Sign in with
Sign up | Sign in
Your question

What happened to upgradeability ?

Tags:
  • CPUs
  • Socket
Last response: in CPUs
Share
April 21, 2004 12:18:30 PM

Socket A, 754, 940, 939 - where will it end?

And now Tom mentions a further possibiliity beyond those 4 as well for the newer DDR2 / 3 modules.

I remember a day when A super socket 7 board would take a chip from Intel or AMD spanning 150MHz chips right through to the K6-2 550. Most of the time a simple bios upgrade would allow the new CPUs to work if your board was good enough to start with.

Now what do we have - the possibility that each CPU upgrade requires a new mobo? OK socket A did last a fairly long time but not by previous standards.

Before I get flamed, I know that major changes in CPU technology mean that they can only be implemented by changing the socket and design that it sits in, but recently building a new PC is getting difficult to recommend with an eye on even the very near future.

I have been building PCs for many years, and before I could bung in say a 900MHz Athlon in the knowledge that later on the end user could install more memory and a much faster CPU in 1 or 2 years time. But now We are facing a time when whole system guts are having to be replaced just to change the CPU - new mobo because of the socket, new memory because of Dual channel and faster DDR, new PSU to cope with the huge power draws of the latest CPUs and more recently the vidoe card as well (2 molex connectors on the 6800 geforce - crazy) - increasingly I am recommending that people just get a whole new machine. Speaking of the new Geforce, even a simple graphics card nowadays requires some users to replace their PSU and get a new motherboard to support the new voltage - even the other day I tried to put an old card in a mates machine for him only to discover that the AGP slot was 3.3 volts and not 1.5 that the GF 3 Ti200 required...


Does anyone think that this trend will continue or do you think that we will arrive at a technology that will provide stability for a while?

I think that the current situation has been caused by advances in memory technology / faster AGP sockets / CPU configurations / serial ATA / PCI express coming all around the same time.

Maybe when all motherboards have DDR 3, PCI express, serial ATA and the new socket type, we can look forward to a few more years of upgradeable PCs.

Rant over. (I have no bias towards AMD / Intel - the examples are JUST examples however, AMD platforms have lasted longer than Intel ones in recent times)

4.77MHz to 4.0GHz in 10 years. Imagine the space year 2020 :) 

More about : happened upgradeability

April 21, 2004 12:37:16 PM

Don't forget that since the inception of Socket A, Intel has gone Through Socket 370 (and there were at least 2 versions of that) -> 423 -> 478, and is now looking at 775. So AMD's socket had an excellent run.

Socket 754 was always going to be an 'interim' socket for AMD systems, much like Socket 423 was for intel. 939 looks to be around for a good while though, although DDR2 becoming more widespread might hasten its ending..

Either way, I would expect 939 to be a very viable socket for at least 3-4 years.

---
Epox 8RDA+ rev1.1 w/ Custom NB HS
XP1700+ @205x11 (~2.26Ghz), 1.575Vcore
2x256Mb Corsair PC3200LL 2-2-2-4
Sapphire 9800Pro 420/744
April 21, 2004 5:11:22 PM

I dont care about socket change. Things are, when you're ready to upgrade, there is new motherboard with new stuff and better performance. Let say that you are ready to upgrade. You just want a faster CPU. But your friend/relatives/...need a computer. Will it be easier to sell the mobo/cpu/ram at the same time to someone that dont need the latest and greatest. Would you prefer sell the MOBO/CPU/RAM at the same time than having him/her shopping for a motherboard that will be maybe faster and more efficient than your old one. Do you really want to put a fast chip on an old motherboard that might not use all what the CPU can give? does that make sense to have a Barton AthlonXP 3200+ on a kt266A mobo just because it fit the socket? Would it be better on a good new nforce2 ultra 400 that can take all what the CPU can give?

When will it end?? I hope it wont. This is EVOLUTION. PROGRESS. I'm still plenty happy with my nforce2/2500+ system. The day I wont be that happy, I will NOT put a 3200+ only because it is faster! My 2500+ is overclocked way over the 3200+. I will dump the mobo/cpu/ram and buy the best a this time.

-Always put the blame on you first, then on the hardware !!!
Related resources
April 21, 2004 6:30:20 PM

This constant platform change is how companies make money. They are always competeing for the fastest and best. My sister-in-law is still using my old SlotA 600. For just net surfing and word processing she thinks it's great. To me it's unbearably slow. To each his own I guess.

"I am become death, the destroyer of worlds. Now, let's eat!
April 21, 2004 8:07:45 PM

Quote:
This constant platform change is how companies make money.

It's wrong, technology changes require socket change. Compare this to car platform, you can change suspension, direction, options, motor and still use the same platform... But if you do this over a long period of time, your car platform will become obsolete because newer supsension will require different attach point or customer will require better handling and stiffness. And when your competitor have all newer platform, people would not buy your car, because it's technology is old and can't compete with newer more technology advance platform.

AMD needed to change socket, mainly because they now have imbedded memory controller, so they need more PINS, because, now, the CPU talk to the Chipset AND to the memory. AMD made 3 socket for purpose :

Socket940 = Dual Channel / REGISTERED DDR memory (Server/Workstation)
Socket939 = Dual Channel / NON-REG DDR memory (Mid/High-end market)
Socket754 = Single Channel / NON-REG DDR memory (Low/mid-end market)
SocketA = No memory controller (Low-end market, will disappear...)

The only non-logical use of these socket are Athlon FX-51/53, AMD wanted to push these CPU to the market as soon as possible, so they did not wait for Socket939 to be ready.

--
<b><font color=red>GO!</font color=red> <font color=blue>HABS!</font color=blue> <font color=red>GO!</font color=red></b>
a b à CPUs
April 22, 2004 2:10:09 AM

Intel got a lot of flack for changing sockets, sure. But you have to remember that there were MANY versions of Socket7 boards that had various support for processors. For example, the first Socket 5 boards supported the Pentium Non-MMX. Some had multiplier adjustments, others were fixed at 1.5x, others went only 1.5x-2.5x, and the best went from 1.5x to 3x. Now, many early Socket 7 boards were based on the same designs as the Socket 5's. Most didn't support adjustable vCore. Some others did. Some only supported 2.8v and 3.3v, others supported 2.8/2.9/3.3v, still others supported 2.5-3.3v, and finally the "Super 7" boards supported lower voltages specified by K6-2's. But my point is, AMD made processors for all these boards, and AMD had more voltage changes at that time than Intel. So you often had to choose a different than specified voltage simply to use an AMD chip on your board.

I just put an XP2000+ in an MSI K7T266-Pro2. It didn't support the Thoroughbred, so I had to modify the board for it. I thought I was getting a Palamino...but surprize, it's a Thoroughbred, had to modify the board to make it work. Use the keyword "C37" if you don't believe that.

And then there's the issues with Intel changing sockets. We already know about all that. And both companies changed bus speeds where older boards didn't support the new bus speed, that was an issue even in the 486 days.

So upgradability has ALWAYS been an issue.

<font color=blue>Only a place as big as the internet could be home to a hero as big as Crashman!</font color=blue>
<font color=red>Only a place as big as the internet could be home to an ego as large as Crashman's!</font color=red>
!