-
We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.
just thought some of you geeks may wanna know...
- Thread starter Mday
- Start date
Buddha Bart
Diamond Member
So whats out there to even start with?
Try looking at some of the flash memory companies web sites - some of the pioneers in the solid state storage market like Sandisk have a lot of good information. If there isn't enough there, then I'm sure if you asked, they'd give you a helping hand.
It's worth looking at several of the different formats of flash to see what the differences are - e.g. Compact flash - an IDE interface to the flash, or MMC a very simple serial interface.
It's not all just as simple as just connecting an interface to a flash chip - there's quite a lot more to it than that (if you want it to work well) - but I'd be ruining your project if I were to tell you right now.
It's worth looking at several of the different formats of flash to see what the differences are - e.g. Compact flash - an IDE interface to the flash, or MMC a very simple serial interface.
It's not all just as simple as just connecting an interface to a flash chip - there's quite a lot more to it than that (if you want it to work well) - but I'd be ruining your project if I were to tell you right now.
Mday, I think you opened yourself wide up for the retort that perhaps you don't know this but Flash is not exactly very slow. Although it would depend on your definition of SLOW, or even VERY SLOW.
It also depends on whether or not you want a NAND structure (fast write) or a NOR structure (fast read) flash memory. Taking numbers from the datasheet for Intel's latest 0.13um Strataflash NOR flash memory which is here: http://ftp://download.intel.com/design/flcomp/datashts/25190204.pdf
Burst read bandwidth: 108Mb/s (54MHz zero-wait burst read mode with 16-bit word size).
Initial read latency: 0.000000085 ms (milliseconds used for comparison with hard disks)
Burst clock-to-clock read latency: 0.000000014 ms
Ok, so the write speed stinks (5 us/byte) but everything else is pretty impressive. And why limit yourself to a mere 16-bit word size. Create a quad-channel type configuration and make it a 64-bit word dize with a lovely 432MB/s theoretical bandwidth. And you can practically flip these numbers around for writing on NAND structure Flash.
Or, to illustrate it even more. Look at this beauty. 200MB/s sustained reads out of Flash memory.
So throw away your budget, buy 10GB of Flash, and buy a big 256MB DRAM buffer for the front-end. Then you can buffer writes into DRAM and read at high-speeds from the NOR Flash. You'd need a controller that would bypass buffered write data that is being requested and this could be a bit tricky, but trickiness only makes the project more fun. A decent FPGA and you could code up a controller pretty quick. Of course, you can do it with DRAM, but doing it all with DRAM is so 90's. 😉
blahblah99
Platinum Member
So you have to gather some ram chips and design a controller for that that interfaces to whatever.. pshh.. 🙂
What are the requirements for the solid state hdd? does it have to meet a data transfer rate? Does it have to be a certain size?
Xilinx, I believe, has application guides on designing ram controllers for use in their SPARTAN fpgas. They may even have blocks of code ready to be used in the fpgas.
What are the requirements for the solid state hdd? does it have to meet a data transfer rate? Does it have to be a certain size?
Xilinx, I believe, has application guides on designing ram controllers for use in their SPARTAN fpgas. They may even have blocks of code ready to be used in the fpgas.
Just a suggestion from the vulgar masses: why not make a dual-capable hard drive; give it a gig or so of flash memory (very feasable) to install WinXP/Linux on, and use the magneto-optical part of the drive to store files like documents, music, etc.
You could take it further and say that there would be no swap file; it could reside in main memory. If you used the technology in the Lexar flash memory cards you could probably produce something with a blazing fast transfer rate that could boot XP in the blink of an eye.
Just a thought. I don't know if this necessarily applies to your thesis.
You could take it further and say that there would be no swap file; it could reside in main memory. If you used the technology in the Lexar flash memory cards you could probably produce something with a blazing fast transfer rate that could boot XP in the blink of an eye.
Just a thought. I don't know if this necessarily applies to your thesis.
It's not as slow as you think! Sure, writes to conventional flash RAM aren't as fast as, say, to a fast HD - but there are plenty of ways of hiding this.
A modern flash IC should easily get 40 MB/s sustained read speed. Write speed might be 10-20% of this, but you could parallel the flash up - have 4, 8 or even more channels and then have a big DRAM buffer. Unless you were using the drive for swap space, you'd be unlikely to perform large volumes of writes frequently. Indeed, a drive required to service numerous writes (e.g. server log files) should not be flash based, due to the destructive nature of the write operation. Again, where necessary, there are ways of hiding this.
Unless, I'm reading this datasheet wrong this chip will manage nearly 120 MB/s in sustained mode. Access time is approx 0.09 ms. However, write performance sucks terribly.
This NAND flash isn't quite as fast on reads (about 20 MB/s), but write performance is better (about 4 MB/s).
This is 100 time better than CDRWs (they are rated for 1000 writes to any section of the disk). This is why UDF and other packet writing formats rotate through the list of which sector gets used next (even if you delete a file) instead of using whatever block is free on the fastest part of the disk, and will fragment a file when you update the middle, rather than rewriting the already-used blocks.
This technique called 'wear-levelling' is essential for media where writes are destructive - this goes equally for flash as it does for CDRW discs. This technique is incorporated into currently available flash based storage (decent devices at least).
Flash has improved significantly in terms of reliability and wear is a considerably smaller problem with flash than with CDRW. Additionally, NAND flash is supposed to be significantly more reliable (with a typical lifetime of 1,000,000 to 10,000,000 erase cycles) than NOR flash (10,000 - 100,000 erase cycles).
You guys might want to keep in mind that an Architect came up with the lithography process for manufacturing CPUs. You should also keep in mind that there are some highly educated people on here who are extremely gifted, creative, and inventive. Just because you go to Yale and pay $37,000US per year doesn't make you any more intelligent or knowledgeable than anybody else. You should all stop fooling yourselves and realize that you may just not know everything about your chosen field. Anyone who claims to know everything about anything is a moron.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 25K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 24K
-
-
