SI
SI
discoversearch

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 : Advanced Micro Devices - Moderated (AMD)
AMD 221.02+6.4%3:59 PM EST

 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext  
To: combjelly who wrote (62744)11/7/2001 7:46:06 PM
From: wanna_bmwRead Replies (1) of 275872
 
Combjelly, you are bringing up two main points. The first is about the speed of SRAM, and how you doubt Intel can purchase the memory at speeds up to 800MT/s (Megatransfers per second - the memory does not have to run at 800MHz), without paying a lot more.

For this, I should point you to my earlier link to Samsung, who claims to be selling 400MHz DDR SRAM, which would be capable of acting like the SRAM for Itanium. They claim they are in volume production of this, so it's only logical to assume that whomever Intel has outsourced for Itanium's SRAM, that they are also in high volume production. Of course, Samsung like many other SRAM providers, does not list their prices on their web site, which makes it hard for me to respond with a quote. However, by looking at contemporary high volume SRAM cells, from TI, for example, we can get an approximate price.

You seem to think that high speed SRAM comes at a hefty premium over standard SRAM, but I don't believe this to be the case. There is nothing innately different about the manufacturing to get these speeds. They are simply double-pumping the data at 2.5ns (which is not unheard of for SRAM cells - and as you said, most very fast SRAMs top out at 400-500MHz and all Intel needs is 400MHz double pumped).

The second thing you address is this.

Re: "Regardless, in the greater scheme of things, Intel has not made very many of these things. As a result, they really haven't had a chance to optimize the process for this product, and they are going to have a higher number of units that bin out at speeds that aren't useful. At ~$2k per wafer, this can add up."

You still don't seem to understand the concept of applying high volume manufacturing to a low volume design. Rather than optimizing the process for new designs like Itanium, Intel actually optimized the design for their high volume processes like P858 (the .18u process that they use for all .18u designs). There is nothing expensive about running under this process - it should effectively be nearly the same as running any other mainstream design (Coppermine, Willamette, etc). The minor differences in manufacturing do not impact the cost.

wanna_bmw
Report TOU ViolationShare This Post
 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext