VOGONS


My first 5x86 build (also motherboard component level repair)

Topic actions

Reply 60 of 71, by feipoa

User metadata
Rank l33t++
Rank
l33t++
133MHz wrote:
I did, it goes like this: […]
Show full quote
feipoa wrote:

Were you able to determine the pin-outs of the serial headers? This is on my list of things to (eventually) do.

I did, it goes like this:

IDC   DE9
1 1
2 6
3 2
4 7
5 3
6 8
7 4
8 9
9 5
10 x

Could you label the pin headers on the motherboard in terms of: RTC, DTR, etc? Does the serial mouse work? Thanks.

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

Reply 61 of 71, by Jolaes76

User metadata
Rank Oldbie
Rank
Oldbie

133 mhz wrote:

[Next I decided to try the 4.00 BIOS. Burned it to a 27C1001 using a standalone programmer, swapped the chip and cleared the CMOS by setting J6 to 3-4 for ~30 seconds as FGB recommended. It came up fine but I didn't get a CMOS Checksum Error or similar - in fact the CMOS settings were intact. When it gets to the boot stage it stops for a few seconds, then displays a white rectangle on the center of the screen and locks up hard. It doesn't even try to boot from a floppy or hard disk.

I had the exact same issue on a v2.11 board with the 4.00a BIOS.

I found some hints on some unoffical BIOS versions which since disappeared from the internet. One of these was exactly aimed to cure the "hang after PnP device detection" bug that affects v2.1 and v2.11 boards. We need to find this file, originally hosted at

levi.cz/pub/support/driver/mb/zidabios/4dps/4dps400e.zip

the very last unofficial BIOS version was 4.01, mostly aimed at v3.1 boards.

"Ita in vita ut in lusu alae pessima iactura arte corrigenda est."

Reply 62 of 71, by Robin4

User metadata
Rank l33t
Rank
l33t

Nice clean build over therem really like it very much.. How you build this system with limited components, i put my hat off for this build!

~ At least it can do black and white~

Reply 63 of 71, by ALEKS

User metadata
Rank Member
Rank
Member

Hi,

I am having the same Tomato 4DPS board in my 486 machine.

The specifications are as follows: Intel 80486DX4/100MHz CPU with Write-Through L1 cache, 64Mb RAM, 256K cache RAM, Matrox Millenium 2 with 4Mb RAM, 4.3Gb Quantum Fireball disc drive.
I have ran the System Speed Test program, and it came out with this information. It is not as fast as a 586/133 CPU but it does well. The BIOS settings are identical to what I saw in the posting of FGB on Page 2, except for the DRAM Write Cycle which is set to 1WS on my machine.

ng7f8.png

However, my problem is that I cannot get this board to start-up with an AMD 486DX4/120MHz CPU. I did all the jumper settings as per User Manual, I inserted the CPU correctly in the socket but it won't do anything. No video outpus, no PC-Speaker beeps. The board refuses to start-up. The CPU is good, I tested it with another board.

Do you have any ideas on this issue?

PS: Don't mind the date, the BIOS has the Y2K bug so it thinks we're in 1998. I'll flash in a BIOS upgrade at some point.

Thank you,
ALEKS

TX486DLC / 40 MHz | 32 Mb RAM | 16-bit ISA Backplane | EXCELGRAPH ET4000/W32i 2 Mb | I/O Interface | Audio Interface | PC Speaker Driver | Signal View Interface
3.5" & 5.25" FDD | 4 x 512 Mb CF | HP 82341D Interface | Intel EtherExpress 16

Reply 64 of 71, by kanecvr

User metadata
Rank Oldbie
Rank
Oldbie
133MHz wrote:

[*]24 MB (16+8MB) FPM RAM (largest FPM sticks I had in my RAM stash - unfortunately this board doesn't support EDO)

Actually it does support EDO - at least my V 1.3 does. It also gets a nice memory read boost running EDO, but it's extremely picky about what you put in it, and it doesn't like it when you mix modules from different brands. The "pickyness" can be sorted by fashing the latest bios (witch also supports HDDs up to 40GB) but my board came with a non-reflashable eprom. I do have some eeprom chips laying around from dead motherboards, I'll probably flash it at one point - if I find it's as fast or faster then the LS486e from lucky tech.

I also found that it will do 60 and 66MHz FSB, and it has a setting for 3.6v CPUs (great for running cyrix chips and overclocking AMD or POD chips).

Reply 65 of 71, by Cga.8086

User metadata
Rank Member
Rank
Member

Hello all, i also have a zida 4dps board with lots of problems,
- it has like four 512k chips of cache installed and one 256k chip in the tag socket.
- i checked bios and cache is both configured as active. i also did a reset of the bios.
- Booted the pc and it detects the 256K cache on the boot quare screen. ....but when i run cachechk.exe the L2 cache is not detected! and it only detects 8k cache!

I even changed all the cache to 256k chips to have 128 cache (One 256k chip as tag and four 256k chips for cache). booted the machine, the cache is found as 128k on boot.

On boot screen i get in the box Am486DX4 (even tho the cpu is actually an intel dx4 100mhz)

This is the picture of cachechk, the cache is not even found:

Whats_App_Image_2018-03-21_at_9.23.40_AM.jpg

i also configured the jumpers correctly so i don´t know whats going on

Reply 66 of 71, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Perhaps there is something wrong with your actual cache chips. You should probably start a new thread topic for this issue.

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

Reply 67 of 71, by byte_76

User metadata
Rank Member
Rank
Member
feipoa wrote on 2012-06-30, 03:11:
It looks like someone bought this board already. The shipping price was a little too steep for me. There is a high likelihood […]
Show full quote
badmofo wrote:

speak of the devil, there's one for sale for 30 bucks here:

http://www.ebay.com.au/itm/Vintage-486-PCI400 … #ht_3076wt_1396

I've bought of this guy before, he's honest.

It looks like someone bought this board already. The shipping price was a little too steep for me. There is a high likelihood that someone here bought it after you posted that link, so for whoever bought it, could you send me an image of the BIOS? I happen to have this board already, but would like to see if your BIOS is newer than mine. I can also send you my BIOS image. It is dated 10/7/96.

EDIT: Looking further at the images on the ebay site for this board, it seems that board's BIOS was dated 1/10/96, which is quite a bit older than the one I have.

The BIOS from the TMF PCI400-4 (sometimes referred to by the brand "Taken") does not work reliably in my Zida Tomato 4DPS 2.1 board. The main difference seems to be related to FSB jumpers and how the BIOS interprets the FSB jumpers. If anyone has the official manual for the PCI400-4, please share it.

In comparing both the Taken and Zida boads, I have determined that the Taken board has L2 and memory throughput equivalent to a fast UMC board, whereas the Zida board was lacking about 20% behind. Both Advanced Chipset settings were set equivalently. I am very interested to see if this 4.00A BIOS will clear this up. For whoever has it, please share it.

I have tested the Taken board with a Voodoo3 3000 PCI and it works very well. It also works well with a Matrox Millennium G200. I am in the process of doing a Matrox G200 vs. Voodoo3 3000 show down using an AMD X5-133. The PS/2 mouse port also works fine in Windows 98SE. The correct pin-outs are (1-5), Data, Clock, Gnd, Empty, Vcc.

As far as RAM is concerned, the manual for the 4DPS board mentions this board will only take 64 MB of FPM RAM, however I have found that it will work fine with 128 MB FPM RAM, at least the sticks I have worked. The BIOS lets you set write-through or write-back for L2 cache and set the TAG bit, so you could run your L2 in write-through mode and cache all 128 MB of RAM if you install 512 KB of cache. I have 512 KB of cache installed on my Taken and Zida boards. CTCM7 properly reports all RAM as cacheable when 128 MB is installed and when the cache is set to write-through mode. The ability to cache 128 MB of RAM and the ability to use a Voodoo 3 3000 are the biggest advantages of this board over a good UMC board. The advantage of the Voodoo 3 over a Voodoo 1/2 is that you do not need a consume a second PCI slot for a 2D card.

It is my overall hope that this board will be the SiS equivalent of a UMC-based Biostar MB-8433UUD. After the Voodoo vs. Matrox show down, I'll need to determine if the Cyrix 5x86 is up to snuff on the SiS. The main drawback of the Taken board is that it requires FPM RAM, no EDO. The Biostar board will take FPM and EDO RAM, however the Taken board has a built-in coin cell battery, whereas the Biostar board has a removable RTC module (a small drawback of the Biostar UMC board).

Some initial benchmark tests have revealed that the Taken board is ever so slightly faster for Cachechk's memory speed. The SiS has 42.8 MB/s, whereas the UMC was at 39.7 MB/s. Perhaps the advantage of using EDO in the UMC board will equalise this effect.

Quake 2 at 640x480 in 3dfx OpenGL mode is 8.7 fps with an AMD X5-133. The frame rate will probably be playable with a POD or fast Cyrix. I have not yet experimented with overclocked FSBs on this board.

One other drawback of the Zida or Taken board is that the I could only get the onboard CPU voltage regulator to output 3.35 V and 4.96 V. It doesn't seem to have a 3rd jumper option for 3.6 V. The Biostar UMC board, on the other hand, had a 3rd voltage regulator resistor option. I will likely modify the Zida board's 3.35 V setting to be 3.0 - 4.0 V variable.

I've just got my hands on a PCI-400-4 and I'm struggling with the CPUJ settings.

I managed to get an AMD DX4-100 SV8B working on the board using the config from page 4 of Rev 2.0 Zida 4DPS manual on TRW but the config in that manual (amendment at the bottom of the document) for Am5x86-133 does not work, neither does the config for Cyrix 5x86 100GP on page 4.
The Cyrix will POST at 66 if I close JP17 pins 5-6.
The Am5x86 will POST at 100MHz with the settings that I used for the SV8B CPU.
I also cannot get my Intel DX4 100 to POST at 100 but it will POST at 66MHz with same settings as the SV8B and JP17 pins 5-6 closed.

None of the manuals seem to have the correct jumper config for all the jumpers on my PCI400-4.

Please help!

Reply 68 of 71, by zapbuzz

User metadata
Rank Oldbie
Rank
Oldbie
byte_76 wrote on Today, 13:28:
I've just got my hands on a PCI-400-4 and I'm struggling with the CPUJ settings. […]
Show full quote
feipoa wrote on 2012-06-30, 03:11:
It looks like someone bought this board already. The shipping price was a little too steep for me. There is a high likelihood […]
Show full quote
badmofo wrote:

speak of the devil, there's one for sale for 30 bucks here:

http://www.ebay.com.au/itm/Vintage-486-PCI400 … #ht_3076wt_1396

I've bought of this guy before, he's honest.

It looks like someone bought this board already. The shipping price was a little too steep for me. There is a high likelihood that someone here bought it after you posted that link, so for whoever bought it, could you send me an image of the BIOS? I happen to have this board already, but would like to see if your BIOS is newer than mine. I can also send you my BIOS image. It is dated 10/7/96.

EDIT: Looking further at the images on the ebay site for this board, it seems that board's BIOS was dated 1/10/96, which is quite a bit older than the one I have.

The BIOS from the TMF PCI400-4 (sometimes referred to by the brand "Taken") does not work reliably in my Zida Tomato 4DPS 2.1 board. The main difference seems to be related to FSB jumpers and how the BIOS interprets the FSB jumpers. If anyone has the official manual for the PCI400-4, please share it.

In comparing both the Taken and Zida boads, I have determined that the Taken board has L2 and memory throughput equivalent to a fast UMC board, whereas the Zida board was lacking about 20% behind. Both Advanced Chipset settings were set equivalently. I am very interested to see if this 4.00A BIOS will clear this up. For whoever has it, please share it.

I have tested the Taken board with a Voodoo3 3000 PCI and it works very well. It also works well with a Matrox Millennium G200. I am in the process of doing a Matrox G200 vs. Voodoo3 3000 show down using an AMD X5-133. The PS/2 mouse port also works fine in Windows 98SE. The correct pin-outs are (1-5), Data, Clock, Gnd, Empty, Vcc.

As far as RAM is concerned, the manual for the 4DPS board mentions this board will only take 64 MB of FPM RAM, however I have found that it will work fine with 128 MB FPM RAM, at least the sticks I have worked. The BIOS lets you set write-through or write-back for L2 cache and set the TAG bit, so you could run your L2 in write-through mode and cache all 128 MB of RAM if you install 512 KB of cache. I have 512 KB of cache installed on my Taken and Zida boards. CTCM7 properly reports all RAM as cacheable when 128 MB is installed and when the cache is set to write-through mode. The ability to cache 128 MB of RAM and the ability to use a Voodoo 3 3000 are the biggest advantages of this board over a good UMC board. The advantage of the Voodoo 3 over a Voodoo 1/2 is that you do not need a consume a second PCI slot for a 2D card.

It is my overall hope that this board will be the SiS equivalent of a UMC-based Biostar MB-8433UUD. After the Voodoo vs. Matrox show down, I'll need to determine if the Cyrix 5x86 is up to snuff on the SiS. The main drawback of the Taken board is that it requires FPM RAM, no EDO. The Biostar board will take FPM and EDO RAM, however the Taken board has a built-in coin cell battery, whereas the Biostar board has a removable RTC module (a small drawback of the Biostar UMC board).

Some initial benchmark tests have revealed that the Taken board is ever so slightly faster for Cachechk's memory speed. The SiS has 42.8 MB/s, whereas the UMC was at 39.7 MB/s. Perhaps the advantage of using EDO in the UMC board will equalise this effect.

Quake 2 at 640x480 in 3dfx OpenGL mode is 8.7 fps with an AMD X5-133. The frame rate will probably be playable with a POD or fast Cyrix. I have not yet experimented with overclocked FSBs on this board.

One other drawback of the Zida or Taken board is that the I could only get the onboard CPU voltage regulator to output 3.35 V and 4.96 V. It doesn't seem to have a 3rd jumper option for 3.6 V. The Biostar UMC board, on the other hand, had a 3rd voltage regulator resistor option. I will likely modify the Zida board's 3.35 V setting to be 3.0 - 4.0 V variable.

I've just got my hands on a PCI-400-4 and I'm struggling with the CPUJ settings.

I managed to get an AMD DX4-100 SV8B working on the board using the config from page 4 of Rev 2.0 Zida 4DPS manual on TRW but the config in that manual (amendment at the bottom of the document) for Am5x86-133 does not work, neither does the config for Cyrix 5x86 100GP on page 4.
The Cyrix will POST at 66 if I close JP17 pins 5-6.
The Am5x86 will POST at 100MHz with the settings that I used for the SV8B CPU.
I also cannot get my Intel DX4 100 to POST at 100 but it will POST at 66MHz with same settings as the SV8B and JP17 pins 5-6 closed.

None of the manuals seem to have the correct jumper config for all the jumpers on my PCI400-4.

Please help!

such old motherboard components struggle at high frequencies because clock crystals capacitors etc deteriorate over time and need to be tested for confidence.
I always have other motherboards to prove the CPU's are healthy.

Reply 69 of 71, by byte_76

User metadata
Rank Member
Rank
Member
zapbuzz wrote on Today, 16:43:
byte_76 wrote on Today, 13:28:
I've just got my hands on a PCI-400-4 and I'm struggling with the CPUJ settings. […]
Show full quote
feipoa wrote on 2012-06-30, 03:11:
It looks like someone bought this board already. The shipping price was a little too steep for me. There is a high likelihood […]
Show full quote

It looks like someone bought this board already. The shipping price was a little too steep for me. There is a high likelihood that someone here bought it after you posted that link, so for whoever bought it, could you send me an image of the BIOS? I happen to have this board already, but would like to see if your BIOS is newer than mine. I can also send you my BIOS image. It is dated 10/7/96.

EDIT: Looking further at the images on the ebay site for this board, it seems that board's BIOS was dated 1/10/96, which is quite a bit older than the one I have.

The BIOS from the TMF PCI400-4 (sometimes referred to by the brand "Taken") does not work reliably in my Zida Tomato 4DPS 2.1 board. The main difference seems to be related to FSB jumpers and how the BIOS interprets the FSB jumpers. If anyone has the official manual for the PCI400-4, please share it.

In comparing both the Taken and Zida boads, I have determined that the Taken board has L2 and memory throughput equivalent to a fast UMC board, whereas the Zida board was lacking about 20% behind. Both Advanced Chipset settings were set equivalently. I am very interested to see if this 4.00A BIOS will clear this up. For whoever has it, please share it.

I have tested the Taken board with a Voodoo3 3000 PCI and it works very well. It also works well with a Matrox Millennium G200. I am in the process of doing a Matrox G200 vs. Voodoo3 3000 show down using an AMD X5-133. The PS/2 mouse port also works fine in Windows 98SE. The correct pin-outs are (1-5), Data, Clock, Gnd, Empty, Vcc.

As far as RAM is concerned, the manual for the 4DPS board mentions this board will only take 64 MB of FPM RAM, however I have found that it will work fine with 128 MB FPM RAM, at least the sticks I have worked. The BIOS lets you set write-through or write-back for L2 cache and set the TAG bit, so you could run your L2 in write-through mode and cache all 128 MB of RAM if you install 512 KB of cache. I have 512 KB of cache installed on my Taken and Zida boards. CTCM7 properly reports all RAM as cacheable when 128 MB is installed and when the cache is set to write-through mode. The ability to cache 128 MB of RAM and the ability to use a Voodoo 3 3000 are the biggest advantages of this board over a good UMC board. The advantage of the Voodoo 3 over a Voodoo 1/2 is that you do not need a consume a second PCI slot for a 2D card.

It is my overall hope that this board will be the SiS equivalent of a UMC-based Biostar MB-8433UUD. After the Voodoo vs. Matrox show down, I'll need to determine if the Cyrix 5x86 is up to snuff on the SiS. The main drawback of the Taken board is that it requires FPM RAM, no EDO. The Biostar board will take FPM and EDO RAM, however the Taken board has a built-in coin cell battery, whereas the Biostar board has a removable RTC module (a small drawback of the Biostar UMC board).

Some initial benchmark tests have revealed that the Taken board is ever so slightly faster for Cachechk's memory speed. The SiS has 42.8 MB/s, whereas the UMC was at 39.7 MB/s. Perhaps the advantage of using EDO in the UMC board will equalise this effect.

Quake 2 at 640x480 in 3dfx OpenGL mode is 8.7 fps with an AMD X5-133. The frame rate will probably be playable with a POD or fast Cyrix. I have not yet experimented with overclocked FSBs on this board.

One other drawback of the Zida or Taken board is that the I could only get the onboard CPU voltage regulator to output 3.35 V and 4.96 V. It doesn't seem to have a 3rd jumper option for 3.6 V. The Biostar UMC board, on the other hand, had a 3rd voltage regulator resistor option. I will likely modify the Zida board's 3.35 V setting to be 3.0 - 4.0 V variable.

I've just got my hands on a PCI-400-4 and I'm struggling with the CPUJ settings.

I managed to get an AMD DX4-100 SV8B working on the board using the config from page 4 of Rev 2.0 Zida 4DPS manual on TRW but the config in that manual (amendment at the bottom of the document) for Am5x86-133 does not work, neither does the config for Cyrix 5x86 100GP on page 4.
The Cyrix will POST at 66 if I close JP17 pins 5-6.
The Am5x86 will POST at 100MHz with the settings that I used for the SV8B CPU.
I also cannot get my Intel DX4 100 to POST at 100 but it will POST at 66MHz with same settings as the SV8B and JP17 pins 5-6 closed.

None of the manuals seem to have the correct jumper config for all the jumpers on my PCI400-4.

Please help!

such old motherboard components struggle at high frequencies because clock crystals capacitors etc deteriorate over time and need to be tested for confidence.
I always have other motherboards to prove the CPU's are healthy.

I also have other motherboards and I know the CPU's are good but my board doesn't want to POST at anything above 100MHz.

Also, my Cyrix 100GP doesn't POST at 100MHz with the same jumper settings that a board on TRW has with the same CPU .

Maybe it's an issue with the BIOS.
I'm using the Taken/J-Bond PCI400 BIOS version 4.51G from TRW.

Reply 70 of 71, by zapbuzz

User metadata
Rank Oldbie
Rank
Oldbie
byte_76 wrote on Today, 17:05:
I also have other motherboards and I know the CPU's are good but my board doesn't want to POST at anything above 100MHz. […]
Show full quote
zapbuzz wrote on Today, 16:43:
byte_76 wrote on Today, 13:28:
I've just got my hands on a PCI-400-4 and I'm struggling with the CPUJ settings. […]
Show full quote

I've just got my hands on a PCI-400-4 and I'm struggling with the CPUJ settings.

I managed to get an AMD DX4-100 SV8B working on the board using the config from page 4 of Rev 2.0 Zida 4DPS manual on TRW but the config in that manual (amendment at the bottom of the document) for Am5x86-133 does not work, neither does the config for Cyrix 5x86 100GP on page 4.
The Cyrix will POST at 66 if I close JP17 pins 5-6.
The Am5x86 will POST at 100MHz with the settings that I used for the SV8B CPU.
I also cannot get my Intel DX4 100 to POST at 100 but it will POST at 66MHz with same settings as the SV8B and JP17 pins 5-6 closed.

None of the manuals seem to have the correct jumper config for all the jumpers on my PCI400-4.

Please help!

such old motherboard components struggle at high frequencies because clock crystals capacitors etc deteriorate over time and need to be tested for confidence.
I always have other motherboards to prove the CPU's are healthy.

I also have other motherboards and I know the CPU's are good but my board doesn't want to POST at anything above 100MHz.

Also, my Cyrix 100GP doesn't POST at 100MHz with the same jumper settings that a board on TRW has with the same CPU .

Maybe it's an issue with the BIOS.
I'm using the Taken/J-Bond PCI400 BIOS version 4.51G from TRW.

So did you tested these CPU's that won't function after you updated the BIOS from TRW? before? both?

If you didn't test before you updated your BIOS, TRW doesn't guarantee the latest BIOS for your motherboard and you could've downgraded it to a version before the CPU's existed thus you'd require a BIOS with microcode update that TRW hasn't found yet.
For if this is a lesson be sure to backup your original BIOS before flashing it with a supposed update AWARD fash programs provide that feature in general using command line switches.
A custom mod with microcode update is not easily done it requires more tools but the correct version of these tools to suit the award modular bios version.
Anyhow you mention the manual from TRW doesn't give correct jumper settings this usually means its for a different revision of motherboard than yours and its possible yours is older
I always confirm the CPU's and motherboard chipsets release dates then I look for compatibility information to see if the chipset itself can support CPU at such speed to determine if a BIOS microcode update is even useful to solve the issue.

Reply 71 of 71, by byte_76

User metadata
Rank Member
Rank
Member
zapbuzz wrote on Today, 19:16:
So did you tested these CPU's that won't function after you updated the BIOS from TRW? before? both? […]
Show full quote
byte_76 wrote on Today, 17:05:
I also have other motherboards and I know the CPU's are good but my board doesn't want to POST at anything above 100MHz. […]
Show full quote
zapbuzz wrote on Today, 16:43:

such old motherboard components struggle at high frequencies because clock crystals capacitors etc deteriorate over time and need to be tested for confidence.
I always have other motherboards to prove the CPU's are healthy.

I also have other motherboards and I know the CPU's are good but my board doesn't want to POST at anything above 100MHz.

Also, my Cyrix 100GP doesn't POST at 100MHz with the same jumper settings that a board on TRW has with the same CPU .

Maybe it's an issue with the BIOS.
I'm using the Taken/J-Bond PCI400 BIOS version 4.51G from TRW.

So did you tested these CPU's that won't function after you updated the BIOS from TRW? before? both?

If you didn't test before you updated your BIOS, TRW doesn't guarantee the latest BIOS for your motherboard and you could've downgraded it to a version before the CPU's existed thus you'd require a BIOS with microcode update that TRW hasn't found yet.
For if this is a lesson be sure to backup your original BIOS before flashing it with a supposed update AWARD fash programs provide that feature in general using command line switches.
A custom mod with microcode update is not easily done it requires more tools but the correct version of these tools to suit the award modular bios version.
Anyhow you mention the manual from TRW doesn't give correct jumper settings this usually means its for a different revision of motherboard than yours and its possible yours is older
I always confirm the CPU's and motherboard chipsets release dates then I look for compatibility information to see if the chipset itself can support CPU at such speed to determine if a BIOS microcode update is even useful to solve the issue.

I have a BIOS programmer. I removed the original chip and used a different chip, the SST PH29EE010

I've tested with both the original BIOS and now today, I've tried various BIOS files from TRW for the Zida 4DPS 2.x but my board simply won't POST at 133 with the Am5x86., however the same CPU works at 100.
It also won't POST at 100 with the Cyrix 5x86 100GP but it does POST at 66 with the same CPU.

Maybe there is a faulty component on my board. (Maybe the clock chip is partially faulty?)