AMD's Rambus licence may be used in Hammer

[ToBeMe]

Interesting........ Didn't find a similar post with a search so I hope this is not a repost........

• IT'S IN THE MISTS of time, but we never should forget that AMD has a Rambus licence and it will use it if it needs to.
  Thanks to Mario, who writes for Van's Hardware, for pointing me to this public lecture by a senior technical bod at AMD, Kevin McGrath, who confirmed the existence of that licence at a public lecture.
  
  McGrath talks about the up-and-coming Hammer X86-64 architecture in the Webcast, which you can find here.
  
  There's also some interesting slides but you can hear him admit the existence of the Rambus licence by forwarding the vid to 1:05:30.
  
  Hammer may support both DDR (double data rate) memory and Rambus memory - shock...
  
  Compaq successfully uses Rambus architecture with some of its Alpha server boxes.
  
  Now Via - which successfully promoted DDR architecture so that now it appears to be the de facto standard for desktop PCs, doesn't have a licence for Rambus. But as it is building chipsets for Hammer (see separate story), it may well have to get one too. SIS, of course, now has a Rambus licence. But Nvidia doesn't... unless something has radically changed, and we missed it.

 

[AndyHui]

We have a screenshot of RAMBUS partners from our RDRAM article from nearly 2 years ago.

While AMD does hold an RDRAM licence, I must admit that they are missing from the current RAMBUS partners page, but so are the other System Controller manufacturers like ALi. Then again, Apacer is a division of ALi.

 

[dullard]

I've seen articles for quite some time that quote AMD people as saying: we will use RDRAM if we need to. The current Athlons use up basically all the DDR bandwidth - there isn't much room to spare. You'd have to be quite naive to think a 50% faster processor wouldn't be bottlenecked by DDR bandwidth. DDR isn't going to get much faster (DDR2 maybe, but not DDR). Thus RDRAM is a perfect solution for AMDs newer and faster processors.

 

[mchammer187]

will hammer's integrated memory controller limit it to only 1 type of memory?

because i doubt they would make two different processors one for rambus other for DDR

and it seems like it would be very difficult to make one that could use both types of memory

 

[WilsonTung]

Yeah. Unless the mem controller supports DDR and RDRAM you need a new chip to switch mem types. Perhaps some sort of MTH coudl be built into a chipset.

I predict that RDRAM will be dominant in a few years.

 

[tigerbait]

this is OT, but related to rambus' future:
what you guys think of rambus from an investment standpoint.... i know the company is a bunch of rotten lawyers, but they have a good product and the stock is only ~ $5-6.

 

[Athlon4all]

Wow! I would be very curious to see how RDRAM does on HAMMER. That could though get complicated though. Even if the Hammer would have both DDR and RDRAM memory controller's, motherboard manufacturer's would need to have different traces. People would have to make sure they buy RDRAM Hammer Motherboards and RDRAM Hammer CPU's but this is interesting none the less.

As for RDRAm's future, I definately think that it has a future for the P4. I wouldn't hesitate to invest in them (but what do I know about Stock Trading?). DDR will prolly remain the memory of choice for AMD platforms, but for the P4, RDRAM will reign.

 

[Bluga]

interesting.

 

[AGodspeed]

<< this is OT, but related to rambus' future:
what you guys think of rambus from an investment standpoint.... i know the company is a bunch of rotten lawyers, but they have a good product and the stock is only ~ $5-6. >>

From an investment standpoint, it's definitely a risk, I don't think you could argue otherwise. Any company that has a majority of their revenue coming from a niche market like Rambus does is risky investment. I currently own stock in both AMD and Intel. AMD is clearly the riskier investment when you look at their financial numbers and market diversification. However, IMO, AMD's current technology and future plans look promising enough to take seriously, as does Intel's.

If you're going to invest in the company Rambus, PLEASE diversify by investing in Rambus's competitors as well (i.e. companies that make DDR-only DIMMs, like Micron, etc.).

Also, PLEASE invest in other industries as well, like the pharmaceutical industry, oil industry, real estate, etc. Technology is a very volatile industry; you don't want all your eggs in one tech basket when your portfolio takes a nose-dive in the future.

 

[Rand]

<< this is OT, but related to rambus' future:
what you guys think of rambus from an investment standpoint.... i know the company is a bunch of rotten lawyers, but they have a good product and the stock is only ~ $5-6. >>



Way too volatile, and their future as a desktop standard is quite dubious.
With even intel having dual channel DDR on their roadmaps now I wouldnt feel at all safe in investing in Rambus.
Most insiders have RDRAM with a decreasing marketshare from 2003 onwards, with it only peaking at about 10-13% in 2002. Even Rambus own roadmap still has DDR with a majority marketshare through 2005.

If your willing to take a chance though it might be worthwhile, it has the potential to pay off huge if everything works out right.
If you do choose to invest, look into diversifying your stock portfolio, it's never wise to put all you cash in one compnay... especially not one as widely disliked and based upon such a volatile financial base.

 

[AGodspeed]

Looks like Hammer won't support Rambus memory.

Not surprising if you ask me...

 

[ToBeMe]

<< Looks like Hammer won't support Rambus memory.

Not surprising if you ask me... >>


Well yeah..............Van would say that. I'm sure it's a long shot, but, it would be really nice to see...................

 

[ST4RCUTTER]

Most likely Hammer will initially support DDR, but if RDRAM can scale as high as RAMBUS is preaching, it may be a better choice for servers from a bandwidth point of view. Of course for some servers, latency can be more important than bandwidth. DDRII is still too much of a question mark. No matter how the dice fall though, AMD should be in a good position simply because they don't have to rely on third party chipset makers to determine their memory architectures.

 

[Remnant2]

What most people who say that the Hammer will need RAMBUS tend to overlook, is that AMD is making a 128-bit (dual channel) DDR interface version of the chip.

Unlike the Athlon, that extra memory bandwidth will not be clamped to a trickle by the FSB -- a DDR333x2 system would have 5.4GB/s bandwidth, which is more than the current dual channel RAMBUS implementations, and coupled with the lower latency of DDR would provide some really nice performance.

 

[BD231]

I wouldent count on seeing anything like RDRAM and AMD together for a LONG time. AMD seems to be trying to take advantage of DDR so we have to see how that works out first.

 

[Snoop]

I agree, people tend to forget DDR is going to get a major speed increase when companies implement dual channel, rambus is already using dual and is only marginally faster than DDR. IMO its only a matter of time till Rambus will be forced into bankruptcy or a major ip selloff. They have burned to many bridges with memory and cpu mfg, as well as wall street.

 

[Athlon4all]

Well, it's hit and miss on Dual Channel DDR. It could be very expensive and yet it could not. Really, we'll have to wait and see, but here's the thing. If someone ever implements a Single Channel RDRAM chipset designed for 32-Bit RIMM's, then there is no doubt about it in my mind, RDRAM will be the supiroior choice for the P4 at least, as for Hammer, well it would be interesting to see how RDRAM does on Hammer, but I think that in all odds, PC2700 will give Hammer enough bandwidth in a single channel for a while, and then, DDR-II should come and hopefully, we'll never see a real need for Dual Channel DDR in the mainstream

 

[KevDODOUBLEG]

I bet AMD will drop Rambus bacause of Intel Dropping Rambus. If you own Rambus stock, it's time to sell, they are going down. (I'm not anti Rambus or anything, I have a P4 with 512mb or RDRAM)

 

[Novgrod]

Yeah, what with intel dropping RDRAM, Rambus will be limited to the ps2 pretty soon

I don't know why everybody but intel has been so reluctant to include RDRAM in their solutions; perhaps latency?

NB: RMBS was up some 23% today because they could demo a new memory piece. Tomorrow, unfortunately for people who bought it, I think everybody is expecting it to tank.

I'm really reluctant to believe anything about rambus' performance gains on anything but the p4. Everybody had said that the p4 was designed to negate rambus' latency problems, and if other cpus aren't, I can't imagine them being so successful.

 

[vedin]

One word...doubtfull. AMD has worked long and hard to push DDR based systems in the last couple of years, I don't see why they would want to suddenly switch to a dying ram type. I call Rambus half dead already because other than P4 mobos, it's got little use. Granted, consoles still use it for some reason, but that's about it. Since Intel is dropping Rambus, I think it'll plummit, and DDR will only get cheaper and faster. Just my two cents.

 

[Azurik]

<< One word...doubtfull. AMD has worked long and hard to push DDR based systems in the last couple of years, I don't see why they would want to suddenly switch to a dying ram type. I call Rambus half dead already because other than P4 mobos, it's got little use. Granted, consoles still use it for some reason, but that's about it. Since Intel is dropping Rambus, I think it'll plummit, and DDR will only get cheaper and faster. Just my two cents. >>



Where do get that Intel is dropping Rambus?

Jack Robertson at EBN changed his story from yesterday evening to today. He did this without noting his correction (as I believe most professional journalists today now practice). In my opinion, any news articles that are changed should have the word "CORRECTION" or "UPDATED" prominently displayed in the heading of the article so that investors can see what changed and know that the article was corrected and/or changed. However, Jack Robertson did not do this. I wonder why?

Here are the two articles, paragraph by paragraph, with the first paragraph being the article that was published last night and the second paragraph, with the changes, from the article that appeared this morning. The most important change for those who do not wish to read the entire post is the ADDITION of this paragraph this morning, which did not appear in the article last night and which I believe is a very important and significant change. Please note the second sentence:

"Intel will continue using Direct Rambus memory with its network processors. Also, although not new products, the next iterations of its 850 and 860 chipsets, supporting a 533MHz front-side, will support RDRAM when they arrive, probably in the second half of this year."


I also want to point out he added the words "next generation" when referring to DDR chipsets in his article this morning. Here is the text from both articles. The changes are in bold:



"http://www.siliconstrategies.com/story/OEG20020226S0040

Intel to drop support of Rambus in new products
By Jack Robertson mailto:jroberts@cmp.com
EBN http://www.ebnews.com
(02/26/02 17:48 p.m. EST)


http://www.ebnews.com/story/OEG20020226S0040
Intel to drop support of Rambus in new CPU products
By Jack Robertson mailto:jroberts@cmp.com
EBN http://www.ebnews.com
(02/26/02, 05:48:54 PM EST)




Intel Corp. in the second half of this year will drop its final Direct Rambus DRAMs support in new products, it was learned Tuesday at the Intel Developers Forum.

Intel Corp. in the second half of this year will drop its final Direct Rambus DRAMs support in new computer products, it was learned Tuesday at the Intel Developers Forum.


The last RDRAMs used in Xeon workstations will be replaced by new chipsets supporting double data rate (DDR memory).

The last RDRAMs used in Xeon workstations will be replaced by new chipsets supporting double data rate (DDR memory).


An Intel workstation roadmap secured by EBN showed a Placer chipset with DDR SDRAMs for dual processor Xeon workstations, and a Granite Bay DDR chipset for uniprocessor Xeon workstations. They will replace the Intel 860 workstationchipset using RDRAM and Intel 850 with RDRAM.

An Intel workstation roadmap secured by EBN showed a Placer chipset with DDR SDRAMs for dual processor Xeon workstations, and a Granite Bay DDR chipset for uniprocessor Xeon workstations. They will replace the Intel 860 workstationchipset using RDRAM and Intel 850 with RDRAM.


The new Prestonia Xeon processor for servers introduced in January already uses DDR memory, supported either by Intel's own E7500 Plumas chipset or a third party vendor chipset from ServerWorks, Santa Clara, Calif. Intel's desktop and notebook lines since last year have virtually dropped RDRAM in favor of single data rate SDRAM or in January going to DDR memory as well.

The new Prestonia Xeon processor for servers introduced in January already uses DDR memory, supported either by Intel's own E7500 Plumas chipset or a third party vendor chipset from ServerWorks, Santa Clara, Calif. Intel's desktop and notebook lines since last year have virtually dropped RDRAM in favor of single data rate SDRAM or in January going to DDR memory as well.


Intel will continue using Direct Rambus memory with its network processors. Also, although not new products, the next iterations of its 850 and 860 chipsets, supporting a 533MHz front-side, will support RDRAM when they arrive, probably in the second half of this year.


When the new workstation DDR chipsets arrive, it will mark the end of a long and torturous episode when Intel tried to mandate Direct Rambus as the next generation memory for the PC and workstation markets.

But when the next generation of workstation DDR chipsets arrive, it will mark the end of a long and torturous episode when Intel tried to mandate Direct Rambus as the next generation memory for the PC and workstation markets.


After a series of embarrassing troubles and lukewarm reception by the PC market, Intel in the last year has reversed course and shifted rapidly to support SDRAM and DDR.

After a series of embarrassing troubles and lukewarm reception by the PC market, Intel in the last year has reversed course and shifted rapidly to support SDRAM and DDR.


A spokeswoman for Rambus Inc. said she couldn't comment on new Intel workstation chipsets supporting DDR, and referred all questions to Intel.

A spokeswoman for Rambus Inc. said she couldn't comment on new Intel workstation chipsets supporting DDR, and referred all questions to Intel.


Intel roadmaps released at IDF also continued to reveal new server processors slated to be introduced in 2004. On Monday, Mike Fister, senior vice president and general manager of the Intel Enterprise Platform group, revealed the existence of a new Montecito 64-bit processor to be built on the firm's 0.09-micron processes.

Intel roadmaps released at IDF also continued to reveal new server processors slated to be introduced in 2004. On Monday, Mike Fister, senior vice president and general manager of the Intel Enterprise Platform group, revealed the existence of a new Montecito 64-bit processor to be built on the firm's 0.09-micron processes.


A new Intel roadmap introduced at IDF Tuesday disclosed a Nocona processor for 32-bit servers to succeed Prestonia, also built on a 0.09-micron process.

A new Intel roadmap introduced at IDF Tuesday disclosed a Nocona processor for 32-bit servers to succeed Prestonia, also built on a 0.09-micron process.


Montecito and Nocona each reportedly may cover both the multiprocessor servers and dual processor servers with the same chip. Until now separate processor versions were needed for the two main server markets.

Montecito and Nocona each reportedly may cover both the multiprocessor servers and dual processor servers with the same chip. Until now separate processor versions were needed for the two main server markets."


=========================

In my opinion, journalists reporting on companies in the press have a significant responsibility to shareholders. Any news article on a company that is changed or corrected should have the word "Correction" or "Updated" very prominently displayed. It seems to me that the powers that be, i.e., SEC, Congress, and other interested parties should be taking a closer look at some of these journalists who repeatedly come out with negative articles about a company (knowing that their articles will be widely repeated in the press), and then, in the dark of night as it were, change their stories with no notice that their article was "corrected" or "updated" It is disappointing that Jack Robertson does not add those words to his changed articles, since his articles have definitely affected Rambus' share price in the past and will undoubtedly do so in the future.