at university right now, and we're talking about this issue as well, the difficulties of programming such applications versus the benefits gained by making use of multiple processes.

Of course its there, but if it makes your development costs double to hire special programmers, and to also take at least twice as long or require 3 times as much stuff (time and the number of staff aren't a 1 to 1 ratio :P) doesn't often pay off.

As the idea of concurrency begins to take off more and more, and specialists begin to emerge in the field, we'll see more of it. I believe that more formal methods for approaching concurrency need to be developed, as well as application development programs either making it easier for coders to implement concurrency, or general knowledge about how to implement general concurrency in the major modern programming languages to be distributed.

Another cool idea would be a consulting company that specifically consults on multi-threaded applications and their implementations on various dual and quad core systems. I bet there's some out there, i've just never heard of them. Also, if someone were to create a mask, or veneer overtop of the dual core or quad core so as to make it look like a single processor, that would be sweet. The program would take requests and process in a manner which identifies instructions that are mutually exclusive and then does what it can to process multiple instructions at the same time by passing them to each core. All the programmer would have to do then is keep in mind to try and smush all instructions together that don't rely on each other. Anyways, just a thought.




RE: The Programmers
By Targon on 9/11/2006 7:06:07 AM , Rating: 2
You only need special programmers because the average programmer isn't familiar with multi-threaded software design. In time, it will become the norm and the extra costs will disappear.


First we need games that use dual cores
By BigFatDuck on 9/10/2006 1:16:10 AM , Rating: 1
There aren't even any games that really use dual core to give a significant performance increase. First we need devs to code 2 threads. 4 threads is far off




By mikecel79 on 9/11/2006 4:08:47 PM , Rating: 2
Once you've made your application multi-threaded it can easily expand to more than 2 cores....It's the process of making it multi-threaded that is difficult.


Pointless....
By Regs on 9/11/2006 8:21:12 PM , Rating: 2
I'll be the first to pee my pants if someone actually adopts and programs for 4x4. It will likely be MS Flight Simulator. So then I look at Crysis...then I look at MS Flight Simulator....and then my head explodes from the stupidity and lost hope AMD has for programmers to actually adopt whatever crap they have coming out in the future.

The K8 has a lot of room to imrpove. Adding another core isn't going to help much in the next 5 years.




"There's no chance that the iPhone is going to get any significant market share. No chance." -- Microsoft CEO Steve Ballmer

DailyTech Poll
Do you use copy/paste on your smartphone? 




17 Comments
Latest By Rahul Sood
NVIDIA's Quest to Become a Platform Company
June 20, 2007, 3:55 PM
Ripping the Competition a New 20"
June 11, 2007, 9:31 PM
ATI Kung Fu better than NVIDIA's?
February 12, 2007, 9:49 AM
Blind Arrogance Developmental Disorder Part 2
February 9, 2007, 7:06 PM
Blind Arrogance Developmental Disorder
February 4, 2007, 11:03 AM
Going Back to Basics
February 1, 2007, 2:39 AM
Why Start a War When You Don't Need to?
January 22, 2007, 2:01 PM
Rahul's Predictions for 007
December 28, 2006, 8:54 PM
What? Me Manipulate You?
December 2, 2006, 1:53 PM
Intel: From Winnie the Pooh to Red Bull
November 25, 2006, 3:28 PM
botimage
Copyright 2010 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki