[1613] in linux-scsi channel archive
Re: Re[2]: Which pci-scsi host adapter should I get?
daemon@ATHENA.MIT.EDU (art s. kagel IFMX x2697)
Tue Mar 25 18:39:09 1997
Date: Tue, 25 Mar 1997 18:01:12 -0500 (EST)
From: "art s. kagel IFMX x2697" <kagel@dg1.bloomberg.com>
To: Kimmy Posey <Kimmy_Posey@usa.racal.com>
Cc: linux-scsi@vger.rutgers.edu
In-Reply-To: <9702258593.AA859318829@usa.racal.com>
TMC-950 sounds right (it seems that the controller chip version was 1680
does that sound right? Of course I could just remember to look at home
before I come in tomorrow). BTW anyone need a cheap ISA SCSI controller?
Art S. Kagel, kagel@ts1.bloomberg.com
On Tue, 25 Mar 1997, Kimmy Posey wrote:
>
> The Future Domain controller you refer to sounds like a TMC-850 or TMC-950.
>
> KimmyD
>
> ______________________________ Reply Separator _________________________________
> Subject: Re: Which pci-scsi host adapter should I get?
> Author: "art s. kagel IFMX x2697" <kagel@dg1.bloomberg.com> at ftl03
> Date: 03/24/97 6:30 PM
>
>
>
> See my comments in context below:
>
> Art S. Kagel, kagel@ts1.bloomberg.com
>
> On Fri, 21 Mar 1997, Gerard Roudier wrote:
>
> >
> > On Fri, 21 Mar 1997, art s. kagel IFMX x2697 wrote:
> >
> > > Linux supports most major SCSI controllers SCSI-I -> SCSI-III Fast/Wide.
> > > I have had Future Domain (RIP), NCR, and BusLogic and I have found the
> > > BusLogic PCI SCSI-III Fast/Wide Bus Master controller to be extremely
> > > fast and reliable. (My FD was ISA bus and Very slow as expected. The
> > > NCR was PCI and apparently had a bad trace so I returned it and replaced
> > > it with the BusLogic.)
> >
> > If you just mean that oranges are not apples, I agree.
>
> No I just mean that I have had three different, Linux supported, SCSI
> cards and that I like the BusLogic best.
>
> >
> > > The driver support for the BusLogic is great with the BusLogic people
> > > directly supporting the maintainer with new products and specs. Also all
> > > BusLogic controllers are driver compatible so that the driver is simple
> > > and supports new controllers quickly. BusLogic even lists Linux support
> > > in their docs and Web pages.
> >
> > If you just mean that oranges mixed with apples have taste of oranges and
> > apples as well, I agree too.
> >
> > Can you please be more accurate?
> >
> > - Which Future domain board?
>
> I do not remember the model # but it was an ISA card with narrow SCSI-II
> support. It actually worked fine with both DOS and Linux but when I
> upgraded from my ISA 386-40 motherboard to a PCI Cyrix-P166 mother board
> I wanted a faster PCI card. The FD was slower than either the NCR or the
> BusLogic simply because it was constrained by the ISA bus.
>
> > - Which NCR board?
>
> It was a Tyan 1365 board with an NCR 825 SCSI chip and Bios. The card was
> PCI SCSI-II Fast/Wide. The Bus Logic is a BL-958 Fast-Wide Ultra
> SCSI-III PCI Bus Master card.
>
> > - Which drivers did you use in which version?
>
> The Future Domain was used with the last Pre 2.0 Linux release from WGS.
> I upgraded to Red Hat 4 when I brought up the new mother board and the
> NCR controller. This may have had some effect on comparing the FD to
> the others, though I suspect that the ISA bus has the greater effect.
> For the Bus Logic I have only used Red Hat 4 with the latest released
> drivers (the one before Flashpoint support was added, 2.0.7?).
>
> > - Which OS/es did you use?
>
> MS-DOS 6.21 and Red Hat Linux 4.
>
> > - In which other boards have been not fast and reliable enough for you?
>
> The Bus Logic has been rock solid. The only problem I have had is that
> one of my disks has > 1024 cylinders and the cylinder remapping that the
> Future Domain performed differed from that used by the Bus Logic and the
> NCR such that I could not install LILO with either until I had backed up
> the DOS partition and performed a low level format (with the Bus Logic) so
> that I could reinitialize the partition table. While the Bus Logic & NCR
> correctly interpreted the partitions and DOS worked fine, Linux fdisk
> complained about the partition table, mkfs hung (for over 24 hours before
> I rebooted), and LILO would not install even on the other drive (with
> <1024 cylinders) or a floppy until I had corrected the partition table to
> use the Bus Logic's cylinder remapping scheme. Since the reformat
> everything has been great.
>
> > - What mean 'fast' for you?
>
> I did some informal I/O benchmarks on the after noticing a subjective
> improvement going from the NCR to the BusLogic card. Anything I/O
> intensive seemed to run 15-40% faster. Again I had no formal timings
> from when the NCR had been installed so these are somewhat subjective but
> I know how long a long compile took and how much finger tapping I did, so...
>
> > - What means 'bad trace'? (I did not find this expression in my dictionnary)
>
> When the NCR card was installed and the machine was on for more than a
> few hours, the machine would hang on any command which accessed the
> drives. In other words I could run command.com or bash built in
> commands and even sometimes list directories that had been in cache
> already but when I tried to load an executable or list a directory not
> read before the machine would hang. In this case rebooting caused the
> machine to hang again in the middle of printing the NCR BIOS copyright
> header. 'Bad trace' means that apparently one of the printed circuit
> trace lines, or one of the solder connections was cracked or 'cold' (a
> term refering to a solder joint in which one of the parts being joined
> was not sufficiently heated to permit a good electrical connection,
> producing a joint whose electrical properties depend upon temperature).
> In other words the card stopped working when it became too hot (though
> the components are properly cooled).
>
> > - What means 'driver compatible' for you?
>
> Leonard Zubkoff, who maintains the drivers for BusLogic support in Linux
> reports, if I understand correctly, that all BusLogic controllers
> essentially use the same driver code. He only had to add code recently
> for the FlashPoint line which does not have an imbedded processor and
> depends on the driver for several functions that are on-board on the
> other BL controllers. The extra code you see below is for the
> FlashPoint line to replace the on-board functions with driver based
> functions. The FlashPoint series is much less expensive than the other BL
> controllers because of the lack of an on-board CPU and this accounts for
> the extra code. The basic BusLogic driver is slightly smaller than the
> aic7xxx driver and much smaller than the ncr53c8xx driver.
>
> >
> > Now, some informations for you.
> >
> > Linux support boards in Wide Ultra-2 mode (80 MB/seconds).
> >
> > And now, just for the fun:
> >
> > Here is the all Buslogic boards compatible driver:
> > (527967 bytes)
> >
> > -rw-r--r-- 1 root root 172611 Mar 21 22:21 BusLogic.c
> > -rw-r--r-- 1 root root 49686 Mar 21 22:21 BusLogic.h
> > -rw-r--r-- 1 root root 305670 Mar 21 22:21 FlashPoint.c
> >
> > Here is a shorter driver that is compatible with the whole AIC7XX chips
> > family and the 53C8XX chips family as well.
> > (449233 bytes)
> >
> > -rw-r--r-- 1 root root 174282 Oct 28 22:21 aic7xxx.c
> > -rw-r--r-- 1 root root 2305 Aug 10 1996 aic7xxx.h
> > -rw-r--r-- 1 root root 15494 Apr 20 1996 aic7xxx_asm.c
> > -rw-r--r-- 1 root root 9261 Oct 13 10:44 aic7xxx_proc.c
> > -rw-r--r-- 1 root root 21170 Oct 13 10:44 aic7xxx_reg.h
> > -rw-r--r-- 1 root root 196906 Jan 17 21:18 ncr53c8xx.c
> > -rw-r--r-- 1 root root 29815 Jan 17 21:18 ncr53c8xx.h
> >
> >
> > Gerard.
> >
>
>
>
