Poll: ZIF-Socket Chipsets on Motherboards, would you pay more for them?!?!?

[chizow]

Yah, so I'm a little pissed. :| Via's recent announcements of 10 different pin-compatible drop-in replacements for their upcoming chipsets as well as nVidia's recent announcement of the "nForce 2 Ultra 400" rebranding for their C1 northbridge boards is getting out of hand.

These chipsets offer marginal improvements over their predecessors, and in many cases, simply deliver options/performance that were implied but never confirmed/delivered with older revisions. Via started the trend of sending out crippled chipsets and then releasing a revised version shortly thereafter, but this leaves a lot of people with neutered/crippled boards and no recourse other than a complete motherboard replacement. /rant off.

For the most part, nothing is changing on these boards other than the chipsets. I'm not sure if this is the case on Intel boards or not. You can replace pretty much anything in your PC by upgrading a particular component, but when it comes to a chipset revision, you're tied to your motherboard. I understand that the board makers probably need these incremental performance boosts to keep selling boards, but would you pay more up front for the flexibility of choosing/upgrading your chipset? I would....now.....on to the poll.

 

[dnoyeb]

Their is no value add. The price difference of the chipset is minimal. The additional testing required to qualify a special part (zif socket) would likely offset any savings.

I would not be interested. Nothing lasts over a year in this industry, chipsets included. A dropin chipset wouldnt mean much since the RAM speed would have increased by then.

I am all for built in video cards since I have to change them anyway when the board is maxed. Too bad the builtins are 2 years behind...

 

[AtomicDude512]

Even if you did add a chipset ZIF the data pathways would have to be fast enough for the new chipset, otherwise no go.

 

[thorin]

There is no advantage to having a socketed chipset. It's completely buyer beware, if you don't do your homework before making a purchase then too f'in bad. There's always going to be change in this market so suck it up and get what's best for you at the time...... Plus that would cause huge test/QA/compatability issues for everyone.

Thorin

 

[Mark R]

I'm trying to imagine the size and cost of a 1000+ pin ZIF socket. Compareing the relatively high cost of the chipset, and low overall cost of a motherboard, and the numerous compatability issues, I can't believe that it would be of any benefit.

 

[Chadder007]

By the way Intel keeps changing Sockets and FSB speeds for the new CPU's, why don't they just integrate them too ??.....based on some of the replies here.

 

[chizow]

Originally posted by: thorin
There is no advantage to having a socketed chipset.

Sure there is, maybe not for you, but I wouldn't mind not having to upgrade my motherboard each and every time a new standard was released, like 333MHz and 400MHz FSB processors, integrated SATA, improved APUs, USB 2.0, etc.

It's completely buyer beware, if you don't do your homework before making a purchase then too f'in bad.

You can do all the homework necessary, but when a chipset/mobo undergo quiet changes without changing the product name or specifications, how would you know what you are getting? If your homework told you that you were getting the best possible chipset today, that wouldn't do you much good if the announcement of a new CPU and chipset were being supported on a pin and feature identical version of your now obsolete board.

There's always going to be change in this market so suck it up and get what's best for you at the time......

What's really changed in the motherboard market? The only major changes are tied to the chipsets and CPU sockets. Sure there are certain features that would make the motherboard PCB itself obsolete, like single vs. dual channel DDR, but for the most part, the ATX standard has been unchanged for some time. The rest of the integrated components are supplementary features that aren't technology enablers or forces that drive sales.

Plus that would cause huge test/QA/compatability issues for everyone.

Not anymore than being able to swap out a CPU. If its drop-in pin-compatible and feature identical, you'd probably just need a new BIOS chip or simple BIOS flash. It explains why ramp-up and time to market occur so quickly once new chipsets or revisions are released.

Originally posted by: Mark R
I'm trying to imagine the size and cost of a 1000+ pin ZIF socket. Compareing the relatively high cost of the chipset, and low overall cost of a motherboard, and the numerous compatability issues, I can't believe that it would be of any benefit.

I'm not sure how much more it would cost to implement a ZIF, but perhaps using a LGA could cut down costs. I know its never going to happen, simply b/c the technology that drives motherboard sales is the chipset, which is tied to the CPU it supports; If you allow replaceable chipsets, you not only take away the only force driving sales, but any revenue from chipset sales isn't going to the boardmaker, its going to the chipset maker and CPU makers (through royalties etc.). Its still something I'd like to see however.

Chiz

 

[thorin]

but when a chipset/mobo undergo quiet changes without changing the product name or specifications

There are never quiet changes that we don't hear about. Whether the manuf. announce them or not, reviewers and enthusiast always find them. If they (the changes) really were 'that' quiet then you'd never know about them and never want them. The only issue is if you're someone that thinks that always have to have the latest greatest (in which case you'd never simply upgrade your chipset you'd get a whole new bigger/better/faster/whatever board). If you are one of those ppl that thinks they have to have the latest greatest then again it's buyer beware, that's the price you pay for trying to stay on the leading edge.

Not anymore than being able to swap out a CPU. If its drop-in pin-compatible and feature identical

If it's feature identical then why would you ever change it?

Thorin

 

[chizow]

Originally posted by: thorin
There are never quiet changes that we don't hear about. Whether the manuf. announce them or not, reviewers and enthusiast always find them. If they (the changes) really were 'that' quiet then you'd never know about them and never want them.

That's what a quiet change is, an unannounced or secretive change, not an irrelavent or undiscovered change. Sure, inquiring minds will always find the differences, but only after they've been released and only after testing or investigation (like examining the chip itself). The reality is that most people would never know the difference until they found out their board was obsolete or incompatible when they go to upgrade. Look at all the "will my motherboard support XYZ CPU" posts in the mobo or CPU forum. What's the typical response? "What board revision is it?". Yah, I think you would want to know about any quiet changes at that point. 400MHz support was implied during early nForce 2 reviews, yet when 400MHz Bartons are released, I guarantee the motherboard forum will be flooded with compatibility issues. What's the real difference between a revision 2.0 and a revision 1.2 board that sells for the same price and specificiations? The memory controller in the northbridge and a different BIOS image.

The only issue is if you're someone that thinks that always have to have the latest greatest (in which case you'd never simply upgrade your chipset you'd get a whole new bigger/better/faster/whatever board). If you are one of those ppl that thinks they have to have the latest greatest then again it's buyer beware, that's the price you pay for trying to stay on the leading edge.

That makes no sense. Being on the leading edge of technology or the tail end of obsolescence doesn't make any difference; at some point the part will become obsolete. The ability to interchange chipsets delays obsolescence equally, whether you are on the leading or tail end of the the technology curve. The majority of motherboard features are static and undergo relatively few changes; the driving reason behind upgrading lies solely in the chipsets and socket types. Until major changes such as integrated SATA, PCI-X, DDR-II, revised AGP specs etc. come about, the rest of the motherboard features will continue to evolve slowly.

If it's feature identical then why would you ever change it?

Feature identical in that all current features will still be enabled. Improving on an existing feature w/out the necessary motherboard ICs would render them useless. You'd still need the proper outputs, traces, inputs, ICs etc to take advantage of the improved features of the board. For instance, if a new southbridge implemented USB 2.0 vs USB 2.1, you'd still need the USB headers on the mobo and backplate. If you integrated a new APU, you'd still need a codec chip and the proper audio outputs to be able to take advantage of it. Implementing 400MHz Dual-DDR support on a board that only had pins and traces for single-channel wouldn't do much good either.

Chiz

 

[CZroe]

I doubt the high-speed buses could make it though a socket

 

[McCarthy]

For a motherboard, no. But I've been wondering this about video cards since Jesus was a boy. At least in the old days video card memory was socketed so you could use it somewhere else, but no more. But alas, it's not feasible there and even less so on a motherboard.

 

[oldfart]

Hmmm....maybe you could remove the VIA mobo chips and install Intel? That would be worth it!!

 

[thorin]

400MHz support was implied during early nForce 2 reviews

Trusting something that's "implied" in a review is not doing your homework. If the manuf spec doesn't say "400MHz Bus Support" then you email the manuf before you purchase, why is this so hard to understand?

Look at all the "will my motherboard support XYZ CPU" posts in the mobo or CPU forum. What's the typical response? "What board revision is it?". Yah, I think you would want to know about any quiet changes at that point.

As I already said that's the cost of trying to be on the leading edge. If you 'have' to be one of the first to get XYZ mobo then you have to live with what it can do. The extra cost associated with allowing a socketable chipset are not balanced by the 0.001% of the population who changes their CPU every 6 (or less) months just to go from 333 to 400Mhz, plus the fact that 50% or more of that 0.001% would want alot more featues and do a full upgrade.

What's the real difference between a revision 2.0 and a revision 1.2 board that sells for the same price and specificiations? The memory controller in the northbridge and a different BIOS image.

And board layout, board components, and manufacturing process, the testing. ALOT is different. (MoBo revision is not the same thing as chipset revision)

the driving reason behind upgrading lies solely in the chipsets and socket types.

The majority of users DO NOT upgrade solely for chipsets and socket types. If you really believe that I'm sorry and I'm not trying to be an a$$hole or anything but you have NFC about the computer industry if you really believe that people upgrade for chipset revisions and socket types.

Feature identical in that all current features will still be enabled. Improving on an existing feature w/out the necessary motherboard ICs would render them useless. You'd still need the proper outputs, traces, inputs, ICs etc to take advantage of the improved features of the board. For instance, if a new southbridge implemented USB 2.0 vs USB 2.1, you'd still need the USB headers on the mobo and backplate. If you integrated a new APU, you'd still need a codec chip and the proper audio outputs to be able to take advantage of it. Implementing 400MHz Dual-DDR support on a board that only had pins and traces for single-channel wouldn't do much good either.

Again the vast majority of computer users do not care about changes that occur between 1.0 of a chipset and 1.1. And the majority of enthusiasts want the new features (ie: USB 2.1, SATA, etc...) when they upgrade not just some kludgy fix that 'might' be present in rev 1.1 of a chipset that isn't in rev 1.0. So for < 0.001% of the computer using population you want to increase cost of motherboards by adding a socket (or two), extending Testing periods, increasing chance of incompatability, etc etc.....

Thorin

 

[sharkeeper]

Take a look at some spec sheets on the different chipset revisions themselves. Sure you can socket the chipset, but then you just have as an obsolete board as you did before with the added cost of the socket. :Q You cannot change signal paths once the board is made.

This hasn't happened and never will. Mainboards are a throwaway item.

-DAK-

 

[vss1980]

In my opinion it will never happen for the following reasons:
1)Cost of implementing such a system..... and I dont just mean of extra socket components, I mean redesign to be able to handle the signal degradation characteristics associated with such components, etc.
2)The fact that CPU manufacturers move goalposts every so often, so in many cases even if the chipset supports it, the power regulators become out-dated and unable to cope, the board tracks would hence not be good enough, etc.
3)Extra development costs related to having to support newer and older motherboards which could possibly accept the chip.

I think people see 'pin-compatible' and get the wrong idea...... this isn't done so that the chip gets dropped into the old board design and opens up new avenues.... at best it would only work as well as what it replaced (or a little better) because the board design itself would not be up to support the extra speed of the signals, or the stronger drive needed for higher power faster chips.
The reason for pin compatibility is to aid the board makers so that when a new chipset comes out, they are already familiar with what needs to go where and design guides to follow.... I think companies such as Asus/MSI/Gigabyte would take an amazingly dim view of a set of chipset chips all using completely different pin layouts.

The other reason to maintain 'pin-compatible' layout is because the connecting points under the chip and on the motherboard are broke into regions..... this is mainly for noise and signal integrity and to 'zone' (if you will) the similar signal paths together..... it would be an amazing error in design to have say a 5V PCI bus signal run right along a 1.5V 100/133/166/200 MHz CPU bus frequency due to the possible noise problems. This idea has been refined and works, so why change it?

I'm all for getting as many options and as much out of something as possible, but if you think some motherboards have stability issues now then this idea is asking for trouble. Sorry, I have to pick 'nah, dont care that much' in the poll because there isn't an option that says 'such an idea is not really an option'.

Edit: And yes, such things are annoying when they happen...... wouldn't liked to have been a KT266 owner when the KT266A came out.

 

[chizow]

Originally posted by: thorin
Trusting something that's "implied" in a review is not doing your homework. If the manuf spec doesn't say "400MHz Bus Support" then you email the manuf before you purchase, why is this so hard to understand?

And I suppose you're supposed to e-mail them and ask if you should wait for a new northbridge revision that fixes memory controller bugs. Or ask them for a roadmap of every revision they plan to issue in the next 6-12 months. I'm sure they'd tell you if you did your "homework" and e-mailed them, considering they don't even release such changes when they make them.

What you still don't seem to understand is that even if you choose to be on the bleeding edge of technology, you upgrade for the best performance possible at that time with upgradeability in mind. The last board I owned before my "bleeding edge" upgrade to an nForce2 board was a KT266A, simply b/c I didn't see any tangible performance benefits in the subsequently released boards. Oh and what I meant by "implied" was the numerous quotes of reviewers asking nVidia whether or not nForce2 would support 400MHz FSB. The answer was of course they would, they didn't say you needed board Rev. XYZ or stepping ZYX however.

As I already said that's the cost of trying to be on the leading edge. If you 'have' to be one of the first to get XYZ mobo then you have to live with what it can do. The extra cost associated with allowing a socketable chipset are not balanced by the 0.001% of the population who changes their CPU every 6 (or less) months just to go from 333 to 400Mhz, plus the fact that 50% or more of that 0.001% would want alot more featues and do a full upgrade.

Cost is a non-issue, as its a hypothetical poll. Its not due to lack of demand or incremental costs that will prevent it from happening, its the simple fact that allowing for socketable chipsets removes one of 2 things that drives motherboard sales, the chipset (the CPU interface being the other).

And board layout, board components, and manufacturing process, the testing. ALOT is different. (MoBo revision is not the same thing as chipset revision)

Just shows how far off the bleeding edge you are I suppose. You need to read some reviews and release notes about current Via and nForce2 motherboards, as they are IDENTICAL, other than the north and south bridges. The ICs and traces are the same, wtf do you think pin and feature identical means? Look at KT266 vs KT266A boards or KT400 and the "new" KT400A when they hit the market. I just received my A7N8X-DLX Rev. 2.0 RMA and its once again IDENTICAL to my Rev 1.4 other than newer dates on some ICs, a different BIOS chip and of course, the upgraded north and south bridges.

The majority of users DO NOT upgrade solely for chipsets and socket types. If you really believe that I'm sorry and I'm not trying to be an a$$hole or anything but you have NFC about the computer industry if you really believe that people upgrade for chipset revisions and socket types.

LoL, you might not be trying, but it just might be your nature. Once again I ask, what "ground-breaking" changes in the motherboard industry over the last year other than socket and chipset upgrades would compel someone to upgrade their motherboard? 8X AGP? USB 2.0? Integrated LAN? SATA RAID controllers? People upgrade for performance and compatibility which are 100% tied into the chipset and CPU support, plain and simple. Not to be an a$$hole (although I admittedly can be), but you've got NFC if you think people upgrade for all the supplementary features on a mainboard. Yeah, you could design a Slot 1 uber-mobo with PCI-X, integrated SATA, USB 2.1, integrated Firewire, dual-channel DDR that accepts PC66 SDR RAM and runs on a 66MHz bus and I'm sure you'll make a killing for all those who buy mainboards for the reasons you cite. Who's going to spend $100+ on a new mainboard for USB 2.1 when they can buy a $10 AIB?

Again the vast majority of computer users do not care about changes that occur between 1.0 of a chipset and 1.1. And the majority of enthusiasts want the new features (ie: USB 2.1, SATA, etc...) when they upgrade not just some kludgy fix that 'might' be present in rev 1.1 of a chipset that isn't in rev 1.0. So for < 0.001% of the computer using population you want to increase cost of motherboards by adding a socket (or two), extending Testing periods, increasing chance of incompatability, etc etc.....

All those features you list will be increasingly chipset dependent in the future as they continue to be integrated into the southbridge, so I'm not sure what point you're trying to make. The vast majority of computer users simply won't know of a revision until there's a problem, but I'm sure they'll care when it comes time to upgrade. When they do, they'll just get a "sorry, your board revision doesn't support that CPU, but all the others do." I'm sure the vast majority of KT266 users were pleased to find out a much improved KT266A was released 2 months later. Or that there's been multiple revisions for nForce 2 boards in the last 6 months since release. As for the majority of enthusiasts, thats obviously not the case judging from the poll results, not to mention real enthusiasts know that features like AGP 8X, USB 2.1, non-integrated SATA are Fool's Gold.

Chiz

 

[Athlon4all]

Such an idea does sound like a good idea, but in reality as others probably have said, is impractical. There are other parts of the motherboard itself which need to be changed in order to be able to "drop in" a new chipset revision.

 

[vss1980]

Actually Chiz I would say real-life is a compromise between you and thorin......

look at it like this:

Assume someone came out with 2 identical spec and component motherboards, except that one has non-removable chipset where as the other one can be removed. Identical processor support, RAM options, USB, ATA, the lot.....
Now lets say the price is $100 for the non-upgradable board, and I'll be generous and say that the price is $150 for the upgradable board..... which will you buy?

Now lets keep in mind how the market has moved and will continue to. Lets say these boards can only support a 166MHz FSB and RAM speed properly, however in 6-8 months time there will be a new chipset out that supports 200MHz FSB and 200MHz DDR RAM, supports a 20% larger power window for up and coming processor cores, has integrated SATA RAID and a few other goodies and boards will be produced in an upgradable and non-upgradable design with identical cost as before.....
There is a drawback though.... anyone who upgrades their old chipset to this chipset cant use the SATA, will not have faster DDR RAM or 200MHz bus speed support, or able to take the newer processor cores (which will mainly be designed for the newer bus speeds). Also, if the upgrade would cost $66 say, who would want to upgrade?

Now which will you buy?

I agree to a point that the 'toys' you get with the motherboard (such as SATA) doesn't effect many enthusiasts, but what one person doesn't deem important is not the same as another. Some people will have a need for USB2 or firewire for example where others may have dismissed them, but at the end of the day these toys are on the motherboards to swing large orders of boards from companies such as (heaven forbid) Packard Bell (eewwww, feel all dirty now I said that). The enthusiast market has a little bit of power in those terms, but if you were Asus and got 10,000 emails from people saying we want ABC on your next board and 1 email from Dell saying we want XYZ on your next board for an order of 100,000 what do you think would happen?
XYZ would definitely be there, with a hint of Z, and maybe Y in the next revision.....

 

[chizow]

Good points vss1980, but I think we disagree on a few fundamental aspects of your analogy. I've been following the AMD side of mainboards closely since I upgraded to Socket A from a PIII, and the platform improvements have been less than spectacular. The trend clearly follows AMD's few changes to their Athlon core and FSB support. So what do mainboard/chipset makers do to spur sales? They release marginal improvements through chipset releases along with some bells and whistles for show.

Only recently, AMD has actually supported higher FSB speeds natively in their chips (2600+ XP's in Oct/Nov.), but there have been 333MHz FSB boards for some time (KT333). Now that AMD is rumored to be launching a "new" 400MHz Barton, the bar is raised again ever so slightly. I also follow the OC'ing scene closely, and its my observation that these incremental speed bumps in FSB support are tied directly to the North Bridge memory controllers. OC'ers have seen for some time that any given generation of a NB will stably support higher than rated FSB speeds. Each time the official standard is raised, these memory controllers need to be "tweaked" for stability. What you end up getting is the multiple revisions and marginal improvements in chipset designations. The only major breakthrough in North Bridge controllers in the past year has been the implementation of dual-channel DDR in nForce2 boards.

The same can be said for the South Bridge. Marginal improvements are implemented through controllers in the SB. If there isn't any improvement needed, the SB will remain unchanged (KT266 to KT266A to KT333). Look at the nForce2's MCP-T for example. The differences between a "Deluxe" or "standard" board is directly tied into the South Bridge used; MCP-T includes the DD APU, integrated firewire, dual MACs, USB 2.0, IDE controllers etc. I agree that the ICs on the board supporting upgraded features are necessary, much moreso than for the NB, but in the examples I give, either the South Bridge goes unchanged or simply doesn't offer any benefit to justify the upgrade.

From my personal observations as well as reviewer comments, I don't think there's as much difference between revisions as people think. As long as the traces and ICs are identical, the only thing that would affect higher frequency support would be power regulation and the number of physical layers required on the PCB itself. To some degree, power regulation is limited to ICs, caps, mosfets etc. on the mainboard, but I still feel that BIOS support is the key factor. For instance, you can put in different Athlons requiring different voltages to operate at stock settings, and you'll get a different window of adjustments. A 1.5V default chip will have a much lower threshold, whereas a 1.65V chip will have a higher threshold. Some boards in the past have allowed for higher operating voltages (over 2V!! :Q), and the IC's themselves are clearly capable of operating at higher voltages as well (look at all the Vmods that can be done). Also, faster chips that were not supported in the past b/c they might have required higher default operating voltages often need a simple BIOS update to become supported.

Again, I agree that any integrated chipset features that would require IC, trace, or PCB changes would not be supported, but if you look at different nForce 2 board changes, they are for the most part, identical. The same can be said between Via boards with A designations, which explains how A revisions ramp up so quickly and how the previous revisions simply vanish from the market. Any increases in FSB speeds or CPU support would be realized (as long as the socket/CPU interface was the same), as the only difference would be the memory controllers supporting the next higher FSB increment.

Using your comparison, it would be nearly a wash in cost by upgrading your chipset vs. buying 2 boards for the same level of performance. The major difference though would be the flexibility in deciding when you wanted or needed to upgrade and the timing of your cash outlay. It would also give you more flexibility to decide if any cosmetic improvements like non-integrated SATA or AGP 8X is worth the upgrade.

Chiz

 

[vss1980]

Well, I was giving a generic example which kinda parodies real-life and one which represented a significant step in terms of FSB and RAM speed.

In the cases of small steps like those given by the KT333->KT400->KT400A, I think these are mainly forced about by the need to be on the leading edge. The onus is on the chipset makers to incorporate any new little change albeit USB2 or official JEDEC certified 200MHz DDR RAM.
As I tried to get across in my post before last..... 'pin-compatible' doesn't mean identical and even though they are compatible and may carry identical signalling, etc., it doesn't mean electrically-identical and this is where the redesigns come.

To make successive chipsets compatible with older boards would mean having to make it compatible with such things. This ties in with my other point, an issue of cost/difficulty... it will eventually just become a pain in the a$$.
Practically the original Thunderbird/Athlon (and to a lesser extent Spitfire/Duron) chip could have been made in Slot versions (and a few were), but these were all incompatible with the leading chipsets at the time (bar AMD's), not because of signalling issues or BIOS support, but due to electrical compatibility. Although this example doesn't directly apply, the idea behind it does.

As I said, 'pin-compatible' basically translates to 'No major redesign needed' as far as the board makers see it, but there are always a few 'minor' changes that happen.

As I said, its not the idea of an interchangable chipset I have a problem with, its the practicality.