AMD working on reverse Hyper-Threading technology

Lord Banshee

Golden Member
Sep 8, 2004
1,495
0
0
Wow if AMD can pull this off it will help alot with the future of CPU Development. This would make more Cores to add performance like nothing before. I know a lot of SMP written apps that only certain parts of the application is multi-threaded where every other thing about the application works better from a faster single core.
If AMD would be able to do this it would also be smart enough to turn off and such whenSMP threads are sent so it doesn't mess up the true performance of SMP. Just think if this tech would be implement in current A64 and it was only good for 40% increase that would make a 2.6Ghz work as fast as 3.64Ghz... GRR!!!... What if it scale to Quad Core at best at 30% per core... 5.72Ghz..

I know these numbers are all made up but it fun to think if it would actually work like this!
 

allies

Platinum Member
Jun 18, 2002
2,572
0
71
Whoa... sounds amazing. Any projected dates for K10? I heard 2007, but this might have been for the K8L, which isn't the same thing I think.
 

Maximilian

Lifer
Feb 8, 2004
12,604
15
81
Wow.... pretty sweet. If thats true, ill be getting one of those! However Intel will probably have somthing similar in the works, they always do.
 

Tangerines

Senior member
Oct 20, 2005
304
0
0
That would be amazing. Just think, apps wouldn't need to be programmed to be multithreaded. The possibilities...
 

BrownTown

Diamond Member
Dec 1, 2005
5,314
1
0
just off the top of my head i can think of many reasons why this would be crazhy hard to get efficient. It is much much more promisiong to have parralelising compilers that automatically parralellise the code before the CPU ever sees it. Only problem for AMD is that if they made one they probably couldnt get anyone to use it.
 

Avalon

Diamond Member
Jul 16, 2001
7,571
178
106
Originally posted by: FelixDeKat
AMD is taking the right stance - if you cant beat 'em, copy 'em! :laugh:

Uh huh. Just like Intel copied the idea of hyper threading.
 

pacho108

Senior member
Jul 14, 2005
217
0
0
Originally posted by: FelixDeKat
AMD is taking the right stance - if you cant beat 'em, copy 'em! :laugh:

i dont see how REVERSE hyperthreading technology is a copy
 

FelixDeCat

Lifer
Aug 4, 2000
30,955
2,670
126
Originally posted by: pacho108
Originally posted by: FelixDeKat
AMD is taking the right stance - if you cant beat 'em, copy 'em! :laugh:

i dont see how REVERSE hyperthreading technology is a copy

It isnt, actually this has to do more with single threading than multithreading, but the OP used the words "hyper-threading" which was too much to pass up on. ;)
 

Avalon

Diamond Member
Jul 16, 2001
7,571
178
106
Originally posted by: FelixDeKat
Originally posted by: Avalon
Originally posted by: FelixDeKat
AMD is taking the right stance - if you cant beat 'em, copy 'em! :laugh:

Uh huh. Just like Intel copied the idea of hyper threading.

:confused:

What? I make the same face almost everytime you post. If you have any questions regarding my comments, feel free to PM me instead of raising your post count with emoticons, and I'll be happy to explain.
 

kknd1967

Senior member
Jan 11, 2006
214
0
0
since the provided info from OP is really limited, I do feel/guess there are many similarities, i.e., instead of relying purely on software parallelism to utilize multicore, exploiting both software and hardware to do so. The "reverse hyper-threading" seems to be an extreme case for the above: when there is no software parallelism hardware is up to achieving parallelism as much as possible.



Originally posted by: FelixDeKat
Originally posted by: pacho108
Originally posted by: FelixDeKat
AMD is taking the right stance - if you cant beat 'em, copy 'em! :laugh:

i dont see how REVERSE hyperthreading technology is a copy

It isnt, actually this has to do more with single threading than multithreading, but the OP used the words "hyper-threading" which was too much to pass up on. ;)

 

Dadofamunky

Platinum Member
Jan 4, 2005
2,184
0
0
Hmmmf. ***MY*** company is already developing a networking device that uses a 16-core processor on each card. Each core runs at 800 MHz.
 

RichUK

Lifer
Feb 14, 2005
10,341
678
126
Very interesting.

This is what I love about AMD, they are so innovative.

Also I wonder if AMD plan to release this new architecture on 65nm, or plan to wait for multiprocessing on 45nm.
 

n19htmare

Senior member
Jan 12, 2005
275
0
0
Originally posted by: RichUK
Very interesting.

This is what I love about AMD, they are so innovative.

Also I wonder if AMD plan to release this new architecture on 65nm, or plan to wait for multiprocessing on 45nm.

Innovative? they havn't really innovated anything in the past few years let alone Reverse hyperthreadeing (check Mitosis)
As far as this news goes, it's been on intels research list for a while. someone posted the link earlier on Mitosis. It basically does the same thing that AMD is just now trying to "innovate".

 

RichUK

Lifer
Feb 14, 2005
10,341
678
126
Originally posted by: n19htmare
Originally posted by: RichUK
Very interesting.

This is what I love about AMD, they are so innovative.

Also I wonder if AMD plan to release this new architecture on 65nm, or plan to wait for multiprocessing on 45nm.

Innovative? they havn't really innovated anything in the past few years let alone Reverse hyperthreadeing (check Mitosis)
As far as this news goes, it's been on intels research list for a while. someone posted the link earlier on Mitosis. It basically does the same thing that AMD is just now trying to "innovate".


I was referring to AMD?s ?actual? implementation of new technology, aka an on die memory controller and the use of HT links, and therefore not carrying on using inferior data transport models, such as a single I/O bus for all communication to the CPU, namely the FSB! Yes Intel may patent quite a few technologies but that doesn?t mean they go ahead and implement them. AMD are innovative as they implemented these technologies and have been leading by example, not following behind in Intel?s foot steps.

My reference to innovation is to AMD?s implementation of such current technologies on their K8?s. In point this has evolved PC, Workstation, Server subsystem?s, where as Intel hasn?t really given us anything other than higher clocks, the inevitable die shrinks and increased level 2 cache, and of course high energy consumption and heat output, all of which hasn?t resulted to much. Meaning that system builders no longer have to build platforms with the ever restricted FSB model (more so for servers) and also Netburst Architecture with its high energy consumption and heatouput. Although the latter is soon to change when Conroe arrives, even though it still uses the FSB model.

I think you are a bit confused, as being innovative in my eyes and by my definition is to actually implement something into your end product which gets produced and utilised by the end user. Not to research and develop technologies that you never end up using then discard, and therefore do not really evolve your products.

There has been no ?revolutionary? technology IMO from either company unless you want to include dual cores as an example, but then again K8?s were designed from the bottom up with dual cores in mind. In my eyes its just more evolutionary technology from AMD that has actually been implemented and been successful.

Don?t get me wrong I have not forgotten about HyperThreading and Intels mobile arena, although HyperThreading was just a substitute and short lived until Dual Cores were released.

I hope this has come across more clear.