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 : AMD/INTC/RMBS et ALL -- Ignore unavailable to you. Want to Upgrade?

To: Charles R who wrote (97)	9/21/1999 11:04:00 PM
From: kash johal	Read Replies (1) \| Respond to of 271

Chuck, Great post. Seems to me that Rambus has lost if Ashok's reading is correct. only 5% of systems in Q4 to use DRDRAM. Now the big white hope is the 840 chipset: Intel faces pressure over new chipset Another article about how much in demand the Camino chipset is! "Intel had planned for the 820 chipset to account for 20 to 25 percent of the chipsets it made, but it has reduced that number to 5 percent, he said." Combine this with a BX shortage and there may not be enough chipsets to go around. But I'm sure Via won't mind that. ;) PB, looks like Intel is doing an AMD launch. More OEMs may announce systems compared to AMD's norm, but based on that quote the quantity offered will be a MUCH lower % than the INtel norm. First it was DELAYED, now not many OEMs systems will use it it. It is amazing how much INtel (heavy emphasis on Intel) has screwed this launch up. Maybe they should have delayed Camino further until Cuontimemine comes out. That way, Intel could have charged less for the smaller processor in hopes it would help balance the cost of DRDRAM. Remember, Intel has Hundred$ of Million$ invested in RamBUST DRDRAM. Oh well, there is always next month/quarter! ____________________________________________________________ Intel faces pressure over new chipset By Stephen Shankland Staff Writer, CNET News.com September 21, 1999, 6:15 p.m. PT The introduction of a new set of chips from Intel isn't going as smoothly as the chip manufacturer would like. On Monday, Intel will debut its new 820 chipset, the set of chips that let a CPU talk to the rest of a computer. But the debut apparently is somewhat tarnished, as some manufacturers shun the new hardware, analysts say it faces a limited market, the Rambus memory system it enables proves to be expensive, and Intel itself reduces its production plans. "Intel has significantly scaled back their short-term expectations," said Ashok Kumar, an analyst with U.S. Bancorp Piper Jaffray. Intel had planned for the 820 chipset to account for 20 to 25 percent of the chipsets it made, but it has reduced that number to 5 percent, he said. The adjustments aren't new to Intel as it works to balance its own plans for the innards of PCs with what the companies that build computers actually desire. For example, Intel abandoned its earlier stance and accommodated manufacturers' desires to use conventional memory running at somewhat higher speed instead of jumping straight to the next-generation Rambus memory technology. And the 820 chipset itself was delayed three months. For its part, Intel says it has billed Rambus not as a miracle cure for PC performance today but as a necessary step in coming years to be able to take advantage of ever-faster processors. "There will be a transition between SDRAM [today's conventional memory technology] and RDRAM [the Rambus technology] over the next couple years," said Intel spokesman Dan Francisco. "Ultimately, the market will decide the rate that transition takes place." Francisco declined to discuss Intel's plans for how many of the 820 chipsets it would make, but said: "The 820 is on schedule to go in late September. System manufacturers want to do systems on their own schedules." Micron avoiding the 820 Micron Electronics has decided to use a rival chipset from Via Technologies, because it allows the PC maker to get the same performance the 820 provides while cutting $200 to $300 off the price tag of a system, said product manager Robert Wheadon. The Via-based systems will show up at the top end of Micron's line in systems using 600-MHz or 533-MHz Intel chips, he said. Micron speed testing has shown that both the Via chipset and the 820 show an increase of 3 or 4 percent over current Intel chipsets in real-world tasks, Wheadon said. Regarding the performance, Francisco said Intel's position has been that customers "aren't going to see some astronomical difference." Instead, Micron is more enthusiastic about a later chipset, Intel's 840 or "Carmel" chipset, due in the first half of 2000. The Carmel chipset will have two channels the CPU can use to talk to main memory, doubling the bandwidth and making the computer better able to utilize Rambus' theoretical capacity. Scaling back 820 production The reason Intel would scale back its production plans is simply lack of demand, said MicroDesign Resources analyst Peter Glaskowsky. "It's going to be very difficult for Intel to sell a lot of those chips" because of the high price of Rambus memory. A system with the 820 chipset and 128 MB of Rambus memory will cost a manufacturer another $300 to make. He estimates that cost will translate to another $500 added to the price the end customer sees. Except for a very few people such as gamers and users or large databases, "that's a pretty much unacceptable price penalty," he said. Another price effect has been that most computer makers are likely to adopt lower-speed versions of Rambus memory, said Shawn Willett of Aberdeen Group. "I would expect most to go with the 600-MHz Rambus because of the costs associated with the 800-MHz chips and the difficulty in getting it," Willett said. Francisco didn't comment on how popular the different speeds would be, saying only that Rambus will be available and used in 600-, 700-, and 800-MHz speeds at the time of launch. news.cnet.com.

To: Charles R who wrote (97)	9/22/1999 12:27:00 PM
From: Charles R	Read Replies (2) \| Respond to of 271

Thread, I have spent a decent chunk of time on the post about RDRAM near-term and long-term markets with the hope that we can get a dialogue going to get some estimate of RDRAM future potential. Can you guys add your thoughts to the subject? Otherwise there isn't much reward to sharing my thinking, is there? Chuck

To: Charles R who wrote (97)	9/23/1999 9:39:00 AM
From: Bilow	Respond to of 271

Hi Charles R; Re your long/short term use of RDRAM &c... I thought I'd comment on the high end graphics stuff, particularly with regard to eDRAM. You said: I can see the appeal but I am unsure how this would compete with embedded DRAM In this I am sure you are correct. Given a display size (in the future) of 2048x2048, with 8 bits per pixel, RGB, the total memory size is 96Mb. Embedded DRAM could certainly put that on a chip, in fact, IBM's standard cell data book gives 128Mb as an example. For a smaller picture size, the memory requirement would decrease. You might want to offer two frames of data, this would double the memory requirement. The average bandwidth requirement for such a display would be 96x60 Mb/sec, assuming a 60 Hz refresh. (Such slow refereshes hurt my eyes, but the resolution is kind of high, anyway.) This is a total bandwidth of 750MB/sec, call it 850MB/sec peak, where I assume a FIFO sufficient in size to buffer a line, but not, of course, a frame. You also need to have some bandwidth to write to memory, and if you want to require that you can rewrite the whole memory in a single frame (1/60 of a second), then this about doubles the bandwidth requirement to 1600MB/sec. This kind of memory bandwidth could almost be provided by a single RDRAM chip, while that single chip would easily provide sufficient capacity. DDR SGRAM could provide that bandwidth only by going to two chips. The total system cost might end up being quite similar to RDRAM, due to lower DDR SGRAM chip costs, and other things like heat sink, termination, etc. I haven't looked at pricing for DDR SGRAM, but usually SGRAM is pretty cheap. The reason for this is that they don't put very many memory cells on a chip, relative to the other processes. So the chips are small, but have a lot of I/O. The I/O savings of RDRAM will not save much money due to the lowered cost of modern high pin count packages. For example, IBM's standard cell allows something like 1700 pins per package. As for the future, the trend in memory capacity and bandwidth has been faster than the trend in display sizes. For example, a 1280x1024 screen has 16 times as much data as a 320x240 screen, but early DRAMs were about 256,000 times smaller than the current generation. Because of this remarkable disparity in long term growth rates, it is likely that embedded DRAM will hold the future for high end display controllers. That is, embedded DRAM will allow sufficient capacity at system prices that will be below system prices that use discrete DRAM. In other words, system on a chip is the wave of the future. -- Carl