The Serpent Rider wrote on 2020-09-12, 12:39:

I wouldn't jump to conclusion yet. Perhaps SSD is just that horrible at linear read/writes. Looks like Patriot low-end SSDs are quite bad performers for such capacity (no cache, bad NAND?).

Sis 962L is just a rebrand of 961

962L was major upgrade, supporting USB 2.0 and Firewire.

What about the hard drive? Despite being a Maxtor, is it normale that it does just 68 mb\s as peak in your opinion?

You're expecting too much from "potato" 160 Gb HDD.

You are probably right, those are the results with another cheap ssd:

Same ssd connected with usb3\sata sabrent adapter to my main rig:

I find it curious that every sata device I connect give a tops the result at around 90 mb\s.
Wd velociraptor:

Again it tops at around 90mb\s...
So or the Ata-133 controller is flaky or the ide-sata adapter... Who knows...

I've tested another sis mb, it is a MSI MS-6719 (MEDION) which use as NB the Sis 651 and as SB the Sis 962, i've been able to install windows xp in a ssd but guess what?
The result are sloppy as before:

Ok, I've tried Gigabyte GA-8SR533P board with cheap chinese adapter and I've got similar results. Around 80-90 mb/s read speed on InnoDisk FiD 2.5 SATA10000. But sometimes I've got spikes up to 108 mb/s in some tests. So either SIS 962L aren't very fond of such adapters or IDE performance is somewhat inconsistent.

Nemo1985 wrote on 2020-09-08, 15:35:

Thank you for the feedbacks, I was wondering if it wasn't detected because it has been formatted from windows 7 so it uses a different cluster size? Could it be?
But now that I think I have a sata velociraptor too I will check if it get's detected or not and report back.

Windows Vista and newer Windows versions use LBA formatting with 1 MiB aligned partitions. They do not care about legacy CHS informations.
This may not compatible with Windows 98 or XP, because those old systems need non-LBA informations to boot.

CHS aligned partitions (DOS, Win98, XP) are compatible with Windows Vista and newer.

But...

Newer disks have 4 KiB physical sectors and may emulate classic 512 byte sectors (4K native with 512e).
When you use CHS, there are 63 sectors reserved in track 0 head 0. This means that the first sector is unaligned on disks with 4K native sectors.

If the sectors are not aligned it may happen that the disk has to read the rest of the physical sector before being able to write sectors to the disk.
If the alignment is proper there is no need to read before write.

Using SSDs and 4k native discs I prefer manual alignment. First partition starts at C=0 H=8 S=1. 504 sectors before partition start.

The Serpent Rider wrote on 2020-09-23, 18:35:

Ok, I've tried Gigabyte GA-8SR533P board with cheap chinese adapter and I've got similar results. Around 80-90 mb/s read speed on InnoDisk FiD 2.5 SATA10000. But sometimes I've got spikes up to 108 mb/s in some tests. So either SIS 962L aren't very fond of such adapters or IDE performance is somewhat inconsistent.
IDE test.png

I'm glad that you have been able to confirm those weak results, to get it definitely sorted, is there any eide device able to actually reach the ata133 performance?

Nemo1985 wrote on 2020-09-24, 11:34:

is there any eide device able to actually reach the ata133 performance?

Doubtful. Maybe some modern IDE SSDs, but they usually have crappy controllers and/or lack cache (and likely also use bridge). I think ATA133, at its initial release, was more of a RAID thing.

I thought a ATA133 was the max ATA speed because it was synchronous with a 32 bit 33Mhz PCI bus. Limited returns running your drive interface faster than your system bus, right?

Test system:

• P5I430TX TITANIUM IB+ with 2003 bios
• MII-333GP (66 MHz bus)

Here's what I got with:

• Promise 150 PCI sata adapter
• 128GB Crucial M4 SSD

I think that's about the best throughput you can hope for on a PCI bus, stock speeds, no?

I liked reading this pata vs sata article from 2003--
https://techreport.com/review/4886/ata-hard-drives-compared/

I thought a ATA133 was selected because it was synchronous with a 32 bit 33Mhz PCI bus

It was primary limitation, yes. But nothing outside of RAID could achieve such speed anyway.

The Serpent Rider wrote on 2020-09-24, 22:37:

I thought a ATA133 was selected because it was synchronous with a 32 bit 33Mhz PCI bus

It was primary limitation, yes. But nothing outside of RAID could achieve such speed anyway.

Yep. Looking at those drives from 2003, they couldn't get past 45MB/s even when operating out of the cache.

I'll probably check later how far I can push similar VIA south bridge with ATA133 support - VT8235.

Ok, so I've tested VIA VT8235 and guess what? Yep, it also can't reach ATA133 peak performance. Although it's mildly better than SIS 962L, with brief peaks around 113 Mb/s.

Ignore write speed, I've used another SSD which is still capable to achieve read speed above ATA100 - Sandisk SDSA6AM-008G

In conclusion, ATA133 peak speed isn't achievable from one device on such old south bridges.

The Serpent Rider wrote on 2020-09-25, 22:11:

Ok, so I've tested VIA VT8235 and guess what? Yep, it also can't reach ATA133 peak performance.
....
In conclusion, ATA133 peak speed isn't achievable from one device on such old south bridges.

I agree that ATA peak speed isn't achievable, but not entirely because of the storage device.

You are never going to see theoretical ATA133 raw throughput when you are connected to a PCI bus on a functioning computer.

ATA over head, ATA to PCI transfer latency, and PCI over head all take their bite before you start to factor in PCI contention.

If you get more than 80 MB/s real throughput on a PCI connected hard drive, you are doing great.

https://documentation.help/The-Hardware-Guide/agp-pci.htm
Due to contention with other devices on the PCI bus and the lack of pipelining, PCI offers roughly 80 MB/sec.

You are never going to see theoretical ATA133 raw throughput when you are connected to a PCI bus

Once again - internal IDE ports were not connected to PCI bus since early 00s. They used proprietary buses like V-Link or MuTIOL.

The Serpent Rider wrote on 2020-09-25, 23:52:

You are never going to see theoretical ATA133 raw throughput when you are connected to a PCI bus

Once again - internal IDE ports were not connected to PCI bus since early 00s. They used proprietary buses like V-Link or MuTIOL.

I had been talking about PCI. My bad.

Maybe we agree on these points?

• Single hard drives prior to 2006 didn't push real world throughput over 50MB/s real throughput
• At some point past 50MB/s, the portion of the total system throughput resistance caused by Ultra ATA/133 or SATA 150 bus begins to increase
• By the time you get to 80MB/s real throughput on a PCI attached device, a significant portion of the resistance is coming from the PCI portion of the system
• By the time you get to 90MB/s real throughput on an Ultra ATA133 attached device or bridged SATA devices, a significant portion of the resistance is coming from the ATA interface.
• By the time you get to 110MB/s real throughput on a native SATA 150 attached device, a significant portion of the resistance is coming from the SATA interface.

Do those sound like reasonable ball park numbers?

By 2006, single drives started getting close to those limits--

https://techreport.com/review/10152/western-d … dfd-hard-drive/

By the time you get to 110MB/s real throughput on a native SATA 150 attached device, a significant portion of the resistance is coming from the SATA interface.

Not really. Intel ICH5 can push somewhere around 135-140 Mb/s on SATA. Or at least 129 Mb/s on that Sandisk, which matches my result from USB 3.0 SATA dock station. But I'm starting to suspect that there was very specific reason why Intel never bothered with ATA133 on ICH5.

I think this may be of interest to IDE to SATA adapter users . Re: Small capacity SSD PATA/SATA benchmarks

I don't understand why you'd want to use a IDE to SATA converter.... A PCI SATA controller should work a treat on your system. Just make sure it's the type of controller that allows booting, because some (like my Promise IDE/SATA controller) don't... Generally VIA chip based SATA controllers allow booting from attached drives. It can still be a bit of a pain in the but to set up, but it works if done right. I'm currently running a 64GB Corsair SSD in my Tualatin build, using a no-name chinese VIA based PCI SATA controller I got off Aliexpress. Unlike my promise controller, the no-name one does not support raid (witch would be useless and would just slow down boot time) but it does support booting by setting the first boot device option in BIOS to "SCSI".

Socket3 wrote on 2020-09-28, 09:04:

I don't understand why you'd want to use a IDE to SATA converter.... A PCI SATA controller should work a treat on your system. Just make sure it's the type of controller that allows booting, because some (like my Promise IDE/SATA controller) don't... Generally VIA chip based SATA controllers allow booting from attached drives. It can still be a bit of a pain in the but to set up, but it works if done right. I'm currently running a 64GB Corsair SSD in my Tualatin build, using a no-name chinese VIA based PCI SATA controller I got off Aliexpress. Unlike my promise controller, the no-name one does not support raid (witch would be useless and would just slow down boot time) but it does support booting by setting the first boot device option in BIOS to "SCSI".

I agree, most IDE to SATA converters seem like temporary solutions, seem somewhat dangerous because of the bulky shape and exposed power pins, and are prone to sub-optimal ATA protocol negotiation on 1990's EIDE controllers. Not ideal for anything you expect to leave together for more than a couple days.

There are cases where I've used the converters though. I had some trouble getting promise IDE controllers to work reliably on my old PCI 2.0 486 motherboards. In those cases, I used a converter just for fun to see how a modern SSD would work in a completely anachronistic configuration. Wasn't a permanent build though. If you get a SATA-IDE converter, I like the bi-directional ones that come in plastic cases. They are worth the extra dollar, since they don't have the exposed pins and they work in either direction for those times you want to put an old CDrom on a SATA cable I guess.

You can tell which PCI IDE controllers are bootable because they have a rom on them. Example: Only one of these three VIA controllers is bootable.