Retrotechtacular: Double For Nothing

If you are in the market for web hosting in 2021 and you sign up with one of the cloud computing providers, you’ll soon see how the different resources are priced. Storage and bandwidth are cheap, while CPU time is expensive. This reflects the state of a modern computer, in which a typical disk drive now holds a terabyte or more and rising by the year while a new processor is becoming a bottleneck whose performance hasn’t increased as much as the manufacturers would like over models from years ago.

Twice As Much Hardware From A Bit Of Software?

In the early 1990s though it was a different matter. A 486 or early Pentium processor was pretty powerful compared to the DOS or Windows 3.1 software it was expected to run, and it was the memory and disk space attached to it that limited performance… and cost an arm and a leg. There was a period in about 1995 when a supposed fire in a chip factory somewhere sent RAM prices into the hundreds of dollars per megabyte, briefly causing an epidemic of RAM raiding in which criminals would break into offices and take only the SIMs from the computers.

A solution to this problem came perhaps surprisingly from the software industry. Disk Doubler was a DOS driver that promised more disk space, achieving this seemingly impossible feat by compressing the disk to fit more data on it. Processor power swapped for disk space was a reasonable trade at the time so it became extremely popular, and eventually Microsoft incorporated their own disk compression into DOS. In some cases it could even speed up a computer with a slow disk drive, as I found out as a student with a 286 packing an MFM drive.

Something For Nothing, Perhaps It’s Too Good To Be True.

If compression could increase disk space then couldn’t it do the same for RAM? The industry came to the rescue once more with an array of RAM doubler products, first applying the disk doubling technique to on-disk virtual memory, and then doing the same with the contents of the memory itself. The first approach worked at the expense of a system slow-down, while the second, not so much. In fact it was little more than a scam, with software products promising much but delivering absolutely nothing behind the scenes.

The video below the break takes a look at Syncronys SoftRAM, and takes us through its operation. There were RAM optimisers for WIndows 3.1 that worked by moving as many items as possible from the 640k of conventional memory in a DOS PC to the extended memory supported by 386 and higher processors. These worked very well in allowing many more processes to run, but while they increased the available conventional memory, they didn’t claim to deliver any extra memory.

SoftRAM by comparison appeared to work in exactly the same way as the disk doublers, and had an impressive dashboard with a meter showing just how much extra memory you had gained by using it. Unfortunately that was the only thing the software did, and investigators soon rumbled the scam. Lawsuits and an FTC investigation followed, and the company finally folded in 1998 with many of its fines and rebates for users unpaid.

At the time, most non-corporate PC users were sitting in front of their first computer, usually bought from a big-box electronics retailer more used to selling appliances. They were sold on wonders of the age such as CD-ROM multimedia discs, and their owners had not yet learned to mistrust software products of uncertain provenance. The term coined for SoftRAM was “Placebo software”, for software that simply reassured the user it was doing something such that they imagined their machine was faster. Given that the company sold 700,000 copies of the software and many of those users never even suspected that there was anything wrong with it, evidently they were on to something.

47 thoughts on “Retrotechtacular: Double For Nothing

  1. It was (and is) a problem with technology literacy. Computers are magic boxes and most people have no idea how they work. They’re unable to pose a simple thought experiment of what-if’s because they have no idea of how to apply cause and effect to something they do not understand. People need to have a model in their head, even if greatly simplified, in order to make rational choices on that topic. Without this we go with gut, intuition, and emotion. And you get products like this.

    Once CPUs got fast, and some low computational cost compression algorithms appears it became practical to compress swap and RAM. (e.g. zram and compcache on Linux)

      1. We still have sham computer products like this today, from USB sticks that protect from 5G to stickers that protect from RF to USB sticks that prevent hackers.

  2. I remember the panic of 1995, but never saw anything but a fire as the cause of the shortage (or higher prices).

    It did get back to normal, in the fall of 1996, I got four 1M SIMMs for my Mac Plus for ten dollars each. That was after being given some 256K SIMMs, taking off the RAM, and replacing it with higher density RAM I took off a scrap board. Seemed to work fine in the Mac Plus, until I let it sit. When I pressed a key or moved the mouse, it would crash. It didn’t refresh properly (actually use made sure it was refreshed), something I then read about.

    When I got a 512K upgrade board for my Radio Shack Color Computer III about 1987, it was because Microware OS-9 made use of it. But the board came with “bonus software”, a print spooler, and RAM disk. It made me realize they wanted to sell the expansion, but it did nothing for the CoCo without OS-9. Hence the utilities to give value to that extra memory.

    1. What a crock disk doubler was. It trashed my disc on a couple of occasions. Iirc it assumed 2:1 compression, and if you failed to get that on a nearly full disc, you could wrap round over the boot sector. Stacker was by far a better product.

  3. It didn’t exactly stop there.

    Anyone remember the ram boosters that were supposed to free up memory? The registry cleaners? None of that actually did anything useful.

        1. A win 9x system that had many programs installed, uninstalled, more stuff installed and removed could end up with a ton of crap left behind in the Registry. I used Regclean and on installs like that it could take out all the junk and reduce the registry size by several megabytes.

          To help even more, after RegClean I’d boot to DOS mode and export, compact, and re-import the registry. That also defragmented it.

          On computers with really “dirty” registries this could shave a lot off boot time in addition to making it work better thanks to freeing up the RAM used by the bloated registry. Win 9x loaded the whole registry into RAM. When a computer has megabytes instead of gigabytes of RAM, every bit saved helps.

    1. Eh. I used to use tools that would free up RAM by killing some idle processes that weren’t started by the system, usually bloatware stuff. I got into the habit of installing it on people’s PCs when I fixed them for cash in high school.

      It worked and made a difference since you no longer had a bunch of bloatware running in the background, but it masked a symptom rather than fixing the issue.

    1. It was often far more useful to do the opposite.
      Create a virtual drive in the RAM. Copy over some essential programs at boot and then feel the “blistering fast” speed.

        1. I recall various schemes to add solid state drives, though most were not long term storage. But every time there was some breakthrough for better density, program size would leap forward and the scheme was barely useful.

          I especially remember about 1984, a multi -part article about using bubble memory as a drive. So yes, itdidn’t forget. But part way through, the developer’s board it was based on dried up, end of the potential. Bubble memory faded soon after.s

    1. The main reason it worked was that macs of that time didn’t really do memory management. They just handed each application a giant fixed-size block of ram, on which it had its own independent stack and heap. Programs would ask for the largest block of memory they could possibly need, meaning that in normal usage, each program would have a ton of empty memory in its allocation. Ram Doubler, I believe, either made all applications use the same heap, or dynamically changed the size of the programs memory blocks. If this couldn’t create enough free space, Ram Doubler could turn to other means, but these would come with heavy performance penalties.

      1. RAM compression worked in the classic Mac System and MacOS because they treated RAM like a disk drive.

        For example, the Mac had 7 megabytes free after booting. Launch a program that needed 3 meg, then one that needed 2 meg, then one that needed 1 meg. Leaves one meg free. Quit the 2 meg program and try to launch a program that needs more than 2 meg and you get “Not Enough Memory” despite having 3 meg free. You’d have to quit and relaunch the 1 meg, then launch the 2+ meg, or quit the 1 meg, launch the 2+ meg, then relaunch the 1 meg. Just like saving, deleting, and re-saving files to disk.

        If your Mac had enough RAM it didn’t need compression. What Connectix RAM Doubler was really good for was its replacement for Apple’s lousy virtual memory. I used it with the compression set to off.

        The other must have software for old Macs was Jump Development’s RAM Charger. That managed RAM by dynamically reallocating all free RAM into one block. It also had dynamic allocation of RAM for programs. They would launch with just a little free space assigned and if/as the software required more, RAM Charger would supply it. Then when the software was done using it, RAM Charger would take it back for the single piece free space. A very small number of programs wouldn’t work with dynamic allocation so RAM Charger had per-program settings to use the default allocation settings.

        Apple apparently did not like being upstaged on memory management. Rather than license a version of RAM Charger or buying it outright (as Apple had done with many 3rd party tools and utilities which became official parts of the OS), Apple decided to break it. Every new update from 7.1 through 9.2.2 made changes to make RAM Charger incompatible so Jump Development would have to figure out what Apple had done and change their software to work again, until the next OS update. What must have really irritated whomever at Apple was responsible for this was Jump was able to make each update compatible with all the previous OS versions so nobody had to use some older RAM Charger for MacOS 7.5. The last version RAM Charger is 100% compatible with was 9.1. With 9.2.x there are changes that broke some features but not the core function of RAM defragmenting and dynamic allocation. Dunno why they never did one final update for full 9.2.x compatibility after Apple was done with the Classic OS. Wouldn’t have surprised me if Apple would’ve released 9.2.3 touting some special improvements but really just another RAM Charger breaker.

  4. SoftRAM 3.1 was a nice trick if you didn’t know any better ways of doing the job (or couldn’t download equivalent freeware from your local BBS). SoftRAM 95 was a scam, plain and simple.

  5. Yeah I remember this on-the-fly compression back in the days of Windows 95. I was 11 at the time and had this transparent compression driver AND WinZip on my computer back then. Had to learn the concept of entropy the hard way when trying to store a 2.5MB zip file on a floppy disk compressed to a virtual capacity of 2.88MB or so.

  6. People were stealing the RAM out of computer rooms in schools and colleges. I might have the time period wrong, but I had a nice product sales ruined when Micron could not compete on the global stage and filed dumping charges against Japanese and Korean memory makers, and succeeded! The action taken was so broad that the special video RAMs my device depended on were caught in the “fair market value” tariffs and quadrupled in price. It affected all RAM pricing for quite a while. Bastards.

    1. It was actually both. Extended memory was any memory above 1mb. Expanded memory was a block of extended memory mapped into a block (typically a 64k block) of lower memory below 1mb. By swapping in different regions of extended memory, DOS applications could access more memory with out having 32 bit support.

        1. I used to have a 12Mhz 80286 with 12 megabytes of RAM. 512K in DIP chips on the motherboard, the rest in three 16 bit ISA cards from Micron. Card 1 backfilled the 512K to 640K then its remainder and the other two I split 50/50 between XMS and hardware EMS. The cards used EEPROMs and a DOS setup program to define what parts of the RAM went to each type.

          The other cards were video, multi-I/O, MFM hard drive controller, Soundblaster Pro, REEL Magic video + Pro Audio Spectrum 16 *minus* all the chips and parts for video and SCSI.

          I had the autoexec.bat and config.sys and all my DOS games setup so they all would automatically use whichever was the best sound card they supported. So if a game supported the PAS 16 it’d use it, otherwise it’d use the Soundblaster because a real Soundblaster was better than the PAS 16’s SB emulation.

          The PAS 16 was a nifty sound card, designed to be able to use two simultaneously. Apparently the REEL Magic card I had was intended for use as a secondary card with the fully stuffed version. A PAS 16 with the Trantor SCSI controller could be used with any SCSI device, hard drives, scanners etc – as long as you didn’t mind a transfer speed topping out at what a 4x CD-ROM could do, and it didn’t support booting.

  7. “Placebo software” reminds me of the time I doubled the speed of a colleague’s PC. For the benefit of younger players the speed display was just a set of seven segment LEDs that were programmed by plugging jumpers. I altered the settings so it showed twice the original value. He was convinced his PC was working faster and *so* grateful.

  8. The Ramdemic, Before it had happened I had upgraded my 486 to 16MB of RAM. Shortly into the Ramdemic my Dad lost his job because Clinton decided to close some military bases and I decided to help out the Family by pulling the RAM back out and sold it for $800.00 and it paid the mortgage for two months. Dad found a new job and the following year I was able to get another four sticks of 30pin simms and got 16MB again for around $100.00

    1. Clinton had some obvious character flaws, but the closure of these bases was one of the best things to happen to the American military for both organizational efficiency and logistics, and for technological reasons.

      1. I remember when he campaigned in Florida right after Hurricane Andrew, promised to get Homestead Air Force Base all fixed up from the damage. One of his earliest acts as President was getting Homestead closed. He was President and could do what he wanted. It wasn’t like he was dependent on just one State anymore, like when he lost re-election for Governor of Arkansas when he didn’t follow through on his promise to the Teamsters Union to get the truck weight limit raised to 80,000 pounds. Next election he lost union support, and the election. So for the election after that, same promise. The unions supported Clinton and he won, and that time he followed through.

  9. That’s because half of cpu time is spent protecting the integrity of the other half and also spying for corporate “statistics”: the antivirus and the “activity report”. Cortana is acting like a nazi prostitute: it eavesdrops on any conversation then passes the messages to the “Gestapo”.
    Get *BSD working and reclaim the other half. Wou win “Double for nothing”: no more money spent for Ms-Sh%t licenses, no more money for antivirus and bonus: no more time spent on Ms-“hotline” talking to tech support. And no more digital prostitution. Simple.

  10. “an epidemic of RAM raiding in which criminals would break into offices and take only the SIMs from the computers”

    Uh, yeah, no. That never happened. I was maintaining about 50 PCs in a corporate setting that year and no one was stealing memory. On the contrary we were handing it out like candy.

    Think about it: someone needs to break in, partially disassemble a computer, steal the SIMM and nothing else and sell it somewhere for less than cost. If there’s a stupider way to commit crime I haven’t heard of it.

    1. I was working in the games/CD-ROM industry at the time. And yes, it did happen. One of the companies we were dealing with came in one morning to find all their PCs gone. They were found about half a block away with all RAM stripped out of them. They were far more worried they’d lost all the game assets and other work on the HDDs.

      Fun fact, they nicked the ROMs from some of the Apple Macs, that were on a daughter card that looked like a SIM.

  11. An MFM hard drive! That brings back some memories. I remember the computer teacher in high school showing me how to do a low-level format of the full height 30 mb drive that used to be in my Dad’s IBM AT into a whopping 45mb! That was living space, man. I could install BOTH Wing Commanders!!

    1. Not entering the defect map when doing a low level format, then doing a full, unconditional format to see if there really were any defects. Odd thing was, the MFM drives I had all worked 100% doing that. Apparently the manufacturers were excessively cautious in defining “defects”.

      What I really wanted back then was an RLL controller to double the capacity of MFM drives. Would’ve been very nice for my first hard drive, a 5 meg, 5.25″ full height Tandon with stepper motor head actuator. I installed DOS and all the software I had and the drive was *half full*. Then I did a full backup onto 360K floppies. A LOT of 360K floppies.

  12. OSX/MacOS has had memory compression since 2013, and IOS since V7 (2009?). On macs with memory pressure (low on available RAM) the OS agressively kills processes not being used, then reclaims application ram marked purgeable, then reclaims any file cache RAM, then begins to compress least recently used RAM typically by 50%. Under additional pressure, OSX will use swap as a last resort. For a number of years this meant that intel hardware running OSX typically required half the physical RAM of an identical system running Windows. Microsoft finally caught up and Windows 10 build 1507 has memory compression for data it was going to swap out. For Linux, it’s always been very memory efficient anyway, and via zram/zswap Linux can compress RAM it needs to swap out when under memory pressure.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.