VOGONS


486 IDE controller performance

Topic actions

First post, by GL1zdA

User metadata
Rank Oldbie
Rank
Oldbie

I decided to compare the performance of the 486 PCI integrated IDE controllers. I'm testing it with Wester Digital WD800BB drive, so that the drive won't limit the performance. I setup DOS 6.22 on it (it created a 2 GB partition on it), disabled SmartDrive and copied several benchmarking tools on it. I'm using SPEEDSYS to benchmark the controller - this is the only DOS program I know which can benchmark IDE performance.

I started with the UMC 8881 based Gigabyte GA486A/MS. The results with nearly default settings:
-33 MHz PCI clock (no divider on a DX4). Note: with an AMD 120 MHz CPU, the PCI bus will be set to 1/2 of the CPU bus speed and this will linearly affect the peroformance of the controller.
-performance PCI settings - I/O Recevory Time - 0 WS and CPU-to-PCI Post Write - 0 WS
-performance RAM settings - Cache Burst Read 2-1-1-1 and DRAM 0 WS
PCI settings and RAM settings seem to have no effect on IDE performance, so I'm not uploading results with default RAM and PCI settings.

gigabyte_default.png

With:
-IDE HDD Block Mode [ENABLED]
-IDE 32-bit Transfer Mode [ENABLED]
-IDE PIO Mode 3 Support [ENABLED]
the results are a bit (20% - 35% for reading) better:

gigabyte_performance.png

I also mesured the performance on a Intel 420EX (Aries) Classic-PCI Expandable Desktop (Ninja) and it seems it works much better (have yet to see if there are any BIOS settings which can improve it). Access Time is halved, Read performance is nearly twice the performance of the tweaked Gigabyte:

ninja.png

Can anyone comment on this? Is the UMC8881 really that bad? Are there any other BIOS settings I have to look for? Are there any other benchmarks for DOS IDE performance?

I will try to test a ASUS PVI-486SP3 later.

Attachments

  • Gigabyte default.png
    Filename
    Gigabyte default.png
    File size
    11.33 KiB
    Views
    4782 views
    File comment
    Gigabyte GA486A/MS - default settings
    File license
    Fair use/fair dealing exception
  • Gigabyte performance.png
    Filename
    Gigabyte performance.png
    File size
    11.35 KiB
    Views
    4782 views
    File comment
    Gigabyte GA486A/MS - performance settings
    File license
    Fair use/fair dealing exception
  • Ninja.png
    Filename
    Ninja.png
    File size
    11.48 KiB
    Views
    4782 views
    File comment
    Intel Classic-PCI Expandable Desktop (Ninja) - default settings
    File license
    Fair use/fair dealing exception
Last edited by GL1zdA on 2012-11-11, 12:37. Edited 1 time in total.

getquake.gif | InfoWorld/PC Magazine Indices

Reply 1 of 25, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Can you not go to PIO-4 mode?

What results do you get if you run the same tests with a (PCI) Promise ATA TX2-100 card in there?

Plan your life wisely, you'll be dead before you know it.

Reply 2 of 25, by GL1zdA

User metadata
Rank Oldbie
Rank
Oldbie
feipoa wrote:

Can you not go to PIO-4 mode?

What results do you get if you run the same tests with a (PCI) Promise ATA TX2-100 card in there?

No, it seems PIO-3 is max for the Gigabyte - there are no BIOS settings for PIO-4, there's nothing about it in the manual. The UM8881F is the 9524-BTA, the UM8886AF is 9525-AYO. I don't have a Promise ATA controller, but I probably could get one. Is there a big difference between FastTrack 100 and FastTrack 100 TX2?

getquake.gif | InfoWorld/PC Magazine Indices

Reply 3 of 25, by elianda

User metadata
Rank l33t
Rank
l33t

So at which setting is the HDD really communicating with the Controller?
PIO or MW-DMA ?

It would also be good to have an atbus output first to see what settings are supported.
ftp://ftp.heise.de/pub/ct/ctsi/ctatb47.zip

Retronn.de - Vintage Hardware Gallery, Drivers, Guides, Videos. Now with file search
Youtube Channel
FTP Server - Driver Archive and more
DVI2PCIe alignment and 2D image quality measurement tool

Reply 4 of 25, by GL1zdA

User metadata
Rank Oldbie
Rank
Oldbie
elianda wrote:

So at which setting is the HDD really communicating with the Controller?
PIO or MW-DMA ?

How can I check it? Speedsys just prints what the HDD supports. I'll download and run CTATBUS. I found in c't archives programs which could set some chipset settings which were hidden in BIOS. Does anyone know something like that for the UM8881F or 420EX?

getquake.gif | InfoWorld/PC Magazine Indices

Reply 5 of 25, by GL1zdA

User metadata
Rank Oldbie
Rank
Oldbie

CTATBUS sucks at printing to file, so I'm only attaching the things I think are relevant:

Datenrate      : 5-10Mbit/s
DMASupport : Ja
IORDY : Ja (what is this?)
IORDY-Disable : Ja
PIOmode : 2
DMAmode : 0
Enh. PioModi : 3,4
Mult-Sektor gesetzt auf 16
Sgl-DMA-Modi : /inaktiv
Mult-DMA-Modi : 0,1,2/aktiv:2
MUL-DMA minimal: 120 ns, empfolen: 120 ns
Ultra-DMA Modi : 0..5/aktiv:inaktiv

So I assume it runs at MWDMA 2.

getquake.gif | InfoWorld/PC Magazine Indices

Reply 7 of 25, by feipoa

User metadata
Rank l33t++
Rank
l33t++
swaaye wrote:

I remember a DOS chipset tweak utility called TweakBIOS. It supports a number of mid 90s chipsets.

An interesting program. It seems to still be available for download from, http://miro.pair.com/tweakbios/
Click on "Download the freeware version of TweakBIOS here! (for DOS or Windows 95)"

I tried it on my UMC 8881F/8886BF motherboard with a Cyrix 5x86-133 (2x66). I have attached some photos from the program. It is an easier to use version of CTCHIP34. I was hoping to be able to set the MB-8433UUD L2's cache into write-through mode since it is not natively included in the BIOS. Unfortunately the board hangs up. I tried many different methods, such as disabling the L2 in the BIOS first, disabling it in TweakBIOS, changing it to write-through, changing the tag length from 7 to 8 first, second, etc. I guess there is not a write-through L2 option in the 8433's BIOS for a reason.

Back to OP, for some reason the UMC IDE controller is not supported, so we cannot change the PIO mode. Images attached.

TweakBIOS also incorrectly identifies the CPU as a 6x86 instead of a 5x86, and as such, the register settings are incorrect. There was one feature of interest under CPU-to-PCI bridge called "enabled resource locking", which claimed enabled is for better performance. Any idea what this does?

You can also alter the PCI latency timer - have fun with that!

One other item of interest - there is a file, winnt.txt, in the zip package which describes a method using INT19.exe (included) to run the TweakBIOS program prior to booting into Windows NT. This could be useful for other hardware configuration programs which do not have a WINNT driver.

Attachments

  • TweakBIOS-5.jpg
    Filename
    TweakBIOS-5.jpg
    File size
    148.43 KiB
    Views
    4644 views
    File license
    Fair use/fair dealing exception
  • TweakBIOS-4.jpg
    Filename
    TweakBIOS-4.jpg
    File size
    74.66 KiB
    Views
    4644 views
    File license
    Fair use/fair dealing exception
  • TweakBIOS-3.jpg
    Filename
    TweakBIOS-3.jpg
    File size
    105.37 KiB
    Views
    4644 views
    File license
    Fair use/fair dealing exception
  • TweakBIOS-2.jpg
    Filename
    TweakBIOS-2.jpg
    File size
    75.9 KiB
    Views
    4644 views
    File license
    Fair use/fair dealing exception
  • TweakBIOS-1.jpg
    Filename
    TweakBIOS-1.jpg
    File size
    86.64 KiB
    Views
    4644 views
    File license
    Fair use/fair dealing exception

Plan your life wisely, you'll be dead before you know it.

Reply 8 of 25, by Anonymous Coward

User metadata
Rank l33t
Rank
l33t

Would it be possible to force your L2 cache to write through using a ROM editor and burning a custom flash ROM?

Maybe using a program like this one?

http://awdbedit.sourceforge.net/

"Will the highways on the internets become more few?" -Gee Dubya
V'Ger XT|Upgraded AT|Ultimate 386|Super VL/EISA 486|SMP VL/EISA Pentium

Reply 9 of 25, by feipoa

User metadata
Rank l33t++
Rank
l33t++
Anonymous Coward wrote:

Would it be possible to force your L2 cache to write through using a ROM editor and burning a custom flash ROM?

Maybe using a program like this one?

http://awdbedit.sourceforge.net/

I also figured that the only way to have a chance of getting L2 write-through working was to have it set in the BIOS before POST. This is assuming there isn't some PCB design limitation with running the cache in write-thru mode. The program you linked looked promising. The layout is easy to follow. There's quite a few options which can be added to the Chipset menu, and in particular, the L2 update scheme.

When I change this feature to Active and Write-through (see photo), it goes right back to disabled after I click on another setting and go back to L2 update scheme. For some reason my input is not getting saved. The documentation on how to use this program is horrible. I tried it with a Socket 7 BIOS image, and same thing happens, I cannot save my changes to the chipset or BIOS options. Pressing Edit Item List doesn't help save changes either. I can, however, change the revision information that gets displayed at boot.

If this program worked, I think it would be of benefit to anyone with an AWARD BIOS. Anyone know how to save changes in it?

Attachments

  • awdbedit_2.png
    Filename
    awdbedit_2.png
    File size
    25.16 KiB
    Views
    4597 views
    File license
    Fair use/fair dealing exception
  • awdbedit_1.png
    Filename
    awdbedit_1.png
    File size
    114.69 KiB
    Views
    4597 views
    File license
    Fair use/fair dealing exception
Last edited by feipoa on 2012-11-12, 22:55. Edited 1 time in total.

Plan your life wisely, you'll be dead before you know it.

Reply 10 of 25, by GL1zdA

User metadata
Rank Oldbie
Rank
Oldbie
swaaye wrote:

I remember a DOS chipset tweak utility called TweakBIOS. It supports a number of mid 90s chipsets.

Okay, I have run it and although it does not support the built-in IDE, there is one suspicious thing: Bus Mastering is 'Disabled' for it and I can't enable it - it allows me to change the value, but after applying (F10) it reverts to 'Disabled'. Is this normal?

getquake.gif | InfoWorld/PC Magazine Indices

Reply 11 of 25, by feipoa

User metadata
Rank l33t++
Rank
l33t++
GL1zdA wrote:

Bus Mastering is 'Disabled' for it and I can't enable it - it allows me to change the value, but after applying (F10) it reverts to 'Disabled'. Is this normal?

I think that is normal if the feature is unsupported. Try playing around with the AWARD BIOS editor. Your UMC chipsets are somewhat on the old side. Mine are all from mid-1996, yours is from mid-1995. Also, I'm not sure what the 8886BF has over the 8886AF. I try to avoid PCI 486 motherboards which don't have at least a 1996 date code.

Plan your life wisely, you'll be dead before you know it.

Reply 12 of 25, by Anonymous Coward

User metadata
Rank l33t
Rank
l33t

Award BIOS Editor is definitely a lemon. Maybe it just doesn't like XP. Maybe it needs a magic DLL. Maybe I'm just not "l33t" enogh. Who knows. I tried MODBIN too. Couldn't get that one working properly either.

Apparently in later versions of AWARD bios you could make hidden settings appear with one of the following cominations:

SHIFT + F1
ALT + F1
CTRL + F1

I have a feeling your BIOS is probably too old, but give it a shot anyway. In the meantime I will find a way to mod your BIOS image.

"Will the highways on the internets become more few?" -Gee Dubya
V'Ger XT|Upgraded AT|Ultimate 386|Super VL/EISA 486|SMP VL/EISA Pentium

Reply 13 of 25, by Anonymous Coward

User metadata
Rank l33t
Rank
l33t

Must be your lucky day. I was able to successfully mod your BIOS using MODBIN for award bios version 4.5x.

http://www.ryston.cz/petr/bios/modbin63.zip

Attachments

  • Filename
    UUD0520S.zip
    File size
    86.13 KiB
    Downloads
    177 downloads
    File license
    Fair use/fair dealing exception

"Will the highways on the internets become more few?" -Gee Dubya
V'Ger XT|Upgraded AT|Ultimate 386|Super VL/EISA 486|SMP VL/EISA Pentium

Reply 15 of 25, by feipoa

User metadata
Rank l33t++
Rank
l33t++
Anonymous Coward wrote:

Must be your lucky day. I was able to successfully mod your BIOS using MODBIN for award bios version 4.5x.
http://www.ryston.cz/petr/bios/modbin63.zip

Hah! It looks like we were working in parallel on this. I too found that modbin63 from that same source. It is the modbin63 which works with AWARD 4.5X PG BIOSes. The newer MODBIN's want AWARD version 6.X.

I've attached some images of MODBIN. This program actually lets you save the changes. I have unhidden everything from the BIOS Feature Setup and Chipset Feature Setup, except for the IR settings.

This program also lets users adjust the default Cyrix register settings. This way you don't need to run some NT4/W2K/DOS driver to enable all those awesome 5x86 enhancements This will take some time to figure out what is what though. The register index names AWARD uses for the Cyrix chips are a little different than the Cyrix documentation. Images attached.

One final comment on that other AWARD BIOS Editor, AWEBEDIT, we are not the only ones who cannot save changes. Check out this post,
http://sourceforge.net/projects/awdbedit/foru … 3/topic/1274666

After activating some things from the "system bios" section, how do you save it? There does not seem to be a save button, and when I clicked "run", everything just froze and the application closed. I want the settings to be permanent, can anyone help me?

There was no substantial reply.

Attachments

  • MODBIN_5.png
    Filename
    MODBIN_5.png
    File size
    15.35 KiB
    Views
    4447 views
    File license
    Fair use/fair dealing exception
  • MODBIN_4.png
    Filename
    MODBIN_4.png
    File size
    16.77 KiB
    Views
    4447 views
    File license
    Fair use/fair dealing exception
  • MODBIN_3.png
    Filename
    MODBIN_3.png
    File size
    11.75 KiB
    Views
    4447 views
    File license
    Fair use/fair dealing exception
  • MODBIN_2.png
    Filename
    MODBIN_2.png
    File size
    15.35 KiB
    Views
    4447 views
    File license
    Fair use/fair dealing exception
  • MODBIN_1.png
    Filename
    MODBIN_1.png
    File size
    13.44 KiB
    Views
    4447 views
    File license
    Fair use/fair dealing exception

Plan your life wisely, you'll be dead before you know it.

Reply 16 of 25, by feipoa

User metadata
Rank l33t++
Rank
l33t++

And here is what the actual BIOS screen looks like now. To confirm write-through L2 works, I added 128 MB of RAM, set SRAM TAG to 8+0, L2 to WT, and ran CTCM7.

It works. To verify further, I booted into NT4 and ran CPUMark99. The score is what it should be for all 128 MB cached!

My plans now are to clean up where the BIOS displays the items, for example, all RAM timings should be together, PCI stuff together, etc.

I have also added my BIOS revision with everything unhidden.

Attachments

  • CTCM7_with_new_AWARD_BIOS.jpg
    Filename
    CTCM7_with_new_AWARD_BIOS.jpg
    File size
    77.43 KiB
    Views
    4443 views
    File license
    Fair use/fair dealing exception
  • New_AWARD_BIOS_2.jpg
    Filename
    New_AWARD_BIOS_2.jpg
    File size
    165.84 KiB
    Views
    4444 views
    File license
    Fair use/fair dealing exception
  • New_AWARD_BIOS_1.jpg
    Filename
    New_AWARD_BIOS_1.jpg
    File size
    148.93 KiB
    Views
    4444 views
    File license
    Fair use/fair dealing exception
Last edited by feipoa on 2012-11-13, 14:34. Edited 1 time in total.

Plan your life wisely, you'll be dead before you know it.

Reply 17 of 25, by Anonymous Coward

User metadata
Rank l33t
Rank
l33t

Excellent. Glad to know everything is working. I'll have to find a way to do this to AMI BIOSes as well. We should probably make this information available in a different thread. I'm sure not too many people would be looking for this under "IDE controller performance"

You must get up insanely early. It looks like you posted this before 4AM.

"Will the highways on the internets become more few?" -Gee Dubya
V'Ger XT|Upgraded AT|Ultimate 386|Super VL/EISA 486|SMP VL/EISA Pentium

Reply 18 of 25, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Newborn babbies are up at odd times and so are the parents - especially when they learn to flip onto their belly!

I've added my hacked rearrangement of the Biostar MB-8433UUD BIOS. All options included. It seems to work well. This is a major accomplishment - 128 MB cacheable range on a MB-8433UUD. The value of these boards just increased!

I agree, this doesn't have much to do with IDE control, but perhaps the original poster can play around with ModBin to see if he can find what he is looking for. If moderator wants to cut this BIOS talk out and create a new topic, that is preferred.

My next goal for this project would be to mod the BIOS to enable, by default, the most important stable Cyrix 5x86 features, particularly LSSER and FP_FAST. I know what register index CD, CE, CF, C1, C2, and C3 are. I am not sure about C0, C5, C6, C8, C9, CB, and CC. I figure C0 might be 20h (PCR0), while C5, C6, C8, C9, CB, and CC might be part of CR0. The ModBin-listed BIOS default HEX data values are not what get reported when checked in DOS.

Attachments

  • Filename
    UUD2012B.rar
    File size
    85.83 KiB
    Downloads
    165 downloads
    File license
    Fair use/fair dealing exception

Plan your life wisely, you'll be dead before you know it.

Reply 19 of 25, by feipoa

User metadata
Rank l33t++
Rank
l33t++

A photo of the added BIOS features and arrangement by category: cache, RAM, PCI, ISA, etc.

Attachments

  • New_AWARD_BIOS_3.jpg
    Filename
    New_AWARD_BIOS_3.jpg
    File size
    167.75 KiB
    Views
    4398 views
    File license
    Fair use/fair dealing exception

Plan your life wisely, you'll be dead before you know it.