As far as I know, the first Pentium 4 processors were less powerful than the later Pentium 3 processors.

But I think in 2001, when someone had a P4 1.4 GHz, it was a good computer back then.

Ok I think that to my Pentium 3 I will choose GF2MX, GF4MX or Radeon 9200 because lower price 😜

The Serpent Rider wrote on 2024-09-18, 13:04:

SDRAM on Pentium 4 was more of an afterthought really. Intel had Tualerons to sell as low-budget offering at that time.

erm....

Im not really sure what that means.

Early P4 processors on Skt 423 launched with a focus on high clock speeds (they didnt start there but they knew they were heading there), but their performance didn’t scale as well as expected when paired with slower memory technologies like SDRAM (there were some boards made that put SDRAM on a P4 and they didnt perform as well when using it).
RDRAM helped improve the performance of the P4 by providing the high bandwidth that the NetBurst architecture required. This made the Skt 423 P4 better suited to RDRAM compared to the Pentium 3, which didn’t need as much memory bandwidth. Dont forget the P3 was optimized for lower latency memory rather than high bandwidth, which means the P3 was optimized for SDRAM and the P4 was optimized for RDRAM.

As time went on technology developed, things like DDR SDRAM came along and things changed, but you cant know the future of somebody elses development. And then you have DDR support in the 845 chipsets and RDRAM is dropped.

Intel probably did not have P4 SDRAM combo in mind initially. They had RDRAM first and DDR somewhere down the road. That explains 130nm refresh for PIII platform as a budget SDRAM offering. But I guess OEM were pushing them for P4 SDRAM atrocity due to shortages.

The Serpent Rider wrote on 2024-09-18, 20:40:

Intel probably did not have P4 SDRAM combo in mind initially. They had RDRAM first and DDR somewhere down the road. That explains 130nm refresh for PIII platform as a budget SDRAM offering. But I guess OEM were pushing them for P4 SDRAM atrocity due to shortages.

Initially?

NetBurst is a bandwidth hungry architecture...
PC133 SDRAM just doesnt have that.

The P3 doesnt need the high bandwidth you get with RDRAM. The pairing makes little sense. Which is exactly why they dropped it on the P3 after just over a year of trying to push it.
Intel was in partnership with RAMBUS, so they had to push it but they quickly realised it just wasnt working and dropped the 820 like a hot brick after the MTH debacle.

You guys are trying to make out like there is some kind of "conspiracy theory" going on where by Intel was deliberately pushing a load of crap on people who didnt want it because it made them more money.
I guess in a way it is kind of true, but it is also a load of old rubbish at the same time, it was more about the technology and the matching of the technologies that went best together than it was about screwing more money out of its customers, which is born out by the fact that Intel dropped the use of RDRAM on the P3 after just one year, they dropped the use of RDRAM as soon as DDR SDRAM emerged as a technology that was capable of offering the bandwidth at a more competitive price.
Dont forget while Intel was taking up the use of DDR on its chipsets it was still in partnership with RAMBUS.

NetBurst is a bandwidth hungry architecture...
PC133 SDRAM just doesnt have that.

Yes, like I said, this combo was most likely just an afterthought, because 845 SDRAM platform was released after Tualatin.

The Serpent Rider wrote on 2024-09-18, 22:42:

NetBurst is a bandwidth hungry architecture...
PC133 SDRAM just doesnt have that.

Yes, like I said, this combo was most likely just an afterthought, because 845 SDRAM platform was released after Tualatin

Exactly! I recall that it was planned to only use RDRAM but the prices for it were very high at that time. Was it due to a factory has been hit by an earthquake, can't remember!? So after the AMD platform came up with DDR-SDRAM they had to offer support for SDRAM with there than crippled chipsets which couldn't feed the bandwidth - as said above. Shortly later (early 2002) DDR won the race because it was much cheaper and more available than RDRAM.

-> https://en.wikipedia.org/wiki/RDRAM, https://de.wikipedia.org/wiki/Rambus_Dynamic_ … m_Access_Memory - the German Wikipedia article is better, especially concerning the SDRAM <-> RDRAM <-> DDR-SRAM topic.

kind regards
soggi

The Serpent Rider wrote on 2024-09-18, 22:42:

NetBurst is a bandwidth hungry architecture...
PC133 SDRAM just doesnt have that.

Yes, like I said, this combo was most likely just an afterthought, because 845 SDRAM platform was released after Tualatin.

Yeah but there was a specific reason for that wasnt there.
There was no "after thought", it was bandwidth hungry, the RDRAM made the chips possible, the mating of NetBurst and SDRAM in the early 845 chipset revisions proved that all to well, it was a shockingly bad combo, there are a few reviews of boards with SDRAM and DDR SDRAM (VIA chipset), this was for later versions based on skt 478 but the idea of starving the chip and bandwidth in the ram is illustrated well there are would have carried over the skt 423.

1998-1999 - Intel gets into bed with RAMBUS and they produce RDRAM and Intel come up with the 820 chipset, but RDRAM is expensive and the 820 isnt that great (and there are good reasons for those technical problems like I said above the P3 is better suited to SDRAM because of the lower latency SDRAM provides, it doesnt require the higher bandwidth RDRAM gives)

2000 - P4 and the 850 chipset are released with RDRAM but the cost is high... RDRAM is a major factor in the cost.

2001 - Intel introduce the 845 chipset, this has SDRAM support but it cripples the P4 performance, the chipset has revisions later than year that support DDR SDRAM.

2001 mid - Tualatin is released the 815 chipset doesnt support RDRAM. (In fact I dont think there were any chipsets, intel or otherwise that supported Tualatin and RDRAM, i could be wrong on that, but they would be an exceptional case and they prove my point not yours if they do exist.)

2002 - The market fully transitions to DDR SDRAM. Intel start to drop RDRAM in favor of DDR SDRAM because it costs less for the performance.

2003 - Intel drops RDRAM. DDR and DDR2 dominate from here on.

I dont see how you are trying to weave Tualatin into this.
The Pentium 3, RDRAM and the Intel 820 plus the MTH debacle is the only time I can start to see your conspiracy theory take shape.
But that it died on the vine because the P3/RDRAM combo was a bad one. The P3 needs low latency ram not high bandwidth ram. By the time the Tualatin was released any idea of mating P3 to RDRAM was dead.

ElectroSoldier wrote on 2024-09-18, 10:36:

Ok fine. I dont see that in the real world going by the artefacts that are left to see but ok we can roll with that as an idea. […]
Show full quote

VivienM wrote on 2024-09-18, 00:42:

They basically starved the market of non-RDRAM chipsets. […]
Show full quote

ElectroSoldier wrote on 2024-09-17, 11:28:

When you say obsession how did that manifest itself in the real world? […]
Show full quote

When you say obsession how did that manifest itself in the real world?

I mean Intel being the ones in charge of their own production schedule did they decide to stop the production of say the 810/815 chipsets to make the 840 chipsets and then go on to force those 840 chipsets onto the world?
Or did they pay off the big OEMs like HP and Dell to sell their 840 chipset systems by offering them cheaper than they should have been? to make their sale more attractive to the OEM?
Maybe starve the market of the 810/815 to force 840 sales?

Or did they just run an aggressive marketing campaign in the hope that people will take up the product they were making?

They basically starved the market of non-RDRAM chipsets.

And they did so in two phases:
1) Phase 1. i820. Basically, if you wanted 133MHz FSB, 4X AGP, UltraATA 66, or any of the other platform improvements, and you wanted an AGP slot, then you needed an i820 with RDRAM.
When the market resisted and 440BX stuck around much longer than expected, they decided to come up with their MTH to 'translate' SDRAM into RDRAM. Then that... turned out to be a snafu and got recalled. So they gave away 128 megs of RDRAM to everybody who had the Intel CC820 board.
Then they were forced to launch the i815, which got you all the nice platform improvements over the 440BX but with SDRAM support. That was the end of RDRAM on the PIII platform - maybe a few high-end Dells got sold with i820/i840 but that's about it.
(As an aside, as someone who would never have touched a VIA chipset back in the day and who is still shopping for an additional retro system, I am continually astounded at how many socket 370 boards/big OEM systems are out there with Via's Apollo 133 chipsets. That's the legacy of Intel not offering a decent non-RDRAM chipset before the i815.)

2) Phase 2. Pentium 4. The Pentium 4 launched with i850 and RDRAM only. I think the MTH was supposed to come back, but didn't. So, again, if you wanted Intel's flagship processor, you needed RDRAM.
After close to a year, they launched the first iteration of the i845, but that only supported PC133 SDRAM, not DDR, and performance was mediocre. I'm googling this and reading old AnandTech articles, apparently it was crazier than I remembered - the i845 could do DDR, but Intel strictly, strictly prohibited motherboard manufacturers from making DDR boards until at least Jan. 1, 2002. And I haven't researched it and I don't remember offhand, but I'm pretty sure the DDR i845 boards remained unappealing alternatives to i850E and it was only with i865/i875 that DDR platforms for P4 really became good.

So basically, for the first 15ish months at least, if you wanted a P4, you had to get RDRAM.

How that manifested itself in the real world? If you weren't VIA-phobic, in late 2001, you were getting a KT266 motherboard with your Athlon XP on socket 462.

Intel had a chipset problem from late 1999 until the launch of the i865 in 2003. There was a gap of about a year where the i815 provided a decent alternative, at least if you could live with the max 512 megs of RAM at a time when 256 megs of SDRAM started costing $60CAD. But otherwise, Intel's chipset options were always flawed because the decent/high-end option was only RDRAM.

(Disclaimer: I had a Willamette P4 with a gig of RDRAM. Never had a real problem with it, although what annoyed me was that the Northwoods that came out a month later used PC1066 RDRAM and mine was PC800... so oops, zero upgrade path for me.)

Ok fine.
I dont see that in the real world going by the artefacts that are left to see but ok we can roll with that as an idea.

I think I would like to seperate the Pentium 3 and the Pentium 4 here, certainly I would agree that intel pushed RAMBUS on us with socket 423 as when you look for an find a socket 423 board all you ever seem too get is RAMBUS, but with the Pentium 3 chipsets there are so many boards out there that use SDRAM (440/810/815 chipsets) then it is hard to see how a company that controlled the supply starved the market because there are so many out there.

I remember the realease of RAMBUS during the Pentium 3 life time, I remember the articles in magazines talking about it and how it was so much more expensive than SDRAM and while it filled a high end neich the majority of computers are not high end (something Intel was all to aware of) and most magazines pushed SDRAM systems.
If for what you are saying was true then you would be assuming that Intel wanted to ignore the majority of its sales and squeeze more money out of buyers, by selling them RAMBUS boards over SDRAM boards, and they did that by deliberately not producing enough SDRAM boards to satisfy demand.
And like I said ok I can work with that as an idea.
But the physical items we are left with is a lot more Pentium 3 boards that use SDRAM than RAMBUS. And if what you are saying was true then that would be the other way around, because in your idea intel starved the market of 810/185 chipsets.

The actions dont fit the words. And the actions are unchangable. So it is the words (your ideas about what they did) that must be wrong as stated.

Because if you were right in the case of the Pentium 3 then you would have a similar situation as you have with socket 423 boards.

I think it's simpler than that:
- for about 6 months in early 2000, Intel did not have a competitive leading-edge SDRAM chipset for Pentium III
- Intel's solution was the i820 with RDRAM; a few Dell XPS Bxxx machines were sold, but that's about it. (I would note that I have seen XPS Bxxx machines on eBay so some have survived) Otherwise i820 was a flop.
- many, many people went to VIA's chipsets, and again, you see a strangely high number of socket 370 (or slot 1) motherboards with VIA's chipsets from this era. You don't see very many VIA chipset PIIs or Katmais, but you do see them in the 133MHz Coppermine era. Including in a number of large OEM systems from HP, etc.
- i810 was also an option, and again, you see a number of surviving systems with nice 933/1GHz Coppermines with i810 boards.
- some of the Taiwanese motherboard manufacturers put together socket 370 440BX boards that were basically intended to be overclocked at 133MHz FSB
- other OEMs explored AMD platforms, and indeed, I have seen some Athlon Gateway boxes on eBay from this era. Dell was one of the very few large OEMs to stick with a 100% Intel lineup.

It's not "Intel restricted sales of the i810/i815". It's "Intel had a gap in their product line until they scrambled to launch the i815." And I see plenty of evidence looking at what Coppermine-era hardware is available on eBay or elsewhere to back that up.

And maybe that's the difference between this era in PIII land and socket 423. In socket 423, you had no options - no VIA/SiS SDRAM chipsets, no Intel options until the i845 PC133 SDRAM chipset launched, etc. If you didn't want RDRAM, you got a PIII or an Athlon. In PIII land, you had options - VIA's chipsets, sticking to 440BX with 100MHz FSB, giving up on AGP with i810, etc. And if you look at PIII motherboards/systems on eBay today, you'll see that motherboard manufacturers and large OEMs resorted to all of those options. But those options were all much worse than a proper Intel-designed chipset which is why when the i815 launched everybody went that way.

ElectroSoldier wrote on 2024-09-18, 21:30:

The P3 doesnt need the high bandwidth you get with RDRAM. The pairing makes little sense. Which is exactly why they dropped it o […]
Show full quote

The P3 doesnt need the high bandwidth you get with RDRAM. The pairing makes little sense. Which is exactly why they dropped it on the P3 after just over a year of trying to push it.
Intel was in partnership with RAMBUS, so they had to push it but they quickly realised it just wasnt working and dropped the 820 like a hot brick after the MTH debacle.

You guys are trying to make out like there is some kind of "conspiracy theory" going on where by Intel was deliberately pushing a load of crap on people who didnt want it because it made them more money.
I guess in a way it is kind of true, but it is also a load of old rubbish at the same time, it was more about the technology and the matching of the technologies that went best together than it was about screwing more money out of its customers, which is born out by the fact that Intel dropped the use of RDRAM on the P3 after just one year, they dropped the use of RDRAM as soon as DDR SDRAM emerged as a technology that was capable of offering the bandwidth at a more competitive price.
Dont forget while Intel was taking up the use of DDR on its chipsets it was still in partnership with RAMBUS.

I don't think we're saying Intel pushed RDRAM because it made them more money. We're saying Intel pushed RDRAM. Something in the late 1990s clearly convinced Intel that RDRAM was the future of their high-end platforms, and they pursued that both on PIII and P4.

And, in an alternative world where Samsung & co were making RDRAM RIMMs at much lower prices, it might yet have succeeded. Or in an alternative world where the price of PCxxx SDRAM wasn't plummeting. Or an alternative world in which Rambus had gotten away with their attempt to insert patented content into the JEDEC specs, forcing the entire SDRAM/DDR industry to pay royalties.

And on both PIII and P4, the attempts to push RDRAM hurt the success of the platform. Intel had to embarassingly back away from RDRAM both times. And as much as everybody loves to bash on HotBurst and Preshot and whatnot today, DDR P4s in 2003-4ish were not that badly regarded by enthusiasts. But most enthusiasts (myself excepted) skipped P4 - either went socket 462 with a VIA/nForce chipset, stuck to PIII/i815, etc - until RDRAM was dead and buried. And it's not just enthusiasts - outfits like Gateway that used to be Intel-only started selling Athlons in the i820 era.

I ended up with an Intel P4 SDRAM motherboard recently, and even for $30.00 CPU included (1.😎 and in mint condition, no one on Ebay wanted it.

VivienM wrote on 2024-09-19, 01:56:

I don't think we're saying Intel pushed RDRAM because it made them more money. We're saying Intel pushed RDRAM. Something in the […]
Show full quote

ElectroSoldier wrote on 2024-09-18, 21:30:

The P3 doesnt need the high bandwidth you get with RDRAM. The pairing makes little sense. Which is exactly why they dropped it o […]
Show full quote

The P3 doesnt need the high bandwidth you get with RDRAM. The pairing makes little sense. Which is exactly why they dropped it on the P3 after just over a year of trying to push it.
Intel was in partnership with RAMBUS, so they had to push it but they quickly realised it just wasnt working and dropped the 820 like a hot brick after the MTH debacle.

You guys are trying to make out like there is some kind of "conspiracy theory" going on where by Intel was deliberately pushing a load of crap on people who didnt want it because it made them more money.
I guess in a way it is kind of true, but it is also a load of old rubbish at the same time, it was more about the technology and the matching of the technologies that went best together than it was about screwing more money out of its customers, which is born out by the fact that Intel dropped the use of RDRAM on the P3 after just one year, they dropped the use of RDRAM as soon as DDR SDRAM emerged as a technology that was capable of offering the bandwidth at a more competitive price.
Dont forget while Intel was taking up the use of DDR on its chipsets it was still in partnership with RAMBUS.

I don't think we're saying Intel pushed RDRAM because it made them more money. We're saying Intel pushed RDRAM. Something in the late 1990s clearly convinced Intel that RDRAM was the future of their high-end platforms, and they pursued that both on PIII and P4.

And, in an alternative world where Samsung & co were making RDRAM RIMMs at much lower prices, it might yet have succeeded. Or in an alternative world where the price of PCxxx SDRAM wasn't plummeting. Or an alternative world in which Rambus had gotten away with their attempt to insert patented content into the JEDEC specs, forcing the entire SDRAM/DDR industry to pay royalties.

And on both PIII and P4, the attempts to push RDRAM hurt the success of the platform. Intel had to embarassingly back away from RDRAM both times. And as much as everybody loves to bash on HotBurst and Preshot and whatnot today, DDR P4s in 2003-4ish were not that badly regarded by enthusiasts. But most enthusiasts (myself excepted) skipped P4 - either went socket 462 with a VIA/nForce chipset, stuck to PIII/i815, etc - until RDRAM was dead and buried. And it's not just enthusiasts - outfits like Gateway that used to be Intel-only started selling Athlons in the i820 era.

I have both PIII with SDR and DDR .. looks like I have to hunt down a RDR board now .. For science and my own curiosity, Tualatin and DDR is already stupidly fast when overclocked. I doubt RDR will beat it but my brain wants to know.

VivienM wrote on 2024-09-19, 01:45:

I think it's simpler than that: - for about 6 months in early 2000, Intel did not have a competitive leading-edge SDRAM chipset […]
Show full quote

ElectroSoldier wrote on 2024-09-18, 10:36:

Ok fine. I dont see that in the real world going by the artefacts that are left to see but ok we can roll with that as an idea. […]
Show full quote

VivienM wrote on 2024-09-18, 00:42:

They basically starved the market of non-RDRAM chipsets. […]
Show full quote

They basically starved the market of non-RDRAM chipsets.

And they did so in two phases:
1) Phase 1. i820. Basically, if you wanted 133MHz FSB, 4X AGP, UltraATA 66, or any of the other platform improvements, and you wanted an AGP slot, then you needed an i820 with RDRAM.
When the market resisted and 440BX stuck around much longer than expected, they decided to come up with their MTH to 'translate' SDRAM into RDRAM. Then that... turned out to be a snafu and got recalled. So they gave away 128 megs of RDRAM to everybody who had the Intel CC820 board.
Then they were forced to launch the i815, which got you all the nice platform improvements over the 440BX but with SDRAM support. That was the end of RDRAM on the PIII platform - maybe a few high-end Dells got sold with i820/i840 but that's about it.
(As an aside, as someone who would never have touched a VIA chipset back in the day and who is still shopping for an additional retro system, I am continually astounded at how many socket 370 boards/big OEM systems are out there with Via's Apollo 133 chipsets. That's the legacy of Intel not offering a decent non-RDRAM chipset before the i815.)

2) Phase 2. Pentium 4. The Pentium 4 launched with i850 and RDRAM only. I think the MTH was supposed to come back, but didn't. So, again, if you wanted Intel's flagship processor, you needed RDRAM.
After close to a year, they launched the first iteration of the i845, but that only supported PC133 SDRAM, not DDR, and performance was mediocre. I'm googling this and reading old AnandTech articles, apparently it was crazier than I remembered - the i845 could do DDR, but Intel strictly, strictly prohibited motherboard manufacturers from making DDR boards until at least Jan. 1, 2002. And I haven't researched it and I don't remember offhand, but I'm pretty sure the DDR i845 boards remained unappealing alternatives to i850E and it was only with i865/i875 that DDR platforms for P4 really became good.

So basically, for the first 15ish months at least, if you wanted a P4, you had to get RDRAM.

How that manifested itself in the real world? If you weren't VIA-phobic, in late 2001, you were getting a KT266 motherboard with your Athlon XP on socket 462.

Intel had a chipset problem from late 1999 until the launch of the i865 in 2003. There was a gap of about a year where the i815 provided a decent alternative, at least if you could live with the max 512 megs of RAM at a time when 256 megs of SDRAM started costing $60CAD. But otherwise, Intel's chipset options were always flawed because the decent/high-end option was only RDRAM.

(Disclaimer: I had a Willamette P4 with a gig of RDRAM. Never had a real problem with it, although what annoyed me was that the Northwoods that came out a month later used PC1066 RDRAM and mine was PC800... so oops, zero upgrade path for me.)

Ok fine.
I dont see that in the real world going by the artefacts that are left to see but ok we can roll with that as an idea.

I think I would like to seperate the Pentium 3 and the Pentium 4 here, certainly I would agree that intel pushed RAMBUS on us with socket 423 as when you look for an find a socket 423 board all you ever seem too get is RAMBUS, but with the Pentium 3 chipsets there are so many boards out there that use SDRAM (440/810/815 chipsets) then it is hard to see how a company that controlled the supply starved the market because there are so many out there.

I remember the realease of RAMBUS during the Pentium 3 life time, I remember the articles in magazines talking about it and how it was so much more expensive than SDRAM and while it filled a high end neich the majority of computers are not high end (something Intel was all to aware of) and most magazines pushed SDRAM systems.
If for what you are saying was true then you would be assuming that Intel wanted to ignore the majority of its sales and squeeze more money out of buyers, by selling them RAMBUS boards over SDRAM boards, and they did that by deliberately not producing enough SDRAM boards to satisfy demand.
And like I said ok I can work with that as an idea.
But the physical items we are left with is a lot more Pentium 3 boards that use SDRAM than RAMBUS. And if what you are saying was true then that would be the other way around, because in your idea intel starved the market of 810/185 chipsets.

The actions dont fit the words. And the actions are unchangable. So it is the words (your ideas about what they did) that must be wrong as stated.

Because if you were right in the case of the Pentium 3 then you would have a similar situation as you have with socket 423 boards.

I think it's simpler than that:
- for about 6 months in early 2000, Intel did not have a competitive leading-edge SDRAM chipset for Pentium III
- Intel's solution was the i820 with RDRAM; a few Dell XPS Bxxx machines were sold, but that's about it. (I would note that I have seen XPS Bxxx machines on eBay so some have survived) Otherwise i820 was a flop.
- many, many people went to VIA's chipsets, and again, you see a strangely high number of socket 370 (or slot 1) motherboards with VIA's chipsets from this era. You don't see very many VIA chipset PIIs or Katmais, but you do see them in the 133MHz Coppermine era. Including in a number of large OEM systems from HP, etc.
- i810 was also an option, and again, you see a number of surviving systems with nice 933/1GHz Coppermines with i810 boards.
- some of the Taiwanese motherboard manufacturers put together socket 370 440BX boards that were basically intended to be overclocked at 133MHz FSB
- other OEMs explored AMD platforms, and indeed, I have seen some Athlon Gateway boxes on eBay from this era. Dell was one of the very few large OEMs to stick with a 100% Intel lineup.

It's not "Intel restricted sales of the i810/i815". It's "Intel had a gap in their product line until they scrambled to launch the i815." And I see plenty of evidence looking at what Coppermine-era hardware is available on eBay or elsewhere to back that up.

And maybe that's the difference between this era in PIII land and socket 423. In socket 423, you had no options - no VIA/SiS SDRAM chipsets, no Intel options until the i845 PC133 SDRAM chipset launched, etc. If you didn't want RDRAM, you got a PIII or an Athlon. In PIII land, you had options - VIA's chipsets, sticking to 440BX with 100MHz FSB, giving up on AGP with i810, etc. And if you look at PIII motherboards/systems on eBay today, you'll see that motherboard manufacturers and large OEMs resorted to all of those options. But those options were all much worse than a proper Intel-designed chipset which is why when the i815 launched everybody went that way.

That lot needs a lot of unpacking.

They didnt starve the market of none RDRAM chipsets, there were two options, the 440 and the 810. One was PC100 and the other supported PC133.

Yes there was no "high end" none 820 P3 option from Intel. They saw the future of the P3 as being with RDRAM, so there was a gap in the market for people who sort of want performance but dont really want to pay for it.
People wanted the cheapness of the 810 chipset but that took barebones all the way down the nothing. Like no agp support.

Now we have established what you said is wrong we have also seen that it is exactly backwards because there are twice as many chipsets that support the P3/SDRAM as there are the support the P3/RDRAM.

People say they want high end but dont want to pay for high end. Which means people want mid range but what to think its high end. Which is not the same thing.

For me the mating of RDRAM to the P3 is a complete mess, I can only think the reason they pushed it for the year they did was because they got into bed with RAMBUS and so had to push it, but when they saw the mating wasnt a good one and that there was no uptake of it due to cost (refer to the above as to why that was), the MTH debacle put it all to bed and Intel came up with the 815, which wasnt a high end option, but it had a veneer of high end so people were happy with that.

The VIA Apollo 133 chipset is proof of that in its own right, because when you look at the features it wasnt great, you yourself said as much in that you wouldnt consider a VIA chipset until that point because of the reasons you gave. It was the cheap out option compared to the Intel offering, but because the Intel offering was to cheap or to expensive you went for the mid range VIA option, which was made even sweeter because of the price.

Also as you say the abundance of the Apollo 133 chipsets out there shows that people like the mid range option but want to think its high end.

The idea that the 440 stuck around longer than they though....THEY WERE THE ONES WHO WERE MAKING IT!
The real problem was that the 810 chipset was to cheap. To stripped to many features that were utter game killers so people couldnt go for it because it didnt even cover the basics they needed. which is as you said AGP and a decent ATA interface. It was a load of old crap and no mistake. When I see those boards on ebay I think of Windows 95 running a pre GeForce 256 video card and some old games like CnC, DooM and ResEvil struggling to play... Its not, but thats what I think of and going by the prices on the fact they are always on there no selling other people think the same too.

Dont forget all this occupied about a year in time. It wasnt long before it was here and gone. The Apollo 133 stuck around because it was cheap and gave nice features.

Yes it was at this time the over clocked 440 chipset boards came out. All with a disclaimer on the box, on a slip of paper inside the box telling you that overclocking might be a problem and that you do it at your own risk.

Athlon Thunderbird 1ghz vs P3 coppermine 1ghz, which one is better?

Depends.

VivienM wrote on 2024-09-19, 01:56:

I don't think we're saying Intel pushed RDRAM because it made them more money. We're saying Intel pushed RDRAM. Something in the […]
Show full quote

ElectroSoldier wrote on 2024-09-18, 21:30:

The P3 doesnt need the high bandwidth you get with RDRAM. The pairing makes little sense. Which is exactly why they dropped it o […]
Show full quote

The P3 doesnt need the high bandwidth you get with RDRAM. The pairing makes little sense. Which is exactly why they dropped it on the P3 after just over a year of trying to push it.
Intel was in partnership with RAMBUS, so they had to push it but they quickly realised it just wasnt working and dropped the 820 like a hot brick after the MTH debacle.

You guys are trying to make out like there is some kind of "conspiracy theory" going on where by Intel was deliberately pushing a load of crap on people who didnt want it because it made them more money.
I guess in a way it is kind of true, but it is also a load of old rubbish at the same time, it was more about the technology and the matching of the technologies that went best together than it was about screwing more money out of its customers, which is born out by the fact that Intel dropped the use of RDRAM on the P3 after just one year, they dropped the use of RDRAM as soon as DDR SDRAM emerged as a technology that was capable of offering the bandwidth at a more competitive price.
Dont forget while Intel was taking up the use of DDR on its chipsets it was still in partnership with RAMBUS.

I don't think we're saying Intel pushed RDRAM because it made them more money. We're saying Intel pushed RDRAM. Something in the late 1990s clearly convinced Intel that RDRAM was the future of their high-end platforms, and they pursued that both on PIII and P4.

And, in an alternative world where Samsung & co were making RDRAM RIMMs at much lower prices, it might yet have succeeded. Or in an alternative world where the price of PCxxx SDRAM wasn't plummeting. Or an alternative world in which Rambus had gotten away with their attempt to insert patented content into the JEDEC specs, forcing the entire SDRAM/DDR industry to pay royalties.

And on both PIII and P4, the attempts to push RDRAM hurt the success of the platform. Intel had to embarassingly back away from RDRAM both times. And as much as everybody loves to bash on HotBurst and Preshot and whatnot today, DDR P4s in 2003-4ish were not that badly regarded by enthusiasts. But most enthusiasts (myself excepted) skipped P4 - either went socket 462 with a VIA/nForce chipset, stuck to PIII/i815, etc - until RDRAM was dead and buried. And it's not just enthusiasts - outfits like Gateway that used to be Intel-only started selling Athlons in the i820 era.

Yeah but that was how it came across, which is why I said what I said and this whole spur of the thread started.

There was a technicnological reason why the P4 used RDRAM.
There was a business reason why the P3 used RDRAM.

You might not agree with them, and in the latter case I dont think it should have been done at all, however it doesnt mean those things were done because Intel are money grabbing *

GreenBook wrote on 2024-09-19, 11:31:

Athlon Thunderbird 1ghz vs P3 coppermine 1ghz, which one is better?

I think Philscomputerlab did a video on that.

ElectroSoldier wrote on 2024-09-19, 09:55:

They didnt starve the market of none RDRAM chipsets, there were two options, the 440 and the 810. One was PC100 and the other supported PC133.

The 440BX didn't support 133FSB processors. The 810 didn't support discrete video.

In October 1999, they launched 133FSB processors. With one decent chipset option - the i820. They did not have a proper, full-featured chipset SDRAM option for the 133MHz FSB processors until mid-2000 with the i815.

Is that not obviously a problem? From October, 1999 to July 2000, you had four options if you wanted to use Intel's flagship 133FSB processors:
- a chipset with no AGP (which is a huge problem for a flagship processor)
- a chipset with RDRAM
- a third-party SDRAM chipset from VIA or maybe SiS
- overclock the 440BX

For eight months, Intel did not provide a chipset that allowed its flagship processors to be paired with SDRAM unless you were willing to give up on discrete graphics.

ElectroSoldier wrote on 2024-09-19, 09:55:

Yes there was no "high end" none 820 P3 option from Intel. They saw the future of the P3 as being with RDRAM, so there was a gap in the market for people who sort of want performance but dont really want to pay for it.
People wanted the cheapness of the 810 chipset but that took barebones all the way down the nothing. Like no agp support.

Now we have established what you said is wrong we have also seen that it is exactly backwards because there are twice as many chipsets that support the P3/SDRAM as there are the support the P3/RDRAM.

Excuse me? What have I said that is wrong?????????

They had a strategy. With the 133MHz FSB Coppermine processors, people were supposed to go to RDRAM on the i820. And then Celerons were supposed to be a 66MHz on i810 and SDRAM.

The high-end of that strategy failed as the i820/RDRAM was not well-received. Their first 'solution' was the MTH; when that flopped, they rushed and launched the i815 and gave up on RDRAM for Coppermine (and Tualatin).

ElectroSoldier wrote on 2024-09-19, 09:55:

For me the mating of RDRAM to the P3 is a complete mess, I can only think the reason they pushed it for the year they did was because they got into bed with RAMBUS and so had to push it, but when they saw the mating wasnt a good one and that there was no uptake of it due to cost (refer to the above as to why that was), the MTH debacle put it all to bed and Intel came up with the 815, which wasnt a high end option, but it had a veneer of high end so people were happy with that.

The i815 was high-end enough - it supported the 133MHz FSB processors, AGP 4X, UltraATA 66, all the 'high end' features expected in mid-2000. Other than the ability to drive more than 512 megs of RAM, what was really missing from the i815 to build a decent gaming/enthusiast system?

ElectroSoldier wrote on 2024-09-19, 09:55:

The VIA Apollo 133 chipset is proof of that in its own right, because when you look at the features it wasnt great, you yourself said as much in that you wouldnt consider a VIA chipset until that point because of the reasons you gave. It was the cheap out option compared to the Intel offering, but because the Intel offering was to cheap or to expensive you went for the mid range VIA option, which was made even sweeter because of the price.

Also as you say the abundance of the Apollo 133 chipsets out there shows that people like the mid range option but want to think its high end.

The "too expensive" chipset, to be clear, was not too expensive because of the chipset. It was 'too expensive' because it required RDRAM.

ElectroSoldier wrote on 2024-09-19, 09:55:

The idea that the 440 stuck around longer than they though....THEY WERE THE ONES WHO WERE MAKING IT!

People like Dell were cranking out mad numbers of 100MHz FSB 440BX systems in June 2000 (they even sold me one!). Those systems should have had 133MHz FSB processors, and indeed got replaced by 133MHz processors when the i815 came out.

That's what this keeps coming down to - if you wanted an Intel chipset, one of those 133 FSB processors and AGP, for 8 months, you had to get RDRAM. And because almost no one was willing to pay for RDRAM, they ended up having to compromise - either go VIA, go 100MHz FSB/2X AGP on 440BX, or give up AGP on i810.

ElectroSoldier wrote on 2024-09-19, 09:55:

The real problem was that the 810 chipset was to cheap. To stripped to many features that were utter game killers so people couldnt go for it because it didnt even cover the basics they needed. which is as you said AGP and a decent ATA interface. It was a load of old crap and no mistake. When I see those boards on ebay I think of Windows 95 running a pre GeForce 256 video card and some old games like CnC, DooM and ResEvil struggling to play... Its not, but thats what I think of and going by the prices on the fact they are always on there no selling other people think the same too.

The i810 was intended for low-end eMachines systems for grandma to get on the Internet. Or for office workers doing word processing and email. Great chipset for that.

ElectroSoldier wrote on 2024-09-19, 09:55:

Dont forget all this occupied about a year in time. It wasnt long before it was here and gone. The Apollo 133 stuck around because it was cheap and gave nice features.

In those days, a lot happened in PC land in a year. It's not like now...

VivienM wrote on 2024-09-19, 21:54:

The 440BX didn't support 133FSB processors. The 810 didn't support discrete video. […]
Show full quote

ElectroSoldier wrote on 2024-09-19, 09:55:

They didnt starve the market of none RDRAM chipsets, there were two options, the 440 and the 810. One was PC100 and the other supported PC133.

The 440BX didn't support 133FSB processors. The 810 didn't support discrete video.

In October 1999, they launched 133FSB processors. With one decent chipset option - the i820. They did not have a proper, full-featured chipset SDRAM option for the 133MHz FSB processors until mid-2000 with the i815.

Is that not obviously a problem? From October, 1999 to July 2000, you had four options if you wanted to use Intel's flagship 133FSB processors:
- a chipset with no AGP (which is a huge problem for a flagship processor)
- a chipset with RDRAM
- a third-party SDRAM chipset from VIA or maybe SiS
- overclock the 440BX

For eight months, Intel did not provide a chipset that allowed its flagship processors to be paired with SDRAM unless you were willing to give up on discrete graphics.

ElectroSoldier wrote on 2024-09-19, 09:55:

Yes there was no "high end" none 820 P3 option from Intel. They saw the future of the P3 as being with RDRAM, so there was a gap in the market for people who sort of want performance but dont really want to pay for it.
People wanted the cheapness of the 810 chipset but that took barebones all the way down the nothing. Like no agp support.

Now we have established what you said is wrong we have also seen that it is exactly backwards because there are twice as many chipsets that support the P3/SDRAM as there are the support the P3/RDRAM.

Excuse me? What have I said that is wrong?????????

They had a strategy. With the 133MHz FSB Coppermine processors, people were supposed to go to RDRAM on the i820. And then Celerons were supposed to be a 66MHz on i810 and SDRAM.

The high-end of that strategy failed as the i820/RDRAM was not well-received. Their first 'solution' was the MTH; when that flopped, they rushed and launched the i815 and gave up on RDRAM for Coppermine (and Tualatin).

ElectroSoldier wrote on 2024-09-19, 09:55:

For me the mating of RDRAM to the P3 is a complete mess, I can only think the reason they pushed it for the year they did was because they got into bed with RAMBUS and so had to push it, but when they saw the mating wasnt a good one and that there was no uptake of it due to cost (refer to the above as to why that was), the MTH debacle put it all to bed and Intel came up with the 815, which wasnt a high end option, but it had a veneer of high end so people were happy with that.

The i815 was high-end enough - it supported the 133MHz FSB processors, AGP 4X, UltraATA 66, all the 'high end' features expected in mid-2000. Other than the ability to drive more than 512 megs of RAM, what was really missing from the i815 to build a decent gaming/enthusiast system?

ElectroSoldier wrote on 2024-09-19, 09:55:

The VIA Apollo 133 chipset is proof of that in its own right, because when you look at the features it wasnt great, you yourself said as much in that you wouldnt consider a VIA chipset until that point because of the reasons you gave. It was the cheap out option compared to the Intel offering, but because the Intel offering was to cheap or to expensive you went for the mid range VIA option, which was made even sweeter because of the price.

Also as you say the abundance of the Apollo 133 chipsets out there shows that people like the mid range option but want to think its high end.

The "too expensive" chipset, to be clear, was not too expensive because of the chipset. It was 'too expensive' because it required RDRAM.

ElectroSoldier wrote on 2024-09-19, 09:55:

The idea that the 440 stuck around longer than they though....THEY WERE THE ONES WHO WERE MAKING IT!

People like Dell were cranking out mad numbers of 100MHz FSB 440BX systems in June 2000 (they even sold me one!). Those systems should have had 133MHz FSB processors, and indeed got replaced by 133MHz processors when the i815 came out.

That's what this keeps coming down to - if you wanted an Intel chipset, one of those 133 FSB processors and AGP, for 8 months, you had to get RDRAM. And because almost no one was willing to pay for RDRAM, they ended up having to compromise - either go VIA, go 100MHz FSB/2X AGP on 440BX, or give up AGP on i810.

ElectroSoldier wrote on 2024-09-19, 09:55:

The real problem was that the 810 chipset was to cheap. To stripped to many features that were utter game killers so people couldnt go for it because it didnt even cover the basics they needed. which is as you said AGP and a decent ATA interface. It was a load of old crap and no mistake. When I see those boards on ebay I think of Windows 95 running a pre GeForce 256 video card and some old games like CnC, DooM and ResEvil struggling to play... Its not, but thats what I think of and going by the prices on the fact they are always on there no selling other people think the same too.

The i810 was intended for low-end eMachines systems for grandma to get on the Internet. Or for office workers doing word processing and email. Great chipset for that.

ElectroSoldier wrote on 2024-09-19, 09:55:

Dont forget all this occupied about a year in time. It wasnt long before it was here and gone. The Apollo 133 stuck around because it was cheap and gave nice features.

In those days, a lot happened in PC land in a year. It's not like now...

I never said the 440 supported a 133FSB.
The only way to get that was to overclock it.

The 810 did support "discrete" video in the form of a PCI addin card, it did not support the AGP.

The idea Intel had and the one you seem to have are not the same thing.
It was Intels idea to have the 820 occupy the top spot and the 810 would be the low end. So if you wanted an office PC then get the cheap option, otherwise buy the other one.

Its only a problem if you are trying to cheap out on the sale.
Wanting the 810 chipset but with AGP and ATA 66 is the idea right, i mean that way you get the best of both worlds, you get all the features you want but dont have to pay for it.
The 810 came before the 820, but they planned to release it, you cant have all things all at once. It wasnt like there wasnt an option to go with the 440 if you wanted AGP in 1999, or you could wait because on its release the 820 was already known.

Im not going to say that putting RDRAM into any P3 board was a good idea. It wasnt, it was a mistake, and in some ways I think it set back development years, but it was what it was. Intel had the idea that they had.
It was a mistake. The P3 doesnt need high bandwidth RAM its architecture needs low latency, which is what RDRAM doesnt provide. Which is why I dont really understand why they did it, I can only imagine it was a business limitation due to getting into bed with Rambus.

The part which is wrong is you said Intel starved the market of SDRAM boards.
They didnt, the made the 440 and the 810 chipsets, both of which support SDRAM.
At the time everybody knew the 820 was coming soon. With the pace of change at the time you couldnt have everything all at once.
I remember the magazines of the time talking about it all.

Nothing is missing from the 815. It was fine, it was made to replace the 820, and it did rather well.

Yes the 820 was an expensive option because of the cost of the RDRAM not the chipset itself. Pedantry, but yeah... When you mated the 820 to SDRAM it was a disaster, so the 820 and RDRAM go hand in glove.
You dont get the one without the other, or you get a disaster.

I wouldnt disagree with any of that part, they created a problem that lasted ages because you got low end with missing features, old or you couldnt afford it.

I remember all this going on like it was yesterday. I remember the 440BX being a year old for instance.
Its not like it was a dinosaur, it was decent and a lot of people had only just bought it.

ElectroSoldier wrote on 2024-09-19, 22:37:

The part which is wrong is you said Intel starved the market of SDRAM boards. They didnt, the made the 440 and the 810 chipsets, […]
Show full quote

The part which is wrong is you said Intel starved the market of SDRAM boards.
They didnt, the made the 440 and the 810 chipsets, both of which support SDRAM.
At the time everybody knew the 820 was coming soon. With the pace of change at the time you couldnt have everything all at once.
I remember the magazines of the time talking about it all.

I'm not even sure what we're arguing about anymore.

In September/October 1999, they launched the 133MHz FSB Coppermines with the i820 chipset. They did not offer an SDRAM option for the 133MHz FSB chips, at least not one with AGP. They expected the middle-to-high-end of the market to adopt RDRAM. But RDRAM was priced out of contention while the price of SDRAM was just plunging.

By late January 2000, they were launching the ugly MTH hack. And recalling it by April.

Finally, in late June, they launch the i815 and put the whole thing behind them. (Until they recreated a similar mess with the P4... starting in November, six months later)

You may want to read the great Anand Lal Shimpi's take on 2000's chipset situation - https://www.anandtech.com/show/693/5 . And actually, Anand wrote that Intel was threatening to cut 440BX shipments to Taiwanese board makers.

How did they not "starve" the market of SDRAM chipsets?!? If they weren't obsessed with pushing RDRAM, they would presumably have launched the 133MHz coppermines with an i815-like SDRAM chipset and have avoided this weird awkward mess for 8 months, not to mention the MTH recall. Not to mention putting some of their most loyal large OEMs like Dell in a bit of a bind - hard to market an RDRAM i820 system against a high-end SDRAM Athlon. And I could be wrong about this, but I think this was the time when other large Intel-only OEMs like Gateway started offering AMD.

Honestly, the whole thing reminds me a bit of x64 vs Itanium - Intel had their dumb stubborn idea, someone else comes out with a more practical alternative, Intel is forced to adopt the more practical alternative, except in this case it's VIA's SDRAM chipsets playing the role of AMD's x64.

And believe me, I remember that period well too. I ordered one of Dell's last 440BX Slot 1 systems in late June 2000. And everybody was talking about this situation, the impending i815 launch, etc.

VivienM wrote on 2024-09-20, 02:22:

I'm not even sure what we're arguing about anymore. […]
Show full quote

ElectroSoldier wrote on 2024-09-19, 22:37:

The part which is wrong is you said Intel starved the market of SDRAM boards. They didnt, the made the 440 and the 810 chipsets, […]
Show full quote

The part which is wrong is you said Intel starved the market of SDRAM boards.
They didnt, the made the 440 and the 810 chipsets, both of which support SDRAM.
At the time everybody knew the 820 was coming soon. With the pace of change at the time you couldnt have everything all at once.
I remember the magazines of the time talking about it all.

I'm not even sure what we're arguing about anymore.

In September/October 1999, they launched the 133MHz FSB Coppermines with the i820 chipset. They did not offer an SDRAM option for the 133MHz FSB chips, at least not one with AGP. They expected the middle-to-high-end of the market to adopt RDRAM. But RDRAM was priced out of contention while the price of SDRAM was just plunging.

By late January 2000, they were launching the ugly MTH hack. And recalling it by April.

Finally, in late June, they launch the i815 and put the whole thing behind them. (Until they recreated a similar mess with the P4... starting in November, six months later)

You may want to read the great Anand Lal Shimpi's take on 2000's chipset situation - https://www.anandtech.com/show/693/5 . And actually, Anand wrote that Intel was threatening to cut 440BX shipments to Taiwanese board makers.

How did they not "starve" the market of SDRAM chipsets?!? If they weren't obsessed with pushing RDRAM, they would presumably have launched the 133MHz coppermines with an i815-like SDRAM chipset and have avoided this weird awkward mess for 8 months, not to mention the MTH recall. Not to mention putting some of their most loyal large OEMs like Dell in a bit of a bind - hard to market an RDRAM i820 system against a high-end SDRAM Athlon. And I could be wrong about this, but I think this was the time when other large Intel-only OEMs like Gateway started offering AMD.

Honestly, the whole thing reminds me a bit of x64 vs Itanium - Intel had their dumb stubborn idea, someone else comes out with a more practical alternative, Intel is forced to adopt the more practical alternative, except in this case it's VIA's SDRAM chipsets playing the role of AMD's x64.

And believe me, I remember that period well too. I ordered one of Dell's last 440BX Slot 1 systems in late June 2000. And everybody was talking about this situation, the impending i815 launch, etc.

The reason why is because there are several "mistakes" that keep coming up.

Ill just concentrate, this time, on the first paragraph and show you what I mean.

"In September/October 1999, they launched the 133MHz FSB Coppermines with the i820 chipset."

Right thats technically true. The rest of the statement is also true.
You make it sound like Intel have done something wrong there in doing that.
I cant find anything that shows PC-133 RAM actually on sale until 2000...

"Anand wrote that Intel was threatening to cut 440BX shipments to Taiwanese board makers. "

Where does he write that?

Because that and what was actually happening isnt the same thing.
The way you say it Intel was threatening to stop sending chips to manufacturers. Where as it was telling manufacturers it will be stopping production, and so supply would dry up.

ElectroSoldier wrote on 2024-09-20, 12:00:

The reason why is because there are several "mistakes" that keep coming up. […]
Show full quote

VivienM wrote on 2024-09-20, 02:22:

I'm not even sure what we're arguing about anymore. […]
Show full quote

ElectroSoldier wrote on 2024-09-19, 22:37:

The part which is wrong is you said Intel starved the market of SDRAM boards. They didnt, the made the 440 and the 810 chipsets, […]
Show full quote

The part which is wrong is you said Intel starved the market of SDRAM boards.
They didnt, the made the 440 and the 810 chipsets, both of which support SDRAM.
At the time everybody knew the 820 was coming soon. With the pace of change at the time you couldnt have everything all at once.
I remember the magazines of the time talking about it all.

I'm not even sure what we're arguing about anymore.

In September/October 1999, they launched the 133MHz FSB Coppermines with the i820 chipset. They did not offer an SDRAM option for the 133MHz FSB chips, at least not one with AGP. They expected the middle-to-high-end of the market to adopt RDRAM. But RDRAM was priced out of contention while the price of SDRAM was just plunging.

By late January 2000, they were launching the ugly MTH hack. And recalling it by April.

Finally, in late June, they launch the i815 and put the whole thing behind them. (Until they recreated a similar mess with the P4... starting in November, six months later)

You may want to read the great Anand Lal Shimpi's take on 2000's chipset situation - https://www.anandtech.com/show/693/5 . And actually, Anand wrote that Intel was threatening to cut 440BX shipments to Taiwanese board makers.

How did they not "starve" the market of SDRAM chipsets?!? If they weren't obsessed with pushing RDRAM, they would presumably have launched the 133MHz coppermines with an i815-like SDRAM chipset and have avoided this weird awkward mess for 8 months, not to mention the MTH recall. Not to mention putting some of their most loyal large OEMs like Dell in a bit of a bind - hard to market an RDRAM i820 system against a high-end SDRAM Athlon. And I could be wrong about this, but I think this was the time when other large Intel-only OEMs like Gateway started offering AMD.

Honestly, the whole thing reminds me a bit of x64 vs Itanium - Intel had their dumb stubborn idea, someone else comes out with a more practical alternative, Intel is forced to adopt the more practical alternative, except in this case it's VIA's SDRAM chipsets playing the role of AMD's x64.

And believe me, I remember that period well too. I ordered one of Dell's last 440BX Slot 1 systems in late June 2000. And everybody was talking about this situation, the impending i815 launch, etc.

The reason why is because there are several "mistakes" that keep coming up.

Ill just concentrate, this time, on the first paragraph and show you what I mean.

"In September/October 1999, they launched the 133MHz FSB Coppermines with the i820 chipset."

Right thats technically true. The rest of the statement is also true.
You make it sound like Intel have done something wrong there in doing that.
I cant find anything that shows PC-133 RAM actually on sale until 2000...

"Anand wrote that Intel was threatening to cut 440BX shipments to Taiwanese board makers. "

Where does he write that?

Because that and what was actually happening isnt the same thing.
The way you say it Intel was threatening to stop sending chips to manufacturers. Where as it was telling manufacturers it will be stopping production, and so supply would dry up.

I don't think I said anything about "stopping sending chips", although Anand (in the article I linked - https://www.anandtech.com/show/693/5 ) suggested that Intel would indeed slow down/stop shipments of the 440BX. (Which, in itself, isn't that nefarious - the 440BX was old, it just... didn't have a proper replacement in the lineup until June 2000)

What chips would they stop sending?

The problem is the chip they didn't launch. They shipped exactly zero chipsets that could support 133FSB, 4X AGP, UltraATA 66 with native SDRAM support between October-November 1999 and June 2000. They didn't stop shipping it because they never launched it.

If you were Dell in December 1999 and you had a nice little shipment of 133MHz FSB Coppermines and you called up Intel and said "hey guys, we'd like to buy some motherboards/chipsets to pair those processors with SDRAM and have some AGP slots", Intel did not have a product for Dell to order until the i815 launched in June 2000. Dell ended up selling a ton of XPS Txxxr 440BX 100FSB systems until June 2000; many others paired their 133MHz Coppermines with VIA chipsets.

Isn't there something a bit "odd" about the fact that Intel, which has dominated the market for chipsets for its own processors since, oh, I don't know, one of the earlier Pentium chipsets like the Triton or maybe even before did not have a passable chipset to offer to go along with its flagship processors for eight months?

I don't understand what is so controversial about this: the market wanted an SDRAM chipset with AGP 4X, 133MHz FSB support, etc to pair those lovely Coppermines with. Intel told people "you're going to use RDRAM and you're going to like it", at least until VIA came along with their own SDRAM chipset (and "coincidentally" at a time when the Slot A Athlons made AMD a player at the high end.). Then they scrambled, launched the i815 in June 2000 with a few weird face-saving quirks trying to pretend it wasn't a high-end chipset, and that was the end of RDRAM on the PIII platform.

And it seems clear enough to me that if it wasn't for i) the VIA chipset, and ii) AMD being a serious competitive threat (and doing so using SDRAM), they probably... would have eventually discontinued the 440BX (and/or the 100MHz Coppermines) without an SDRAM replacement. And really, the 440BX was becoming obsolete, the world was going to move towards the P4 anyways, etc., so whether it remained available or not was less important...

As I said earlier, the more I think about this, the more this reminds me of Itanium vs x64. Same story - Intel has some bold aggressive plan to just abruptly end the evolution of the existing stuff and force people to something new and different and expensive. And AMD/VIA/etc come along and undermine that plan, forcing Intel to adopt the lower-priced, lower-disruption path instead.