Talk:SSE5

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

AMD's instruction sets and Intel?[edit]

Will Intel ever use SSE5, MMX+, 3DNow! and 3DNow!+ instruction sets in its processors? Please, add that info also to the main article. Urvabara (talk) 11:43, 16 February 2008 (UTC)[reply]

"128-bit"[edit]

Just so we're absolutely clear here, this instruction set extension operates on 128-bits at a time, but does not make the processors supporting it "128-bit". For example, the Pentium III is a 32-bit processor despite the fact it supports SSE instructions (which work on 128-bits of data at a time). Similarly, the Pentium 4 (prior to it supporting AMD64) supported SSE2, which also operated on 128-bits at a time. We do not call the Pentium 4 a "128-bit CPU". There's nothing wrong noting that these instruction operate on 128-bit sized data, but wording that suggests this is an extension (similar to AMD64) are just misleading. —Locke Coletc 04:36, 25 February 2008 (UTC)[reply]

I don't totally agree, your change simply discards the fact that the SSE5 instruction set operates on 128-bit operands. The processor is not 128-bit, but the SSE5 instruction set is. — Rayson 04:43, 25 February 2008 (UTC)[reply]
My change removes the text implying that this is another extension like AMD64, which it most certainly is not. —Locke Coletc 04:49, 25 February 2008 (UTC)[reply]
You should then *improve* the article rather than discard the information that it operates on 128-bit data. — Rayson 05:00, 25 February 2008 (UTC)[reply]
I did improve the article. I removed the text that gave the impression this was an actual 128-bit extension similar to AMD64, which it is not. You chose to revert war rather than try to improve the article yourself. And FWIW, I was trying to edit it when you reverted me again, and I lost the changes I'd made. Next time, instead of simply reverting, maybe consider rewording the article? —Locke Coletc 05:10, 25 February 2008 (UTC)[reply]
First of all, where do you find the definition of 64-bit or 128-bit processor?? Are you sure that your definition is the only one commonly used and it is correct?? Depending on who you talk to, the definition is different. The 8086 can address 1MB (20-bit) of memory, but operates on 16-bit operands -- can you tell me whether you define it as an 8-bit, 16-bit or 20-bit processor??
From 16-bit, "the Intel 8088 was program-compatible with the Intel 8086, and was 16-bit in that its registers were 16 bits long and arithmetic instructions, even though its external bus was 8 bits wide."
In the context of the SSE5 article, "128-bit" is the max. width of the operands supported by the instruction set, as the SSE5 registers are 128-bit long, and the SSE5 functional units operate on 128-bit data. If you are so sure that mentioning 128-bit SSE5 instruction set is wrong here, please also change all the articles in wikipedia that mention Intel 8086 as 16-bit (which you think is wrong) to your correct definition, thanks.
Otherwise, I will change it back to "128-Bit SSE5 Instruction Set", as it is in the AMD documentation. Or, may be you think the AMD engineers are wrong as well?? — Rayson 05:36, 25 February 2008 (UTC)[reply]
See 64-bit and 128-bit. The latter in particular discusses SIMD instruction set extensions such as this. FWIW, my primary concern was the language indicating this was an extension similar to AMD64 (an AMD128, was basically how I read it). That is not the case, this is just another set of extensions similar to MMX/SSE/SSE2/SSE3/SSE4. It's not an overhaul of the instruction set like AMD64 was.
BTW, WP:3RR applies to reversions only, you can still edit the article, just don't do a revert (to the same exact language). I'm more than willing to compromise on the language used right now. —Locke Coletc 06:08, 25 February 2008 (UTC)[reply]

128-bit is not a term for anything. bit is just a UNIT. From bit article, "A bit is a binary digit, taking a value of either 0 or 1. For example, the number 10010111 is 8 bits long, or in most cases, one modern PC byte. Binary digits are a basic unit of information storage and communication in digital computing and digital information theory." So it can represent the width of the registers/ALU and the memory address width and other things. So from 128-bit article, "Sony's Playstation 2 CPU Emotion Engine is advertised as a 128 bit processor. It has 128-bit SIMD registers, like many processors, but is only a 64-bit processor in the traditional sense as it can only addresses 64 bits of memory. Most high-end CPUs today have 64-bit memory address, there are only very few true 128-bit supercomputer chips." I don't see any confusions in the article referring 128-bit extension as processors have 128-bit registers/ALU but they can only address 64-bit memory. Maybe it's someone being confused about the memory address width and the register/ALU bit width, which the two are NOT related at all, but I don't see the confusion at all.

P.S.: see this in x86-64 article: "SSE instructions: The original AMD64 architecture adopted Intel's SSE and SSE2 as core instructions. SSE3 instructions were added in April 2005. ..." So I think it's correct it's a correct reference to say that it's an "instruction extension" but not "AMD128"... --202.40.157.145 (talk) 08:31, 26 February 2008 (UTC)[reply]

The current revision has the objectionable content excised. To see the article prior to my edit, click this link (diff link, so you can see the changes I made). Specifically, the introduction read as follows: "...is a new 128-bit extension to the AMD64 instruction set (itself a 64-bit extension to the 32-bit Intel x86 instruction set) ...". Which of course this isn't an extension at all like AMD64 (it is not a new operating mode ala protected mode/real mode/"long mode"). It's just more instructions like previous SSE. BTW, my problem wasn't so much with the "128-bit" as the implication the article text made (that this was a full fledged AMD128 basically). —Locke Coletc 08:42, 26 February 2008 (UTC)[reply]