• Please review our updated Terms and Rules here

CF card vs HDD

The "DOM" is not a disk, it is a flash card in another form factor. We didn't differ between them when constructing microITX routers and telephony systems back 20 years ago, running Linux.
Once you minimize the number of writes (/var and /tmp are memdrives, periodically synced to the CF partitions) it can have a long life.

The issue here is Windows which are horrible in their structure and their patterns of working. It is not possible to do.

But the main factor is uptime at that worst case. I have Windows on the CF cards for my retro comp(s) because the comps are occasionally powered on. In this case, a Sandisk/Transcend card and any OS will work "fine". If it fails, you had the benefit of easier imaging e.g. backups with CF so it should be less of a problem. If it fails, it is not a primary hdd of your main computer that died suddenly, but a storage device for ancient PC...

I used to run FreeBSD from the thumbdrive in an almost 'read only' mode, had a fileserver in a 1U case with Intel Atom board and 2 WD Greens. It ran for few years before being blown up by power surge, there was no issue that OS is on USB, in this context.
 
bsdphk said:
No, the flash itself does not, but the microcontroller running the FAL does.
Click to expand...
Reading the flash without using that microcontroller is... well. I treat both as a single unit, same as IDE or SATA hard drives.

For retro purposes, using flash drives is perfectly fine and the advantages (no seek time, relatively easy to buy, easy to replace, no mechanical failures, silent, ...) far outweigh the disadvantages (writing kills, may not work without XUB, ...) in almost all environments. I don't see any reason why stressing a 20 year old mechanical drive would be any better if it can be avoided.

My old systems don't hold important data, and imaging a 1-2 GB CF card means nothing in an era where my main machine keeps a few LLMs at dozens of GB each. Sure, a 486 running Windows 95 will likely do a ton of paging to disk, but it's still a slow system - it simply can't write sufficiently large amounts of data per hour.
 
Svenska said:
My old systems don't hold important data, and imaging a 1-2 GB CF card means nothing in an era where my main machine keeps a few LLMs at dozens of GB each. Sure, a 486 running Windows 95 will likely do a ton of paging to disk, but it's still a slow system - it simply can't write sufficiently large amounts of data per hour.
Click to expand...

I use those "video rated" (IE, designed for continuous video streaming and overwriting) SD cards in an SD-to-PATA converter with my homemade XT-CF controller in my Tandy 1000s, and I'm reasonably confident, based on some back-of-the-envelope calculations, that even with a fairly mediocre wear leveling algorithm I will probably not live long enough to wear out one of those cards even if I set that fire-breathing 7.16Mhz V-20 specifically to the task of trying to torture the disk to death 24/7/365. That combined with the fact that nothing living on those file systems is ever going to be "valuable" in any objective sense makes me pretty comfortable not worrying about the media not being "archival quality".

(I mean, let's face it, no computer media is "archival quality". If it's not the magnetic domains fading or the iron physically flaking off the plastic tape or metal disk, it's that a suitable drive to stuff that tape or disk in will simply not exist ten, twenty, or fifty years from now. If you want your data to last forever engrave it on a stone tablet and bury it under a pyramid, and even then all bets are off, really.)

When the Raspberry Pi came out depending on an SD card as its main storage medium there was endless pearl clutching about how unsuitable that was for a "real computer", and I know it's *possible* to total a cheap SD in that kind of environment in a few months if you really put the effort in (IE, disable noatime, etc), but for the most part these things are fine for casual, non-mission-critical use. (And probably more reliable than old/cheap spinning rust will be at this point. I especially wouldn't trust most laptop drives anymore; they used to embed a PATA laptop drive in a lot of old high-end HP printer/plotters, and it was extremely common for them to throw in the towel in less than five years of standby use.)
 
I mean the OP has wrong POV to begin with, comparing 486 to a modern file server. But it evolved to an interesting discussion.

Just install it on CF and that's it. You won't be sitting at that computer gaming or working with Windows 3.11 applications 5-6 hours every day.
 
Eudimorphodon said:
I mean, let's face it, no computer media is "archival quality"
Click to expand...

Paper-tapes and Punch Cards are acid-free paper and will last until you let something eat it, but the data density is what it is.

The jury may still be out on "punch a hole with the laser" style WORM optical media.

Right now my best suggestion would be QR(-like) codes printed, single-sided, on archive quality paper, with weather resistant black ink-jet. (= how sign-shops print inkjet.)

At 75dpi, you get approx 400k pixels per sheet, which, after error correction, comes out to approx 20kB per sheet in practice.

One megabyte will take approx 50 sheets and weigh 250g[1]

An standard EU-pallet will probably store 1.5GB and weigh 500kg.

In practical terms, it is probably a better idea to use rolled paper: A typical roll for a HP DesignJet will hold 20-25MB, but bulk they may take up more space.

So, yeah: There are no realistic archive quality data storage media.

[1] Remember to use a very wide interleave, so that a few missing sheets does not cause loss of data. None of the projects I have seen does that.
 
gertk said:
Maybe glass is a solution... They claim 10000 years retention...
Click to expand...
First:
That's obviously speculative.

Second:
High power lasers are banned in consumer devices, so if consumers are involved, it is the PhotoCD model, where only professionals can write.

Third:
Nobody has ever been able to stitch a business-case together, because the market for "store forever" is a lot smaller than most people imagine.

In big companies "records retention", much to almost everybody's surprise, is focused entirely on destroying documents, as soon as the laws and regulations permit, to prevent them being hit by discovery in lawsuits.

That leaves collection-libraries such as Library of Congress, The Vatican's Library, the chronically underfunded national libraries in other countries of the world, probably also NASA, CERN and a few other scientific institutions.

The LTO market is much bigger, precisely because LTO does not store forever.
 
How about India ink on sheepskin tucked in a terra cotta jar and left in a cave. Some have lasted over 2 millennia.
 
Quagmire said:
How about India ink on sheepskin tucked in a terra cotta jar and left in a cave. Some have lasted over 2 millennia.
Click to expand...

Meh. You just made me remember the end of a Dudley Moore movie, and not a good one. (Is there a good one?)

Al Kossow said:
How about we lock the thread, since this has deteriorated to the inevitable peanut gallery shitpost level?
Click to expand...

Fair enough. I mean, it was mostly on point until suddenly it veered into this weird debate about how likely it is that the copy of Jazz Jackrabbit you intend to play on your old 486 will still be intact when the far-off descendants of man (or the cockroaches which inherit the earth after our passing) witness in horror the Sun's final ripening into a red giant.
 
Al Kossow said:
How about we lock the thread, since this has deteriorated to the inevitable peanut gallery shitpost level?
Click to expand...
To bring it back to the point you were making earlier in the thread,
The JEDEC standard says that SSDs, after having their write-endurance fully exhausted and then powered off, only have to retain the data for three months at 40 Celsius. (https://www.dell.com/support/kbdoc/...powering-off-systems-for-a-prolonged-duration)

That Dell document recommends that SSDs be "powered-up once every 2.5 months for a minimum duration of 3 weeks".

I feel like this isn't talked about that often?

So I'm not sure whether these are such worst-case assumptions it doesn't even matter (are consumer SSDs going to do better, or worse than enterprise ones?), or if the fine print is basically telling us that unlike pulling the data off some 90s IDE hard drive that still works, leaving an SSD off for 20 years then powering it back on just isn't going to be a thing.
 
Contracts are one
bsdphk said:
As I said: Flash/SSD is not an suitable archive media.
Click to expand...
Those estimates are a worst case situation. Don't count on long stored flash not retaining its data*. I have done some impromptu tests with obsolete flash drives that show they still had data completely correct about 5 years after the previous usage. Even had a phone retain all of its data after about 15 years in a box in the storeroom. Newer flash should not have the same longevity but I don't know anyone willing to spend the millions of dollars needed to get a good understanding of the likelihood of data loss over time on SSDs.

* Only drives that contain data one wishes to go away will still have data decades later.
 
bsdphk said:
As I said: Flash/SSD is not an suitable archive media.
Click to expand...
Nobody cares. That's not the topic of this thread. And if your definition of "archive media" is literally forever, then nothing is archive media.
 
allenm said:
I assume the CF card is faster than the Maxtor HDD, but I don't know that for sure.
Click to expand...
Clearly there is no seek time with flash media, but that doesn't necessarily mean faster.
Multiple Sectors per Interrupt (Block Mode) gives a nice speed boost to IDE PIO mode--by cutting out interrupt overhead.
For reasons that are unclear, CF cards, SD-IDE adapters, and SATA SSDs don't support it, while spinning IDE HDDs (and SATA HDDs through a SATA-IDE adapter) do. There's no technical reason they couldn't support it, and I believe there a few Intel Sata SSDs that do.
 
Plasma said:
Nobody cares. That's not the topic of this thread.
Click to expand...
We have reached the end of the thread topic quite a while ago.

krebizfan said:
I have done some impromptu tests with obsolete flash drives that show they still had data completely correct about 5 years after the previous usage. Even had a phone retain all of its data after about 15 years in a box in the storeroom.
Click to expand...
Flash technology is a race to the bottom, and the quality in recent years had a very steep drop-off in terms of raw flash reliability.

krebizfan said:
Newer flash should not have the same longevity but I don't know anyone willing to spend the millions of dollars needed to get a good understanding of the likelihood of data loss over time on SSDs.
Click to expand...
Aging is a well-known and well-studied technique, and reputable manufacturers should do that. (Does not apply to fake SD cards, obviously.)

We generally trust the manufacturer word, and some modern flashes specify sufficient (= expected device life-time) power-off data retention only for temperatures below 40°C. In some of our markets, that gets uncomfortable close to the expected room temperatures if nobody's home. To deal with that, we introduced a background service doing constant (but slow) read accesses to refresh the cells at runtime. Yes, it is that bad.
 
It's amazing what can be retrieved from 'temporary' storage sometimes. A year or so ago, I found on the E place a RAM storage device for some semiconductor manufacturing equipment my work had built back in 1993. It was clearly in NOS condition (it apparently sat on a spares shelf at National Semiconductor for ages until it was turned over to a reseller). Since all the ICs were socketed, it was a simple matter to pop the Dallas chip out and attempt a read in an EPROM reader / programmer. It still had a PERFECT copy of the factory diagnostic cartridge's contents in it! On the other hand, I also had a CPU card from the late '80s for the same system that also had a Dallas chip. I dumped that as well, but just looking at the buffer output, I was almost positive the chip had already reached 'bit rot' stage.
 
You must log in or register to reply here.
Back
Top