Sadler2010 wrote on 2025-07-01, 01:00:

I found this one on ebay with the Marvell 88i8030-TBC chip on it, but I don't see any jumpers. Oh well it arrives I will see how it goes..

I don't think this adapter is unidirectional, which means it will convert PATA to SATA, not the other way around.

ux-3 wrote on 2024-06-30, 12:30:

I had to set the DVD as master and the SATA as slave on some machines to get both to work.

i have the identical adapter. same issue, but with ide hdd instead of dvd-rom. same solution worked on a kt133a board.

mockingbird wrote on 2025-07-01, 03:20:

Sadler2010 wrote on 2025-07-01, 01:00:

I found this one on ebay with the Marvell 88i8030-TBC chip on it, but I don't see any jumpers. Oh well it arrives I will see how it goes..

I don't think this adapter is unidirectional, which means it will convert PATA to SATA, not the other way around.

Confirmed no SATA to IDE happening with it so I could use it to get stuff off IDEs onto SATA systems to archive it. Which would be great to reuse spare ID drives.

There's a new Marvell-based SATA PATA adapter called Xenium XATA made primarily for the old XBox but compatible with PC as well.

Otherwise the best ones I've used are Startech IDE2SAT2 which are also Marvell-based.

I've used a generic/unbranded adapter recently but it didn't work reliably - it was similar to the OG picture but had no jumbers for Master/Slave etc. I ended up changing it for the Startech one, which worked fine (Apple G3 Blue and White btw)

nfraser01 wrote on 2025-08-26, 11:20:

I've used a generic/unbranded adapter recently but it didn't work reliably - it was similar to the OG picture but had no jumbers for Master/Slave etc. I ended up changing it for the Startech one, which worked fine (Apple G3 Blue and White btw)

I know you are writing for Apple, but yesterday I did some troubleshooting for PIIX4, and it I discovered that it is best to NOT use Marvell, but rather to use JMB363.

Intel SSDs should be aligned to 512 during partitioning for the JMB363, NOT 4k, or else you will experience write errors.

mockingbird wrote on 2025-08-26, 15:37:

I discovered that it is best to NOT use Marvell, but rather to use JMB363.

Intel SSDs should be aligned to 512 during partitioning for the JMB363, NOT 4k, or else you will experience write errors.

That's an interesting edge case to look out for. I never heard of that one.

Do you know if affected all models of Marvell Pata<->Sata bridge chips or was it just a specific chip model?

Do you know if it was a specific intel SSD firmware on a specific line of SSD's ?

I can't find anything in the white paper from Intel regarding partition offsets:
https://www.intel.ph/content/dam/www/public/u … -tech-brief.pdf

2.1 Partition Offset Windows* up to and including Server 2003* uses a first partition offset of 63 sectors to comply with the 63 hidden sectors reported by disk hardware. Assuming 512 Byte sectors gives an offset equal to 31.5KB. This works fine on traditional 512B sector HDDs, but does not work well for Intel SSDs due to misalignment. It is possible to manually align the partition offset to 64 sectors, or 32 KB. This configuration is better than the misalignment caused by 63 sectors, but is not recommended for Windows* servers. Windows* Server 2008 and newer use an offset of 1024KB (1 MB), which is the recommended offset for Intel SSDs. Partitioning tools in Linux, such as GPARTED, also use an offset of 1 MB in newer versions.

If these SATA-IDE things really do try to interpret the MBR/partition table or change behavior based on where partitions begin and end, as opposed to being a dumb pipe that feeds sectors back and forth, it sure explains a lot of the problems with them...

douglar wrote on 2025-08-26, 17:41:

Do you know if affected all models of Marvell Pata<->Sata bridge chips or was it just a specific chip model?

Do you know if it was a specific intel SSD firmware on a specific line of SSD's ?

That's a very good question -- and I thought about this yesterday and decided that I was going to approach the matter in an extrapolative fashion rather than diagnose the cause of it. Mainly because I already spent years in limbo with this drive and merely thinking about what causes the incompatibility never solved anything in the short term. I don't know how 512 to 4k AF translation works, that's why I say this, and there are so many factors at play (does the bridge chip need to be compatible with it?, do different NAND SATA controllers play differently?, different behaviour amongst southbridge IDE controllers, etc...).

But I can tell you how to replicate it -- just boot to dos and run an XCOPY of a large amount of data. Before long, you'll get a square box in the upper left corner of DOS informing you of a serious write error (this may also happen in Windows setup, but is much easier replicated in DOS) with xcopy).

When creating your partition for your vintage PC, use AOMEI Partition Manager and make sure "SSD Optimization" is checked, and then select 512 from the drop down box when creating your partition.

Wes1262 wrote on 2024-06-30, 13:10:

https://m.media-amazon.com/images/I/71tOJSwDI+L._AC_SL1500_.jpg I like this in particular. Costs 20 euro ish. […]
Show full quote

I like this in particular. Costs 20 euro ish.

After trying many different adapters to no avail on my Powermac G4 Sawtooth, the one that looks like this one is the only one that worked fine.

mockingbird wrote on 2025-08-26, 19:40:

But I can tell you how to replicate it -- just boot to dos and run an XCOPY of a large amount of data. Before long, you'll get a square box in the upper left corner of DOS informing you of a serious write error (this may also happen in Windows setup, but is much easier replicated in DOS) with xcopy).

When creating your partition for your vintage PC, use AOMEI Partition Manager and make sure "SSD Optimization" is checked, and then select 512 from the drop down box when creating your partition.

Do you think this is explicitly about 512 vs. 4k, or DOS wanting partitions to be aligned on a cylinder boundary, which modern stuff ignores even when using MBR style partitions?

jakethompson1 wrote on 2025-08-26, 20:05:

Do you think this is explicitly about 512 vs. 4k, or DOS wanting partitions to be aligned on a cylinder boundary, which modern stuff ignores even when using MBR style partitions?

I think the former rather than the latter with the caveat that we're referring to FAT32, and that it is specific to the combination of controller and NAND controller. I think Intel had their own in-house controllers for their drives. I also suspect that Silicon Motion and Marvell NAND controllers of olde had better compatbility and tolerance for odd setups like this. If you need FAT16 DOS, I state with ignorance that you are probably introducing more factors in the equation.

Again, the way I deal with drives and controllers with vintage systems is to test it and see what works. I am not smart enough to understand the intricacies of all these subsystems nor do I have the time to study them.

"Intel SSDs should be aligned to 512 during partitioning for the JMB363, NOT 4k, or else you will experience write errors."

I have not been able to find any first hand reports of this happening or vendor warnings about this. If you find any links to descriptions of this issues, please post them. I enjoy reading up on things like that.

I'm digging at this because it goes against several bits of storage knowledge that I accept as true--
1) My understanding is starting the first partition at 512 bytes into the SSD storage is not the correct thing to do. It will reduce performance and shorten the life span of the device. Most SSD's produced since 2004 use 4k cells and you want your file system allocation units to line up with the cells, not be off by 512 bytes so that each allocation unit hits two cells.
2) I'm not sure how an pata-sata bridge could change sector mapping. The bridges don't have enough smarts to do that. If they did, you would know because your data would be become unreadable if you removed the pata-sata bridge.
3) There were some early SSD's that did use 512 byte flash cells, but since 512 divides 4k, it's aligned at 512 and at 4k, so alignment was not an issue on these.
4) Even if you don't have the partition alignment right, it should just make your drive slower, it shouldn't cause data corruption unless there's a serious bug in the SSD firmware.

Intel & Microsoft both appear to recommend setting the first partition at the 1 MB point on the drive, which would be aligned with a 4k boundary ( and a 512 boundary). Apple doesn't have a recommendation, but their diskutil tool has always aligned partitions on 1MB boundaries, so maybe it's never required clarification. There's no caveat about pata-sata converters.

douglar wrote on 2025-08-27, 01:41:

I have not been able to find any first hand reports of this happening or vendor warnings about this. If you find any links to descriptions of this issues, please post them. I enjoy reading up on things like that.

They won't exist, none of these NAND vendors anticipated that people would use their drives with that Taiwanese bridge chip. The SM2246EN has support for legacy IDE, so maybe it's more forgiving than an Intel controller.

1) My understanding is starting the first partition at 512 bytes into the SSD storage is not the correct thing to do. It will reduce performance and shorten the life span of the device. Most SSD's produced since 2004 use 4k cells and you want your file system allocation units to line up with the cells, not be off by 512 bytes so that each allocation unit hits two cells.

Shorten the life, maybe. Slow down the drive -- not in my use case. I tested both 512-aligned and 4k-aligned with ATTO, the results were identical. Remember, I am using this with PIIX4, so I'm stuck at UDMA2. I also run RLOEW's Trim every once in a while. It also helps that the Intel 530 is MLC and not TLC. I don't see any issues in the long term, in terms of endurance.

4) Even if you don't have the partition alignment right, it should just make your drive slower, it shouldn't cause data corruption unless there's a serious bug in the SSD firmware.

Intel & Microsoft both appear to recommend setting the first partition at the 1 MB point on the drive, which would be aligned with a 4k boundary ( and a 512 boundary). Apple doesn't have a recommendation, but their diskutil tool has always aligned partitions on 1MB boundaries, so maybe it's never required clarification. There's no caveat about pata-sata converters.

Hey, I'm just reporting my results... There are other reports of chicanery with Intel SSDs in these legacy setups... This may or may not be unique to Intel. I don't know what the default behaviour of AOMEI is when "SSD optimization" is left un-checked and you create a FAT32 partition. I can test that, but you would have to tell me how I can check how it aligned it ex-post-facto.