Intel Deadends

Maybe it is just my perception , but too many of Intel's advances , even small ones require entire system changes . When I bought my current rig HT was about to come out and I couldn't afford to buy a board and then find out later that it was a deadend for upgrades . At that time the T'bred 2100+ chips weren't out yet and the northwood 1.6A and 1.8A were among the best oc'ers on just air . But I chose the AMD Nforce 2 route which has kept me current for about 9 months ( by AMD standards ) , at a moment's notice I can drop in the latest and greatest AMD cpu or oc my 2600 to those speeds .

What kept me away from Intel last year ( deadends and price ) will likely keep me away again late this year or early next year when I'd like to upgrade . Too many chips that need a totally different board/voltage setup . In recent memory the P4 series has just done too much wholesale changes, buy a highend system this month only to find out 2 months later that it cannot be upgraded . Just look at some of them :

1/ Socket 423 ( at least people knew about this one early )
2/ Socket 478 at 400 FSB
3/ Socket 478 with 533 FSB
4/ Socket 478 for the 3.06 , HT and new voltages
5/ Socket 478 for 800 FSB

6/ Now it looks like there will be a new version of socket 478 for the 478 Prescotts .

7/ And guess what , if you buy a 'high end ' rig now , there may well be no upgrades beyond the 3.2 Northwood ( unless you call the D1 Celerons upgrades ).

8/ If you wait a few months and get a Prescott and one of the new boards that is supposed to support it good for you . But just like the 423 to 478 switch , the 478 Prescott is likely to be a 6-7 month deadend at most . 478 Prescotts are not supposed to last too far into 2004 and will make a switch to a totally new pin layout . I just hope that this is the same time as the shift to the .09 process or that may be another deadend in the making .

I forgot to mention the changes in Xeons ; pincount and HT and soon 800 FSB .

For some of you this may not matter but it was brought home to me by this latest Prescott story as Intel waited until a host of 865and 875 and granite bay boards have been sold to now mention their incompatibility . That is exactly the reason that kept me back last year .

Some may mention that different manufacturers got around some of the things I listed with bios releases and so on . But the problem is that it makes buying risky . You are forced to guess which manufacturers will be able to get beyond the deadend with bioses in a few months time . A very risky buisness indeed .

/end rant

I don't think everyone's worries are valid. Everyone knows cpus come out faster every 6-12 months. You really can't count on upgrading to the latest and greatest every 6-12 months. Most systems will last a few years. Remember, you're buying a great board that will last 90% of the people 2-3 years at least. You can play the waiting game but then you'd be waiting forever and sitting on a 10 yr old system "hoping" that you'll get the best computer.

When you buy a car, do you expect them to give you a new engine because the next model comes with one that has 20hp more?

-Bobby

The problem is not that stuff gets faster , that's good . The problem is not being able to upgrade to things which can often be quite simple . Along with planned obsolescence such as the socket 423 to 478 change which brought no increased performance or quality . It just meant that a whole lot of people were stuck with a deadend board .

The other problem is that sometimes Intel only lets you know that you have a deadend too late . Imagine someone who spent the big dough on a 2.53 or 2.8 last year when they were the fastest , all in the hope that when Intel ramped up to greater speeds they could swap in a newer CPU in their $150+ motherboards only to find out that the 3.06 needed more voltage just after they bought them , and that they can't support HT . Or better yet imagine somone who got an 845PE board and paired it with a 3.06 only to find that to be the last CPU the PE would support . Something Intel told nobody until relatively close to the advent of 800 FSB ! If they always tried to tell you beforehand then you could work to suit . But they don't .


Back in the day a friend of mine put down some big $$ on a Pentium2 400 Compaq , 2 years later he could upgrade it with a Pentium3 1 GHZ ( 100 FSB ) a geforce 2/radeon and maybe one more stick of ram . this would have given him performance on par with anything selling at the time . Or 3 years after the initial purchase he could have done the same thing to turn his lowend Pentium2 into a midrange P3 based rig .So after 2 years he could upgrade it to highend again or after 3 years upgrade it to midrange level ( of that time period ) . Now that is upgradeability , which Intel is now sorely lacking .

I don't know if I agree with you here. I certainly understand though. I usually upgrade every 2-3 years and I pretty much expect to get at least a new motherboard and cpu. To think that a future cpu will work with a current chipset is a bit too optimistic. Of course it would be nice but then you compromise what the new cpu can do or is capable of doing.

I buy stuff for what it can do now and don't think too much about what I'll be missing out on in the future. If it really matters to me, I'll upgrade but even now, my new systems is much more than what I need now and I don't see having to upgrade in the near future. I got a great system now and I don't regret it. I had a P4 1.6 (williamette) with PC133 so pretty much anything I bought now would've been a huge improvement.

My next upgrade will definitely include a motherboard and cpu at least. But that won't be anytime soon

-Bobby

Bobby .................. I normally keep my rig pretty long and do a few upgrades ( eg. incresed memory or harddrive or new vcards ) until I actually have to change the whole thing .

But I like to have the option to do a decent partial upgrade like the one I described earlier . As I said , I fully understand the need for progress . But I'd really rather not see any more than one dead end path a year ( I'd prefer 18 months actually, with good warning ) but Intel is looking to have 2 to 3 a year if the Prescott thing continues in the direction that it is currently going . To me that is too many and removes some of our options , it forces you to either keep the same CPU mb combo for a long time or to do a total upgrade , no incrementals .

But I understand where you are coming from .

Well, unless you buy a 3.2 and a 875/865 board now, you still have an upgrade path, albeit not much of one. I almost never buy the fastest cpu b/c I think it's a waste of money. I usually don't spend more than $200 on the cpu. That seems to be the sweet spot for price/performance. After that, the principle of diminishing returns kicks in.

Back in the 486 days, it was easy to upgrade to an overdrive cpu, which I did but it was almost not worth it unless you had a really slow cpu. I've noticed that now, I've needed to upgrade my computer less and less as time goes on. I had a P2 400 before the P4 1.6.

For most people, the computer systems they buy now is what they keep for a few years. It wouldn't make economical sense for Intel to just cater to the o/c crowd when a majority of their revenue comes from corporate and regular home users.

Of course I'd love to be able to just buy a cpu and increase my system performance but I know better and I'm resigned to just buying motherboards, cpus and memory each time I upgrade. I kinda like it... more toys to buy and play with

-Bobby

Intel has not stated that 865/875 boards are not Prescott ready. The only issue is whether or not mobo makers ignored the voltage requirements for Prescott. Some boards are prescott ready. Don't forget the AMD boards have gone through the same series of progressions requiring upgrades in many cases. Lets see - if you started out with a 100fsb socket 462 you had to upgrade to:
133fsb
133fsb with ddr support
166fsb
and now 400. In the socket 478 platform there was one less change since it has only gone from 100,133,200.

Not really, for a while in the AMD switch to 133 you could get 100 FSB processors and I don't recall there being such short notice . 133 FSB with DDR support was a personal choice ( a better one , but personal ) , since you could still run the 266FSB Athlons on a KT133 board with SDRAM . So you could still do a CPU swap , save $$ and get better performance in the interim . The performance would not be as good as getting an DDR board but it was an option .

The switch to 166 FSb was a bigger deal but when we were getting ready for that switch several of the curent chipsets already fully supported the 166 FSB ( 333 ) CPUs , the KT333 and SIS chipsets at the time AFAIK only neede bios upgrades. And in fact several KT266A boards could do it as well . Futhermore a 133 FSB ( 266 ) 2600 + was released for such people who still wanted an upgrade . Infact today you can find MP versions of the higher end thoroughbreds which because they have to run on the old MP chipset , run by default at 133 meaning there were upgrades for people with now quite old AMD boards right up to at least 2700 + levels .With 400FSb , the chipsets which support it were both out before the CPU and in the case of the Nforce 2 , for almost a year . Most if not all Nforce2 ver1 boards can run the 3200 + with a bios update . At least I know the Asus boards do and they have been out since last August .

So the changes with AMD were much less likely to result in sudden deadends .

Well much the same could be said of the intel platform as well. There weren't any surprises for those that followed their road map, and boards were available for each step as soon as the cpu was out. The switch from 100 to 133 was easy too, as almost all boards would easily do 133. Even rdram boards that only officially ran 100 would easily run 133

I'm not sure what you mean about the Nforce 2 supporting 400 front side though. I had 2 A7N8X's and neither of them would run 400 front side without serious aftermarket work, and they never would run dual channel much faster than 170fsb. A local computer store returned their entire stock when they could not get the 3200+ to run on any board they had, including the Asus.

I don't really see any of this as a dead end though. It doesn't really matter if a board is out the same day as new chip, or 6 months before - you have to have the board that supports a new chip. Socket 462 has kept the same number of pins, but there have been a number of changes along the way. Whether the board came out before the chip doesn't make much difference. You've got to have that board to support the new chip. It doesn't make any difference to me whether a board that supports a new Intel chip due out a year from now is available to me now or a week before the chip comes out.

Well I couldn't try the 400 fsb thing on my Asus . The thing is though that getting a 400 FSB CPU to run is different from getting a 266/333 CPU up to 400 and Asus has bioses for the version 1 boards that claim 3200 + support . I don't have one to try it , so I don't know if they are lying or have a poor implementation . I have a good friend who has gotten 170 FSB on his 1700+ T'bredA with a version 1.04 A7N8X and I can't recall seeing many people complaining of a 170 barrier except with some Palamino steppings .

Look carefully at what you said " boards were available for each step as soon as the cpu was out. " this is the opposite with AMD at present ( except maybe for the 400 FSB depending on which of us is correct ) . AMD had boards AFAIK before the release of any of their recent new chips ,not at release . Even if new boards came out as well the CPUs could often still work in the mainstream and highend boards of the day ( Intel's may not even work in the highend ones , as may be the case with Prescott ) . secondly AMD often released and supported a few of the older designs for a while . Examples include the 100 FSb (200 DDR ) Thunderbirds which went right up to the 1400 model . This allowed users to keep their older but otherwise decent motherboards and memory .

This is not an Intel vs AMD post for me , since it was not always this way with Intel . Previously when Intel made abrupt changes or produced deadends they would try to offer upgrade support . Such as the old Pentium overdrives and more recently the 100 FSB Pentium3 line coexisting with the 133 FSB line right up to the 1Ghz parts ( beyond if you consider the celerons ) . It is a disturbing trend that today's highend , top of the line Intel motherboards can often within 3 to 4 months and 1/2 chips then have absoloutely no upgrade path . We are not talking about year old bargain or midrange boards but the best of the best , $150+ .

Finally we still have to wait and see if when the dust clears the situation is the same or not . And if it turns out that all or some boards are not supported then we need to find out who is at fault , mb-makers ignoring specs or Intel changing specs . Until then I'll stop speculating and hope that there won't be another deadend this year .

Well... Its a two edges sword sometimes. True, Intel keeps changing things, but its often neccesary to make things faster.

423 to 478 ? needed for higher cache of future CPU's (Northwood)
400 to 533 ? faster
533 to 800 ? faster
478 to Prescott ? OK this one DOES suck, but to me, I would have wanted the Grantsdale with DDR2 and PCI Express in Q2 2004 anyhow, so I would have had to buy all new stuff anyhow.

The Alternative, AMD has sat on the Athlon for a long time without really getting much better. Remember when the p4 got its but kicked by the Athlon ? Now look... The P4 is at 3.2ghz, many can OC to 4ghz, while AMD falters and cant keep up.

BTW, the Hammer (the chip that was supposed to be released more than 2 years ago) will be using a new pin count, and a DDR333 controller (integrated on the chip), It will be replaced shortly by a DDR2 controller that will require a CPU/Ram/Mobo upgrade within 6 months. And given that the CPU has a built in RAM controller, it will require a CPU/Mobo/Ram upgrade every single time there is a newer faster RAM standard (ddr2 400, 533, then 600/666, then 800 etc...) thats really gonna suck for upgrades.

It depends on how it will be implemented , I will be watching it carefully as well . But hwta is good so far is that it seems that the Athlon 64 will be setup to work on Opteron boards eg the Nforce 3 and current opteron workstation boards .

FWIW

Late last summer I bought an Asus P4B533-E (845E). Contrary to info in this thread, that Asus 845E will support all the way up to 3.06GHz and will suport Hyperthreading with BIOS updates. Of course I sold it to get a 845PE so I never got to find out myself. But there are folks at www.abxzone.com running that 845E with HT CPUs.

Secondly, I then went on to an Asus P4PE (845PE). Contrary to info in this thread, Asus did relase BIOS updates for the P4C 800MHz CPUs. They will work in that Asus 845PE. Again I didnt keep it long enough to find out, but people are using the P4Cs in P4PE mobos.

Thirdly, I spent a few years with AMD systems. Sure, they didnt change the socket. But I cant tell you how many times Id buy a certain chipset mobo to find out that within six months you needed a new mobo for CPU support or optimizations. The AMD world plays the platform obsolesence game too. Intel changes sockets. AMD related mobos just stop CPU support to encourage people to upgrade. They are two different paths to the same end.

And of course ask any person with a new nForce2 who doesnt read roadmaps how happy they are that their brand new socket A is already capped out thanks to A64 coming.

My point(s)?

1. CPU support for Intel boards isnt as bad as indicated in this thread. It all depends on if the specific brand of mobo you are using is from a company that is known for updating CPU support frequently (a strong point of Asus)

2. Note that in points 1 and 2 above I state, I never kept the motherboard long enough to find out. Many enthusiasts are the same way. They buy a motherboard, drool over its anticipated CPU support life. But once CPU buses change, or features change, or 6 to 9 months go by; these said entusiasts buy a newer motherboard anyway.

Point being: When P4 C 800FSB was announced, what was everyones question in Asus forums?

It was, "Will it run in my 845PE? The answer was, "Yes it will."

The answer made everyone happy at the time. But how many people do you see running around with 845PEs and P4C CPUs? Almost none! The same people that last winter demanded their 845PEs run the new CPUs, were the same folks who bought the Canterwood the first 12 weeks it came out!

---------------------------------------

So....

In practicality, people tend to rant and cry for longer CPU support. But the majority of enthusiasts dont even use all of the CPU support life once its given to them. They end up going on to the next greatest motherboard even if their current one has a longer CPU life. And as such for the majority of people, the entire topic is moot.

JMHO

Based on your info , that is very good stuff from Asus . But that is somewhat missing the point I'm making . There are alot more people than you who would have goten burned . And even with Asus / Abit etc you are still buying based on a gamble/potluck , buy a board and then hope that it is one of the ones which will get an enabling bios upgrade . It boils down to guessing , which is what we are all now doing with this current Prescott rumour .


Secondly , as far as I know , you had to get a new revision 845PE in order to reliably get 800FSB CPUs to run and in some mobos to get it to run at all . The original 845PE is not validated for the P4 C . The othwer problem was that when these 800 FSB capable 845 PE boards came out , it didn't make sense to buy them unless you were on a budget since there was Granite Bay and then Canterwood and Springdale .

Many times that you upgrade you can lose a feature of your new CPU or whatever because the board doesn't support it . ( both Intel and AMD ) . As I explained , that was for example a tossup when the DDR Thunderbirds came out . If you ran them in a KT133A board you'd save money but you'd miss out on the new ability of the DDR mobos . But more importantly you would still see a very good performance increase over your previous lets say thunderbird 850 . So the there is often a toss up . Both companies because of progress do inevitably have to make some things obsolete but how sudden is the difference between the two . At present AMD is the more gradual you normally have advanced warning so for example it has been known for years that the Hammer designs would be on totally different sockets and boards . Depending on how long good speed Bartons/Thorntons are released someone buying an AMD board right now is taking a deadend risk . When looked at in context you would see however that the Nforce 2 has been out for almost a year . The KT400 for over a year and the KT400A for several months now . Whereas the 875 and 865 Intel chipsets are spanking new and may already be deadends . ( I hope not because i want to have as many options as possible .) I think it to be a fact that Intel has had more mobo obsolescences in the past 2 years , and if this Prescott thing is true they are looking to speed up this trend . Some people may not mind it , but I like my options .

Come on Cowboy. Your being a bit too dramatic. Its not really a big deal. an I865/875 mobo pairs nicely with a 2.4 to 3.2 ghz nowrthwood. If (and its still to be determined) it wont take a Prescott , are there any apps that dont run well on a 3.2 ? or a 2.4 for that matter? It fast enough.

No of course not , this is after all , the age of overkill . But of course what ever performance the 3.2 gives you'd truly better be happy with it . Because when its performance begins to drop off you're stuck .


What I really want is to have several Canterwood boards Prescott approved . That way I can watch and see what happens when the Athlon64 lands . If the AMD soloution is better ,I'll get that . If the Prescott is better , then I'll get a 2.4C and a Canterwood/Springdale and next year be able to put in whatever is the most powerfull Prescott available for socket 478 .

Don't forget also with the Athlon 64, the chipset/motherboard makers can bypass the built in memory controller and use a traditional North/south bridge design.

But look at this, when AMD was competing with PIII it started using the current Socket A.

It's still using it....

The way AMD has is set up you can choose to upgrade your motherboard, btu still run the slow CPU, or upgrade the CPU, and still run it, sure there are troubles running the new CPUs on old old boards, but almost any design that's at least a year old will run anything on the 333 FSB.

The P4 will go through 3 socket changes in the time it takes AMD to go through 1.... and you can expect that there won't be a huge amount of Socket changes with the Atlon 64 line.


now we are users that use top end stuff... we buy new stuff every 6-8 months anyway... so it doesn't matter.... I have been considering using a P4 setup for my next system simply cuz it's different from my current machine...

Not cuz it's better... or worse... just different. We all know both platforms have there advantages and weaknesses, but you learn more about both if you've used both.... I'm happy with my current setup.. but I could always use a change...

Just remembered one thing from my time with AMD. Very often you bought a new mobo with the A chipset, only to find out a couple months later that the B chipset version would have everything the A should have had in the first place, so you either feel taken, or upgrade How many times did Via pull that....

TC said:

Just remembered one thing from my time with AMD. Very often you bought a new mobo with the A chipset, only to find out a couple months later that the B chipset version would have everything the A should have had in the first place, so you either feel taken, or upgrade How many times did Via pull that....

Click to expand...

Not to mention the Nforce2 , what must be the most unstable platform in years.

Besides the KT266 rubbish , I can't think of another chipset where the 'A' version was poor . Importantly the AMD 761 based boards were there as an alternative , they were faster and were available before the slower KT266 .

My Nforce 2 is very stable so I can't agree with Retrospooty on that one .

Finally I am trying to make sure that we don't degenerate into an Intel versus AMD argument . I am certain I can mention all the failed Intel chipsets and third party failed chipsets too . But that woukldn't have much to do with what I was concerned about when I started this thread . This is not an 'Intel sucks' thread , this is a thread about premature obsolescence and deadend upgrade paths .