• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

AMD's Secret Weapon

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Molester

Member
Joined
Jul 8, 2003
Location
Dallas, Tx
http://theinquirer.net/?article=32589
TheInquirer said:
THE LAUNCH OF Intel Conroe is scheduled to go on July 23rd, when partners will start rolling out the new products. The NDA for reviews is scheduled to expire around July 28th.

However, AMD has a counter-attack weapon hidden in its Socket AM2 infrastructure.

It seems that all AM2 CPUs were outfitted with a support for Reverse-HyperThreading, an architectural change which enables software to think that it is working on a single-core alone. By combining two cores, the company has been able to produce the six IPC "core" that will go head to head against four IPC "core" from Conroe/Merom/WoodCrest combo.
I know, I know, it's the Inquirer, but it's looking to be true.
 
Last edited:
Typically AMD,

if that is true, Conroe will at least have a much harder competition than thought!
 
Yes, thats all good and great. However, single core software will benefit but today is a time when all software is being developed and adapted to multicore and SMP. So, in theory AMD has a little shock and awe going with this Reverse Hyperthreading Technology but if all newer software is tailered to take advantage of multicore processing I dont see how this would benefit in the long run. Maybe I am wrong?

PS- AMD Fans, look in my signature.:beer:
 
Molester said:
Why would software developers add all that code if the cpu can divide it up itself?

At the introduction of Dual Core CPU's it was thought as a look toward the future and an investment. Where we wouldnt achieve the uber speed increase right away. We would need software to take advantage of multicore technology. Thats the only point I am making here.

I guess your point is "if the CPU can simply split single threads without having software actually plot it out to benefit performance wise" This would definitely be awesome and I agree. But if this is the case than why would everyone be tailoring software to take advantage of multicores today?
 
Well it would be sweet if this were true, which its likely not to be.

However if it is, then its garentied that Intel is working on a version of thier own.

The article also states that RHT would be turned off for non single thread programs.
 
This sound good. If this technology succeeds, then it would scale amazingly, with even more processors added to the core.

Another article said that AMD appears to be flushing out the product channel of the 939, to start pushing the AM2. Which they might feel that 939 is no longer a viable product with Conroe coming.

http://theinq.net/?article=32603
 
Last edited:
Intel will probably also have a version of this shortly. It's not hard to implement and would be pointless if programmers did their jobs correctly. A properly written dual core driver, such as the ones by ATi and nVidia, would accomplish the same thing. This just does it at a lower level. All in all its one of those things that'll make some synthetic benchies look nice. It'll only help single-threaded applications, and the main reason that most purchase multi core processors is for multitasking.
 
how did people miss this on the front page:? its already been spoken of reverse hyper threading, and how its not going to do much to match conroe
not trying to be flamish or intel/amd but... if its justa little chip gimmick you can be damn sure all intel has to do is do the same thing, then they still rule
in cpu performance. and if its just close ? then theyll fly miles ahead.
 
This is not the first time I have heard this rumor. I have never seen a credible source for it, but you do have to wonder why it keeps popping up if there is not some truth at least to the fact that it is in development.

Gautam said:
and the main reason that most purchase multi core processors is for multitasking.

Actually I think the reason that most people buy dual core CPUs is that it is new and must be better. Oh and 2 > 1. Informed people on this site might not buy it for this reason (although I have seen plenty of posts that want to do multiple non cpu intensive tasks at once and think that the multicore cpu will be the magic answer to make their internet connection and CD burner faster).

For the vast majority of the public that is buying dual core CPUs, this would be very helpful. Developers are certainly writing software with multi-threading in mind, but as a developer myself I can tell you that writing multi-threaded applications properly is HARD. Keeping your threads in sync can be difficult and buggy. Some applications are just not suited to this type of work. If you can slice up a task into pieces and those pieces can be worked on seperatly (example: encoding a movie, you can just encode different portions) then multiple threads works fine. It works much better when the work can be quantified before it begins. In the example of the movie, you know when you start example how much work there is to be done.

Other applications that take place in real time like gaming, transactional database work etc, all are more difficult. I would not expect the majority of software to be multi-core friendly for a long long time. Also you have billions of dollars of software written by corporations that handle their warehousing, accounts payable etc etc. Nobody is going to go back and rewrite all of that software, but if a magic new technology plus a new driver suddenly extends the life of some servers that can be a huge cost savings.
 
I'll grant you that many buy dual cores for no reason...myself included, I get them just because they're high end. I don't think I'm most people though.

Also you have billions of dollars of software written by corporations that handle their warehousing, accounts payable etc etc
Most of this sort of software is implemented on multi-way systems as it is and unless I'm way off base is already, and has been multithreaded for many years. SMP has been around forever, it's just recently that its gone mainstream, as I'm sure you know...just putting that out there.

I don't like the precedent that this little tidbit of technology will set, but it looks like both AMD and Intel will indeed adopt it. Makes enough sense anyways I guess.
 
I think this is just a temporary solution to draw more people to dual core/AMD until more the majority of programs take advantage of it.

If AMD can demonstrate that its processors are faster in productivity applications, many companies might buy them.
 
Well I am excited about this; there are plenty of applications which are still single threaded. In particular Matlab, which I will be utilizing soon to solve some design optimization problems through brute force computation.
 
dlavrenz said:
I think this is just a temporary solution to draw more people to dual core/AMD until more the majority of programs take advantage of it.

If AMD can demonstrate that its processors are faster in productivity applications, many companies might buy them.
AMD has been the frontrunner in the enterprise arena for at least the past four years, and neither Conroe nor "reverse HT" is going to change that.
 
^^^^ Agreed the place where amd really has the lead is with HT and 4 and 8 way opteron servers.They will still have the lead there and will for some time to come.

As far as AMD's secret weapon goes intel already has the same thing, its already in the latest beta bios for the Bad Axe intel board, called multiplexing. I don't forsee either company gaining an advantage since they both will have the same technology released at the same time. AMD will likely make some noise about it and try and steal some of the spotlight from Intel, which is a pretty good marketing strategy.
 
Last edited:
Back