feipoa wrote:

Those sound like motherboard BIOS settings, not SCSI BIOS settings.

Yes, they are. Must have been tired when I read your post. Didn't change any SCSI BIOS settings.

How much did the score for the Samsung IDE drive on AVA-2825 EIDE VLB card increase with these new motherboard BIOS settings?

You can use a LBA bios extender which sits in the MBR of the hard drive to gain access to the space normally out reach on larger disks. Though I have no idea what might happen if you combine it with a disk cache like this.

Here is something I did a number of years ago:
Benchmarked some SSDs in old laptops

Out of curiosity, did you try benching against software-based caches?

feipoa wrote:

How much did the score for the Samsung IDE drive on AVA-2825 EIDE VLB card increase with these new motherboard BIOS settings?

The results from the EIDE controller was with the same settings. I had to reset BIOS settings when switching the cards and I forgot to switch them to manual and optimize them when posting the first SCSI results.

You can use a LBA bios extender which sits in the MBR of the hard drive to gain access to the space normally out reach on larger […]
Show full quote

You can use a LBA bios extender which sits in the MBR of the hard drive to gain access to the space normally out reach on larger disks. Though I have no idea what might happen if you combine it with a disk cache like this.

Here is something I did a number of years ago:
viewtopic.php?f=46&t=40062#p368626

Out of curiosity, did you try benching against software-based caches?

I know it's possible to use a LBA bios extender, but this has to be setup when the disc is newly formatted, which I had not done in this test case. I also don't want any extra "layers" which could cloud the results.
Haven't benched against software caches. I'm actually having trouble finding a good bench for that. ATTO and HD Tune is going haywire with the cache controller. I need to test more benchmarks.

OK, thank you for the clarification. I am looking forward to some results which indicate some advantage of using VLB SCSI over VLB IDE when newer HDDs are being used. I figure some benchmark which draws heavily on simultaneous CPU + HDD usage would be beneficial.

Well, hardware is in place. Lets get some benches up and running:

Any news?
Interested in cache controllers too.
I've done some tests, short results are here: Bought these (retro) hardware today
While full discussion is on russian https://translate.google.com/translate?sl=ru& … 8429&edit-text= (google translated)
Last few month I've received another 2 cache controllers, for IDE bus and for PCI, still untested.

easy_john wrote:

I've done some tests, short results are here: Bought these (retro) hardware today

quite frankly, isn't that the worst case scenario for cache controllers? afaik usually they accelerate the random access write operations mostly;

also i'd expect scsi to be a little bit slower when it comes to max values, as it is far more complex computationally than ide, on the other hand i would expect it to hold the transfer rate much better than ide;

last thing, none of you guys seem to be using a cache controller with battery backup - these were nonexistent back then?

ps really last thing, if you have a pci system to play this on with, fetch a compaq smart controller for comparison (pentium era stuff), just 2 words about this one: _damn_fast_ (and you may have to disable onboard ide in order to use it)

mrau wrote:

quite frankly, isn't that the worst case scenario for cache controllers? afaik usually they accelerate the random access write operations mostly;

Yes, but this is quite fast test and I need to know, where is bottleneck. What is a maximum speed we can get from this controller ram+chip+bus.
To see a real cache work we need a more complex tests, like batch install win95 and measure time.

mrau wrote:

also i'd expect scsi to be a little bit slower when it comes to max values, as it is far more complex computationally than ide, on the other hand i would expect it to hold the transfer rate much better than ide;

Maybe yes, but scsi version released 1-2 years before ide one, so it's can be just limited by slower chips or less optimized firmware.

mrau wrote:

last thing, none of you guys seem to be using a cache controller with battery backup - these were non-existent back then?

In consumer market - no. Only in server raid models. BBU become "a must" options somewhere in early 2000, all of our servers in 90-s has a raid controller, but most of them don't have a BBU.

mrau wrote:

ps really last thing, if you have a pci system to play this on with, fetch a compaq smart controller for comparison (pentium era stuff), just 2 words about this one: _damn_fast_ (and you may have to disable onboard ide in order to use it)

I've see a lot of raid controllers - all of them is quite fast.
But, only in 90-s there was just "cache controller", without a raid options, and we use them back in time, to speedup our disks. I'm personally use ISA(!) ide tekram even on a Pentium1 system - linear (max) transfer is limited to ISA speed ~5mb/s, but under random load (in windows 9x mostly) I've got a higher overall performance and less system latency, comparing to onboard (pci) ide controller.

vetz wrote:

Well, hardware is in place. Lets get some benches up and running: […]
Show full quote

Well, hardware is in place. Lets get some benches up and running:

vetz, well, we've waited more than a year for the results, and as such, have passed your test of patience. Can we see the results now?

feipoa wrote:

vetz wrote:

Well, hardware is in place. Lets get some benches up and running: […]
Show full quote

Well, hardware is in place. Lets get some benches up and running:

vetz, well, we've waited more than a year for the results, and as such, have passed your test of patience. Can we see the results now?

I started on the project, but never finished it due to technical issues. Some of the controllers required reinstallation of Windows due to drivers, or I had issues with multiple disks/controllers being used in the same PC at the same time which stopped boot or crippled performance. Benching multiple controllers in a 486 system is a test of patience.

Could you take a photo of your AVA-2825, vetz?

It's looking like a really good controller. The VLB market was so brief that Mode 4 support seems to turn up somewhat rarely. I have a lot of questions. 😀

Does it require a driver to operate fast in DOS? (wrt. 12 MB/s linear in Speedsys.) What linear read rates is it limited to without a driver? How much conventional memory does it take up; does the BIOS take up a small chunk of memory similar to Promise?

vetz wrote:

feipoa wrote:

vetz wrote:

Well, hardware is in place. Lets get some benches up and running: […]
Show full quote

Well, hardware is in place. Lets get some benches up and running:

vetz, well, we've waited more than a year for the results, and as such, have passed your test of patience. Can we see the results now?

I started on the project, but never finished it due to technical issues. Some of the controllers required reinstallation of Windows due to drivers, or I had issues with multiple disks/controllers being used in the same PC at the same time which stopped boot or crippled performance. Benching multiple controllers in a 486 system is a test of patience.

Do you think you'll get back to this? I'm particularly interested in the AVA-2825 IDE vs. conventional VLB IDE controllers, like offerings from Winbond and UMC. Also really interested in AVA-2825 IDE vs. AHA-2842A, and AVA-2825 SCSI vs. AHA-2842. If there really is no benefit in having VLB SCSI, then it might be better to use the AVA-2825 IDE controller because I could also use the POD and Cyrix 5x86 chips with L1 in WB mode. If I recall correctly, only the Am5x86 would work with L1 WB when the AHA-2842A was installed.

try to benchmark and compare the similar age controllers, with authentic hdd from the same year. you will get the right result.

I made some tests of the
Promise DC4030VL-2 with firmware 4.3 BIOS 4.4
Tekram DC-680C (VLB)
Generic Promise 20230 based non-cache controller (VLB)
Some Promise 20230 based controller with it's own BIOS (adds Fast and Turbo mode, although Turbo mode hangs DOS during boot).

I tested it with 1GB Transcend Industrial CF card and IBM 520MB IDE drive.

Results below.

Non-cached controllers for comparison

I also tested the DC4030VL-2, the result is very bad.
His advertisement promises 14.3mb/s

Its linear reading is slightly higher than 1mb/s (according to other sources, 1.3mb/s), i.e. even worse than his tekram brother.
Writing is about the same (1.1mb/s).
The cache is certainly more interesting - writing 2.4mb/s, and reading as much as 15.17mb/s!
But if you write a lot, then the speed drops even less than 1mb/s (for example, we write 32mb with a 16mb cache)
I drove with a flash drive at 512mb (9.5 mb/s exactly gives), a little later I will test with a normal HDD.
And, if the bios on the motherboard does not know the secondary channel, then the bios of the controller does not add it 🙁

It is really hard to benchmark storage devices under Dos becuase in most cases, the bench mark is just measuring how fast your BIOS does real mode polled IO over the ISA bus at ATA-0 speeds, even when you are using a VLB controller and a fast storage device. Most motherboard bios routines are written for compatibility and don’t take advantage of 32 bit transfers, eide transfer modes, or bus mastering.

You might see much better performance with speedsys if you run it under windows 98 with a protected mode driver that has bus master enabled or maybe you can find a udma driver for dos that works with your card, either on an option rom or as something you put in your config.sys.

You're wrong.
First there are fast adapters -
QD6580 https://www.phantom.sannata.org/viewtopic.php … =615622#p615622
appian adi/2 https://www.phantom.sannata.org/viewtopic.php … =631309#p631309
Promise EIDE2300Plus Re: My Big Red Switch 486
And secondly, why do I need Win98 on this machine? I can run it on P4-3.2/i915 too.

Do those fast adapters come with option roms that replace your motherboard int 13 bios call?