VOGONS


Reply 20 of 32, by Marco

User metadata
Rank Member
Rank
Member

Hi, thanks.

- It is only a Two-ended Cable (1 Connector to Adapter, Other remaining to the HDD. And the HDD is terminated)
- I just enabled SCAMP and Plug&Play ID'ing and the error now went away.

So it should be ok 😀

Anyway I will get a "standard" 4-connector cable including terminator. I will test this as well while deactivating HDD onboard termination.

Last edited by Marco on 2023-01-25, 10:30. Edited 1 time in total.

1) VLSI SCAMP 311 / 386SX25@29 / 16MB / CL-GD5434 / CT2830/ SCC-1&MT32 / Fast-SCSI AHA 1542CF + BlueSCSI v2
2) SIS486 / 486DX/2 66(@80) / 32MB / TGUI9440 / LAPC-I

Reply 21 of 32, by Marco

User metadata
Rank Member
Rank
Member

My performance results so far:

Procom Disk Bench Random Xfer Rate:
IDE160GB@8GB HDD: 2749kb/s
SCSI HDD: 2247kb/s

Sisoft Sandra Drive Index:
IDE160GB@8GB HDD: 1548
SCSI HDD: 1444

Real World Copy 100 Files:
IDE-to-IDE: 1:13
SCSI-to-SCSI: 1:19

Real World copy 100 Files MSDOS w/o cache:
IDE: 0:36
SCSI: 0:57

Time-till-Winword-starts-while-copy-100-files:
IDE: 45sec
SCSI: 39sec

Copy 100 files while Turbo95 bench running:
IDE: 38sec
SCSI: 32sec

Attention: Especially the real world Win95 benchmarks are hard to trust, since on the 386SX running multi-tasked applications (1x app + 1x copyjob) is not reliable. As e.g. the app is put on hold / hanging or something.

Questions:
1. Do you have any idea to better bench further? Apps like WInbench etc. wont work as they mostly require something more on my pure basic Win95 setup
2. Are the results suprising to you?

Last edited by Marco on 2023-01-25, 11:19. Edited 1 time in total.

1) VLSI SCAMP 311 / 386SX25@29 / 16MB / CL-GD5434 / CT2830/ SCC-1&MT32 / Fast-SCSI AHA 1542CF + BlueSCSI v2
2) SIS486 / 486DX/2 66(@80) / 32MB / TGUI9440 / LAPC-I

Reply 22 of 32, by Disruptor

User metadata
Rank Oldbie
Rank
Oldbie

Ah, well, I even can burn CDs with 2x on my 386SX/20 SCSI.
On a faster 386SX (40 MHz + cache) I'd try 4x too, since I was not far away from that border.

Reply 23 of 32, by Marco

User metadata
Rank Member
Rank
Member

Ok so get a stable and continious 300Kb/s burst rate including keeping buffers filled.
Thats ok I would say. But not quite sure if I'd say thats amazing. Ahhhh sorry. 😀

1) VLSI SCAMP 311 / 386SX25@29 / 16MB / CL-GD5434 / CT2830/ SCC-1&MT32 / Fast-SCSI AHA 1542CF + BlueSCSI v2
2) SIS486 / 486DX/2 66(@80) / 32MB / TGUI9440 / LAPC-I

Reply 25 of 32, by Marco

User metadata
Rank Member
Rank
Member

One thing is here definitely my ISA/DMA Bus Speed. If I lower that to normal I drop down to below 2mb/s for diskbench whereby scsi performance remains.

Interesting but in my „overclock“ setup the IDE performance with faster IDE HDD seem to be just the better option. And I don’t even talk about noise 😀 the IDE just uses 12MHz whereby SCSI sticks at 5,7MHz and won’t go higher. No one could really explain why that is btw (to me) 😀

PS: anyway it still can be that interacting in windows while doing file transfers feels faster. But It can simply be imagination.

1) VLSI SCAMP 311 / 386SX25@29 / 16MB / CL-GD5434 / CT2830/ SCC-1&MT32 / Fast-SCSI AHA 1542CF + BlueSCSI v2
2) SIS486 / 486DX/2 66(@80) / 32MB / TGUI9440 / LAPC-I

Reply 27 of 32, by Marco

User metadata
Rank Member
Rank
Member

Always with aspi4dos.
Pls also remind that access time of the 160gb ide might be (much?) better. And its internal cache is larger.

Pls also don’t compare dos copy with windows copy as the set of files was different

Update:
Just want to inform all that I also tested with:

- the new scsi cable: same results
- shadowed scsi rom: same results
- ensuring sync neg and fast scsi: was done already
- disabling scsi parity: same results.

1) VLSI SCAMP 311 / 386SX25@29 / 16MB / CL-GD5434 / CT2830/ SCC-1&MT32 / Fast-SCSI AHA 1542CF + BlueSCSI v2
2) SIS486 / 486DX/2 66(@80) / 32MB / TGUI9440 / LAPC-I

Reply 28 of 32, by Sphere478

User metadata
Rank l33t++
Rank
l33t++

How many ns do you get in speedsys?

2mb/sec seems
About right. These aren’t going to be speed demons. Seek time is where you can pick up some performance maybe. I get 8ns on my 140gb drive.

This might be about as fast as you can get on isa. Don’t forget. You are doing this without loading the cpu much. So don’t undercut your success here. 😀

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 29 of 32, by Marco

User metadata
Rank Member
Rank
Member

Benches added here
Faster HDD= IDE
Slower HDD= SCSI. Seems all limited at 3mb/s

Attachments

1) VLSI SCAMP 311 / 386SX25@29 / 16MB / CL-GD5434 / CT2830/ SCC-1&MT32 / Fast-SCSI AHA 1542CF + BlueSCSI v2
2) SIS486 / 486DX/2 66(@80) / 32MB / TGUI9440 / LAPC-I

Reply 30 of 32, by maxtherabbit

User metadata
Rank l33t
Rank
l33t

The "DMA transfer speed" setting on the 154x series cards makes all the difference. Try setting it higher if you can, but you must test to make sure it doesn't crash or corrupt data

Reply 31 of 32, by Marco

User metadata
Rank Member
Rank
Member

Hi,
Sure thanks. Already tried out. Cannot go higher than 5,7! All self Test fail and real-time boot as well. Ideas on how to make it possible?

1) VLSI SCAMP 311 / 386SX25@29 / 16MB / CL-GD5434 / CT2830/ SCC-1&MT32 / Fast-SCSI AHA 1542CF + BlueSCSI v2
2) SIS486 / 486DX/2 66(@80) / 32MB / TGUI9440 / LAPC-I

Reply 32 of 32, by maxtherabbit

User metadata
Rank l33t
Rank
l33t
Marco wrote on 2023-01-26, 22:21:

Hi,
Sure thanks. Already tried out. Cannot go higher than 5,7! All self Test fail and real-time boot as well. Ideas on how to make it possible?

Short of "use a different motherboard" not really

5.7 is pretty poor, seems like 154x cards and that motherboard are just a poor match