In spite of the fact that Advanced Micro Devices unveiled its streaming SIMD instructions 5 (SSE5) recently, the technology may not be as forward looking as previously thought. Intel Corp. believes that a superset of SSE4 will satisfy all the current requirements, whereas Application Targeted Accelerators (ATAs) available in the future chips will boost performance dramatically, perhaps, without SSE5.
The introduction of SSE4 instructions along with 45nm code-named Penryn family of processors seems to be the last major update to SSE technology from Intel. The code-named Nehalem chips will feature SSE4.2, which will add several new instructions to make accelerated string and text processing more efficient. That seems to be the final set of SSE from Intel and already in the code-named Westmere processors Intel will implement its first application targeted accelerator, which will speed up AES encryption/decryption.
Application targeted accelerators extend the capabilities of Intel architecture by adding performance-optimized, low-latency, lower power fixed-function accelerators on the processor die to benefit specific applications. Such accelerators are the start of a natural evolution of adding advantageous implementations of fixed-function capabilities to the processor, Intel said. Just as the evolution of silicon technology from 65 nm to 45 nm to 32 nm will enable more transistors for additional cores and cache, so too will it also enable these fixed-function on-die implementations. The benefit will be greater performance – and superior energy efficiency – in processing specific applications.

Software makers will also have to use new instructions to take advantage of Intel’s ATAs, therefore, from software perspective ATAs hardly differ from SSE. At the same time, application specific accelerators is a natural evolution of multi-core processors: both Advanced Micro Devices and Intel Corp. said in the past that homogeneous cores of multi-core central processing units will evolve into arrays heterogeneous cores, some of which may be tailored to process certain types of data.
Intel did not say whether it plans to support SSE5 or not, however, the company may never implement the AMD-proposed set of streaming SIMD instructions in order not to confuse software makers who will have to choose whether to use SSE5 or ATAs.
Comments currently:
7
Discussion started: 09/21/07 01:44:36 AM
Latest comment: 09/24/07 03:08:30 AM
Expand all threads |
Collapse all threads
[1-5]
1.
If you listen closely... you can hear several AMD fanboys whining on the background...
[
Posted by: chips

|
Date: 09/21/07 01:44:36 AM]
2.
Seems like a terrible idea.
An instruction for accelerating *one* specific encryption algorithm?
The thing about encryption algorithms is that they tend to change. The ones used 20 years ago aren't really good enough today. And 5 or 10 years from now? We'll just have a bunch of useless instructions, suitable for implementing obsolete algorithms.
[
Posted by: grumpy

|
Date: 09/21/07 05:26:20 AM]
3.
So will it be a repeat of AMD's 64-bit instructions.. when Integrated Electronic goes something like.. *we shall not use the AMD 64 bit instructions* ;P
[
Posted by: Silver

|
Date: 09/21/07 04:39:12 PM]
4.
That is a shame that Intel is dropping SSE SIMD instructions for in place instructions to speed up special encryption/decryption algorithms. Like someone said previously, encryptions and decryptions now will be obsolete tomorrow. It is best to keep on including more SIMD instructions to increase processor efficiency. IMHO, AMD is on to something about SSE5 providing more efficiency to the end user. Present and new encryption/decryption can use future SIMD instructions to speed them up. Adding encryption/decryption instructions in ROM will be a waste of space which will rarely be used in the future.
Intel needs to re-think.
[
Posted by: linuxnerd

|
Date: 09/21/07 08:03:35 PM]
+ expand thread (2 answers)
- collapse thread
Intel has already implemented the CRC intsruction in SSE4 and specific string search/scan functions.
The ATA is a great idea in terms of optimizing speed. And when the encyption/decryption algos become obsolete in 10 years, well, the CPUs would have became obsolote a LOT before that.
Adding SSE will just confuse developpers even more. Let's face it, most applications out there are still not using any SSE at all!
[
Posted by: Berako

|
Date: 09/23/07 12:47:44 PM]
The obsolete processors are used by robotic engineers and embeds. Including encryption and decryption algorithms is a waste. SSE5 and above can be used to decrease the time the processor takes to encrypt and decrypt data. SSE can be used for future un-invented encryption and decryption algorithms. Adding encryption and decryption SIMD just makes the processor slightly bigger and more costly.
SSE is used to decrease the amount of time a processor to calculate data for video, sound, graphics, or any mathematical problem. In simpler terms, it converts a very complex algebraic formula into a simpler formula like E=mc^2.
[
Posted by: linuxnerd

|
Date: 09/23/07 02:37:32 PM]
5.
They are doing this to win benchmarks what else?
[
Posted by: Joker

|
Date: 09/24/07 03:08:30 AM]
[1-5]