VOGONS


First post, by Marco

User metadata
Rank Oldbie
Rank
Oldbie

Dear all,

In vogons Storage Wiki I read: Few CF's support ATA block transfer sizes > 1.

This means to mean that there ARE CFs with larger block transfer sizes.

According to Wikipedia the CF5.0 standard includes this: „Capacity points beyond current limitation of 137GB (up to 144PB) & more efficient data transfer (32MB per transfer versus 128K per transfer)“

Question: do you know which models (CF or SD) to support this? Or can with be sure that all CFs with rev5 or higher to support this?
Addon: I couldn’t find any Cf card that had its VF revision labeled anywhere so far 🙁

Thank you

1) VLSI SCAMP 311 | 386SX25@TI486SXLC2-50@63 | 16MB | CL-GD5428 | CT2830| SCC-1 | MT32 | WDC160GB/7200/8MB | Fast-SCSI AHA 1542CF + BlueSCSI v2/15k U320
2) SIS486 | 486DX/2 66(@80) | 32MB | TGUI9440 | LAPC-I

Reply 1 of 10, by jakethompson1

User metadata
Rank Oldbie
Rank
Oldbie

I assume you're referring to supporting more than one sector per interrupt in PIO mode.

I haven't seen one that supports it.

Given those huge sizes (128KB and 32MB) those almost certainly are referring to UDMA transfers, so it doesn't sound relevant to what you're asking here.

Reply 2 of 10, by Marco

User metadata
Rank Oldbie
Rank
Oldbie

Indeed your assumption is right I’m referring to the PIO capabilities. Hm ok sad but thanks

1) VLSI SCAMP 311 | 386SX25@TI486SXLC2-50@63 | 16MB | CL-GD5428 | CT2830| SCC-1 | MT32 | WDC160GB/7200/8MB | Fast-SCSI AHA 1542CF + BlueSCSI v2/15k U320
2) SIS486 | 486DX/2 66(@80) | 32MB | TGUI9440 | LAPC-I

Reply 3 of 10, by jakethompson1

User metadata
Rank Oldbie
Rank
Oldbie

A spinny SATA drive through an IDE converter, when it works, can do better on some benchmarks due to multi-sector transfers. And it's not like they make much noise. If a 1TB one is so large that it hangs your Award BIOS for example, you can set a Host Protected Area on another machine to simulate the 32GB jumper and hide the extra capacity.

Reply 4 of 10, by Marco

User metadata
Rank Oldbie
Rank
Oldbie

Thx. Do Sata ssd Like Samsung 830 Support multi sector transfers?

1) VLSI SCAMP 311 | 386SX25@TI486SXLC2-50@63 | 16MB | CL-GD5428 | CT2830| SCC-1 | MT32 | WDC160GB/7200/8MB | Fast-SCSI AHA 1542CF + BlueSCSI v2/15k U320
2) SIS486 | 486DX/2 66(@80) | 32MB | TGUI9440 | LAPC-I

Reply 5 of 10, by jakethompson1

User metadata
Rank Oldbie
Rank
Oldbie

Not in my experience and I suspect it's because by the time SSD firmware was being written, they only needed bare bones PIO support, since it would be only used by things like the BIOS and bootloader before getting a full OS running with UDMA support.

Reply 6 of 10, by Marco

User metadata
Rank Oldbie
Rank
Oldbie

Thanks. Than I will cancel that exercise and stick to my last gen PATA HDD (see signature).

I already tested DOMs and SD cards w/o multi sector transfers all and lot of benches were slower than with my HDD - especially linear ones.

I was hoping to find fast flash drives with multi sector transfer capabilities but as these don’t seem to really exist I have to stop it here unfortunately 🙁

Thanks a lot anyway

1) VLSI SCAMP 311 | 386SX25@TI486SXLC2-50@63 | 16MB | CL-GD5428 | CT2830| SCC-1 | MT32 | WDC160GB/7200/8MB | Fast-SCSI AHA 1542CF + BlueSCSI v2/15k U320
2) SIS486 | 486DX/2 66(@80) | 32MB | TGUI9440 | LAPC-I

Reply 7 of 10, by jakethompson1

User metadata
Rank Oldbie
Rank
Oldbie
Marco wrote on 2025-06-04, 20:04:

I was hoping to find fast flash drives with multi sector transfer capabilities but as these don’t seem to really exist I have to stop it here unfortunately 🙁

My hope is that as ZuluIDE continues to develop, this will be something that can be easily added to the firmware for the disk drive support if it's not there already

Reply 8 of 10, by Marco

User metadata
Rank Oldbie
Rank
Oldbie

great. I will try to follow up on this product as well. Thx.

Btw I found following products to support multi sector read / write in PIO mode:

- Innodisk iCF 4000 and higher
- Transcend PSD330 SSD
- Innodisk iCF 1IE2 CompactFlash
- Cactus Technologies 300/300-P Series Industrial Grade SSD

All in all the information are not very clear and very specific in each case. But these are my findings so far.

Maybe I will check the transcend as another member here was already proposing to test at least

1) VLSI SCAMP 311 | 386SX25@TI486SXLC2-50@63 | 16MB | CL-GD5428 | CT2830| SCC-1 | MT32 | WDC160GB/7200/8MB | Fast-SCSI AHA 1542CF + BlueSCSI v2/15k U320
2) SIS486 | 486DX/2 66(@80) | 32MB | TGUI9440 | LAPC-I

Reply 9 of 10, by douglar

User metadata
Rank l33t
Rank
l33t
Marco wrote on 2025-06-04, 20:49:

- Innodisk iCF 4000 and higher

I was able to test an 8GB Innodisk iCF 4000. XTIDE reports that it is willing to do block mode transfers = 2. I guess that's better than 1.

I did a little testing on a Promise PDC20630 based VLB card, 486@133 and promise driver v3.3. I got the best results with mode D8. ~7000KB/s reads, 2600KB/s writes. Maybe 5% faster than a Transcend CF220i on reads, but only 1/2 the speed on the writes. I only had a little time to test though and there's a lot of variables to pin down when testing storage.

Initial attempts to get the Innodisk iCF 4000 to work with Promise mode M8 didn't work out. This is what chewed up the time that I had. Seems like the newer the firmware, the less likely that Promise's hybrid MWDMA2 transfer mode works. My old blue "Data Partner 512MB Industrial" CF's with the ~2008 firmware get ~10,000KB/s in MWDMA mode on that system. The blue CF’s get the most out of MWDMA. Now things would be really different on a UDMA5 system. The 8GB devices would likely do a lot better, but I don’t see a lot of UDMA on a 486.

The current version of XTIDE universal BIOS likes the PDC20230 better, so I should go back and test with that.

Might makes a bigger difference on ISA systems where the bus is completely saturated and every interrupt is expensive.

Last edited by douglar on 2025-06-17, 03:15. Edited 1 time in total.

Reply 10 of 10, by Marco

User metadata
Rank Oldbie
Rank
Oldbie

Thanks. Same results here - the icf4000 allows for 2 instead of 1 block transfers. but going back from 2 to 1 reduces the performance just about 0,5% thus very marginal. Read performance is still not on par with my hdds allowing 8 block transfers. (4400kb/s vs 3850kb/s). Next test maybe Transcend PSD330 SSD

1) VLSI SCAMP 311 | 386SX25@TI486SXLC2-50@63 | 16MB | CL-GD5428 | CT2830| SCC-1 | MT32 | WDC160GB/7200/8MB | Fast-SCSI AHA 1542CF + BlueSCSI v2/15k U320
2) SIS486 | 486DX/2 66(@80) | 32MB | TGUI9440 | LAPC-I