View Full Version : 232-pin RDRAM (RIMM 4200)?

July 12th, 2014, 05:03 PM
I have an old ASUS P4T533/R system that uses 232-pin 32-bit RDRAM modules. There are only two memory slots on this motherboard and the manual states that the system will support 2 GB of ram (MAX). I am fortunate to have both ram slots populated with 512 MB RDRAM modules for a total of 1024 MB RAM (232-pin 512 MB chips are hard to find). 1 GB chips may not have been produced.

I know that Samsung, Kingston, and Elpida were working on prototype 1 GB Chips but when Intel decided to move away from RDRAM in favor of DDR, no 1 GB chips (to the best of my knowledge) ever hit the store shelves.

1. Can anyone confirm that no 232-pin 1024 MB RDRAM chips were ever sold.

2. Is there any way to determine if prototype chips were produce of this capacity (i.e., 1 GB 232-pin RDRAM)? Perhaps I could write the memory manufacturer and obtain two chips that they otherwise have no use of... ?

512 MB chips, when I have seen them on eBay, often sell for $150+ each.

Just curious.

F. Fota
Stafford, VA

July 14th, 2014, 07:15 AM
They don't exist. After the disaster of an unveiling of the chipset and Rambus pretty much hanging everyone by the balls on royalties nobody wanted to use RDRAM and jumped ship once their contracts were up.

July 14th, 2014, 08:14 AM
I never did understand the advantages of RDRAM over, say, SDRAM. I think I still have a P3 FIC KC19+ motherboard with 2 256MB RDRAMs on it. It doesn't seem to be any faster than SDRAM-equipped boards; the RDRAM interface is very complex and the memory is expensive.

July 14th, 2014, 08:41 AM
Samsung probably had prototype RIMMs of GB capacity when they shipped the 512MB; development cycles being what they are. Contact Samsung marketing and you might be treated to a copy of the press release done 12 years ago. No, I did not find any record of an announcement. I have no idea if the prototypes would have worked with PC4200 systems. My experience with RDRAM was the systems were less than ideally stable even with supported memory.

RDRAM was focused on maximizing memory throughput. A Pentium 4 doing video encoding was one of the few times the increased bandwidth of RDRAM helped. In most cases, the extra latency of RDRAM made the actual performance lower. Eventually, had Intel continued their planned improved Pentium 4s and RDRAM not crashed in the market like it did, RDRAM would have been very useful. Still would have been too expensive anyway and Intel needed to get some good chipsets before I would bought one.

July 14th, 2014, 08:52 AM
Sure, maximize throughput by reducing the bus width to 16 bits? High-latency compared to other technologies, power-hungry and daisy-chained signal paths (forget that CRIMM for an empty socket and you've got problems).

None of it ever made any sense. My 820-chipset board is easily outrun by a 440 one using SDRAM.

Mostly, I remember Rambus for their tendency to sue anything that moved.

June 13th, 2015, 06:22 PM
512Mbit RDRAM chips and 1GB 16-bit RIMMs actually was produced for AlphaServers, but they don't work in i850/i860 unfortunately and are not 232-pin. I wonder if 1GB 32-bit RIMMs based on 32 devices would have been possible. Probably rare, but I also wonder if SiS R658/R659 would have supported 512Mbit RDRAM devices.

June 13th, 2015, 07:09 PM
I think the Samsung MR18R326GAG0 are the 1GB RIMM modules that were used on AlphaServer ES47/ES80/GS1280

( 32Mx18 ) * 16pcs RIMM Module based on 576Mb A-die, 32s banks, 32K/32ms Ref, 2.5V
Each RDRAM device has 32 banks, for a total of 512 banks on 1.1GB module