Fully working and tested SSD solution

Started by t-rexky, February 29, 2012, 03:23:32 pm

Previous topic - Next topic

t-rexky

I spent some more time experimenting with adapters and different SSD combinations and I now have fully working and tested (for over a week) solution.  I have not benchmarked it, but my Turbo Color feels slightly "snappier" than with the already fast Seagate ST318418N.  My main motivation for this, however, was to eliminate the reliance on aging and slowly disappearing 50 pin SCSI drives.

I used an ACARD SCSI to SATA converter model ARS-2000SUP with firmware version 1.14P.  Do not upgrade the firmware to the latest 1.15P as it appears to have some compatibility issues that prevent the black hardware from booting. With 1.15P my Station reads in the boot block and then waits forever for the drive to become ready.  Firmware 1.14P on the other hand appears to work perfectly.  More information about the adapter can be found here: http://www.acard.com/english/.  I purchased it from Mars Technologies, since they had the best price: http://www.mars-tech.com/ars-2000su.htm.

For the SSD I settled on a 40GB Intel 320 series drive.  I initially tried the OCZ Vertex 2, but it had some issues on shutdown that made the fsck run on every restart.  The Intel 320 SSD has the current firmware version of 0362 and it works perfectly.  The absolute maximum physical size of a drive available to NS/OS is 28GB (seven 2GB partitions on NS or seven 4GB partitions on OS).  I therefore used the HDAT2 DOS utility on a PC to activate the HPA on the drive by limiting the usable size to 58720256 blocks.  This will substantially improve the life of the drive, as the HPA area is used by the drive for wear levelling.  In fact this is exactly what Intel recommends for server applications of the 320 series drives.  The HDAT2 utility can be obtained here: http://www.hdat2.com/.

The final step is creation of a disktab entry for the new setup.  Because we are using an SSD, the actual physical geometry of the media is pretty meaningless from the perspective of the host OS.  It undoubtedly can make a difference in terms of the SSD write amplification, but I have not analyzed this aspect.  I plan on testing the drive on a PC after several months of normal use to see how much life has been used.  Here is the /etc/disktab entry that I created for the setup:

#------------------------------------------------------------------------
# 40GB INTEL SSDSA2CT04 via ACARD ARS-2000SUP, 512-byte sectors
# use ACARD firmware 1.14P since 1.15P fails to boot the kernel (not ready)
# drive is secure erased and set to 58720256 sectors with HPA using HDAT2
#
#       /usr/etc/scsimodes tells us:
#               Drive type: ATA INTEL SSDSA2CT04
#               512 bytes per sector
#               255 sectors per track
#               64 tracks per cylinder
#               0 cylinder per volume (including spare cylinders)
#               32 spare sectors per track
#               0 alternate tracks per volume
#               58720255 usable sectors on volume
#
# (however, we will use fake geometry below since it is an SSD)
#
# this creates seven 2 GB partitions
#
ATA INTEL SSDSA2CT04|ATA INTEL SSDSA2CT04-512:\
       :ty=fixed_rw_scsi:nc#14336:nt#8:ns#512:ss#512:rm#10000:\
               :fp#320:bp#0:ng#0:gs#0:ga#0:ao#0:\
               :os=sdmach:z0#64:z1#192:hn=localhost:\
               :ro=a:rw=b:rw=c:rw=d:rw=e:rw=f:rw=g:\
       :pa#0:sa#4194304:ba#8192:fa#1024:ca#4:da#4096:ra#10:\
               :oa=time:ia:ta=4.3BSD:\
       :pb#4194304:sb#4194304:bb#8192:fb#1024:cb#4:db#4096:rb#10:\
               :ob=time:ib:tb=4.3BSD:\
       :pc#8388608:sc#4194304:bc#8192:fc#1024:cc#4:dc#4096:rc#10:\
               :oc=time:ic:tc=4.3BSD:\
       :pd#12582912:sd#4194304:bd#8192:fd#1024:cd#4:dd#4096:rd#10:\
               :od=time:id:td=4.3BSD:\
       :pe#16777216:se#4194304:be#8192:fe#1024:ce#4:de#4096:re#10:\
               :oe=time:ie:te=4.3BSD:\
       :pf#20971520:sf#4194304:bf#8192:ff#1024:cf#4:df#4096:rf#10:\
               :of=time:if:tf=4.3BSD:\
       :pg#25165824:sg#4194304:bg#8192:fg#1024:cg#4:dg#4096:rg#10:\
               :og=time:ig:tg=4.3BSD:
#------------------------------------------------------------------------


While exploring the setup I noticed that the ACARD adapter was somewhat picky about the termination.  I ultimately placed jumpers on both, the termination power pins and the termination enable pins of the adapter.  This works reliably in all scenarios including attaching external devices to the external SCSI connector.

I added some photos of the setup: http://www.flickr.com/photos/t-rexky/sets/72157629117909298/

t-rexky

gtnicol

Sweet! I think you can run 4GB partitions on OpenStep... I know I have my ST15150N drives configured that way (might be killing myself...). The only issue I've seen is that dump appears constrained to 2GB partitions.

t-rexky

Indeed!  The interesting thing is that restore works, just dump does not.  So you can clone a 2GB partition onto a 4GB partition, but not the other way around.  I set-up my i386 OS machine with 2GB system partition 'a' and all subsequent partitions set to 4GB.

I also tried porting a newer version of dump from BSD sources but gave up pretty early.  That was a very long time ago though, and I have learnt a lot since.  That would be a viable solution for all seven partitions set to 4GB.

As a side note, ACARD also make much faster LVD SCSI adapters such as ARS-2160 and ARS-2320, that could be used on i386 with the Adaptec U2W on OS42.

I am also currently using a really nifty external SCSI DVD-ROM that internally uses a TEAC DV-516GA IDE DVD drive with an ACARD 7720U.  It works perfectly and the whole thing is black!  Oh, and the black NeXT reads the DVD file system as well with no issues, after it throws up a console message about disc size...

Quote from: "gtnicol"Sweet! I think you can run 4GB partitions on OpenStep... I know I have my ST15150N drives configured that way (might be killing myself...). The only issue I've seen is that dump appears constrained to 2GB partitions.

t-rexky

The slab with the SSD storage has been up since I set it up.  I have done numerous compiles of the whole gcc-3.2.3 tree, which is pretty IO intensive.  Everything has been 100% solid, even with an external CD-ROM...  No errors or strange behaviour of any kind, so I give it both thumbs up.

mikeboss

hi everybody

since t-rexky's glowing review I also wanted an ACARD ARS-2000SUP. first I tried it in my Macintosh Quadra 700 as system disk for installing A/UX (the CF AztecMonster is incompatible with A/UX). the ARS-2000SUP worked like a charm. in my NeXTcube (turbo; N1000A) the CF AztecMonster works flawless since I got the right firmware. but the ARS-2000SUP does not work reliable in the NeXTcube! the system only sees the drive in about one of four power-on procedures. it seems to work in the NeXTstation though without any issues. will investigate further and let you guys know if I come to any helpful conclusion.

just beware if you're planning to mount the ACARD ARS-2000SUP into a cube!

@t-rexky: what revision is your PCB? mine's VER:R3.5
you'll find the revision right next to the 50pin SCSI connector,
bottom (no-components side) of the ARS-2000SUP. it's NOT neccessary to
take the drive apart.

regards,
michael


P.S.
yes, I've downgraded the firmware to 1.14P just as t-rexky suggested. it didn't work for me either with version 1.15P. also the jumpers are fine (termpower ON, terminator ON).
October 12, 1988 Computing Advances To The NeXT Level

t-rexky

@mikeboss:

Glad that you got the setup working in the station.  It is really strange that the setup does not work in your cube.  I first thought that it might be related to a different revision of the NCR SCSI controller, but you did mention the cube is a turbo, so it should be identical to a station...  Have you tried it with an SSD and with a mechanical drive as well?

I will have a look at the board revision the next time I have my station open.  At the moment it is all "pegged out" working on a new gcc port.

Regarding the firmware - I contacted acard and got a response indicating that they will pass my observations to engineering.  Not sure if they can really troubleshoot it without access to the black hardware, but perhaps it is something obvious in the code.  I would suggest that you contact them as well - perhaps that raises the priority a bit.

mikeboss

it's really behaving strange, until now I was not able to see a 100% reliable pattern. I tried three different SSDs (Crucial m4 64GB; Transcend 16GB, intel 80GB) and a real HDD from WD. also tried with the jumpers set to disable parity and delay start. right now the acard's in the turbo-station. I have to check again in the non-turbo station as well. mostly the computers refuse to see the acard after a few hours powered off. if I shut down and power on again (within minutes) it works most of the time. and *if* the machine's booting there's no trouble at all afterwards. I will have to investigate further, will maybe try with even older firmware revisions...

regards,
michael
October 12, 1988 Computing Advances To The NeXT Level

mikeboss

update: I bought a second ARS-2000SUP and this one works flawless in all of my NeXT systems.
October 12, 1988 Computing Advances To The NeXT Level

t-rexky

Hi Michael,

I finally was able to spend a little bit of time with my station after a long hiatus.  I can confirm that the PCB version of my ARS2000SUP is 3.5 as well.  If only one of your two cards is acting flaky, I wonder if it might perhaps be defective?

On a separate note, I just checked the Acard web site and it appears that they pulled firmware 1.15.  Version 1.14 is now the latest one available for download...


Quote from: "mikeboss"@t-rexky: what revision is your PCB? mine's VER:R3.5
you'll find the revision right next to the 50pin SCSI connector,
bottom (no-components side) of the ARS-2000SUP. it's NOT neccessary to
take the drive apart.

regards,
michael

mikeboss

thanks for checking the PCB revision of your acard adapter.

the hang on power-up occured at least once or twice with my second ARS2000SUP too... I didn't start my cube recently to investigate this any further.
October 12, 1988 Computing Advances To The NeXT Level

t-rexky

Resurrecting this old thread.  I cannot find my copy of the 1.14P firmware version anywhere, except it's loaded on my adapter.  Does anyone have a saved copy?

I am asking because I noticed that Acard now have 1.19P on their web site, but I will not try it unless I can revert back to the one that I know works reliably.

Many thanks!


t-rexky

Aaah, beautiful!  Thank you!  I have not even thought to look at the Wayback Machine  :oops: .  I contacted Acard a few days ago but have note hear anything back from them yet.  So this is perfect.

rooprob

t-rexky, are you back online with black hardware ?

I hope to be pest on some software compilation niggles. Your tenacity and skills are much needed!
:O2: r12 400 mapleleaf :Indigo2IMP: r10 195 IRIS :SlabMonoTurbo: NeXT
New Zealand

t-rexky

Hey.  Not at the moment, unfortunately.  I am finishing my home network, reworking my home theatre setup and doing some 3D printing at the moment.  All while balancing my time with my 4 year old son.  So at the moment I have zero time for all my vintage computers.  Once all the other projects are under control I will likely start tinkering again.  Sorry...  :?