Before you build a P4 read this!

[Guest]

"Also, I would like to say that some people out there believe that the DDR solution to the P4 may be limited to PC1600, since Intel chips require memory to be at the same speed as the FSB."

But wasnt intel going to introduce 133 fsb with Nortwood ? That would require PC2100 in a DDR configuration..

 

[SammyBoy]

Actually, there is a very high demand right now compared to supply. Intel sold its 1 millionth Pentium 4 during Q1 of this year and sales keep increasing.

Well, my only questions then are these: Although Intel has sold the 1 millionth unit, were they mostly to OEM and resellers in bulk, therefore, not nessicarily assembled yet, or is the one million referring to the total amount of P4-based machines built and sold to end-users?, What is 1 million compared to sales of P3s, Celerons, and AMD's lines in the same period (in other words, what was the total % of P4s sold compared to total CPUs sold)?, and Wasn't the expected arrival of the P4 on the market supposed to be a big deal, therefore, making the need for RDRAM chips great? Now, if any of those questions point to a situation where the actual need for RDRAM is great, then my assertion about RDRAM being a technology that manufactures are loath to accept would be correct.

It would seem then that the memory manufactures out there would rather continue to build SDRAM and DDR-RAM instead of paying royalties and licensing fees to Rambus (SDRAM and DDR-RAM being open source, RDRAM being patented by Rambus). Also, it seems that Intel's marketing department either failed to sell the P4's reliance on RDRAM to other companies, or they grossly underestimated the selling power of the P4, the latter being the least likely. Usually it is very rare for a company, be it Intel, AMD, IBM, or what not, to not be overly optimistic about a product, since it's their job is to give the most positive spin possible.

So, if those statements are true, it would seem that RDRAM is going to have an extremely tough time at becoming the standard, and actually, as long as Rambus keeps RDRAM technology proprietary (sp?) then it will never become a standard, as by the time the patent expires and Rambus is no longer collecting licensing fees and royalties, the next advancement in RAM will come along, rendering RDRAM and DDR obsolete.

-SammyBoy

Without Evil, there can be no Good. Therefore, without an Intel, there can be no AMD.

 

[SammyBoy]

Well, I don't know how good the source is, but <A HREF="http://freespace.virgin.net/m.warner/RoadmapQ102.htm" target="_new"> here</A> is a link to this quote:

"<b> Intel i845B chipset (Brookdale DDR),</b> for use with the Pentium 4 'Northwood', is expected to be released in Q1 2002. The DDR version of the Brookdale chipset may only be able to support PC1600 DDR SDRAM rather than the faster PC2100 variety, possibly due to complications with asynchronous FSB memory access and Intel's leaning towards RDRAM. It should be noted that if DDR Brookdale does not have support for PC2100 or higher, its performance is likely to be little better than the PC133 SDR version due to DDR's higher latencies. The DDR version of Brookdale is expected to use the ICH3 south bridge."

So, if the source is good, the technology unchanged, the analysis good, and everything else falls the same way, then no, Northwood will still be limited to PC1600 due to a 100MHz FSB... or at least that's what I think the reason would be.

-SammyBoy


Without Evil, there can be no Good. Therefore, without an Intel, there can be no AMD.

 

[Sojourn]

I believe the reason RDRAM is more thoroughly tested is because it has to be tested as an assembled RIMM, unlike SDRAM. This is one of the reasons it is more expensive.

Memory manufacturers ramped up RDRAM production last year and into this year expecting P4 sales to skyrocket, providing platforms for their memory. Intel failed to deliver the sales promised, however, and now the market is glutted with RDRAM memory. There is a lot of fab space that has been changed over to RDRAM production in the last year that is now sitting idle, and OEMs have large stockpiles of RDRAM (which is why they have been begging Intel to stop packaging RDRAM with the CPU.) The current competativeness of RDRAM pricing is a symptom of the market and misallocated fab resources. I don't see it lasting through the end of the year, especially as RDRAM alternative P4 platforms appear.

I wonder exactly how much of RamBus' revenue comes directly from Intel's RDRAM support? If Intel chooses to abandon RDRAM for DDR SDRAM, I wonder if the company will still float?


-= This is our wading pool.
Stop pissing in it. =-

 

[Raystonn]

"Intel failed to deliver the sales promised"

Not even close to true. They are selling very well. We hit 1 million P4 CPUs sold back in Q1.

-Raystonn

= The views stated herein are my personal views, and not necessarily the views of my employer. =

 

[Raystonn]

"Now, if any of those questions point to a situation where the actual need for RDRAM is great, then my assertion about RDRAM being a technology that manufactures are loath to accept would be correct."

I do not understand you here. If the need for RDRAM is great, manufacturers are all over it.

"it would seem that RDRAM is going to have an extremely tough time at becoming the standard"

In my opinion, they will become the standard for the next whole generation of processors. However, this is purely speculation, as is most of your post.

-Raystonn

= The views stated herein are my personal views, and not necessarily the views of my employer. =

 

[beans]

sojourn -

You said "There is a lot of fab space that has been changed over to RDRAM production in the last year that is now sitting idle, and OEMs have large stockpiles of RDRAM..."

How do you know that?


beans

 

[SammyBoy]

I would think that his speculation comes from a few things. First, Intel hasn't sold as many as they wanted to. Second, there were announcements weeks or months ago from some memory fabs that they were increasing productions. Third, it is safe to say that the current bottom falling out for RDRAM prices is caused not by decreased demand, but a glut of supply.

-SammyBoy

Without Evil, there can be no Good. Therefore, without an Intel, there can be no AMD.

 

[SammyBoy]

"Now, if any of those questions point to a situation where the actual need for RDRAM is great, then my assertion about RDRAM being a technology that manufactures are loath to accept would be correct."

I do not understand you here. If the need for RDRAM is great, manufacturers are all over it.

Well I don't blame you for not understanding that part of my post. I don't understand it either now that I reread it. But, I think that idea of that part was that manufactures were loath to accept that RDRAM had a future, and slow to jump on the bandwagon. I just wanted to point out that even with success of the P4, manufactures seemed to be hesitant to convert some of their lines over to RDRAM production. This is because I think that the RDRAM technology is seen as a stab in the back to the standards committee, developed during a time where Rambus was on that committee when DDRwas being developed. I'm sure that only the largest, most established fab plants and manufactures are able to get away with pushing aside the SDRAM and DDR-RAM and produce RDRAM chips. The standards committee is composed of all the large chip makers (except Rambus) and to produce a competeing product requires a lot of guts, finacial stability, and power by that manufacture.

-SammyBoy

Without Evil, there can be no Good. Therefore, without an Intel, there can be no AMD.

 

[beans]

Sam -

You say "...it is safe to say that the current bottom falling out for RDRAM prices is caused not by decreased demand, but a glut of supply."


In the absence of hard evidence of a glut, I'm not sure it's safe to say that. RDRAM prices may be down from previous levels, but they are still not competitive with prices of other technologies.

Rather than "the bottom fell out," from my perspective RDRAM prices dropped from "out of sight" to "out of reach." If I'm a typical customer, that's not going to clear any glut.


beans

 

[SammyBoy]

RDRAM will never be price competetive with the current techonology, as the making of RDRAM is much more intensive and costly. Now, in comparison to the old prices for RDRAM, it is possible to say that the bottom has fallen out, as the prices have dropped a lot, much as the P4 prices have dropped. But, this in not due to decreased demand (it has gone up over the last quarter and a half), so the only safe conclusion is that there has been an increase in supply capabilites, resulting in a glut on inventory.

-SammyBoy

Without Evil, there can be no Good. Therefore, without an Intel, there can be no AMD.

 

[beans]

Sam -


I think "never" is a rather strong term to use here, even if you're right about the relative complexities. Manufacturing technologies improve all the time, and production cost isn't the only consideration in setting prices anyway.


Regarding what constitutes the "bottom falling out," I think that implies the kind of panic selling you see on Wall Street from time to time, not just any large price drop. But, that's just my opinion. Would you say the bottom has fallen out of P4 prices?


RDRAM production capacity almost certainly is up, I agree. That may account for the price drop, because it generally makes sense for manufacturers to set prices at a level that will sell their production. That does not mean they will wait until there's a glut -- the warehouse fills up with RDRAM -- before making a move, however.

Finally, there could be some long-term strategy at work here that involves neither current production costs nor demand. So, I persist -- absent hard evidence of a glut, it is not at all safe to say that current pricing is the result of oversupply.


beans

 

[Guest]

"They are selling very well. We hit 1 million P4 CPUs sold back in Q1."

Just curious.. how many P3's and Celerons did they sell in Q1 ? I have no clue if 1 million is a lot or not..