Bye-Bye RAMBUS!

[Penguin4x4]

980 is pretty damn good for PC800 RDRAM. The most 1066 can muster is @~1200. Do the math: 1200/1066 = X/800 X=900.

 

[Wallacebw]

Thanks all...

The atricle was rewritten by overclockers.com and (i believe) is less clear than the origional.

Here is a link to the original article (before the rewrite) that may clarify some things...

http://wallacebw.homelinux.com/oc/dualddr.html

If you have any qustions feel free to email me at:

[email protected]

Brian

 

[jdmcnudgent]

980 is pretty damn good for PC800 RDRAM. The most 1066 can muster is @~1200. Do the math: 1200/1066 = X/800 X=900.

no, its not. that means that the max fsb i could do with it is 160, and i dont think the chips on my board will go any higher than 150. now 150 is not bad, but i wanted something higher, like 180.

 

[Penguin4x4]

Well, get a "B" processor,(200 with a 50% OC)

 

[Emericana]

Firstoff, might i ask why rambus leaving is so good? what did rambus do in the first place (i have been in my own little AMD world which has seemed forever but i am about to make my dad a 2.0Aghz Intel... first time ever buildin intel).

here is a sneak peak at dual channel ddr... it looks really good:
http://www.hardocp.com/article.html?art=MzIz

in a year or so when i upgrade my athlon 1900mhz i will prolly go with intel dual ddr.

 

[CrystalMethod]

Actually, Rambus ain't going anywhere anytime soon. Intel intends to offer both. A DDR solution and a rambus solution. For performance, dual channel will be limited to the rambus.

 

[Penguin4x4]

Thanks for the link, Emericana. Whoa, dual 2100 @~ 1066 speeds? Think of dual 2700 or 3200!!

 

[Daovonnaex]

If they would just relase 64-bit RAMBUS......no no thats not going to help any. Think about it like this:

RAMBUS=Microsoft=IBM=Intel=U.S.S.R
DDR=Apple=Linux=AMD=VIA=U.S.A

DDR is "open-source" memory, so its free to use anyway you want it

RAMBUS is liscenced memory, so you must pay Rambus, Inc. to use thier technologies.

This argument is completemly illogicaly. Apple is severely anticompetitve in its approaches--notice how only Apple sells hardware for the Macintosh (the clone era, from 96 to 98, was killed by Apple). AMD is hardly on the "good side", as they're possibly the strongest backers of Microsoft's Palladium effort. As for the U.S.A., it's increasingly draconian and plutocratic.

By the way, the Apple I was not the first desktop computer, the Altair was (1971).

Anyhow, I don't think one should exactly be celebrating a demise of RDRAM technology, as it is probably the most advanced DRAM architecture available. Latency in RDRAM decreases as speeds increase, unlike DDR SDRAM, where they increase.
The introduction of Intel's dual-channel DDR SDRAM chipset, code named Granite Bay, does not necessarily spell doom for RDRAM (by the way, that efficiency figure is largely the result of the fact that RDRAM is dual-channel at the moment, and furthermore, the i850 and i850E are not exactly steller memory controllers). [H]ard|OCP| tested a beta Granite Bay equipped board from Asus , and it was beaten by the Intel D850EMV2 equipped with PC1066 RDRAM. While the margins were slim, recall that the Intel D850EMV2 is the poorest performing i850E board by a fair deal.
32-bit RDRAM is nearly here. Asus is about to put the P4T533, designed for 32-bit RIMM4200 modules, into the retail channel. The majority of tests show this board completing demolishing the P4T533-C with two 16-bit PC1066 RIMMs or the Gigabyte i850E board (Gigabyte's naming scheme is confusing, sorry) similarly equipped. This solution will outperform the Granite Bay board, though the memory is more expensive (the Granite Bay chipset is more expensive than the i850E chipset, so costs may be offset).

As to RDRAM's future, soon 64-bit RIMM9600 modules will be available (PC1200, but 64-bits wide). Furthermore, by later 2003 ultra low-voltage RIMMs with ODR (octal data rate) may be available. That means that a 600MHz RIMM, instead of be PC1200, would be effectively PC 4800. A 64-bit PC4800 RIMM would provide 38.4 GB/second of theoretical bandwidth, and be low voltage to boot. Therefore, I wouldn't count RDRAM out just yet. It may have a rough year now, but it will definitely be back, and will likely replace current SDRAM derived technology (SDRAM, DDR SDRAM) and its future technology (DDR II, DDR III).

 

[Emericana]

ya i asked this before and no one answered so i will ask again. what exactly did rambus do to make everyone hate them?

 

[Penguin4x4]

Stiff Licsensing Fees

 

[---X---]

Anyhow, I don't think one should exactly be celebrating a demise of RDRAM technology, as it is probably the most advanced DRAM architecture available. Latency in RDRAM decreases as speeds increase, unlike DDR SDRAM, where they increase.
The introduction of Intel's dual-channel DDR SDRAM chipset, code named Granite Bay, does not necessarily spell doom for RDRAM (by the way, that efficiency figure is largely the result of the fact that RDRAM is dual-channel at the moment, and furthermore, the i850 and i850E are not exactly steller memory controllers). [H]ard|OCP| tested a beta Granite Bay equipped board from Asus , and it was beaten by the Intel D850EMV2 equipped with PC1066 RDRAM. While the margins were slim, recall that the Intel D850EMV2 is the poorest performing i850E board by a fair deal.
32-bit RDRAM is nearly here. Asus is about to put the P4T533, designed for 32-bit RIMM4200 modules, into the retail channel. The majority of tests show this board completing demolishing the P4T533-C with two 16-bit PC1066 RIMMs or the Gigabyte i850E board (Gigabyte's naming scheme is confusing, sorry) similarly equipped. This solution will outperform the Granite Bay board, though the memory is more expensive (the Granite Bay chipset is more expensive than the i850E chipset, so costs may be offset).

As to RDRAM's future, soon 64-bit RIMM9600 modules will be available (PC1200, but 64-bits wide). Furthermore, by later 2003 ultra low-voltage RIMMs with ODR (octal data rate) may be available. That means that a 600MHz RIMM, instead of be PC1200, would be effectively PC 4800. A 64-bit PC4800 RIMM would provide 38.4 GB/second of theoretical bandwidth, and be low voltage to boot. Therefore, I wouldn't count RDRAM out just yet. It may have a rough year now, but it will definitely be back, and will likely replace current SDRAM derived technology (SDRAM, DDR SDRAM) and its future technology (DDR II, DDR III).

Thank you for some anti-zealot clear thinking here We need more of this and less product hating, hell, if AMD was where Intel is now, think about Palladium

 

[Daovonnaex]

Stiff Licsensing Fees

Actually, that's not really what vexes most. Rambus stole specifications from the JEDEC committee, including technology required for all DRAM chips. They then proceeded to launch one lawsuit after another against DRAM manufacturers, winnning some and losing others. They now face an FTC lawsuit over their unethical activities.

EDIT: Browse through some older articles at Tom's Hardware. The good Doctor Thomas Pabst wrote a few things regarding Rambus's business tactics and ethics (or lack thereof).