Sorry Rambus lovers.

[HolyGrenade]

The price is on the rise...


<font color=red><i>Tomorrow I will live, the fool does say
today itself's too late; the wise lived yesterday

 

[Raystonn]

Not according to Pricewatch, where RDRAM prices are either the same or lower. There isn't a single upward red arrow in the lot.

-Raystonn


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

 

[FatBurger]

That's only for one day though. If they go up $250 for 128meg one day, then back down $1 the next, you'd say the prices are going down.

Rambus is trying to recoup from their lawsuits.



<font color=blue>Quarter pounder inside</font color=blue>
<font color=red>Change the Sig of the Week!!!</font color=red>

 

[Raystonn]

Rambus doesn't set the prices on RDRAM. They only receive royalties from RDRAM products that others, such as Samsung, sell. These royalty rates were agreed upon in contracts quite a long time ago and cannot be changed.

-Raystonn


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

 

[FatBurger]

Doh! You're right, my bad.



<font color=blue>Quarter pounder inside</font color=blue>
<font color=red>Change the Sig of the Week!!!</font color=red>

 

[jc14all]

Seems to me that Rambus is really in need of those royalties at the moment for their <b><A HREF="http://biz.yahoo.com/prnews/010831/phf019.html" target="_new">Class Action</b></A> settlement on SDRAM. "Oh, what an evil web we weave, when we set our heart to deceive."

JC-------<*){{{>{~~~~~
Fisher of men

 

[flamethrower205]

OOO, Rambus really done it now.

When I rule the world, Apple will only mean the fruit.

 

[Guest]

Great post! What an article. I suppose that every investor of RAMBUS or perhaps buyers of RAMBUS technology that read the Rambus SEC report may have grounds for joining the class action. Rambus will probably have no money to pay if it loses. I am sure Intel will be dragged in and this will play out over several years.

I prepare SEC filings sometimes. If you lie to your auditor or put misleading things in your 10Q or 10K, the SEC can make your life miserable. You can get criminal or civil actions. Not pleasant at all!

 

[Raystonn]

"I suppose that every investor of RAMBUS or perhaps buyers of RAMBUS technology that read the Rambus SEC report may have grounds for joining the class action."

It's only for stockholders. The lawsuit is in regards to SDRAM only, which is disappearing soon anyway. (It cannot scale to the bus speeds soon to be released.)


"I am sure Intel will be dragged in and this will play out over several years."

There's no reason for Intel to get involved. This is solely between Rambus and their stockholders.

-Raystonn


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

 

[Guest]

You absolteuly right Raystonn. Lawsuit is against Rambus. There are over 12 lawsuits class action now. This will keep lawyers busy AND RICH.

I think Intel has secret conspiracy to CORNER memory market with RAMBUS.

INTEL has most powerful and stable CPU.

Intel has best Chipset in the world.

INTEL use to make memory but got out. Does INTEL wish go back in?

does Intel own shares in Rambus? IS THERE ANY CONTROL BY INTEL? HOW COME THINGS SECRET?

WHY DOES INTEL ONLY USE RAMBUS IN P4 MAINBOARDS? It does not make sense!

What happen if RAMBUS is bankrupt? Who will RAMBUS memory makers pay royalties? Will makers GIVE UP technology once company stops R&D? Will memory price of RAMBUS go up?

 

[Raystonn]

Do you by chance read conspiracy theory papers?

-Raystonn


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

 

[Arbee]

Really?

No. Intel would never reach an agreement with a proprietary memory IP company to displace the royalty-free JEDEC supported SDRAM for the royalty paying memory technology of the IP company. Nor it would get a large chunk of the that company shares at special price and free licence for the use of certain patents in exchange for active promotion of the tech and prohibition of using alternative high bandwith memory technologies just to make sure it would stay comited.
Nope, Intel would never do that.
Would it?

Of course not - imagine if the technology really doesn't brings the expected improvements in its begining interactions due to, lets imagine, high latency, or, even worse, proves to be a complex and expensive thing to produce, being much more expensive than the alternate (JEDEC sponsored) high bandwith memory tech? This would really hurt intel CPU sales, puting the company future on a third party ability to deliver.

Nope, definitely intel would never do that.



How terrible is wisdom when it brings no profit to the wise

 

[Raystonn]

"Intel would never reach an agreement with a proprietary memory IP company to displace the royalty-free JEDEC supported SDRAM for the royalty paying memory technology of the IP company."

Sure would. Especially if this newer form of memory has greater future potential in light of how quickly speeds will be ramping up over the next 2 years.


"Nor it would get a large chunk of the that company shares at special price and free licence for the use of certain patents in exchange for active promotion of the tech and prohibition of using alternative high bandwith memory technologies just to make sure it would stay comited."

That sounds like a good deal to me. When two companies each have something the other wants, a trade is to the mutual benefit of both parties. Trades are a very common thing in the technology industry.


"imagine if the technology really doesn't brings the expected improvements"

If you're talking about RDRAM here, it most certainly has. 533MHz bus speeds that support PC1066 RDRAM will likely be out in the second half of 2002. SDRAM (SDR or DDR) simply cannot keep up. DDR SDRAM requires a massive number of pin counts which makes it nearly impossible to create a chipset design with more than one channel at a low price. RDRAM is capable of far more bandwidth per pin than SDRAM will ever see.

-Raystonn


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

 

[Arbee]

"Intel would never reach an agreement with a proprietary memory IP company to displace the royalty-free JEDEC supported SDRAM for the royalty paying memory technology of the IP company."

Sure would. Especially if this newer form of memory has greater future potential in light of how quickly speeds will be ramping up over the next 2 years.

Really? what a surprise... (hint - was I saying that intel would never do that - face value - or insinuating that it did so two years ago? Only God knows... :tongue: )

"imagine if the technology really doesn't brings the expected improvements"

If you're talking about RDRAM here, it most certainly has. 533MHz bus speeds that support PC1066 RDRAM will likely be out in the second half of 2002. SDRAM (SDR or DDR) simply cannot keep up. DDR SDRAM requires a massive number of pin counts which makes it nearly impossible to create a chipset design with more than one channel at a low price. RDRAM is capable of far more bandwidth per pin than SDRAM will ever see.

1 - Rayston, when quoting me please don't cut my phrase too short. You removed an important part of it (initial...)

2 - Yes I was talking about RDRAM and no, it didn't. It is quite obvious that RDRAM has so far failed to deliver its promised benefits (for intel, but not only) - specially due to its price.

3 - 2.1GBs in a year isn't impressive. DDRAM DIMMs are getting 2.7GBs now. Granted it is easier to implement multiple channels with RDRAM, but nonetheless... Also the cost advantage of the cheaper chipset is quite moot if your memory is less performing (per channel) and quite more expensive. And there is already DDRAM working at 230/460Mhz(not in DIMM format) - the graphic card market hold by DDRAM gives this technology manufacturing advantages.

4 - We'll see what the future holds in terms of memory tech, but IMHO so far RDRAM is more suited for systems requiring low memory (not really possible with DDRAM modules) and high bandwith - ie consoles, not personal computers.


How terrible is wisdom when it brings no profit to the wise

 

[Raystonn]

"It is quite obvious that RDRAM has so far failed to deliver its promised benefits"

I disagree completely. The technology is obviously superior to SDRAM, either in SDR or DDR form. The basic unit of measurement is the amount of data transferred per second per pin. (You can think of a pin as a wire.) PC2100 DDR SDRAM uses 64 pins to transmit data at 2.128GB/s. That is 33.25MB/s of bandwidth per pin. PC800 RDRAM uses 2 channels of 16 pins each to transmit data at 3.2GB/s. That is 100MB/s of bandwidth per pin. Given an equal number of pins, which would take up an equal amount of space on a motherboard, the RDRAM leaves SDRAM completely in the dust. The technology is far superior and need only be developed further to add more width (pins). The only reason DDR is even remotely in the same area of performance is because RDRAM has not yet been increased to the same number of pins. RDRAM definately has the best potential for the future.

-Raystonn


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

 

[Arbee]

Rayston, which part of the "so far" you don't understand?

And even if the Rambus memory TODAY was superior to DDRAM - which isn't - it is an inferior product (performance, price/CO, availability and compatibility are key factors) - it would still not have fulfilled its promises - it was suposed to be completely superior...

Granted, using the pin bandwith as measure of superior tech is innovative. If someone brought forth a "great" new tech with a 512MB/s per pin but that only use one pin, would it be superior? (BTW RDRAM does not uses 2 channels - i850 chipset does. The PC800 RDRAM bandwith is 1.6GBs).

And have you ever thought why is it being quite difficult to increase RDRAM pin count or frequency (DDRAM went from 100 to 166Mhz - RDRAM stayed the same)?

Does RDRAM has a future as the superior memory tech? Perhaps. But for that it will have to improve as a product.


How terrible is wisdom when it brings no profit to the wise

 

[Raystonn]

"it is an inferior product"

It certainly outperforms all current forms of DDR SDRAM. Just what are you comparing it to, SRAM?


"If someone brought forth a 'great' new tech with a 512MB/s per pin but that only use one pin, would it be superior"

It most certainly would be. It's not too difficult to expand to multiple pins. It just takes a bit of time. SDRAM has had plenty of time to be tweaked.


"why is it being quite difficult to increase RDRAM pin count or frequency"

It's no more difficult than doing so for SDRAM. 400MHz parts are currently available. 533MHz parts will be available by 2H 02. That's a 133MHz jump. Greater pin counts are also on their way.

-Raystonn


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

 

[Guest]

JUIN,

How come you have so many handles? It's better to have one handle and to stick with it.

That was a great post by the way. You really think deep about the issues. Raystonn and Arbee are good too!

 

[Arbee]

It certainly outperforms all current forms of DDR SDRAM. Just what are you comparing it to, SRAM?

Don't even need to compare it to SRAM. Your regular Geforce3 has 7.2GBs bandwith through DDR SDRAM (you've said "all current forms"). RDRAM has 1.6GBs (or 3.2 with two channels). Which is more performing?
(hint 7.2>1.6 ).


How terrible is wisdom when it brings no profit to the wise

 

[Arbee]

And, BTW, having superior performance doesn't make a technology necessarily a superior product market wise - example: gallium arsenide has a much better performance than silicon but isn't used much in cpu manufacturing - perhaps intel and AMD don't want faster processors (hint - it is much more difficult and expensive to manufacture...).


How terrible is wisdom when it brings no profit to the wise

 

[FatBurger]

Raystonn, you're the first person I've ever seen measure RAM by 'per-pin' bandwidth. What about RAM on video cards? Can't be measure that way, unless you count the pins on the AGP/PCI card, which obviously would make no sense.

Let me say this: DDR as the current technology (not even counting DDR333, which exists), is better.
RDRAM as a PLATFORM is better.

If true dual-channel DDR comes out, it will be better.
RDRAM just doesn't have the bandwidth yet. The 16-bit channel is hurting them.



<font color=blue>Quarter pounder inside</font color=blue>
<font color=red>Change the Sig of the Week!!!</font color=red>

 

[Raystonn]

Your GeForce3 uses 128 pins to achieve those 7.2GB/s of bandwidth. That's about 56MB/sec of bandwidth per pin. Any low-grade engineer can come up with a design that extends the number of pins in order to increase bandwidth. It requires a higher level of thinking to develop a new technology that will increase the bandwidth of the base component, the pin. While the video card with 128 pins has more bandwidth than a single channel (16 pins) of RDRAM, this is only because of the number of pins used in the layout, not any great technology. Stick RDRAM on a video card with the same 128 pincount (8 channels with current RDRAM pin technology, but not a problem since the memory is soldered to the board) and you'd get 12.8GB/s of bandwidth, far surpassing the capabilities of that DDR SDRAM memory. Remember that the base unit is the pin, not the channel. It's the pins that require the space on the motherboard, not the channels.

-Raystonn


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

 

[Raystonn]

"What about RAM on video cards"

RAM on video cards have pins as well. Every pin requires a pathway to the memory controller, taking up valuable real-estate on either the video card or the motherboard (depending on if this is video RAM or main RAM). One pin takes up the same amount of space regardless of which memory technology you use. This is why bandwidth per pin is the basic unit of measurement for any memory technology. If you can fit 128 pins on your video card, that's great. You can now choose any memory technology to fit to those pins. You will obviously want the one that gives the best performance per pin. The current form of RDRAM (PC800) will deliver 100MB/s of bandwidth per pin. That far exceeds the 56MB/s per pin that DDR is currently offering the GeForce3 card. If nVidia would develop an RDRAM version of their video cards they would get much better memory bandwidth.


"If true dual-channel DDR comes out, it will be better."

Dual-channel DDR in main memory would use 128 pins. Those pins would be best served being attached to the memory technology that offers the most bandwidth per pin. With 100MB/s of bandwidth per pin, RDRAM is currently ahead by a factor of 3. You must remember that increasing the pin count, either by increasing the number of channels or increasing the count on the individual memory modules, is a memory-technology-independant way of increasing overall bandwidth. You can still attach any memory technology to those pins. You will undoubtedly want to attach the best one, and that is not DDR SDRAM.

-Raystonn


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

 

[Arbee]

Your GeForce3 uses 128 pins to achieve those 7.2GB/s of bandwidth.

Really? - I thought it was called 128 bit DDR because of some nice stickers over it or something similar... :lol:
(BTW I don't have GF3)

Any low-grade engineer can come up with a design that extends the number of pins in order to increase bandwidth.

That easy? good - by that reasoning then you must think that Rambus doesn't have even low grade engineers because so far they have been unable to increase the pin count. And they have been trying... what grade are Rambus engineers? Pointy hair level?

Remember that the base unit is the pin, not the channel. It's the pins that require the space on the motherboard, not the channels.

:lol: :lol: :lol: This deserves a technology award...
Channels are traces in the mobo. They take realestate. The reason that it is far easier to implement multiple channels with RDRAM comparing with SDRAM is that as RDRAM has a lower pin count it also requires less traces => thiner channels require less space => easier implementation.



How terrible is wisdom when it brings no profit to the wise

 

[Arbee]

This is why bandwidth per pin is the basic unit of measurement for any memory technology.

AFAIK for you only. Everybody else seems to prefer total bandwith. And latency.


How terrible is wisdom when it brings no profit to the wise