There is no good platform!

[Crashman]

OK, so the Tualatin Celeron looks like a good deal for overclockers like myself. But by the time I get around to upgrading my system (probably late winter/early spring) it will be outdated even at an overclocked speed of 1600MHz. AMD garuntees their platform to stick around a while, but to what end? Is AMD going to release any fast processors by then? And what of the P4? Is Intel going to price me out of their market? People keep talking about two chipsets, the latest ones from VIA for AMD and Intel. Now there's a company who's lost my trust. What to do? If you live with a slow computer, I suggest the ECS K7S5A and an AMD 1.4GHz T-Bird. That would not be a significant upgrade for me, so I wait. For what, I don't know.

Back to you Tom...

 

[peartree]

Yeah, AMD will bring out new cpu's next spring. You can go to their website and look at their 'processor roadmap' that tells you when to look for new silicon. Personally, I want a clawhammer! Barton looks like an Athlon4 on steroids. And the TBird goes 1.533 GHz this month.

As for the P4, it will move to the .13 die process and get a LOT faster. I've seen hints of 3 GHz by sometime next year.

What does it all mean? It means that whatever nice, shiny new upgrade you just installed will be last year's s==t by this time next year . As usual .

 

[FatBurger]

Don't worry, Crash has looked at AMD's roadmap many times, I'm sure.

By the way, the Tbird isn't actually going to 1.53. It's a new core, Palamino.

<font color=green>I post so you don't have to!
9/11 - RIP</font color=green>

 

[Crashman]

Well the T-Bird is getting close to it's max speed. I susupect that the current version of the XP won't go much farther if it's on the same die process. So when are they shrinking to .13?

Back to you Tom...

 

[peartree]

I don't know as how I regard the Palomino as so much of a new core. More evolutionary as revolutionary. And we'll see if it's worth it!

 

[peartree]

Ah, well, it appears that it won't be until "1H02" as it says in the Processor Roadmap. Personally, I won't buy a Palomino processor on the old core. The Thoroughbred looks intersting for early next year, but the Clawhammer really looks like the winner. But will Intel sit still long enough for AMD to catch up? I wonder...
And then we should really deal with the question of how good AMD's CPU's really are. Yes, they run fast, but do I want to buy a CPU that will suffer core meltdown if the fan quits for over 1/2 second? I'm not sure at all!

 

[FatBurger]

Someone unplugged their fan for about 15 mins and the CPU was fine. Then the motherboard (or software, can't remember), shut it down. As long as you're not a huge idiot and take off the heatsink while running a Quake III demo...

<font color=green>I post so you don't have to!
9/11 - RIP</font color=green>

 

[girish]

Very true!

Intel processors dont have very good and stable platform. just look at this:
I buy a P-III Katmai slot1 processor and board
They turn back to sockets.... change the board!
Coppermine arrives... change the board!!
Tualatin is launched... change the motherboard!!!

For a while there is no good chipset for P-IIIs with only i810 around, and BX phasing out. Then the i815 arrives, but limits its memory to only 512 MB. BX is already out of question since it cant take FCPGA chips and cant officially handle 133 MHz FSB.

summary: transition from Katmai->Coppermine->Tualatin that is PPGA->FCPGA->FCPGA2 required a motherboard change each time.

P4 platform it still worse. Still, there is no viable ecnomical platform for the P4. There is skyhigh RDRAM or sloppy SDRAM solutions. Plus Intel is making a stock clearance sale of 423 pin P4s with insane price-cuts of as high as 54%, and bringing in a new 478 pin processor!

What do people do who already have a bought a 423 pin P4 at earthly prices not even a year ago? Limit 2 GHz! Hope the 478 pin socket stayes atleast till people start feeling their 2.0 GHz systems too slow. then they might venture into some socket 536 and lo! yet another board change!!!

on the other hand, AMD has maintaines a comendeble stabiility with the 462 pin platform. with a small hiccup at 100->133 transition, most boards bought even a year ago will support upcoming breed of Athlon processors. With Palomino and Morgan already on the scene and Thoroughbred and Appaloosa on the horizon there is still no sign of a platform change. Just upgrade the BIOS and thats it! Too bad most boards come with the damned VIA southbridge. One needs a solution line the ALi Magik1!!! Highly scalable, flexible, decent performing, omni-compatible and virtually bug free!

Intel does have such stable chipsets but platform stability is virtually non-existent!

girish

<font color=red>No system is fool-proof. Fools are Ingenious!</font color=red>

 

[girish]

those damn PR ratings are back!

AMD actually needed to project their processors running at that "relative" speed now comparing to P4...! Actually AMD does not need such stunts. Overclocking enthusiasta already proved that Athlon Tbirds can be clocked at atleast 10~20% higher than the rated. It could simply remark the processors and lock the multipler. Tbird could really close at 1.7 GHz or so, still far from 2.0 GHz of P4.

Athlon architecture is now quite mature and with 0.13 micron transition with the Thoroughbred it will return to its golden days of high clock as well as overclocking. Just recently the 0.13 micron Celeron showed that it could be overclocked to as much as 25%. Athlon Thoroughbred would do even better!

And this Thoroughbred is supposed to arrive later half next year. By that time much water would have flown under the Intel bridge! [sigh]

girish

<font color=red>No system is fool-proof. Fools are Ingenious!</font color=red>

 

[girish]

Well, there is yet yet yet another board upgrade expected!

Newer Tualatin processors have a differential clock that enables higher FSB speeds at low voltages. and that suggests Intel would bring out faster Tualatins maybe upto 1.6 GHz or so. May not be higher but maybe with higher FSBs like 166 MHz or even 200 MHz! thats yest another board upgrade...!

Pentium-III will stay for a while in its Tualatin incarnation that would compete well with the fastest P4 available.

girish

<font color=red>No system is fool-proof. Fools are Ingenious!</font color=red>

 

[FatBurger]

Right...I know that...am I missing the reason you said that to me?

<font color=green>I post so you don't have to!
9/11 - RIP</font color=green>

 

[girish]

It came to you? Sorry abt that Fat!

I just posted to this thread. You see, I dont reply to my own posts!

girish

<font color=red>No system is fool-proof. Fools are Ingenious!</font color=red>

 

[Crashman]

Actually up until recently Intel has been BETTER than AMD when it comes to platforms! You didn't REALLY need to change from Slot 1 to Socket 370, you could just get a converter. Could you do that with AMD Slot A to Socket A? NO!
So what of the older LX motherboards taht only suport 66MHz FSB? You can still put a PPGA Celeron up to 533 in them using a Slotket. Could you do that with a Super 7 to Duron? NO!
I mostly ignored the P4 socket change, since the Willy was a cut down version of the Northwood and I figured I'd never use it anyway. Intel was nice enough ot let us know BEFORE the Willy was released that the Socket 423 would NOT be in their future. So anyone looking toward future upgrades were advised by Intel themselves to wait for the S478! So for those of us who really wanted an upgrade path can ignore that change because we knew about it before the platform hit the market.
But this FC-PGA2 change really anoys me. I have had my PIII 933 up and running for over a year and it's still faster than what I need. I'm looking forward to upgrading simply so I can play future games! None of the current games use up what I already have!
And I am extrememly disappointed in AMD for NOT realeasing faster Athlons. The Pally should be up to 2GHz in .13 micron trim already. What's holding them back, the cost of changing? Everything comes at a price, sooner or later they HAVE to spend the money, and by waiting until later, they are loosing sales NOW.
So that leaves Intel, by the time I upgrade they should have the Northwood out at 3.0GHz But they will undoubtedly want more for that processor than any of my cars cost me! Can't justify the expense, AMD is falling behind, so where does that leave me? With no reasonable upgrade options!

Back to you Tom...

 

[Crashman]

Oh, and about the i815-I need to run Win9x to make some of my cards work, so I'm already limmited at 512MB. Blame MS, Intel only limmited the i815 to the same limmit as Win9x so it would be used as a ocnsumer board instead of a high-end workstation board. I see a lot of 820/Rambus and 840 boards going quite cheaply if your into that sort of thing (Rambus boards do not have MTH). And the BX still works, except with the Tually, which is one reason I'm pissed off.

Back to you Tom...

 

[girish]

Well, the i815 sure misses out on memory, but where are i820 boards? forget MTH but I have seen the last one - the Asus P3C-E or something like that (of course here in India) almost a year ago.

I'd rather expect Intel to come out with a 440BX supporting 133~166 MHz and Tualatin. that'll be the day!

slot to slotket to powerleap convertors, these are just workarounds for the root problem - incompatibility with newer processors and yes it does work, so nobody should really complain about them. just that intel made them do it.

Duron wasnt supposed to get into socket7, the AMD K6 series already did by expanding into Super7 <i>and</i> maintain a value platform that intel let go long ago to some decent perfromance levels.

Duron is still my favorite processor, I recommend it to almost everybody on budget. Save on the processor, get a decent board and you have a system that could go all the way from Duron Spitfire 600 to Athlon Palamino 1.53 (i wont call it 1800, sorry) right away. the later one Thoroughbreds are also supposed to stick to s462 and so will the Bartons. noody knows about them yes so these are just speculations.

So wont you call a platform supporting such a wide range of processors <i>Stable</i>?

girish

<font color=red>No system is fool-proof. Fools are Ingenious!</font color=red>

 

[girish]

with a LX chip and a Celeron 533 to "upgrade" ,I wont really call it an upgrade. and C533s arent avaialable today anyway. that warrants a compulsory board change. first you needed a board that would support 100 MHz, then suddenly 100 MHz FSB chips go out of market and you need a 133 MHz boards. add to that the socket change problem. i am really sick of this "boardom".

one of my clients who regularly updates his system had to change his board thrice a year when he moved from P-II 350 on P2B-F to P-III 700 socket370 on i810 and then to P-III 933 on CUSL2! and now he is expecting a 1.2 GHz P-III! Irony is that I had recomended the CUSL2-C to him not more than 4 months ago when the TUSL2-C had just been announced and unavaiable in India until recently. yet another boardchange. and I am not going to recommend a P4 just yet. wait till mid 2002. Heck he prefers Intel platformand not willing to buy AMD.

It was very nice of Intel to tell their customers about not buying their 423 pin processors, but what about their clearance sale with staggering 54% price cuts? keeping the s478 platform high and s423 to the dirt and then saying dont buy 423, get those 55 more is a cool marketing strategy!

yeah, I am too disappointed with AMD for not getting their Tbirds or Palominos faster. releasing two Morgans within 3 weeks is no big deal. they sure need a fab upgrade and do it fast, else they will have more than a GHz to catch up. PR ratings arent going to help them, at least I wont be telling my clients "it is a 1800 MHz processor"!

and I dont agree that AMD has no reasonable upgrades. basically the platform (I mean the chipsets and boards) are quite suitable for Athlon line of processors at least a GHz down the line! and 0.13 micron will surely get more out. that does mean you buy one today, what remains is just a processor switch, alongwith a routine BIOS update, thats it. your reasonable upgrades are still a quarter away but nevertheless they will actually cost you reasonably.

and thats really cute calling workarounds like slotkets and powerleap adapters for platform stability.

girish

<font color=red>No system is fool-proof. Fools are Ingenious!</font color=red>

 

[Crashman]

"and thats really cute calling workarounds like slotkets and powerleap adapters for platform stability." Ok, so what do you suggest for people with Super 7 platforms and Slot A platforms? At least there WAS a workaround available for Celerons and PIII's.

Back to you Tom...

 

[Crashman]

Intel released Slot 1 and AMD Super 7. YOu have to rememober that the orignal Socket 7 was not considered adequate for K6-2's. New voltages, bus speeds, and multipliers were required. Which meant a new board. Intel did the same thing by switching to Slot 1.
So now you had two new platforms, Super 7 and Slot 1. The score is even. But wait! Now Intel introduces Socket 370 and AMD introduces Slot A. So their still even. But wait! Now AMD introduces the Socket A. Intel already HAS the Socket 370 so score one for Intel. AMD at this point has done more platform changes. Intel releases Socket 423, which doesn't really matter since we didn't fall for it-Intel TOLD us not to. So finally Socket 478 brings them even. Intel only looses to AMD on this one after introducing FC-PGA2.

Back to you Tom...

 

[girish]

as regards socket7 or super7, Intel had already abandoned the platform at 233, AMD (and Cyrix) actually pulled the string to higher speeds with AMD getting off with Super7 with K6-II and K6-IIs and cyrix still continuing till 433. This was a viable upgrade option although those cyrix chips needed to be supercooled. Still the Cyrix M-II and M-III sells a lot in India just on the old TX boards or the VIA TX-100 boards. of course i dont even touch them. it has Cyrix printed on them.

even the newer K6 can work well with lower FSB giving unprecendented performance on older platform, like the K6-II 450 I tried on a 430TX board at 83x5.5= 456.6 MHz!

This I would call platform stability, people with 430VX/TX boards can also boast of 500 MHz clock speeds, thats more than double what Intel thought was enough.

The idea of putting a Duron on a Super 7 board is interesting, but thats like putting a Celron into a Socket 7 board. It does support 66 MHz FSB!

Slot A platforms were there for a relatively brief period to cause such unprecedented mass upgrades. I would call it a nightmare for their purchasers. But then, its still a stable platform! there is no adapter for using a socket A chip, so you wont really part that CPU+board combo! they will go together whereever they go. I am expecting a K7M with Athlon 550 MHz for my lab. I wont be using it as regularly, but will sure make a nice couple for my museum.

They could get a socket A board and a socket A processor for cheap, and as long as its cheap I wont complain about the upgrade.

If there were no workarounds for Celerons and P-IIIs imagine what fortune Intel would have made selling virtually millions of boards with every new core of P-III!


<font color=red>No system is fool-proof. Fools are Ingenious!</font color=red>

 

[girish]

i dont really understand why you are comparing super-7 to slot 1! super7 was 5th gen better-than-P5-Pentium while slot1 was 6th gen P6 core processor. Super7 is not comparable to even the P6 Pentium-Pro. Intel and Cyrix picked up where Intel left off. thats it.

With P6 generation of processors, Intel managed to put out the competiton for a while by changing to slot1 (and everybody knew they would eventually return to the sockets), AMD did a foolish thing playing follow-the-leader and making slot-A. but AMD wasnt in a position to challange this move by the chipzilla. people would go for a slot because intel called it a slot. AMD <i>had</i> to follow the suite. Good thing was it returned to sockets sooner than intel did, almost at the same time intel made socket370. on this score, both had slots, both had sockets.

Now Intel changed the original socket370 twice, in all three versions with little compatibity between the first two and a complete revamp with the third! on this front, with just <i>one</i> aberration of 133 MHz switch (which is really not too significant since most 100 MHz boards work well at 133 or the processor can be set to work at its rated speed even with a lower FSB). I dont need to change my board the clumsy Luckystar K7VTA with KT133 chipset even after more than a year to use the latest 1.2/266 GHz Tbird even today.

And I dont recall Intel saying dont buy the 423 pin P4, but saying 478 is coming, and 423 will end at 2G! If people cannot think and make out what it really means its their fault of course. It was people, experts and analysts like those around the world and those unknown at the THG told us not to.

girish

<font color=red>No system is fool-proof. Fools are Ingenious!</font color=red>

 

[Crashman]

There is no 5.5x multiplier on the earlier platforms you speek of! There is of course the 2x=6x trick, but that only works on the LATER K6-2's. And it WAS a platform change to go from Socket 7 to Super 7. It required a new board!

Back to you Tom...

 

[girish]

well, there is 2x=6x and 1.5x=3.5x on later K6-IIs.
Actually the multiplier pins are directly connected to the processor's FID pins, so although board does not document it sometimes using forbidden combinations work!

I had a San-Li 430TX board that allowed such setting. and it the BIOS update did show up the processor speed correctly, and Wcpuid verified it! the board did not have that combination silk-screened, it had only upto 4.5x!

afterall, this higher multiplier was implemented on the CPU itself, the board actually did support it after all. it is a minor platform change in a way, but no board change just a BIOS update did it. the board was not running at 100 MHz and yet ran procesor at 450+ MHz.

yes it required a new board for proper implementation of this Super7 platform, but it was backward compatible with older one. it only had a problem recognising newer processors and running it at their full speed with straddled FSB wasnt possible, but most processors, upto the 500 MHz workup at 83x6=498.

Stable platform means the board should stay for as long as possible taking most of the processors coming its way. it should be said that Super7 processors maintained the compatibility with its parent platform, unlike the FCPGA2->FCPGA->PPGA changeover. it dint need a board change. that is why I dint raise the issue of 100->133 FSB change requiring replacement of even larger number of motherboards than the PPGA->FCPGA would possibly have. it atleast had a workaround, 100->133 simply did'nt. and there was no multiplier to save you, like the Super7 so your processors ended up running at 75% of their rated speed.

girish

<font color=red>No system is fool-proof. Fools are Ingenious!</font color=red>

 

[Crashman]

FC-PGA is backwards compatable with PPGA CPU's just as Super 7 is with Socket 5 and Socket 7 CPU's. And most Socket 7 platforms did NOT have the necessary third jumper for getting the 4x and above multipliers. You needed to buy a NEW motherboard for that, Super 7!
Oh, and about the bus speed issue, same thing happened with Socket 7, and Slot A!

Back to you Tom...

 

[girish]

socket7 wasnt backward compatible with socket5, socket5 was 273 pins while the socket7 was 296. just like P4 s423 and s478, different socket for the same processor!

PPGA and FCPGA incompatibility was just regarding voltage support that most boards managed by BIOS update, while others required an adapter. older chipsets couldnt recognise the FCPGA and this was the bad thing. in fact, they needed an upgrade anyway to support higher speed and lower voltage. At least it was not that radical as Tualatin changeover is.

and as a matter of fact, I havent yet seen a board with just two jumpers to set the multiplier. all boards were made to support at least 4.5 when they actually needed only 3.5 (for 66x3.5 Pentium 233) that makes 7 combinations (1.5, 2, 2.5, 3, 3.5, 4 and 4.5), 3 jumpers support 8! and only 6 were implemented at the time. the multiplier 1.5 was used by only two processors, P-90 and P-100 runing at 60 and 66 MHz FSB respectively, and later the same combination was reused for the value 3.5. That was the missing combination that made it work at 5.5x! the other missing combination that was later named as 5.0 should also work on that board.

so as of now, those 8 combinations support a total of 10 multiplier from 1.5 to 6.0, with 1.5x=3.5x and 2x=6x!

girish

<font color=red>No system is fool-proof. Fools are Ingenious!</font color=red>

 

[Crashman]

"socket7 wasnt backward compatible with socket5, socket5 was 273 pins while the socket7 was 296."

Damn I never knew you were an idiot! Do you want me to show you PICTURES of a Socket 5 and a Socket 7, and then show you MORE pictures of a Socket 5 processor in a Socket 7? I do these things all day long. ALL the Pentium (ie Classic) procesors from 75-200 were Socket 5. ALL the Pentium MMX processors were Socket 7. So your going to tell me I cannot put a Pentium Classic into a socket 7 board? I refurbish systems for a LIVING, and one of the BEST DEALS I get is on used Socket 7 systems from a state auction. They ALL have Pentium Classics in them, usually 100MHz. I throw a Pentium 200MMX CPU in them and sell them cheap.
If you want me to send you pics to prove this, give me an address, or better yet, a website to post them on so the rest of the forum can see!
As for our previous discussion, I have not been bothered too heavily by Intel's platform changes up until NOW. A second change in the Socket 370 spec was two much. Prior to that they were comparable to AMD in their platform changes (diregarding S473 becasue most of us took Intel's warning on that one).

Back to you Tom...