• Please review our updated Terms and Rules here

questions about SCSI hard drive size and speeds.

You potentially gave up half of your sequential performance (streaming reads and writes) when you put that adapter on.

And as Chuck points out, that's probably not the best VLB SCSI card out there. Adaptec and Buslogic made the best VL BUS SCSI cards at the time. SIIG was better known for parallel ports.
 
VL-bus wide-SCSI controllers are probably very thin on the ground. CSC had a wide-SCSI VL bus Fastcache controller, but those would be hard to find.

I suspect that most serious wide-SCSI users went EISA or PCI.
 
Thanks for all the good info guys. I'll keep an eye out for a wide bus vlb card as unlikley as finding one will be. Eisa isn't an option for me short of a motherboard replacment. I see a few buslogic cards for reasonable prices so I may try one of those.
 
In 486s the best disk performance is going to come from a proper DMA/bus mastering SCSI controller. I don't think any VLBus IDE controllers supported DMA transfers, just faster PIO modes. The advantage of DMA is faster transfer rates and lower CPU utilization for I/O. 50pin narrow (8bit) SCSI-2 devices support 10MB/sec operation (some rare UltraSCSI ones will do 20MB/sec). You generally won't hit near that speed on a 486 without DMA.

I am going to attempt to run benchmark comparisons of a CF card connected directly to an ISA IDE controller (don't have any VLBus ones hanging around) and connected to an Adaptec AHA-2740W EISA card's narrow bus. The CF card will be adapted to SCSI using an Adcom CF to SCSI bridge board. It is powered by a 53CF96 SCSI controller, so it should be able to do 20MB/sec UltraSCSI transfers in theory. I'm mainly doing this to check the overhead of this adapter.
 
The buslogic card I'm looking at has DMA/bus mastering, and 10mb transfer rates, I may pick it up this Friday and see if its any faster then this SiiG card. I am intrested in what that adaptec can do. A shame mines scrap.
 
The caching controllers got around DMA by using their own CPU to do the work instead of the 486 (286-20's were common on the boards).

No idea what theoretical bandwidth is for VLB but it is shared with any other VLB cards (video mostly) on the BUS and CPU to cache access so I don't think you are going to be able to get the best of an UWSCSI HD even if you had the correct port.
 
I dont expect to be able to use the full potential of a newer drive. I'm just running on the assumption that a new drive even being gimped is still going to be more reliable and faster then a drive of the period.
 
Also, if not mentioned, a controller capable of doing synchronous transfers with a drive similarly equipped can cut out quite a bit of handshaking overhead.
 
FWIW, my Adaptec AHA-2740W card is just as picky as its VLBus cousin. I couldn't get my SCSI-to-CF adapter working on it as the card refused to detect it. It worked fine with a PCI based AHA-2920C. The EISA and VLBus Adaptec cards are based on the same AIC-77xx series host controllers.
 
well guys, my Buslogic BT-445S card came in today which supposedly has all the bells and whistles, async transfer, DMA, all that proper SCSI stuff. hooked it up to the same drive and....well i dunno. The numbers seems a bit better then the SiiG controller but the IDE drive still seems to be far faster in some areas. speedsys also scores it way lower a 95 where on the SiiG it scored a 410 i think though i'm starting to think that score is actually meaningless...i dunno, it seems to compare it to different drives now and i don't quite understand how it judges.

also this new controller only detects 1GB of my 18GB SCSI drive where the SiiG detected 1.4GB. I have no idea why this is, I assumed they would both detect 2GB at least

i'll re post my screenshots and the new ones and you guys can see for yourselves and let me know if it looks correct.

my IDE drive on the motherboard IDE controller.
vpai.jpg


My SCSI HDD on the SiiG SCSI VLB card
ch8w.jpg


My SCSI HDD on my BusLogic BT-445S VLB card
o7bi.jpg


My SCSI HDD on my BusLogic BT-445S VLB card after I enables async transfer and increased the Mbyte transfer rate from 5/MB to 10/MB via switches on card (seems like a very slight boost in a few areas)
lk4r.JPG


the biggest change I see is the Linear Verify speed is way better on the BusLogic then the IDE and even the SiiG controller.....though to be honest I have no idea what this means
 
Last edited:
The Adaptec cards have large drive translation, so others may too. If the option "Use Translation for drives larger than 1GB" is enabled, it changes the CHS values of the drive. When enabled, it allows 255 heads and 63 sectors per track for drives larger than 1GB. The SiiG card appears to be using this, while the Bus Logic isn't. Looking at the manual, the Bus Logic card uses a dip switch to enable that translation.

http://ftp.nchu.edu.tw/Hardware/mylex/manuals/bt445s.pdf (see page 2-5)
 
I ran the full test on my Cx486-DX2 system. Systest gives the CPU speed as 22MHz, which is bizarre.

At any rate for my AHA2840VL with a Seagate 4GB Hawk, running narrow SCSI, here are the numbers:

Random Access Time: 16.04ms
Buffered Read Speed: 5177KB/sec.
Linear Verify Speed: 1681KB/sec.
Linear Read Speed: 4478KB.sec.

If any of that means anything to you.

(TopBench gives the system a score of 200)

Here's a comparison. A 386/20, 13MB RAM and a DTC3280 with a Seagate ST410800 9GB drive.

Random Access Time: 19.99ms
Buffered Read Speed: 1892KB/sec
Linear Verify Speed: 11856KB/sec.
Linear Read Speed: 1964KB/sec.

(Topbench gives the system a score of 48 )

Have you considered running CORETEST to benchmark your disk?
 
Last edited:
For the sake of interest, attached are the Speedsys results from an over optimized 486. It uses an Adaptec 2940U2W on a PCI bus with a U160 or U320 HDD, I forget which now. The CPU is an IBM 5x86-133 (2x66). FSB at 33 MHz.

It seems this forum really reduces the quality of form-attached png and gif images, here are the numbers,

Average/Max seek time: 7.94 / 13.18 ms
Random seek time: 6.97 ms
Track-to-track seek: 0.79 ms
Random access time: 8.42 ms
Buffered read speed: 33137 KB/s
Liner verify speed: 4705 KB/s
Linear read speed: 30292 KB/s

[EDIT: Image re-attached as a jpeg seems to have the best results]
IBM-5x86C-133.jpg
 
Hi, I don't know if your still having trouble with this. I once read somewhere in the PC guide, i think in the section on RAID, that if you put an Ultra2 drive on a UltraWide bus, the *internal* data transfer rate halves as well. I may have misremembered this, or the source. It sounded like faulty reasoning to me at the time but if it is true, (At least on some drives) it could be contributing to your problem. If it wasn't the PC Guide it was another very good source of info, but then there can always be mistakes.

Say you have an Ultra Wide drive, 40MB/s, with an internal transfer rate of 12MB/s If you put it on a fast-wide controller at 20MB/s you'd get 6MB/s, according to the logic above. If you then put it on a fast narrow channel at 10MB/s, it would half the speed again to 3MB/s.

I think its a bit odd to design a drive like this, but it could be designed as such for simplicity. I know of at least one drive that had a jumper to set it to UltraSCSI from Ultra2. Perhaps setting it to UltraSCSI set some divider with respect to the internal transfer rate. "internal transfer rate" is a murky term, as we are dealing with a complex drive, so don't flame me.

Have you also checked termination? I had some IBM Ultra2 drives that don't like anything other that termination fit for ultra2 or better. If you put them on an UltraWide channel they worked poorly, strange lockups in bios scan etc. This could have been due to something else, but they worked fine with an ultra 2 controller and cable with terminator. If you factor in the narrow connection as well, this complicates things further. If you have:

Terminated Controller---------Adapter & 68pin drive--------50pin terminator
Where --- is 50pin cable.
it should work, but as i understand it violates termination rules. The bus is terminated but the drive is only half terminated.

The correct way is:

68pin terminator=======50pin adapter & unterminated Controller======68pin drive===68pin terminator
where === is 68pin cable.
I think the drive also has to provide the termination power.

Check if there are any other jumpers on the drive to set sync speed, sometime there are some itty bitty ones on the bottom.

One long shot is also checking that the card is set at the correct speed for the VLB. Over or under clocked, the behavior is hard to determine. Performing a "Low level format" on the drive, when the bios cant determine its correct size is not something I would do. If the bios is telling the drive to do an actual low level format i don't think its going to make any difference, and if the bios can somehow do the format itself i think this is going to cause more problems.

There is a contemporary review somewhere in InfoWorld, about about VLB SCSI cards and they were not blown away
with the performance compared to EISA. Bios updates may have fixed this.
 
Hi, I thought i replied to this thread a few days ago but it didn't show up.

I don't know if your still having trouble with this, but I once read somewhere in the PC guide (i think in the section on RAID), that if you put an Ultra2 drive on a UltraWide bus, the *internal* data transfer rate halves as well. I may have mus-remembered this, or the source. It sounded like faulty reasoning to me at the time. If it is true, (At least on some drives) it could be contributing to your problem. If it wasn't the PC Guide it was another very good source of info, but then there can always be mistakes.

Say you have an Ultra Wide drive, 40MB/s, with an internal transfer rate of 12MB/s If you put it on a fast-wide controller at 20MB/s you'd get 6MB/s, according to the logic above. If you then put it on a fast narrow channel at 10MB/s, it would half the speed again to 3MB/s.

I think its a bit odd to design a drive like this, but it could be designed as such for simplicity. I know of at least one drive that had a jumper to set it to UltraSCSI from Ultra2. Perhaps setting it to UltraSCSI set some divider with respect to the internal transfer rate, as well as putting it into SE mode. "internal transfer rate" is a murky term, as we are dealing with a complex drive, so don't flame me.

Have you also checked termination? I had some IBM Ultra2 drives that don't like anything other that termination fit for ultra2 or better. If you put them on an UltraWide channel they worked poorly, strange lockups in bios scan etc. This could have been due to something else, but they worked fine with an ultra 2 controller and cable with terminator. If you factor in the narrow connection as well, this complicates things further. If you have:

Terminated Controller---------Adapter & 68pin drive--------50pin terminator
Where --- is 50pin cable.
it should work, but as i understand it violates termination rules. The bus is terminated but the drive is only half terminated.

The correct way is:

68pin terminator=======50pin adapter & unterminated Controller======68pin drive===68pin terminator
where === is 68pin cable.
I think the drive also has to provide the termination power.

Check if there are any other jumpers on the drive to set sync speed, sometimes there are some itty bitty ones on the bottom.

One long shot is also checking that the card is set at the correct speed for the VLB. Over or under clocked, the behavior is hard to determine. Performing a "Low level format" on the drive, when the bios cant determine its correct size is not something I would do. If the bios is telling the drive to do an actual low level format i don't think its going to make any difference, and if the bios can somehow do the format itself i think this is going to cause more problems.

There is a contemporary review somewhere in InfoWorld, about about VLB SCSI cards and they were not blown away
with the performance compared to EISA. Bios updates may have fixed this.
 
You must log in or register to reply here.
Back
Top