Welcome Guest ( Log In | Register )

Outline · [ Standard ] · Linear+

 AMD kill itself by killing 939 so quickly

views
     
c38y50y70
post Apr 16 2007, 09:57 AM

Getting Started
**
Validating
140 posts

Joined: Dec 2005
From: R&D Center & Home



QUOTE(kmarc @ Apr 16 2007, 08:39 AM)
Ask around and see how many LYN "monkeys" would like to see AMD produce a faster 939 or a quad core 939........ You'd be surprise..... Nowadays, we still go for speed but most if not all would prefer multi-core.....
*
C'mon dude, this is a transition of technology. We can't stick to DDR forever. AM2 is a good move to prepare end user to transition from DDR to DDR2 and the same platform can be used with a quad core. There is not much gain from DDR to DDR2 for K8 because of its internal architecture. It was designed based on the BW available from DDR400 only. Like ikanayam pointed out earlier, multicores need high memory BW, DDR can never achieve that.

If the industry leaders dont push technology and forced the transition, we will not be achieving this level today, we probably are still using something as fast as Pentium only.

Besides, ppl in LYN community doesnt judge the demand world-wide. Intel and AMD do market research and get feedbacks from many customers. They need to satisfy the majority, not the minority. If we don't have money to upgrade or overhaul the system, just wait longer, stick with the older system and be more patient with the speed. smile.gif
c38y50y70
post Apr 16 2007, 06:16 PM

Getting Started
**
Validating
140 posts

Joined: Dec 2005
From: R&D Center & Home



QUOTE(kmarc @ Apr 16 2007, 05:46 PM)
I'm not saying that we are backward people who doesn't want change. New technology is definitely good for everybody. What I'm saying (which is the main topic), is that AMD stopped producing 939 CPUs too early!
*
you got your point there. IMHO, AMD stops it so quickly to have a faster transition from DDR to DDR2. This kept their manufacturing cost lower and increases the production volume because AMD doesn't need to assemble the CPUs into different packages. (as you know, AMD is suffering from high demand months ago). Besides this, not all are as educated as folks in LYN. Many people think DDR2 > DDR, and disregard Athlon64 because they uses DDR only. So the marketing people has no choice but to push DDR2 quickly. Thirdly, DDR2 consumes less power and has higher potential. This make their platform power consumption even lower and AMD can continue to advertise on their "low power platform".
c38y50y70
post Apr 16 2007, 07:17 PM

Getting Started
**
Validating
140 posts

Joined: Dec 2005
From: R&D Center & Home



QUOTE(ruffstuff @ Apr 16 2007, 08:01 PM)
This is what happened when greed is over your conscious.  At that time you want it all HT1000, Dual-channel, you have to go 939. If your are an enthusiasts but you don't have enough paper, you end up regretting yourself. The least you can do is to refer the road map.  It'll give you the idea and estimation of the product life cycle, so you can planned your upgrade. Yeah, AMD kinda suck introducing crippled socket each time. But we cannot do anything. 

Don't you guys missed those socket A days?
*
Well, please state your reason why do u think AM2 is a crippled socket? I don't really understand it.
c38y50y70
post Apr 16 2007, 08:59 PM

Getting Started
**
Validating
140 posts

Joined: Dec 2005
From: R&D Center & Home



QUOTE(ruffstuff @ Apr 16 2007, 08:28 PM)
You should know that AM2 doesn't have HT 3.0.  And it cannot be use for future AMD X4 chips(It might be, but no HT3.0 support). That you'll need a completely different socket AM2+.
*
The socket itself isn't crippled, but the platform technology. I dont think HT3.0 is very important for desktop, its real target is multiprocessor system since HTT is used mainly for inter-CPU coherency communcation. Since we have only 1 CPU on desktop, HT1.x is more than enough to for us unless you lke the power management feature on HT3. AMD already confirmed AM2 platforms are compatible with its new quadcore. I am not sure about AM2+, the socket should be the same as AM2 but the chipset and bus are different.

About your complain on fast moving sockets, both Intel and AMD has no choice because of the rat race. A few years back, a new architecture is released every 5 years (K7), 3 years (K8) but now the life cycle is cut to 2 years only (K8L onwards). That is why your socket A can last you very very long, but not with the S939 or S754. Intel is able to remain LGA775 from P4 Prescott up till now because of the same FSB protocol and memory module used. Since the interconnection busses and RAM is a severe bottleneck for multicore CPU, the best way is to constantly change to higher speed busses and memory module so that the multi-core monster can get enough data for processing.

Once upon a time my parents were complaining why Intel was introducing the CPU so quickly from Pentium onwards, because the lifecycle of 386 and 486 were really really slow. What i would say is - technology is moving faster and faster, we can't compare it to oldern days.

Why would Intel and AMD doing this? All because of us, the consumers - we demand high performance processing power and they gave us as quickly as possible. New tech always comes with great price, u cant avoid it.


Added on April 16, 2007, 9:15 pmhttp://en.wikipedia.org/wiki/Socket_AM2+

You are confusing yourself. AM2 and AM2+ is exactly the same socket. For example, it is just like a socket A running on different chipset, one VIA KT600 and one VIA KT333. Can you say socket A is crippled because the newer socket A can go up to 400MHz FSB with a lot more features, while the older one only 333MHz and probably no SATA support at all?

This is exactly the same scenario for AM2 and AM2+.

This post has been edited by c38y50y70: Apr 16 2007, 09:15 PM
c38y50y70
post Apr 16 2007, 10:05 PM

Getting Started
**
Validating
140 posts

Joined: Dec 2005
From: R&D Center & Home



QUOTE(ruffstuff @ Apr 16 2007, 10:55 PM)
I am aware of that. Transition to AM2 and AM2+ is aight. Both AM2 and AM2+ CPU can be use on both platform interchangeable.  Same as socket A era, same socket different chipset with different bus speed and features. No big deal with that.  As long as you have the choice of CPU upgrade which doesn't require you entirely to change motherboard and memory.  The sad thing about 939 is, DDR II kills it. It can't be helped. Just that I think AM2 to AM2+ is not as bad as 939 to AM2.  You just loose the HT3.0.  AM2+ might be have the same fate with 939 after the socket AM3 come into the scene with DDR III support. But still not as bad as 939 because newer AM3 cpu is backward compatible with AM2+.

The point is, you need to know the right time to purchase a system which have longer life span and upgrade choices. 754 to 939 dudes just had to suffer more. If you're true enthusiasts, you don't have to worry about this.  Just go crazy.
*
ha.. got your points now wink.gif
Yes, it is a true pain, but as i mentioned in previous post, this might be AMD's move to reduce production cost for multiple sockets. Nowadays "converge" is the word which helps to increase the efficiency and make things simpler, so a move to AM2 to unified S754 & 939 is a must i'd say.

Actually AMD can make the DDR2 version using existing 939 socket because 30 over pins on the older K8 are reserved. I am not sure why, but i guess maybe AMD is afraid of careless users who drop the old DDR K8 into a new DDR2 platform... During socket A, dropping a barton into an SDRAM platform doesn't harm because the mem controller is in the NB, but now the mem controller is in the CPU. Dunno the consequent though, but it *might* burn, hahahah. I guess the upcoming AM3 version of Barcelona will have some detection mechanism to prevent this scenario happens, when dropping the AM3 K8L into AM2+ platform.

This post has been edited by c38y50y70: Apr 16 2007, 10:06 PM
c38y50y70
post Apr 17 2007, 07:54 AM

Getting Started
**
Validating
140 posts

Joined: Dec 2005
From: R&D Center & Home



QUOTE(cks2k2 @ Apr 17 2007, 12:45 AM)
Those unused pins are usually there for mechanical purpose (strength/warp/crack) and have no signal or just grounding.
Typically they'll just change the pin-out i.e. pin A2 used to be VCC is now VSS etc. Worse thing that could happen is the machine not booting up. It has nothing to do with IMC or NB.
You got the point there. I've read somewhere (cant recall) that AMD did mention about the reserved/unused pins on their K8 S939. They said it is necessary to add an extra pin for some future designs purpose (i guess it is for Barcelona support). Well, maybe I was wrong.

QUOTE(ikanayam @ Apr 17 2007, 02:23 AM)
Yay, same socket, but you still have to change everything. Also does "reserved" necessarily mean "unused"? They could be there for some kind of testing.
775 has lasted a long time. But intel have been complete pricks about their chipsets. Practically every new chip needs a new chipset, or vrm, or just something trivial like that. I think AMD did a pretty good job with 939. It's just time to change.
Yes exactly. Socket A has so many version of chipsets and VRM. The latest AthlonXP is not guaranteed to run on much older socket A which uses SDRAM platform.

 

Change to:
| Lo-Fi Version
0.0185sec    0.85    6 queries    GZIP Disabled
Time is now: 21st December 2025 - 05:04 AM