SCSI2SD v6 and large drives

[AC/DC HACKER!]

Originally Posted by Futaura

Haven't found a way to use synchronous mode reliably and it looks like a problem with SCSI2SD v6, as synchronous mode works fine with my SCSI HDD.

I haven't attempted, yet, with my other SCSI boards. Have you commented to the maker of SCS2SD? I haven't. I've been really focused on OS 3.1.4 compatibility with games and Async. Compact Flash is becoming more pricey, and more rare from my searches..so I've been checking my SD's a bit with SCSI2SD V6. I really dig the speeds vs V5. Your WarpOS DT's really seem faster with V6, I know they aren't but showing friends, and teens I know, really gets replies and conversations. "This was before Windows became a 'thing' it is now?? Wow..!! That's pretty fast." I still adore, big time, discussing the Pioneer of Computing. It's fun to help them explore it.

[Futaura]

I haven't contacted him, but will be doing so. I'm trying to figure out if there is a good test case or even some debug output possible. I know it will not be much help if I simply tell him enabling synchronous mode causes write corruption.

Is termination enabled on your board? I disabled it on mine as I have a CD-ROM drive on the end of my SCSI chain with the termination jumper on. Actually, those are the only 2 devices normally connected, unless I've connected my old HDD via mobile rack.

IIRC, normally Phase 5 SCSI throws a wobbly (and debug output) if parity checks fail, but that is not happening with this issue. The first time I noticed the issue was some corruption in the dir listing, when I was comparing the original drive content (lots of softlinks that I needed to remake), which was when I then tried diskvalid. I think I might try with FFS also, as that may trigger error requesters when the files are written (unlike PFS3). The other possibility is that the correct data is received by SCSI2SD v6, but it is mangling it when writing to the SD.

I'm using a legit SanDisk Extreme Pro 64Mb (Amazon black Friday deal ), so that should work fine.

[Futaura]

Quick update from me - Sync 5Mb/s mode certainly seems more reliable, but I'm sure I got errors with it yesterday. My old HDD works fine in 10Mb/s mode and I get around 8Mb/s reads. However, I notice in unitcontrol that my HDD has FWC Mode On, whereas the SCSI2SD has FWC Mode Off (cannot be turned on). Obviously, the SCSI2SD doesn't really need a fast write cache, but I do wonder if a device says it has one, then the Phase 5 SCSI controller performs writes differently somehow (AFAIK, it would be up to the SCSI device to handle the cache itself anyway).

I could not get any debug output to show up, but I did try FFS instead of PFS3, and corruption occurs there too. Obviously in a different way due the structure used by the FS. With FFS, files themselves tend to be corrupt, which can be tested with "DirDiff 2 -b", for example. With PFS3, sometimes files are corrupt, but since it stores a directory tree (unlike FFS), if you get corruption there, you tend to get weirdness in directory listings which ultimately messes up the entire partition. PFS3 also sometimes gives errors about reads outside the partition because of this.

Anyway, I've mailed the SCSI2SD maker to see what he says. Is there anybody we can ask about the SCSI controller on the B/CSPPC? I would love to know if it itself handles writes differently if the device supports FWC mode or not.

[AC/DC HACKER!]

Termination is active. The V6 is all I have on my CSPPC SCSI. I decided, before I read these replies, to check Sync 5MB again. Using my current SD config with OS3141 updated from OS 3.5, and I use WOS where possible. Was recognized by A4000 quicker than with ASync. My CF is hanging off its reader, I wanted it disconnected for clarity. I did enable SCSI-2 just for fun...before I stated this little test.

SysInfo v4.x is showing about 3.3MB. Payback PPC, Shogo, and Wipeout 2097 all load quicker.

I made an new test partition for Sync, and Cloned files to it from my GB SYS:. That was pretty quick. I used DiskDoctor to check it. All is good.

I powered off, changed to Sync 10MB/s, and repeating..

I'm using PNY Elite 64GB... Which is getting different results than my other posts above. Hmm...

[AC/DC HACKER!]

Something is funky. I just noticed a problem. The Copy I made of my CF to SD SYS..has problems. ASync or Sync, whatever selection..Uropa 2's intro is having graphical problems via ADOS. The copy of partition was done with:

Copy OS314: SD_314: Clone All Quiet

So...some checking to do. But also some family stuff. I'll report back when able.

[Futaura]

I would suggest also using http://aminet.net/package/disk/bakup/DirDiff to check integrity of the files: "DirDiff 2 -b OS314: SD_314:" with async and sync. You can use the P5 unitcontrol command to switch async/sync without needing a reboot.

If you make the copy in async mode, I'm guessing if you then switch back to sync 10mb mode, all will be ok for reads.

[Futaura]

You are using v6.2.7 or newer, right, if I read previous comments correctly?

Apparently, I've heard that v6.2.5 is ok, but data corruption occurs with newer versions. I haven't checked this myself yet, but hope to do so tomorrow night.

[Ferry]

Originally Posted by Futaura

Quick update from me - Sync 5Mb/s mode certainly seems more reliable, but I'm sure I got errors with it yesterday. My old HDD works fine in 10Mb/s mode and I get around 8Mb/s reads. However, I notice in unitcontrol that my HDD has FWC Mode On, whereas the SCSI2SD has FWC Mode Off (cannot be turned on). Obviously, the SCSI2SD doesn't really need a fast write cache, but I do wonder if a device says it has one, then the Phase 5 SCSI controller performs writes differently somehow (AFAIK, it would be up to the SCSI device to handle the cache itself anyway).

...I did try FFS instead of PFS3, and corruption occurs there too.

I found this thread after having problems with my just received SCSI2SD. It’s attached to a Blizzard 1260+SCSI kit. Synchro mode doesn’t work, no matter if you copy to a FFS or PFS partition, the led of the unit stays on and transfer stops, computer must be turned off and on again; FWC cannot be turned on; also, it doesn’t matter if the unit is powered by the SCSI cable or by an external source.


Btw, is it true that an older firmware was better regarding data corruption? I just noticed there’s a new version of the firmware published just a week ago:


http://www.codesrc.com/files/scsi2sd-v6/v6.2.9/


Will try tomorrow…


Saluditos,


Ferrán.

[AC/DC HACKER!]

For some reason if I copy some stuff with DOpus 5.81 or 5.91 I'm noticing some flaws with some stuff. Like Uropa 2's Intro. If I copy the directory via Workbench drag'n'drop to a SCSI2SD V6 PFS3 partition the Intro plays well from my personal boot menu. I'm going to do some more checking on this soon. I'll check with FFS. It doesn't matter if I'm using OS 3.1.4.1 or OS3.5.

I was going to look into DirDiff, and the UnitControl I've left the SCSI menu at Auto. I was switching it via USB. I just downgraded to 6.2.5 and it seems to be doing well with Sync 10MB/s. SysInfo 4 shows 3.27MB/s. So, still not the speed I want, but is a little faster than my IDE CF adapter. More checking to do.

[pipper]

I am using SCSI2SD on a Blizzard 1230IV and 1260. I have to limit the transfer rate to 5mb/s, synchronous; otherwise SysTest will lock up when performing disk access benchmarks.

[Ferry]

Originally Posted by pipper

I am using SCSI2SD on a Blizzard 1230IV and 1260. I have to limit the transfer rate to 5mb/s, synchronous; otherwise SysTest will lock up when performing disk access benchmarks.

Are you getting 5MB/s? I’m getting 3MB/s with async, not bad because it’s 3x of what I’m getting from my SD2IDE, but 5x would be even better

Also, are you getting consistent data integrity? Of course, that’s more important that all the speed one can get. You know, “power without control…”

Saluditos,

Ferrán.

[pipper]

Originally Posted by Ferry

Are you getting 5MB/s? I’m getting 3MB/s with async, not bad because it’s 3x of what I’m getting from my SD2IDE, but 5x would be even better

Also, are you getting consistent data integrity? Of course, that’s more important that all the speed one can get. You know, “power without control…”

Saluditos,

Ferrán.

SysInfo Attests around 4mb/s with the 5mb/s synchronous setting. That setup seems to be stable.
When setting the SCSI2SD to 10mb/s synchronous, SysInfo Reports up to 6mb/s, but SysTest will lock up during the disk benchmarks.

[AC/DC HACKER!]

Originally Posted by Futaura

You are using v6.2.7 or newer, right, if I read previous comments correctly?

Apparently, I've heard that v6.2.5 is ok, but data corruption occurs with newer versions. I haven't checked this myself yet, but hope to do so tomorrow night.

Since the degrade to v6.2.5 it's doing well. Copies from Directory Opus 5.91 are also in tact. I'm running programs that were visibly showing problems.

[Futaura]

6.2.5 seems ok to me too. 6.2.9 definitely still has the data corruption.

I prefer to use http://aminet.net/package/disk/misc/DTA100 to measure real read speeds on a 10Mb or so file. On an A1200 built-in IDE is around 2mb/s, SCSI2SD async around 2.5mb/s and sync 6-8mb/s.

I've been in touch with the SCSI2SD maker - Michael is currently investigating the issue. I've been trying out some test firmwares for him, so hopefully he can find and fix the bug introduced after 6.2.5.

Another symptom I found is when using the BUF parameter of the copy command. It defaults to 64K, but if I increase to 128K or 256K, the SCSI bus dies and the BPPC SCSI controller throws up an unexpected disconnect message, requiring a power off/on. Indeed, any program that tries to write more than 64K in one go would either trigger this or increase data corruption.

I think CSPPC users need to take extra care regarding termination. With the CSPPC having a 68pin connection, you probably should not use the active termination on SCSI2SD. Instead you need to place active terminators at the start and end of the chain (as the CSPPC does not self terminate host end). For BPPC, it is a little easier since it is 50pin and AFAIK the BPPC does self terminate host end, so you just need to terminate the device at the end of the chain, or place a separate terminator there. Active terminators only for both CSPPC and BPPC, of course.

[Futaura]

Originally Posted by AC/DC HACKER!

I was going to look into DirDiff, and the UnitControl I've left the SCSI menu at Auto. I was switching it via USB. I just downgraded to 6.2.5 and it seems to be doing well with Sync 10MB/s. SysInfo 4 shows 3.27MB/s. So, still not the speed I want, but is a little faster than my IDE CF adapter. More checking to do.

If you have the SCSI menu set to Auto, Async/Sync is selected by the Sync flag in the RDB of the drive. Not all tools give the option of changing it and it will default to Async. So, are you sure you are not using Async mode? You should get much faster speeds than that, even with 6.2.5, when using Sync mode.

Unitcontrol mimics the SCSI boot menu, but any changes you make are temporary. After a reboot, any changes you made revert to the saved SCSI menu settings.

[Ferry]

Originally Posted by Futaura

If you have the SCSI menu set to Auto, Async/Sync is selected by the Sync flag in the RDB of the drive. Not all tools give the option of changing it and it will default to Async. So, are you sure you are not using Async mode? You should get much faster speeds than that, even with 6.2.5, when using Sync mode.

Unitcontrol mimics the SCSI boot menu, but any changes you make are temporary. After a reboot, any changes you made revert to the saved SCSI menu settings.

Best thing to do is to use SCSIConfig, set the drive to Syncron and save, that way changes will be stay even after a reboot, because the setting is written in the RDB.

Saluditos,

Ferrán.

[Ferry]

Originally Posted by Futaura

6.2.5 seems ok to me too. 6.2.9 definitely still has the data corruption.

Good news! Rolling back to it right now, thanks for the info.

I've been in touch with the SCSI2SD maker - Michael is currently investigating the issue. I've been trying out some test firmwares for him, so hopefully he can find and fix the bug introduced after 6.2.5.

If I can be of any help, count on me and my Blizzard SCSI kit.

Saluditos,

Ferrán.

[Futaura]

Originally Posted by Ferry

Best thing to do is to use SCSIConfig, set the drive to Syncron and save, that way changes will be stay even after a reboot, because the setting is written in the RDB.

Indeed, although do take extreme care if you used another program to set the drive up. I used HDToolbox from OS3.9, which has no option to set the synchronous flag, but SCSIConfig then had problems reading the RDB correctly. In the end, I used http://aminet.net/package/disk/misc/RDBFlags-1.3 to switch synchronous on/off which didn't mangle my RDB.

It has been 10+ years since I messed with the drives in my A1200 and I had so many different tools installed that I couldn't remember which I should use .

HDToolbox (OS3.1)
HDToolbox (OS3.9)
HDInstTools
SCSIConfig
Media Toolbox

I figured out from my old drives that I must have used HDToolbox from OS3.9, especially as both drives were larger than 4 Gig. The OS3.1 version obviously should be avoided and I couldn't figure out how to work HDInstTools - it didn't seem it wanted to read the current RDB and just wanted to start from scratch. Media Toolbox is I think from early OS4 betas and that can switch synchronous on, but I soon discovered it wrote corrupt RDBs!

[Ferry]

Originally Posted by Futaura

Indeed, although do take extreme care if you used another program to set the drive up. I used HDToolbox from OS3.9, which has no option to set the synchronous flag, but SCSIConfig then had problems reading the RDB correctly. In the end, I used http://aminet.net/package/disk/misc/RDBFlags-1.3 to switch synchronous on/off which didn't mangle my RDB.

I think you can safely use SCSIConfig if you use it to change only that flag, and then save RDB. Of course, much better to use a modern program to partition drive, IIRC SCSIConfig doesn’t work well with HDs higher than 4GB.



Oh, and thanks for the hint about the program.


Saluditos,


Ferrán.

[AC/DC HACKER!]

Originally Posted by Futaura

Indeed, although do take extreme care if you used another program to set the drive up. I used HDToolbox from OS3.9, which has no option to set the synchronous flag, but SCSIConfig then had problems reading the RDB correctly. In the end, I used http://aminet.net/package/disk/misc/RDBFlags-1.3 to switch synchronous on/off which didn't mangle my RDB.

It has been 10+ years since I messed with the drives in my A1200 and I had so many different tools installed that I couldn't remember which I should use .

HDToolbox (OS3.1)
HDToolbox (OS3.9)

HDInstTools - Yup...I won't run this anymore.

SCSIConfig
Media Toolbox

I figured out from my old drives that I must have used HDToolbox from OS3.9, especially as both drives were larger than 4 Gig. The OS3.1 version obviously should be avoided and I couldn't figure out how to work HDInstTools - it didn't seem it wanted to read the current RDB and just wanted to start from scratch. Media Toolbox is I think from early OS4 betas and that can switch synchronous on, but I soon discovered it wrote corrupt RDBs!

Something I just considered. I was using OS 4's Partition program setup SD's.. I think I saw Sync there. I'll check into that. It's more "Modern".

Originally Posted by Futaura

6.2.5 seems ok to me too. 6.2.9 definitely still has the data corruption.

I prefer to use http://aminet.net/package/disk/misc/DTA100 to measure real read speeds on a 10Mb or so file. On an A1200 built-in IDE is around 2mb/s, SCSI2SD async around 2.5mb/s and sync 6-8mb/s.

I've been in touch with the SCSI2SD maker - Michael is currently investigating the issue. I've been trying out some test firmwares for him, so hopefully he can find and fix the bug introduced after 6.2.5.

Another symptom I found is when using the BUF parameter of the copy command. It defaults to 64K, but if I increase to 128K or 256K, the SCSI bus dies and the BPPC SCSI controller throws up an unexpected disconnect message, requiring a power off/on. Indeed, any program that tries to write more than 64K in one go would either trigger this or increase data corruption.

I think CSPPC users need to take extra care regarding termination. With the CSPPC having a 68pin connection, you probably should not use the active termination on SCSI2SD. Instead you need to place active terminators at the start and end of the chain (as the CSPPC does not self terminate host end). For BPPC, it is a little easier since it is 50pin and AFAIK the BPPC does self terminate host end, so you just need to terminate the device at the end of the chain, or place a separate terminator there. Active terminators only for both CSPPC and BPPC, of course.

It has been many years since I've played with the termination. I'll check some physical terminators I have. Update: I haven't noticed any differences between my terminator and V6's. Glad to read about the contact.

Originally Posted by Futaura

If you have the SCSI menu set to Auto, Async/Sync is selected by the Sync flag in the RDB of the drive. Not all tools give the option of changing it and it will default to Async. So, are you sure you are not using Async mode? You should get much faster speeds than that, even with 6.2.5, when using Sync mode.

It has been a long time since I've looked for that in a HD prep program. You're correct, most don't support it. I located HDInstTools and it does, but it fried my SD RDB, and I've been using PFSDoctor and PFSDoctorBeta4GBPlus to repair it. You just have to make sure PFS-3 isn't "holding onto a partition" or either version of PFSDoctor hangs. I've been using WinUAE to do it, with real A4000 I'm having challenges repairing it. Another of many Thumbs up to WinUAE.

Unitcontrol mimics the SCSI boot menu, but any changes you make are temporary. After a reboot, any changes you made revert to the saved SCSI menu settings.

Stay away from MyUnitControl..I figured I'd test it and it's what zapped my SD. Everything is a backup, so not a worry..but note to others.

I've been playing with Sync and ASync with the V6 board and CSPPC's Boot Menu, Sync 10MB/s gets me a 7MB/s transfer but then the Amiga locks. Synch 5 MB/s seems stable. So, perhaps Michael will find a further solution. Still checking combos. Soon I'll check the Unit Control from Phase 5.