Yeah, not the fastest, but very reliable and sturdy.

compared to contemporary HDDs of that size or the S7 era, 21 MBps virtually without any access time is awesome.
Those DOMs are the cleaner solution, just a little expensive, AFAIK, compared to flash and SD adapters.

I have a couple of these and I like them

Sphere478 wrote on 2022-05-11, 07:10:

First I tried it on my tyan s1564D
No dice, system hang every time it tries to verify DMI pool data. Even hangs when installed on a add in card. Ideas?

It's been written before by me and confirmed by another user later that these specific DOMs (from 4GB and up) just don't work on older IDE controllers. We were guessing on controllers that lack UDMA, but if it's not working on tyan s1564D maybe it's not that.

Other than that they have been proved reliable.

My first guess would be ATA IDENTIFY https://wiki.osdev.org/ATA_PIO_Mode#IDENTIFY_command returns something BIOS doesnt like and gets confused. Someone would have to slap a logic analyzer on the IDE bus to get to the bottom of the problem.

I wouldnt call flash storage reliable and sturdy per se, but its silent and you get almost unmeasurable access time compared to mechanical drives.

These drives are fine for anything older than 98-era. Small, quiet, modern replacements for fat failing hard drives. They also don't involve finding a PCI SATA adapter, which costs a lot on its own.

Seems like you are close to the max transfer rates for your ATA modes -- Ultra ATA/66 <= 66.7 MB/s

https://en.wikipedia.org/wiki/Parallel_ATA#Sp … _transfer_modes

Couple quick notes on storage benchmarks that I'm sure you have heard before--

1. DOMs usually won't negotiate ATA modes faster than UDMA2 unless you short pin 34, use an 80 connector cable , or connect to an IDE controller that doesn't look at pin 34.
2. Most BIOS won't use UDMA unless you install UDMA drivers or add a card with an option rom that enables it, so if you are benchmarking in plain DOS, you are likely further limited
3. UDMA is usually turned off by default in Win98 unless you go in and enable it
4. It is not uncommon for early LBA BIOS implementations to get unhappy when they look at drives > 32GB and just freak out. Forcing CHS or ECHS in the bios setup can let the drive work, but probably at a smaller capacity. Then you can install drive overlay software to get full capacity.

These are great for when you need them in tight spaces or if you really want to minimize cabling. I’ve used this brand but the 44 pin ones for thin clients with Win9x and it was plenty fast. That’s about the same speed as the SD card adapters which top out at 25mb/sec and those are pretty responsive as well.

douglar wrote on 2022-05-11, 17:28:

Seems like you are close to the max transfer rates for your ATA modes -- Ultra ATA/66 <= 66.7 MB/s […]
Show full quote

Seems like you are close to the max transfer rates for your ATA modes -- Ultra ATA/66 <= 66.7 MB/s

https://en.wikipedia.org/wiki/Parallel_ATA#Sp … _transfer_modes

Couple quick notes on storage benchmarks that I'm sure you have heard before--

1. DOMs usually won't negotiate ATA modes faster than UDMA2 unless you short pin 34, use an 80 connector cable , or connect to an IDE controller that doesn't look at pin 34.
2. Most BIOS won't use UDMA unless you install UDMA drivers or add a card with an option rom that enables it, so if you are benchmarking in plain DOS, you are likely further limited
3. UDMA is usually turned off by default in Win98 unless you go in and enable it
4. It is not uncommon for early LBA BIOS implementations to get unhappy when they look at drives > 32GB and just freak out. Forcing CHS or ECHS in the bios setup can let the drive work, but probably at a smaller capacity. Then you can install drive overlay software to get full capacity.

Sounds like you know your stuff.

I’m benching it in xp. Hd tach.

So what should I try to get it working on the tyan? A bios setting?

What am I shorting pin 34 to? Ground? 5v?

The tyan has worked with 120gb magnetic drives before.

I had similar experience with 8 Gb DOM. Share your pain there.

Here are some pictures of the mod-- UDMA 5 with DOM's & CF's

Sphere478 wrote on 2022-05-12, 02:50:

So what should I try to get it working on the tyan? A bios setting?
The tyan has worked with 120gb magnetic drives before.

A 120GB magnetic drive is most likely ATA-5 with LBA28 addressing. You can verify that with any tool that pulls your ATA info, like sysinfo, hwinfo, or speedsys.
All of the hyperdisk doms I've seen have been ATA-6 with LBA48 addressing.
Ideally, during post, the DOM should fall back to ATA-5 with LBA28 addressing if that's all the mobo knows, but that doesn't always happen. When an old BIOS gets faced with an addressing mode that has values larger than it can handle, stuff that shouldn't be overwritten gets overwritten and there's a crash. Or data corruption if you are unlucky.

If you look at this screenshot from https://www.adminxp.com/hardware/index.php?aid=129

You can see the spot where you can force an addressing mode--
Normal = CHS (Cylinders / Heads / Sectors per track ) that goes up to 528MB / 504 MiB
Large = ECHS that goes up to a max of 8.4GB but it depends. (7.9 GiB)
LBA = Would be 28-bit LBA that goes up to 137.4 GB

I have a 486 with an ATA-3 controller than just doesn't like hyperdisk controllers (ATA-6) but does work with lei disk controllers (ATA-4). It also has Phoenix BIOS 4.04, which crashes if an IDE has a capacity > 3277 MB , but that's a different story.

More info on BIOS limits here: https://www.vogonswiki.com/index.php/Storage# … age_Limitations

Here's a chart for how things benched out with speed sys on my Nforce2 motherboard sorted by min write speed.

All the number should be considered +/- 5% depending on cold boot vs warm boot, unexpected evictions in the on-device cache, solar flares, etc.

The 8GB Hyperdisk dom is about twice as fast as the 4GB hyperdisk because it has 2x 4GB flash chips instead of one

douglar wrote on 2022-05-12, 19:50:

Here's a chart for how things benched out with speed sys on my Nforce2 motherboard sorted by min write speed.

Wow, great results, thanks for the info!
Then I guess it's all about the controller which they're connected to (obviously). Because in my DOS rig (tucson 430hx) it's probably working in PIO mode, albeit slow-ish. If I turn on DMA in W98 device manager, can't boot anymore and have to reinstall Windows. But doesn't really matter much for DOS usage, still way snappier than any old hdd.

Jura Tastatura wrote on 2022-05-17, 20:59:

Wow, great results, thanks for the info!
Then I guess it's all about the controller which they're connected to (obviously). Because in my DOS rig (tucson 430hx) it's probably working in PIO mode, albeit slow-ish. If I turn on DMA in W98 device manager, can't boot anymore and have to reinstall Windows. But doesn't really matter much for DOS usage, still way snappier than any old hdd.

You guess your storage is PIO? That's like guessing at your tire pressure when you have a pocket full of gages. Get one of those lovely free tools that reports your ATA info and check it out.
And you had to reinstall windows after enabling DMA? That's harsh. I'm sure you tried F5 for safe mode, first yes?

Disclaimer: if you have a decent software drive cache with plenty of ram, you can mask most weaknesses of mechanical storage.

That said, here are what I consider to be the 4 main things that I look for to get a "good" storage performance feel, in order of importance:

Sub 1ms Seek time - Getting the first block back fast does wonders for responsiveness, especially when there are a lot of small random io requests.
DMA - Enabling DMA does wonders for the feel of a multi tasking operating system and PIO tasks can devour large chunks of available CPU when you are < 100Mhz.
Buffered writes - Doesn't feel like someone slamming on the breaks if your storage can buffer a 50KB write
Max linear read - Sometimes you are going to need to read through a lot of data and in those cases, you can't "fake the high throughput funk" so to speak.

Just checked my mbo manual (intel tc430hx with p200mmx), IDE controller supports up to PIO4 (16MB/s) with "enhanced fast DMA controller". Picture shows my speed results without DMA enabled. If I enable it then I cannot boot W98, can't remember but probably did try F5 during boot last time I crashed W98. Even Phil on his page mentions you have to reinstall W98 unless you can fallback from UDMAx to PIO mode in BIOS and then boot to W98 to untick DMA. Which I can't.
Not going to risk and try it anymore because in reality this is my DOS rig and I use W98 only occasionally and mostly for network transfers. Sure they could be faster and not hog the pc while copying stuff, but what the heck. 😁

Jura Tastatura wrote on 2022-05-18, 21:18:

Just checked my mbo manual (intel tc430hx with p200mmx), IDE controller supports up to PIO4 (16MB/s) with "enhanced fast DMA controller". Picture shows my speed results without DMA enabled. If I enable it then I cannot boot W98, can't remember but probably did try F5 during boot last time I crashed W98. Even Phil on his page mentions you have to reinstall W98 unless you can fallback from UDMAx to PIO mode in BIOS and then boot to W98 to untick DMA. Which I can't.
Not going to risk and try it anymore because in reality this is my DOS rig and I use W98 only occasionally and mostly for network transfers. Sure they could be faster and not hog the pc while copying stuff, but what the heck. 😁

Usually you can restart in safe mode and once in windows delete the offending drive.
After restart windows will install it again and DMA should not be ticked.

bofh.fromhell wrote on 2022-05-18, 21:55:

Usually you can restart in safe mode and once in windows delete the offending drive.
After restart windows will install it again and DMA should not be ticked.

Now I will have to try it again and if it works wonder how could I be so stupid. 😁
Or I could just stick one of the Adaptec PCI IDE controllers and get done with it. Already did that, actually, but didn't like the fact that it disabled mbo IDE controller.
Honestly, I didn't even think I had a speed "problem" until I saw that spreadsheet from post above, and now it starts bothering me. 😁 😁

Jura Tastatura wrote on 2022-05-18, 21:18:

Just checked my mbo manual (intel tc430hx with p200mmx), IDE controller supports up to PIO4 (16MB/s) with "enhanced fast DMA controller". Picture shows my speed results without DMA enabled. If I enable it then I cannot boot W98, can't remember but probably did try F5 during boot last time I crashed W98. Even Phil on his page mentions you have to reinstall W98 unless you can fallback from UDMAx to PIO mode in BIOS and then boot to W98 to untick DMA. Which I can't.
Not going to risk and try it anymore because in reality this is my DOS rig and I use W98 only occasionally and mostly for network transfers. Sure they could be faster and not hog the pc while copying stuff, but what the heck. 😁

TC430HX doesn't support any UDMA mode anyway. Best it can do is to use PCI Bus Mastering with MWDMA. I don't know whether W98 actually supports PIIX3 bus mastering, the only driver I know of was for W95. Bus Mastering and MWDMA modes were so buggy in the 90s that barely anyone bothered to implement them before UDMA took over. With Bus Mastering you could get the same speed as PIO4 but with much lower CPU utilization.

Jura Tastatura wrote on 2022-05-19, 08:19:

Honestly, I didn't even think I had a speed "problem" until I saw that spreadsheet from post above, and now it starts bothering me. 😁 😁

I wouldn't focus too much on max throughput. When it comes to day-to-day uses, I find that low seek times helps DOS feel more responsive and boot faster. For windows, enabling DMA dramatically smooths out the annoying stutters and hitches caused by back ground IO. Everything after that is for bragging rights and theoretical studies in system bottlenecks. Thinking back, some of those early 3rd party bus master drivers were pretty touchy. You are probably wise for exercising some caution with those. That said-- http://www.benchtest.com/udma.html

To be fair, I picked that Nforce2 motherboard for the benchmarks because it was the fastest pony in my stable. I wanted something fast enough that I could see the differences in fast storage devices. What's the point of benchmarking fast devices if your controller is working like a high-cut filter ?

Also, never assume that you are getting the fastest transfer mode available for your controller, especially when using contemporary storage in a legacy device. It's not uncommon for contemporary devices to immediately drop back to ATA-0 if they don't feel comfortable with the the controller.