Nakamichi CD changer, nice 😁

Please show some pictures of your rig 😀

The chipset on that motherboard can only cache up to 64 MB of memory. Using more memory than 64 MB will result in a performance hit in Windows. I beleive 64 MB is cacheable if you are using 512 KB of L2 cache. If you are using only 256 KB of cache, then perhaps only 32 MB is cacheable? Does your motherboard have a COAST slot? If it does, and if it is empty, then you may only have 256 KB of cache.

Hello, Arctic! Welcome onboard!

As feipoa stated, you should probably reduce your RAM. You'll notice the difference. I remember playing Populous: TB having 128MB RAM on a similar machine. It was a huge pain until I reduced RAM.

You can also try to run your P166MMX as if it was P200MMX or P233MMX. These chips have the same design and a lot of them passed for 233 but were marked 166 to fill middle-end market. Eventually Intel locked multipliers in 166MMX and 200MMX CPUs to prevent people from using them as 233MMX. Worth a try, who knows, you might be lucky.

edit: Please post a lot of pictures! It is a good tradition on this forum to show others a lot of pictures of your hardware. We all like to watch!

Kamerat wrote:

Nakamichi CD changer, nice 😁

Please show some pictures of your rig 😀

Thanks 😁
Pictures uploaded! 😎

feipoa wrote:

The chipset on that motherboard can only cache up to 64 MB of memory. Using more memory than 64 MB will result in a performance hit in Windows. I beleive 64 MB is cacheable if you are using 512 KB of L2 cache. If you are using only 256 KB of cache, then perhaps only 32 MB is cacheable? Does your motherboard have a COAST slot? If it does, and if it is empty, then you may only have 256 KB of cache.

I did not know about that, I haven't noticed any performance problems yet. The Board seems to have support for a COAST module, but the socket is not soldered on. I need a new board... 😵

RacoonRider wrote:

Hello, Arctic! Welcome onboard! […]
Show full quote

Hello, Arctic! Welcome onboard!

As feipoa stated, you should probably reduce your RAM. You'll notice the difference. I remember playing Populous: TB having 128MB RAM on a similar machine. It was a huge pain until I reduced RAM.

You can also try to run your P166MMX as if it was P200MMX or P233MMX. These chips have the same design and a lot of them passed for 233 but were marked 166 to fill middle-end market. Eventually Intel locked multipliers in 166MMX and 200MMX CPUs to prevent people from using them as 233MMX. Worth a try, who knows, you might be lucky.

edit: Please post a lot of pictures! It is a good tradition on this forum to show others a lot of pictures of your hardware. We all like to watch!

Thank you for your help and the awesome welcome here @ all 😁

Is there a good memory benchmark program that I could run under a pure win98SE?
Photos are done and uploaded 😁

You don't need a new board, 64MB is plenty for that machine 😵 That was a HUGE amount of memory in 1997, a lot of machines were still running (and even being sold!) on 8 and 16 MB.

I had the same Miro HiScore 3D. The Miro uses some kind of round connector (head is mush today, can't remember the name, same format as PS/2) for the VGA passthrough, and I found the picture quality in passthrough mode was very bad. The Voodoo 2 that replaced it (with standard VGA connectors) was way better. How's the picture quality on yours?

IIRC, the HX chipset was considered the professional one by Intel and didn't have the 64 MB cache restriction, which was rumored to be an artificial limit by Intel anyway. HX boards were more expensive than VX or TX boards.

I always wondered, if you had more than 64MB installed, did only the first 64MB get cached, or would that disable the L2 cache completely?

oerk wrote:

You don't need a new board, 64MB is plenty for that machine :dead: That was a HUGE amount of memory in 1997, a lot of machines […]
Show full quote

You don't need a new board, 64MB is plenty for that machine 😵 That was a HUGE amount of memory in 1997, a lot of machines were still running (and even being sold!) on 8 and 16 MB.

I had the same Miro HiScore 3D. The Miro uses some kind of round connector (head is mush today, can't remember the name, same format as PS/2) for the VGA passthrough, and I found the picture quality in passthrough mode was very bad. The Voodoo 2 that replaced it (with standard VGA connectors) was way better. How's the picture quality on yours?

IIRC, the HX chipset was considered the professional one by Intel and didn't have the 64 MB cache restriction, which was rumored to be an artificial limit by Intel anyway. HX boards were more expensive than VX or TX boards.

I always wondered, if you had more than 64MB installed, did only the first 64MB get cached, or would that disable the L2 cache completely?

Thank you for your reply oerk!
I would like to test it!
Should I use memtest86?
I wanted to max the system out around the P166 in every way. I even thought about 256MB! 😁

I tried to rebuild the system I started with in 1997.
My first PC was:

Pentium 166@200
64MB EDO RAM
S3 Trio 64V+ 2MB
Creative 3D Blaster Voodoo 2 8MB PCI
Soundblaster 16 PCI (I know...)
Seagate 1.7GB HDD
Seagate? 6.4GB HDD
Miro Media Connect 34 ISA
15" Siemens Nixdorff MCM1503 -What an awesome monitor, everything has it´s own button! 😀
etc.

The Image Quality with the Miro pass-through cable is great! Check out the "Windows 98 splash screen".
I first wanted to compensate for it by choosing a Matrox Card as the 2D part.
Then I gave it up with the Matrox Mystique, because my driver was not able to play the FMVs in Tomb Raider. No intros, no cutscenes.

Finally - after a long quest - I was able to find the perfect 2D/3D card for the system and the Voodoo 1:

Only to find out that it´s broken:

My dream is to fix it one day and add it to the system. 😢

oerk wrote:

I always wondered, if you had more than 64MB installed, did only the first 64MB get cached, or would that disable the L2 cache completely?

Only the first 64 MB get cached. MS-DOS addresses memory from bottom up, and hardly anything uses more than 64 MB, so this problem is irrelevant under DOS. Windows however addresses memory from top down, so most of the stuff lands in uncached memory, hence the slowdown.

@5u3
Thank you for your help!

I guess I will take the memory out again then...
Weird, because I think the speed is fine.
Does anyone here have reference benchmark results to compare the system to?

I wish the board had a COAST Slot...
Is it possible to solder one onto the board? Since I have contacts for it, right above the CPU.
Maybe the traces are there?

Arctic wrote:

I wish the board had a COAST Slot...
Is it possible to solder one onto the board? Since I have contacts for it, right above the CPU.
Maybe the traces are there?

It's possible, but I personally wouldn't do it since there's probably other things you would have to solder in before you can actually get it working (capacitors and the like). Plus, like you mentioned in the specs, there should be some cache soldered on to the motherboard which is plenty enough for a machine like this.

@Sutekh94
I see. Of course they wouldn't just save on "the plastic", but also the other parts.
As it is with tuning, there is never enough, but I'm ok with 64MB if it's faster.

It's more important to me to have a Rendition V1000, than 128MB of RAM.
Does anyone think the card can be repaired / revived by just replacing the missing C30 capacitor after a no-vga-beep?
I had no luck in finding a second card so far. 😢

Can anyone help me with the USB 1.0?

I know, I know I have lots of questions 😁

@everyone
Thank you! You have been very helpful so far!

Seeing a '97 build, I wouldn't imagine it'd be ATX, very good!

USB 1.0 is painfully slow. I didn't even make it work on one of my old mobos that has i430VX chipset. Only 430TX or newer has always worked for me. Did win 98se install the appropriate drivers on the control panel?
You need to measure the voltage on each pin and compare it to PSU ground (found eg on a molex connector).
It may be:
+ Data Data -
+ Data Data -

or
+ Data Data -
- Data Data +

or even
+ Data Data NC -
- NC Data Data +

or any other crazy combination that the manufacturer implemented.
But I don't think it's gonna work; MSI doesn't even mention the USB pinout on your mobo manual.

If the COAST slot was never soldered in place by the manufacturer, it is possible that they used 512 KB of pipeline burst cache in your board revision. You should be able to see how much cache you have at POST or by using cachechk or speedsys. You can run CTCM7 to check how much memory is being cached. The best benchmark program to observe a hit from uncached memory is Ziff-Davis CPUMark99 (the stand-alone version is best in this regard).

With the usb header I think that the board manufacturer didn't feel like excluding pin 10, but that pin isn't connected anyway.It's a standard header, so you can put a 9 pin usb header in there. You just have to drill out a hole (on the usb mobo cable) for the 10th pin.

Are you guys interested in screenshots of Control Panels, Games, AWEAmp?
Is it correct to keep the CPU Clock below 200MHz for old DOS Games? I read something about a Turbo Pascal Bug with >200MHz. 😕
Otherwise I could install the 200MHz P1-MMX 😁

@feipoa

I found this spec sheet:
http://museum.ttrk.ee/th99/m/M-O/33572.htm

I checked the PC with ZD CPUmark 99 Version 1.0. My Result is: 7.26
Apparently that is too low:

Pentium Pro 200 - 17.5
K6 200 - 14.9
Pentium 266 MMX - 12.3
Pentium 200 MMX - 12.2
Pentium 200 (166) - 10.2
Pentium 166 - 9.1
Pentium 133 mobile - 7.8
Pentium 75 - 5.79

@Dropcik

could this be the correct pin setting? I found it in some old forum post:

| 6 | 5 |
| 7 | 4 |
| 8 | 3 |
| 9 | 2 |
| a | 1 |

1 + 6: VCC
2 + 7 D-
3 + 8 D+
4,5,9,a GND

According to the Ultimate 686 Benchmark Comparison, a P166 MMX with all memory cached should score 11.3 with CPUMark99. This result is for a 430TX chipset with SDRAM. The score for a 430VX chipset might be a 0-10% less.

This is my result with CTCM7

As expected - only 64 MB of your 128 MB of memory is being cached. You have 512 KB cache, which I beleive is the max for your chipset. Remove 64 MB of RAM for a total of 64 MB and your CPUMark99 score will increase.

Yes, i found the same forum you got that pin out from. it is correct and what you will need for the usb header is a dual usb 1-5 cable. but since pins 1 and 6 are at different ends the wrong pin would be getting the 5 volts. In my opinion I would just buy a pci usb 2.0 card. Its much faster and it works with almost all operating systems (w95 sr2 added usb support) but since you are running w98se, you will be fine.

feipoa wrote:

The chipset on that motherboard can only cache up to 64 MB of memory. Using more memory than 64 MB will result in a performance hit in Windows. I beleive 64 MB is cacheable if you are using 512 KB of L2 cache. If you are using only 256 KB of cache, then perhaps only 32 MB is cacheable? Does your motherboard have a COAST slot? If it does, and if it is empty, then you may only have 256 KB of cache.

I thought motherboards had to have the 11-bit "tag" to cache beyond 64 MB, regardless of the cache size.

Arctic wrote:

Only to find out that it´s broken: […]
Show full quote

Only to find out that it´s broken:

My dream is to fix it one day and add it to the system. 😢

It should still actually work even without the ceramic capacitor at C30 (the other pads were never populated). It's *very* easy to solder another one on. Just pop one off of an existing card or order some surface mount ceramic capcaitors.

GeorgeMan wrote:

Seeing a '97 build, I wouldn't imagine it'd be ATX, very good!

USB 1.0 is painfully slow. I didn't even make it work on one of my old mobos that has i430VX chipset. Only 430TX or newer has always worked for me. Did win 98se install the appropriate drivers on the control panel?
You need to measure the voltage on each pin and compare it to PSU ground (found eg on a molex connector).

IBM just transitioned over to ATX in '97 for their consumer computers, so it should be expected. USB 1.0 is fine for transferring files on a 66 Mhz bus. And yeah, just get a multimeter and probe the voltages (will have to be done while the computer is running-- DO NOT short any pins by accident).