Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Rambus (RMBS) - Eagle or Penguin -- Ignore unavailable to you. Want to Upgrade?

To: Joe NYC who wrote (40705)	4/25/2000 5:50:00 AM
From: John Farrell	Read Replies (3) \| Respond to of 93625

6.4 GigaByte/Sec 200 MHz (PC400) DDR SDRAM supported by ATI's new GPU, the RADEON 256. ati.com Nvidia's GeForce 2 GTS being announced tomorrow should also support this type of bandwidth. It would have been interesting to have seen PC Magazine have not tilted it's Winbench 3D benchmark comparisons last month of the 1 Gigahertz Coppermine vs. Athlon comparisons by using the 64Megabyte DDR GeForce cards on the two Dells with the 820 chipsets versus Compaqs/Gateways/HP's with the Athlons (PC100/PC133) which had 32 Megabyte GeForce cards (some SDR and some DDR). I think this and AGP Pro (4x) support on the Dell's versus the AGP 2x support on the Athlon machines are much more responsible for the benchmark differences between than any RDRAM/SDRAM main memory performance differences in these tests. Matrox also announced their G450 at WinHEC supporting DDR SDRAM so another round of graphics accelerators are announced without the RDRAM support that is "supposedly" (according to RMBS bulls here) "perfect for graphics" applications. I doubt that any of the graphics boards are more than 6 layers. -John

To: Joe NYC who wrote (40705)	4/25/2000 1:05:00 PM
From: Tenchusatsu	Respond to of 93625

Jozef, <This still leaves me confused (as I posted earlier) why Intel would prefer DDR to Rambus in the server area, where bandwidth is very important.> 1) DDR will cost less than RDRAM per megabyte. Even a 20% savings can translate into a lot since servers typically require gigabytes of memory. 2) RDRAM has a 32-device limit per channel. DDR SDRAM can go all the way up to 128 devices per channel by using four double-sided double-stacked DIMMs. So the capacity per channel of DDR is better than RDRAM, and servers need all the capacity they can get. (Actually with RDRAM, you can squeeze 128 devices per channel by utilizing repeater hubs, but that carries a performance penalty, just like the RDRAM-to-SDRAM translator hub.) 3) It's easier to implement chipkill on DDR SDRAM than on RDRAM. And as you may know, chipkill is an essential part of server RAS. 4) Intel experienced a lot of pushback from its server customers regarding DDR vs. RDRAM. Basically, none of Intel's server customers wanted RDRAM. (Makes me wonder whether Intel's other customers didn't mount as much of a pushback. Maybe they too see RDRAM's benefits on the desktop.) <Regarding Athlon, I am not sure if the dual DDR will be necessary. The single channel provides all the bandwidth the (future) Athlon chip is able to receive: 133 MHz x 8 bytes x 2 = 2.1 GB/s. Current Athlons on DDR platform are limited to DDR 200 anyway, with bandwidth of 1.6 GB/s. Which incidentally matches a single RDRAM channel.> First, AMD has plans to push the FSB of Athlon from 200 to 266 MHz. This is necessary in order to more easily match up with DDR's 266 MHz speed. And second, there is a chipset being developed that will support two Athlons and two DDR channels, all on one north bridge. Sounds like such a chipset will be part of a low-end server platform. (Incidentally, that north bridge will have more than 1,000 pins. That's not a trivial amount.) Tenchusatsu