Comments on: The difference is as plain as the ear on your face! http://dansdata.blogsome.com/2009/07/06/the-difference-is-as-plain-as-the-ear-on-your-face/ the blog that is not dansdata.com Sun, 06 Dec 2009 21:03:19 +0000 http://wordpress.org/?v=1.5.1-alpha by: Alex Whiteside http://dansdata.blogsome.com/2009/07/06/the-difference-is-as-plain-as-the-ear-on-your-face/#comment-5213 Tue, 07 Jul 2009 01:12:11 +0100 http://dansdata.blogsome.com/2009/07/06/the-difference-is-as-plain-as-the-ear-on-your-face/#comment-5213 I'd still consider that "block access" a subset of random access. You can access addresses arbitrarily, with additional latencies appearing for traversing the heirarchy in certain ways. Maybe it's a side effect of coming in from the "sequential access" side and putting up with tape decks. It's probably not a taxonomically relevant distinction these days when every form of storage does random access in some fashion. My concern is that if you're going to say that "block access" isn't random access, then it's a slippery slope into saying that <i>nothing</i> is really random access. (I recall a CU Amiga article where the author found CD-ROM to be a huge misnomer, preferring CD-RAD (Random Access Disk). I'm not that bad, thankfully.) I'd still consider that "block access" a subset of random access. You can access addresses arbitrarily, with additional latencies appearing for traversing the heirarchy in certain ways. Maybe it's a side effect of coming in from the "sequential access" side and putting up with tape decks. It's probably not a taxonomically relevant distinction these days when every form of storage does random access in some fashion. My concern is that if you're going to say that "block access" isn't random access, then it's a slippery slope into saying that nothing is really random access.

(I recall a CU Amiga article where the author found CD-ROM to be a huge misnomer, preferring CD-RAD (Random Access Disk). I'm not that bad, thankfully.)

]]> by: Popup http://dansdata.blogsome.com/2009/07/06/the-difference-is-as-plain-as-the-ear-on-your-face/#comment-5212 Mon, 06 Jul 2009 23:57:04 +0100 http://dansdata.blogsome.com/2009/07/06/the-difference-is-as-plain-as-the-ear-on-your-face/#comment-5212 &gt; Well, by definition, it's either got to be "random" or "sequential", Does it? I have heard (and used) the term 'block access' to mean the kind of devices where the latency is more-or-less independent of the size of the read request (up to the 'block size'). A prime example would be a hard disk, where the average latency is (as defined by the spin speed) on the order of 10ms, but the extra time required to read a single byte (as defined by transfer speed) is on the other of 30ns! > Well, by definition, it's either got to be "random" or "sequential",
Does it?

I have heard (and used) the term 'block access' to mean the kind of devices where the latency is more-or-less independent of the size of the read request (up to the 'block size').

A prime example would be a hard disk, where the average latency is (as defined by the spin speed) on the order of 10ms, but the extra time required to read a single byte (as defined by transfer speed) is on the other of 30ns!

]]> by: Alex Whiteside http://dansdata.blogsome.com/2009/07/06/the-difference-is-as-plain-as-the-ear-on-your-face/#comment-5211 Mon, 06 Jul 2009 23:37:26 +0100 http://dansdata.blogsome.com/2009/07/06/the-difference-is-as-plain-as-the-ear-on-your-face/#comment-5211 Linear and additive, at that. Linear and additive, at that.

]]> by: Alex Whiteside http://dansdata.blogsome.com/2009/07/06/the-difference-is-as-plain-as-the-ear-on-your-face/#comment-5210 Mon, 06 Jul 2009 23:36:08 +0100 http://dansdata.blogsome.com/2009/07/06/the-difference-is-as-plain-as-the-ear-on-your-face/#comment-5210 Well, by definition, it's either got to be "random" or "sequential", and it sure as hell isn't "sequential" access, as we had back in the old tape days, when the latency was linear with the distance between the memory locations. It's not completely address-independent random access, but nothing is. Even a perfectly crafted ROM will give you different (miniscule) latencies depending on where the address is physically stored on the silicon. Well, by definition, it's either got to be "random" or "sequential", and it sure as hell isn't "sequential" access, as we had back in the old tape days, when the latency was linear with the distance between the memory locations. It's not completely address-independent random access, but nothing is. Even a perfectly crafted ROM will give you different (miniscule) latencies depending on where the address is physically stored on the silicon.

]]> by: Popup http://dansdata.blogsome.com/2009/07/06/the-difference-is-as-plain-as-the-ear-on-your-face/#comment-5209 Mon, 06 Jul 2009 23:28:39 +0100 http://dansdata.blogsome.com/2009/07/06/the-difference-is-as-plain-as-the-ear-on-your-face/#comment-5209 As a matter of fact, I wouldn't call a NAND flash memory 'random access', as the latency of a read is strongly dependent on previous reads. (For two reasons: - as the hardware will have to read an entire 4kb block at a time (to do error-checking/correction) odds are that sequential reads will already be available. - If the system is clever enough it will automatically pre-fetch the next block, and have it available in some kind of cache. In fact, you'll be hard-pressed to find true 'random access' anywhere, certainly not SDRAM, and in most application not even SRAM qualifies - even in those cases sequential accesses often lead to shorter latency, thanks to pre-fetching by the hardware, or caching at a higher level. In fact, the only 'random access' memory I can think of is 'ROM' i.e. good old-fashioned EPROM or the equivalent... As a matter of fact, I wouldn't call a NAND flash memory 'random access', as the latency of a read is strongly dependent on previous reads. (For two reasons:
- as the hardware will have to read an entire 4kb block at a time (to do error-checking/correction) odds are that sequential reads will already be available.
- If the system is clever enough it will automatically pre-fetch the next block, and have it available in some kind of cache.

In fact, you'll be hard-pressed to find true 'random access' anywhere, certainly not SDRAM, and in most application not even SRAM qualifies - even in those cases sequential accesses often lead to shorter latency, thanks to pre-fetching by the hardware, or caching at a higher level.

In fact, the only 'random access' memory I can think of is 'ROM' i.e. good old-fashioned EPROM or the equivalent...

]]> by: Popup http://dansdata.blogsome.com/2009/07/06/the-difference-is-as-plain-as-the-ear-on-your-face/#comment-5208 Mon, 06 Jul 2009 23:14:29 +0100 http://dansdata.blogsome.com/2009/07/06/the-difference-is-as-plain-as-the-ear-on-your-face/#comment-5208 While I ought to know better than to pick a fight with someone who buys pixels by the barrel... Would you also call a hard disk RAM? While I ought to know better than to pick a fight with someone who buys pixels by the barrel...

Would you also call a hard disk RAM?

]]> by: Daniel Rutter http://dansdata.blogsome.com/2009/07/06/the-difference-is-as-plain-as-the-ear-on-your-face/#comment-5207 Mon, 06 Jul 2009 23:00:56 +0100 http://dansdata.blogsome.com/2009/07/06/the-difference-is-as-plain-as-the-ear-on-your-face/#comment-5207 I've never really thought about this in detail, but have instantly ginned up the following justification: Is flash memory a <a href="http://en.wikipedia.org/wiki/Sequential_access" rel="nofollow">sequential-access</a> device? Clearly not, except in a very restricted sense of the term. Is it therefore a <a href="http://en.wikipedia.org/wiki/Random_access" rel="nofollow">random-access</a> device, if we define that as storage which takes the same amount of time to access data anywhere on the device? Yes, it is. I don't think the block-erase limitation and resultant background controller trickery disqualifies it as a random-access device. So it's random-access, and it's memory, and thus it's RAM. (Of course, if someone made a reel-to-reel tape drive that worked at a billion miles an hour and could whip data under the heads as fast as a hard disk can, with no noticeable extra delay for accessing data on any part of the tape, then that might <i>also</i> qualify as RAM under this loose definition.) I've never really thought about this in detail, but have instantly ginned up the following justification:

Is flash memory a sequential-access device? Clearly not, except in a very restricted sense of the term. Is it therefore a random-access device, if we define that as storage which takes the same amount of time to access data anywhere on the device? Yes, it is.

I don't think the block-erase limitation and resultant background controller trickery disqualifies it as a random-access device. So it's random-access, and it's memory, and thus it's RAM.

(Of course, if someone made a reel-to-reel tape drive that worked at a billion miles an hour and could whip data under the heads as fast as a hard disk can, with no noticeable extra delay for accessing data on any part of the tape, then that might also qualify as RAM under this loose definition.)

]]> by: Popup http://dansdata.blogsome.com/2009/07/06/the-difference-is-as-plain-as-the-ear-on-your-face/#comment-5206 Mon, 06 Jul 2009 22:54:54 +0100 http://dansdata.blogsome.com/2009/07/06/the-difference-is-as-plain-as-the-ear-on-your-face/#comment-5206 Off topic: &gt; ...Flash-RAM... I have seen you use that phrase a couple of time, and I can't help wondering why you would use that word... It's true that some flash memories (NOR) can be read in a more-or-less 'random-access' way, (At least as random as SDRAM, with its <a>row, page and bank</a> restrictions...) But the fact that you can only erase it in blocks (of 64 to 256kb) at a time disqualifies it, in my opinion. With NAND flash, it's even more apparent, as erasure has to be done in blocks of (at least) 256kb and read accesses have to be of entire blocks (often 4kb each). And when we talk about Flash today, we typically mean the NAND variety. Still,calling them flash <a>ROM</a> would be even more wrong (even though it's more common). I have also seen them being called Flash PROM, Flash EPROM and Flash EEPROM (sometimes written as "flash e2prom"), but the most common seems to be simply 'Flash Memory'. I decided to ask gogole, and the following are the number of hits for each combination: -"flash memory" 193,000,000 -"flash ram" 321,000 -"flash rom" 2,400,000 -"flash prom" 31,000 -"flash eprom" 161,000 -"flash eeprom" 242,000 -"flash e2prom" 10,300 Clearly in common web-use the term "Flash memory" is the winner. I also decided to look at the worlds no.1 flash manufacturer (Samsung), and it appears that they occasionally used "flash rom" - but not since 2003. It's all "flash memory" now. --- And now back to the audiophile discussion... Off topic:

> ...Flash-RAM...

I have seen you use that phrase a couple of time, and I can't help wondering why you would use that word... It's true that some flash memories (NOR) can be read in a more-or-less 'random-access' way, (At least as random as SDRAM, with its row, page and bank restrictions...) But the fact that you can only erase it in blocks (of 64 to 256kb) at a time disqualifies it, in my opinion. With NAND flash, it's even more apparent, as erasure has to be done in blocks of (at least) 256kb and read accesses have to be of entire blocks (often 4kb each). And when we talk about Flash today, we typically mean the NAND variety.

Still,calling them flash ROM would be even more wrong (even though it's more common).

I have also seen them being called Flash PROM, Flash EPROM and Flash EEPROM (sometimes written as "flash e2prom"), but the most common seems to be simply 'Flash Memory'.

I decided to ask gogole, and the following are the number of hits for each combination:

-"flash memory" 193,000,000
-"flash ram" 321,000
-"flash rom" 2,400,000
-"flash prom" 31,000
-"flash eprom" 161,000
-"flash eeprom" 242,000
-"flash e2prom" 10,300

Clearly in common web-use the term "Flash memory" is the winner. I also decided to look at the worlds no.1 flash manufacturer (Samsung), and it appears that they occasionally used "flash rom" - but not since 2003. It's all "flash memory" now.

---
And now back to the audiophile discussion...

]]> by: Nilbog http://dansdata.blogsome.com/2009/07/06/the-difference-is-as-plain-as-the-ear-on-your-face/#comment-5203 Mon, 06 Jul 2009 17:38:29 +0100 http://dansdata.blogsome.com/2009/07/06/the-difference-is-as-plain-as-the-ear-on-your-face/#comment-5203 ...<a href="http://www.members.optusnet.com.au/pennywyatt/Interests/FlandersSwann/DropOfaHat/At%20the%20Drop%20of%20a%20Hat02.html" rel="nofollow">Flutter on your bottom!</a> ...Flutter on your bottom!

]]> by: Nilbog http://dansdata.blogsome.com/2009/07/06/the-difference-is-as-plain-as-the-ear-on-your-face/#comment-5202 Mon, 06 Jul 2009 17:31:53 +0100 http://dansdata.blogsome.com/2009/07/06/the-difference-is-as-plain-as-the-ear-on-your-face/#comment-5202 Childhood memories... <a href="http://www.members.optusnet.com.au/pennywyatt/Interests/FlandersSwann/DropOfaHat/At%20the%20Drop%20of%20a%20Hat02.html" title="Flutter on your bottom" rel="nofollow"></a> Childhood memories...

]]>