Barcelona and "SSE4a"

P

pxc

Extremely [H]
Joined
Oct 22, 2000
Messages
33,063
About 6 weeks ago, AMD updated the programming manual vol 4: http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/26568.pdf

The 4 instructions of SSE4a are (yeah, that's it):
EXTRQ
INSERTQ
MOVNTSD/MOVNTSS

While the 16 instructions added to Conroe in SSSE3 (the first S stands for supplemental) were somewhat significant, the 4 minor instructions added to Barcelona are nowhere near worthy of a whole new designation on the next number level that Intel is making with 54 new instructions in SSE4.

wikipedia said:
New SSE instructions named as SSE4a: combined mask-shift instructions (EXTRQ/INSERTQ) and scalar streaming store instructions (MOVNTSD/MOVNTSS). These instructions are not found in Intel's SSE4
Click to expand...

I guess this is part of the strategy to "destroy Intel branding" that AMD stated they intended to do. :rolleyes: IMO, this will probably backfire when it gets around that AMD's "SSE4" doesn't work since it's not SSE4 or even a subset of SSE4.
 
so intel created SSE4 and implemented those instructions
and AMD now creates SSE4a and implements the 4a instructions and does not implement the SSE4 intel instructions? yeah. Thats bullshit. Call it SSE3a or something. But wow.
 
At the same time they make SSE5 to make intel look slow/bad.

SSE4a != SSE4

I don't know if these instructions were apart of SSSE3 or SSE4 or if there just something new for AMD to piss on intel's X86 parade.

AMD is making some scummy moves if you ask me in the name of PR.
 
So when AMD does it, they're "scummy"..but when Intel does it it's market dominance?
 
Eickst said:
So when AMD does it, they're "scummy"..but when Intel does it it's market dominance?
Click to expand...
Does what exactly? Like when they create a full instruction set and give it a unique name?

Intel has a new major feature coming up (SSE4) that offers some very good benefits, especially in video encoding. AMD doesn't have it and it will take years for AMD to add it. So in order to spoil it, AMD gives 4 lame instructions a name to create confusion, and/or fraudulently mislead customers into thinking AMD supports that major enhancement. Pretty pathetic, IMO.
 
Eickst said:
So when AMD does it, they're "scummy"..but when Intel does it it's market dominance?
Click to expand...

Even if intel was perfect in every way, even if everyone who worked for them went to church and never said any bad words, with there marketshare they will still be more evil than AMD.

I stand with my previous statement of "scummy" because this is a cheapshot for AMD, and does very little damage other than to say "fuck you intel" with middle fingers in the air.
 
Ya. More instructions are bad! Damn amd for adding more!
/sarcasm

How is this scummy? It's just marketing that's been going on since day 1 of each companies existence.
 
Simpson5774 said:
and does very little damage other than to say "fuck you intel" with middle fingers in the air.
Click to expand...
Naw, I think "dead eye" Ruiz has his foot in clear line of sight, as usual. Blam!

I still think this will probably backfire.
 
SSE5 is another full load of lameness... stealing the name and not including backwards compatibility.
 
Just like how intel kept pushing clocks in netburst when it screwed customers and was only a marketing ploy. Just like they use bribery and bullying in the corporate world to screw AMD.
Besides, all AMD is doing is burning intels marketing scheme where they were going to release SSE4 and claim it makes this chips faster even though nothing or very few programs make use of it.
 
Jakalwarrior said:
Just like how intel kept pushing clocks in netburst when it screwed customers and was only a marketing ploy. Just like they use bribery and bullying in the corporate world to screw AMD.
Besides, all AMD is doing is burning intels marketing scheme where they were going to release SSE4 and claim it makes this chips faster even though nothing or very few programs make use of it.
Click to expand...

So the whole practice of selling cpu's by there frequency and not a "model number" is deceiving the marketplace?

AMD deceived the customer when it started with the whole 'number'+ when there processors could no longer match the clockspeed of intels processors and had to add an arbitrary number which meant nothing in order to compete with the frequency number of intels processors.

Is intel screwing customers by calling their quad core processors quad cores ?

AMD IS screwing customers by calling these instructions SSE4a and SSE5 when there isn't any backwards compatibility, something the genuine SSE1-4 all are.
 
Simpson5774 said:
AMD deceived the customer when it started with the whole 'number'+ when there processors could no longer match the clockspeed of intels processors and had to add an arbitrary number which meant nothing in order to compete with the frequency number of intels processors.
Click to expand...

??

COULD no longer match? There simply wasn't a NEED to have higher frequency. AMD showed the world that they could achieve better design using existing clock speed and technology while Intel kept upping their clockspeed to match AMD's performance.

Sounds to me like Intel was the one in trouble at this time frame, not AMD.

the Number+ just tells everyone that their lower clock processors are about the same speed as Intel's speed indicated by the Number+

Why would people buy a 1.8ghz AMD when they could get a 3.0ghz Intel? The Number+ tells the customer that they're NOT getting the slower processor and that clockspeed isn't everything.

Nothing scummy or underhanded about that.
 
Simpson5774 said:
So the whole practice of selling cpu's by there frequency and not a "model number" is deceiving the marketplace?

AMD deceived the customer when it started with the whole 'number'+ when there processors could no longer match the clockspeed of intels processors and had to add an arbitrary number which meant nothing in order to compete with the frequency number of intels processors.

Is intel screwing customers by calling their quad core processors quad cores ?

AMD IS screwing customers by calling these instructions SSE4a and SSE5 when there isn't any backwards compatibility, something the genuine SSE1-4 all are.
Click to expand...

They had to because intel kept pushing clocks instead of making better cores since to the common customer the consensus of the day was higher mhz = more speed. Would be like selling cars and rating them by the RPM their engine can turn instead of the horsepower. Customers go... ohhh... big number... must be better. Even though motor doing 6000rpms and making 450hp isnt any better than one doing 5000rpms making 450hp.
Amd started rating theirs on a ratio scale that actually allows you to compare one chip to another where as intel was still rating their based on a misleading technical aspect... so who was BSing.....
 
Azhar said:
??

COULD no longer match? There simply wasn't a NEED to have higher frequency. AMD showed the world that they could achieve better design using existing clock speed and technology while Intel kept upping their clockspeed to match AMD's performance.

Sounds to me like Intel was the one in trouble at this time frame, not AMD.

the Number+ just tells everyone that their lower clock processors are about the same speed as Intel's speed indicated by the Number+

Why would people buy a 1.8ghz AMD when they could get a 3.0ghz Intel? The Number+ tells the customer that they're NOT getting the slower processor and that clockspeed isn't everything.

Nothing scummy or underhanded about that.
Click to expand...

Simple, people didn't know how fast a 3000+ was, so that number in effect was useless since you had to explain to them what core it was, what speed it ran at, and how much cache it had (since if i remember correctly, they had to change the numbering systems in between thoroughbred and barton), and then again when they released AM2 A64s.
So the number actually didn't tell you anything.

Atleast with intels numbering schemes, they made sense and didn't change at random.
 
Seems like we are fighting an old war here, boys. It also seems to me someone with an axe to grind against AMD decided to come on in to the AMD forum and stir things up. Hey, here's a clue. If you are an Intel loyalist go post over on their forums and quit trying to start something in here.
We got it - you don't like AMD - well, I don't like Intel, but I've never gone over to the Intel side and posted something that would be inflammatory (or anything else for that matter.) If seeing AMD whiz in Intel's Wheeties gets your dander up then you must be a flaming fan boi IMO. Get over it and move on. I foresee a thread lock coming on this one sooner than later.
 
Why should Amd's list of SSE4 instructions match Intels huge list anyways. Its SSE4a for a reason.... Its AMDS FREAKIN SSE4!. So Intel needs a huge as set of instructions..... What if Amd feels that there fourth set of instructions doesnt need to be 54 freakin lines long!
 
Besides, Intel is baisically doing the same thing with their SSE4.1 And their SSE4.2 designations...... and you thought they were fully implementing SSE4 in one chipset? no.
SSE4.1 will be in Intels Penryn..... and later in 2008 SSE4.2 will be implemented in Intel's Nehalem...... whos creating scamming now?
 
Simpson5774 said:
Simple, people didn't know how fast a 3000+ was, so that number in effect was useless since you had to explain to them what core it was, what speed it ran at, and how much cache it had (since if i remember correctly, they had to change the numbering systems in between thoroughbred and barton), and then again when they released AM2 A64s.
So the number actually didn't tell you anything.

Atleast with intels numbering schemes, they made sense and didn't change at random.
Click to expand...

What the crap is a E4400, T7600, E6600. What does those names mean? I have no clue. How about you? :confused::rolleyes::confused:
 
Precisely. At least AMD's nomenclature was more or less used to define a relative Intel ghz speed rating. AMD cpu's did more work per clock cycle which allowed them to run their cpu's at a slower rate while equalling or besting Intel's "faster" cpu's. They needed a way for John Q. Public to compare their cpu's to Intel's. If compared by ghz alone, John Q. Public would go for the "faster" one every time so AMD just created a marketing scheme whereby people could see AMD's general overall speed. I don't see that as "scummy." Intimidating system builders and bribing them so they won't use your competitor's product to give yourself an unfair competitive advantage - now tha'ts what's "scummy."
 
besides theres six instructions.. not four

The AMD Family 10h processor has been enhanced with the following new instructions:
•LZCNT, POPCNT—Advanced Bit Manipulation (ABM) instructions operate on general purpose registers.
•MOVNTSS, MOVNTSD, EXTRQ, INSERTQ—SSE4a instructions operate on XMM registers.
Support for these instructions is implementation dependent. See the CPUID Specification, order# 25481, and the AMD64 Architecture Programmer’s Manual Updates Application Note, order# 33633, for additional information.
 
LOL at the justifications. What i've learned:

1. Selling a xGHz processor that runs at xGHz is somehow dishonest because it lagged a slower clock AMD chip. So when AMD sells chips now that lag lower clocked Intel chips, is it also scamming? Of couse not, because that argument only works one way. :rolleyes:

2. Intel's naming scheme is somehow dishonest while AMD had put the same speed rating on up to 5 different chips that all have different performance. And that's not even counting that Intel's new naming scheme most closely resembles a cross between Opteron and desktop chips. :rolleyes:

3. Intel announces in advance (2006 fall IDF) that it is splitting the 54 SSE4 instructions between 2 cores, giving 47 that covers the bulk of the acceleration to the initial 45nm chip. That is obviously the same as AMD labeling 4 junk instructions "SSE4a." Yeah right, if you need to tell yourself that. :rolleyes:

And geez, don't be so thinned skinned. AMD is pulling a scam with this and it's pretty obvious that desperate times are calling for desperate measures. Pretty sad.
 
Jakalwarrior said:
What the crap is a E4400, T7600, E6600. What does those names mean? I have no clue. How about you? :confused::rolleyes::confused:
Click to expand...

what I ment was, intel never changed the number of a product once its been produced, case in point, 2.2 ghz 512 cache socket 939 was a 3500+, intill AMD one day decided to call it the 3400+ because there was an AM2 processor, 2.2ghz 512kb cache 90nm so they call it the 3500+ then downgrade the 939, even tho nothing has changed. 3400+ = socket 754, 2.2 ghz 1mb L2 or thats what it used to be.

AMD changes those numbers around because they don't mean anything and are 100% arbitrary. Atleast i Know a E4400 2ghz, 800 fsb 2mb L2 cache allendale core, the T7600 = 2.33 ghz, 667fsb 4mb L2 merom core, and E6600 = 2.4 ghz 1066fsb 4mb L2 conroe, and I didn't have to look that up. The great thing is, those are product numbers and those products aren't going to change or be designated to another processor with different specs.

I don't have a bias against AMD, look at the sig, AMD processor, ATi graphics card (I don't like ATi now however), but they have been pissing me off as of late. They killed off socket 939 (still 100% as good in every way as AM2), now there making up there own instructions but packaging it under intels name, screwing up ATi and its marketshare with R600, and the rest of there PR bs.

Atleast SSE4.1 and 4.2 and w/e else is backwards compatable, don't blame intel for a problem they didn't cause.
 
Ace123 said:
besides theres six instructions.. not four

The AMD Family 10h processor has been enhanced with the following new instructions:
•LZCNT, POPCNT—Advanced Bit Manipulation (ABM) instructions operate on general purpose registers.
Click to expand...
Let me correct my above statement then:

That is obviously the same as AMD labeling 6 junk instructions "SSE4a."

Correction noted. :p

edit: the 4 I listed are still the only "SSE" instructions, the kind that operate SIMD. The other 2 instructions are GPR instructions, which makes their inclusion in "SSE4a" even more ridiculous.
 
Jakalwarrior said:
They had to because intel kept pushing clocks instead of making better cores since to the common customer the consensus of the day was higher mhz = more speed. Would be like selling cars and rating them by the RPM their engine can turn instead of the horsepower. Customers go... ohhh... big number... must be better. Even though motor doing 6000rpms and making 450hp isnt any better than one doing 5000rpms making 450hp.
Amd started rating theirs on a ratio scale that actually allows you to compare one chip to another where as intel was still rating their based on a misleading technical aspect... so who was BSing.....
Click to expand...

They were not lying...holding everything else constant, if you increase the speed of a CPU, performance improves. They sold CPUs that were faster than their previous CPUs, but really only because they ran faster (frequency). What's wrong with that? If a customer buys a car based on where the redline is, no one is BSing, the customer is simply an idiot. What happened is that Netburst did not scale with frequency as Intel had hoped and they basically hit a road block. Recall around the same time that AMD was heavily marketing AMD 64 to the average PC consumer. w/o a 64-bit OS, it did nothing. Even with a 64-bit OS, unless you are into heavy scientific computing or run more than 4 GB of ram...you don't need it.

I stopped paying attention to extra instruction sets after MMX which did not seem to do that much... Oh well, the more instructions the better, as long as you don't write compilers.
 
AMD sucks! No, Intel sucks! No, ATI sucks! No, Nvidia sucks! Oh yeah, well Microsoft sucks! No, Mac sucks more!

* YAWN *

They ALL juggle the numbers and specs people. And why? Because...they're CORPORATIONS that's why and they want your goddamned money!

How long have we known this?? :rolleyes: :rolleyes: :rolleyes:
 
Simpson5774 said:
AMD changes those numbers around because they don't mean anything and are 100% arbitrary. Atleast i Know a E4400 2ghz, 800 fsb 2mb L2 cache allendale core, the T7600 = 2.33 ghz, 667fsb 4mb L2 merom core, and E6600 = 2.4 ghz 1066fsb 4mb L2 conroe, and I didn't have to look that up. The great thing is, those are product numbers and those products aren't going to change or be designated to another processor with different specs.
Click to expand...

Yeah right, can you explain why the E6400 is a 2.13 GHz 1066FSB but with only 2MB cache? Intel use the name E6420 for the 4MB version but the E6600 also has 4MB cache, not 2MB and it is not named as E6620. You didn't have to look that up because you are into Intel CPUs and have remembered the names by heart just like other people who are into AMD CPUs.
 
pxc said:
About 6 weeks ago, AMD updated the programming manual vol 4: http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/26568.pdf

The 4 instructions of SSE4a are (yeah, that's it):
EXTRQ
INSERTQ
MOVNTSD/MOVNTSS

While the 16 instructions added to Conroe in SSSE3 (the first S stands for supplemental) were somewhat significant, the 4 minor instructions added to Barcelona are nowhere near worthy of a whole new designation on the next number level that Intel is making with 54 new instructions in SSE4.



I guess this is part of the strategy to "destroy Intel branding" that AMD stated they intended to do. :rolleyes: IMO, this will probably backfire when it gets around that AMD's "SSE4" doesn't work since it's not SSE4 or even a subset of SSE4.
Click to expand...



You made a fatal mistake !! You dared to speak the truth in the holy house of AMD ! :eek: No dissenters allowed in the most honoured house of Hector.
 
alg7_munif said:
You didn't have to look that up because you are into Intel CPUs and have remembered the names by heart just like other people who are into AMD CPUs.
Click to expand...

But we shouldn't buy either of them because they both SUCK. :mad:
 
A few things. How are these 4/6 instructions junk? Do they not work? Do they not make the core more effective? Has ANY SSE ever been backward compatible? I always see 1-3 listed on any given chip implying that ALL are present and none do the job of the previous. Hasn't AMD always lagged behind in the SSE department? I seem to recall that it took ~1 year to get 3 onto amd chips. And has SSE ever given a decent advantage?

As I see it AMD is giving out some extra instructions that it thought would be good to add. I assume they plan on implementing the full SSE4 whenever they get around to working it into their chips. As for SSE5, I would speculate that they are tired of being behind the curve and want to show that they have the technical capacity to further x86. That or they have had it under development for a while and just slapped a name on it after intel named the SSE4 extensions.
 
So wait, I'm confused.

Barcelona will have an instruction set called SSE4a, but that SSE4 has got nothing to do with Intel's SSE4, is not compatible, and really doesn't introduce any new "Wow!" features?

Talk about shooting themselves in the foot. I can't wait to see people wondering why the Intel guys are getting the advertised 115% performance boot from SSE4-enabled DivX, but the AMD guys don't get the boost even though they have "SSE4."

There's gotta be a law against this - using your competitor's name for your product. Microsoft went crazy when that Linux distro used the name "Lindows," and this is even more blatant of a name rip-off than Lindows was. :rolleyes:
 
Simpson5774 said:
what I ment was, intel never changed the number of a product once its been produced, case in point, 2.2 ghz 512 cache socket 939 was a 3500+, intill AMD one day decided to call it the 3400+ because there was an AM2 processor, 2.2ghz 512kb cache 90nm so they call it the 3500+ then downgrade the 939, even tho nothing has changed. 3400+ = socket 754, 2.2 ghz 1mb L2 or thats what it used to be.
Click to expand...

Yeah, I believe there was a 3400 at 2.2Ghz with 1mb L2. I have a socket 754 3400+ processor at 2.4GHz with 512kb L2. There is also a 939 processor with the same specs as mine but labeled 3800+. Like with the 3400, there are different processors that have the same rating. There was a 754 3700+ at 2.4GHz, and a 939 3700+ at 2.2Ghz I think. Wasn't there a 3200+ running at 2Ghz with 1mb L2, and then a 3200+ at 2.2GHz and 512kb L2? These ratings mean nothing it appears.

Referring to the ratings, where does "6000+" come from? It certainly doesn't mean it is compared to a 6GHz Intel chip. Of course I am not as technical as many on this board are.
 
TonyLee said:
Yeah, I believe there was a 3400 at 2.2Ghz with 1mb L2. I have a socket 754 3400+ processor at 2.4GHz with 512kb L2. There is also a 939 processor with the same specs as mine but labeled 3800+. Like with the 3400, there are different processors that have the same rating. There was a 754 3700+ at 2.4GHz, and a 939 3700+ at 2.2Ghz I think. Wasn't there a 3200+ running at 2Ghz with 1mb L2, and then a 3200+ at 2.2GHz and 512kb L2? These ratings mean nothing it appears.

Referring to the ratings, where does "6000+" come from? It certainly doesn't mean it is compared to a 6GHz Intel chip. Of course I am not as technical as many on this board are.
Click to expand...

The performance rating's haven't meant anything since the introduction of Athlon 64x2 as they don't have anything to compare to.

All we see is that +400 Points on Athlon 64x2's means 200 more MHz on the Dual Core.
 
BigMacAttack said:
I was waiting to hear from you, Manny. Go grind your axe anywhere else but here :rolleyes:
Click to expand...


I'll grind whatever axe I have to grind,where I see fit to,but I currently have no axe to grind with AMD.Still its obvious some here dont like to hear the truth,no matter who's mouth it comes out of. :D
 
I don't mind the truth as I'm not a dyed in the wool, mouth frothing AMD maniac. True, I like them but if offered the chance for a nice C2D I certainly wouldn't turn it down. I just tire of the critical attitude you bring to every thread. One can state the truth without all the vitriole and venom you seem to spill everywhere you go. Your sig says a lot about your feelings toward AMD. I'd say you're a little more than "disgruntled."
And as for the "truth" you speak of - are you sure you speak the "truth" or just your own slanted opinion? I would believe your version of the "truth" if you were some highly placed "deep throat" within the organization, rather than an enthusiast such as myself.
 
I'm always blown away at how Intel fanbios manage to say that AMD adding additional features is a bad thing. Quite amazing, actually.
 
You must log in or register to reply here.
Back
Top