AMD Sempron - Thoughts & Rumors

Hehe...good one CIWS, but everybody knows that AMD doesn't have a Marketing Dept.I'm sure they've just hung a "Marketing" sign on the janitor's closet.

No, the janitor would do a much better job. :D

There is really no cause to bitch about socket 940 at this point IMO.
Don't want an Opteron? Don't buy one, they're not made for you. In a perfect world all CPUs would still be socket 7 compatible and Celerons would be quad processor SMP capable.

Socket 754 transforming into the budget/mainstream platform that we all knew it would become seems like giving the consumer more processor for less money. OEMs don't need matched DIMMs so costs are cut and savings are passed on. A 128bit main memory bus is less "necessary" when your RAM latencies are so small.

It will be interesting to see if the L2 cache gets reduced to 256KB. The Athlon64 isn't as cache hungry as other architectures so a 32bit only 256KB L2 K8 could still be damn good performer, particularly in apps that aren't cache intensive.

Whoever came up with the name Sempron needs to be shot. My dyslexic self read it as Spermon. Hey, that actually might work. Spermon your box?

:D From the people who brought you the Duron comes the new and improved Spermon! All the protection and none of the greasy feeling!
edited to add smartassness
 
Whoa - Bob Muglia said back on the 27th of last month that 32-bit apps running on a 64-bit OS picked up an average of 8%. So the penalty for losing 64-bittitude isn't just giving up some down-the-road 64-bit apps. [Of course, they way M$ is snailing along, Desktop Win64 ain't going to appear next week, either.]
 
I'm betting the performance increase on 32bit apps in WinXP64 was due to the additional registers. Those registers are only enabled when in native 64bit mode.
So yes, we'll lose that.
 
Pheno.....


<------ Forest Trees ------->


Point being the earlier models were all overpriced junk, the gold came when it reached mobile bartons. Learning curve is a biitch, why pay for AMD's just to be "on the forefront".... Ill let you be the guy pulling back the foreskins of technology.
 
Alright, I learned a lot about Sempron last night at the tech tour but I was too tired last night so I'm posting it now. They didn't talk about Sempron as a main part of the show, but I had time to talk to AMD's most senior applications engineer, who was the main AMD technical guy at the show. He knew a lot about Sempron and was willing to tell me a lot about it. This information is completely reliable, directly from AMD.

1. The Sempron line includes, but is not limited to, the chip code-named Paris (see below)
2. Paris is a K8 based chip, with the on die memory controller
3. Paris will not have AMD64 extensions, it is 32bit only
4. Paris has 256k of cache.
5. Paris will not have the NX feature (enhanced virus protection)
6. I didn't find out about pricing
7. I couldn't get a straight answer from anybody about the name, but it was hinted that Semp might be a greek prefix (anyone here know greek?

The other big news is that Socket A is NOT dead. The Sempron line will include both the socket 754 Paris chip, and Socket A athlon XP parts re-named and included under the Sempron name. When he talked about the Socket A parts he mentioned the T-Bred core, but I didn't have time to clarify whether he meant Barton, so it is quite possible that the Sempron Socket A CPUs will be 256k cache chips as well.

So AMD is not phasing out Socket A quite yet, and the Paris chip will have quite a few features disabled as well as less cache.

That should set some of the questions to rest.
 
Yogi said:
7. I couldn't get a straight answer from anybody about the name, but it was hinted that Semp might be a greek prefix (anyone here know greek?
Actually it comes from the latin "semper" which means always [EDIT : got a confirmation from AMD France just now]. There's some sort of analogy with the Duron where the "Dur" prefix was all about duration, something that would last and how it was great value. Now it just last forever.

Only problem I see is, considering the naming, ppl will expect a lifetime guarantee ;)

PS : AMD marketing dept. has a name, it's called nVidia
 
These processors have actually been out for a while and are available in at least one line of notebooks from both hp and compaq

compaq website (athlon xp 3000+ M at 1.6 ghz) on socket 754

Link to product at circuit city with some more details about the chip

as you can see the processor has 256 kb L2 cache, is 64 bit incapable, and runs on socket 754. Oh and the market team really butchered the name on this one, as another version of this laptop that has a true athlon64 rated as 3000+ runs at 1.8 ghz has 64 bit extensions and a larger L2 cache.
 
Socrilles17 said:
as you can see the processor has 256 kb L2 cache, is 64 bit incapable, and runs on socket 754. Oh and the market team really butchered the name on this one, as another version of this laptop that has a true athlon64 rated as 3000+ runs at 1.8 ghz has 64 bit extensions and a larger L2 cache.


AMD just gets better and better at secretly launching chips.
 
Personally i dont like the name. The athlon was OK, the duron was suspect, but sempron? It sounds like a disease. Er wait! I get it!

Sem is always, and PRON!

Always pron! Go AMD! :p
 
Wixard said:
Sem is always, and PRON!

Always pron! Go AMD! :p

It all makes sense now! They've finally figured out that saying a chip will "make the internet run faster" is about as useful a marketing tactic as "free dead cat", so now they're targeting the one thing the internet just can't do fast enough. (aside from dling Linux & game demos/patches, but only broadband can help that)

Now I finally understand why they're adding some new instructions to the new "non-gaming-enthusiast home user" chip. I think it was called Sempron Processor Optimized Online Graphics Extensions.
Intel's already got their own comback in the works for the first crop of Pentium-M based Celerons though. It'll be called Celeron Universal Multimedia Instructions.
 
Any word on whether the Sempron might have some good modding potential, like enabling the 64-bit extensions/enabling more cache?

I'm wondering if they'll use an existing core, just modified for the market sector. Maybe something like the Applebreds or whatever it was where you could unlock more cache.
 
C_Wizard said:
Actually it comes from the latin "semper" which means always
Indeed... You will remember the USMC motto: "Semper Fi!" from the longer semper fidelis, which means "Always Faithful".

All I have to say is that I hope these chips are such a great value that we see an army of budget gamers chanting - "Semperon Fi, Do or DIE!"
 
0ldman said:
I'm betting the performance increase on 32bit apps in WinXP64 was due to the additional registers. Those registers are only enabled when in native 64bit mode.
So yes, we'll lose that.

The R*** registers would still be availible to system functions (like DX, OGL, GDI ect, that would be compiled for 64 bit) that are running the CPU in long mode. I presume that's where the modest improvements from 32 bit code on a 64 bit OS come from.
 
FreiDOg said:
The R*** registers would still be availible to system functions (like DX, OGL, GDI ect, that would be compiled for 64 bit) that are running the CPU in long mode. I presume that's where the modest improvements from 32 bit code on a 64 bit OS come from.
I was hoping someone would reply to that. Thats basically what I'm saying, even if it didn't come across that way.
Any further comments? You know more about it than I do...
 
Well, its possibly a k7 based core. But i would sort of doubt that. I bet its a k8 with the disabled leftovers. I do not think its going to be modable. If only because the core isnt exposed, and exposing it would be an issue in itself.

It should however benefit from the k8 speed improvements. So, if you compare a 32bit k8 to a 32 bit k7, the k8 based core should come out ahead. Everything about this chip is good, except for the name.
 
Wixard said:
Well, its possibly a k7 based core. But i would sort of doubt that. I bet its a k8 with the disabled leftovers. I do not think its going to be modable. If only because the core isnt exposed, and exposing it would be an issue in itself.

It should however benefit from the k8 speed improvements. So, if you compare a 32bit k8 to a 32 bit k7, the k8 based core should come out ahead. Everything about this chip is good, except for the name.

It is a K8 based core, I was told that specifically by AMD

Now that I think of it, I probably should have asked if it includes the extra registers, but I kinda doubt it. The registers are for 64bit mode and this 32bit chip shouldn't even boot an 64bit OS.
 
I don't understand how AMD thinks it can charge for something it's competitor gives away. Why does AMD think it can charge extra for dual-channel 939 processors when it's just standard with Intel.
 
Jonsey said:
I don't understand how AMD thinks it can charge for something it's competitor gives away. Why does AMD think it can charge extra for dual-channel 939 processors when it's just standard with Intel.
Nothing is given away by any of them.
 
wrong thread

Sorry, I really didn't mean to take the thread off topic. It just seems like Intel is really having trouble, and now's the time for AMD to really step in and gain some market share. Instead, they're content to offer extremely expensive processors and too many sockets. It just seems like AMD has a very good product, and they're squandering their chances.
 
Sempron is definitely K8 based since it has an integrated memory controller.
 
I'm going to the AMD Tech Tour tomorrow. One question I want answered is, what is the die size?

Anyone else have anything else to ask?


I since a great D2OL chip here, which is all I care about. :p
 
Just a rough list...
.13 or .09
new core or based on the K8
SSE2? SSE3?
rough clock speed range

lots more, just I know AMD won't answer them... :D
 
I wish it was 64bit but as long as its a worthy sucessor to the Duron and cheap its all good.
 
0ldman said:
new core or based on the K8

It is based on the K8. The other questions are really good though. I just assumed .13 and SSE2, but making sure can't hurt.
 
Yogi said:
It is based on the K8. The other questions are really good though. I just assumed .13 and SSE2, but making sure can't hurt.
I would guess as much too. It would go hand in, hand with a less expensive chip. The only thing I see it as, is a way to be more appealing to the OEM's, by making a good chip for less money.
64bit as a standard, is quite aways off. Or should I say having enough programs and, games for 64 will be the hold up. The 32bit will be around for a long time and, many people in the general population, are still going to be in the older OSes. Look at how many people are still using Win95,98 now. Hell I just got my cousin to up-grade to XP last week. Now Im being plagued with questions on how do you do this and, that. Which is why alot of people will stay with 32. They figure it will be more complicated. When in reality it isnt but, that perception is prevalant.
So yes doing a 32 bit 754 is smart. Now as, has been said give the Ad Dept a good enema. I want to see my stock go up more or, alot more would be nice.
 
Yogi said:
It is based on the K8. The other questions are really good though. I just assumed .13 and SSE2, but making sure can't hurt.
What I mean by that, is it a new core based around the same architecture or an Athlon64 with its balls chopped?

Loosely based or exact copy neutered?
 
Well I am sitting here at the hotel waiting for the doors to open. $10 for a daypass wireless connection!

Things we can already assume to be true:

.13 micron process. They have not gotten .09 done right just yet.
SEE2. Not going to see SEE3 until the .09 shrink for sure.
Yogi already said it was K8, since it has the integrated controller.


So I am down to clock speed and/or speed rating?
Die size?
Soocket A?
 
0ldman said:
What I mean by that, is it a new core based around the same architecture or an Athlon64 with its balls chopped?

Loosely based or exact copy neutered?

Some of each?
http://www.chip-architect.net/news/Opteron_1600x1200.jpg (waring large image ~1mb)

I link to that just to illustrate that physically removing the L2, ala Newcastle, would still be a major benifit to reducing the die size.
While returing to a 32 bit ALU saves little space, fewer transistors, and it's all in the center of the die. It doesn't make alot of sense to physically remove those parts.
 
It follows, even more after looking at the core, that just removing a portion of the L2 and disabling 64bit at the chip level would be the simplest course of action.

Another thing I'm wondering, the xbit link says there are suppose to be Sempron Socket A processors...
That, I assume, would be based on the K7 core.

I would like to know I have an upgrade path for my current system... but it doesn't seem like it would be best for AMD to do so. Moving to 754 and letting socket A die would free up the fabs for the newer lines. 5 processors identical at the core level would be much easier to produce than keeping socket A alive.
 
They would not comment much on it. I got the impression it would be Socket A though.

Nothing else answered, Frank would not say much.

O well wait and see.
 
doh

worth a try, but its early to expect any of the answers we want anyway...

added to the list, what voltage, what max voltage ;) , what fsb, locked or unlocked... hehe
 
OK, Now I have a really good question. Sempron, If it's a k8 Core, Dosen't that mean it has the onboard memmory controller? Would it be disabled for the socket A platform, there by killing off one of the biggest attractions of moving to that core?

This just has me really confused and I think AMD just threw a monkey wrench into the mix with this one, personally.

I think I'll just save the money for 939 and a 3800+, seems to me like the best choice at this moment.

I was once told by a mechanic "How fast do you wanna go? Well, that depends on how much you wanna spend." Kinda rings true with computers too.
 
0ldman said:
I would like to know I have an upgrade path for my current system...

I doubt the Sempron is going to outperform an overclocked Mobile Barton. No way. Socket A has reached its pinacle.
 
SulSeeker said:
OK, Now I have a really good question. Sempron, If it's a k8 Core, Dosen't that mean it has the onboard memmory controller? Would it be disabled for the socket A platform, there by killing off one of the biggest attractions of moving to that core?

The Sempron line will consist of two different processors. The socket A version will be similar to current socket A processors with a K7 core and no intergrated memory controller. The socket 754 version will be made from the Paris core, which has the integrated memory controller but no AMD64 or NX. Both will fall under the Sempron brand name.
 
If it's a different processor, down to the core, then how can they call it Sempron... I just don't get it. they should let socket A die.

I doubt the Sempron is going to outperform an overclocked Mobile Barton. No way. Socket A has reached its pinacle.

Yeah, I'm going to buy a mobile Barton before upgrading to the 64Bit AMD's. I think they need to figure out what socket they really wanna push before too long, or they
are going to fumble the ball with this one. Remember, Intel has a 64bit chip with AMD's extensions in the works.

Every time Intel changed a socket design, they dropped the old one like the plague... least from what I can remember.
 
SulSeeker said:
If it's a different processor, down to the core, then how can they call it Sempron... I just don't get it. they should let socket A die.
I'm sure you've heard of the Celeron... 8 different cores, 3 different architectures... I kinda get the cheap line keeping the name... if you're shooting for a cheap computer, you know a Celeron is going to be it. If you are a techie and you know of a nice Celeron (ie; Tualatin) you know which one to aim for.
Yeah, I'm going to buy a mobile Barton before upgrading to the 64Bit AMD's. I think they need to figure out what socket they really wanna push before too long, or they
are going to fumble the ball with this one. Remember, Intel has a 64bit chip with AMD's extensions in the works.

Every time Intel changed a socket design, they dropped the old one like the plague... least from what I can remember.
While I tend to agree, if AMD cranks the clock high enough on Sempron Socket A's, I'll be upgrading sooner than I planned, which equals to another sale for AMD. I already plan on building a 64bit box soon, but if I can get to 3GHz with a Sempron on socket A in the meantime, I'm there. I don't really see it eating into their sales much, it will just give me another upgrade in the meantime. If the chip is worth anything, I'm sure that will be the case for many other cheapo techs out there.
 
Yogi said:
The Sempron line will consist of two different processors. The socket A version will be similar to current socket A processors with a K7 core and no intergrated memory controller. The socket 754 version will be made from the Paris core, which has the integrated memory controller but no AMD64 or NX. Both will fall under the Sempron brand name.

This is the exact impression I got last night. Still to early but we shall see someday.
 
Back
Top