"
....
How much worse would 1024x768x16bit be compared to 800x600x24? I could probably live with 5 MP/s w/80 MHz MCLK since I intended to use the Voodoo2 for Windows 3D.
...
"
This was only a guess by me. Maybe 24bit mode is always bad. So I mean worse than the data transfer rate would suggest.

"
....
COffeeOne, did you run into the issue whereby shutting down Windows 95 when using 800x600x24 caused a black screen? Maybe it is only an issue in w95? I suspect it is related to using the ARK Tools. Looks like this:
....
"
No I did not have this effect under Windows 98SE.

feipoa wrote on 2023-11-03, 00:52:

Would you trust it for all datecodes, or just the datecodes from after that spec. sheet was produced? In my IC photo there are […]
Show full quote

mkarcher wrote on 2023-11-02, 09:47:

feipoa wrote on 2023-11-01, 23:53:

I then went to looking at the datasheet, and saw a curious statement indicating that 50 ns timings are also valid for 60 ns. Do we trust this?

I would trust it. Likely Alliance got production controlled well enough that all their chips conform to 50ns timing. On the other hand, a lot of products are specified to use "60ns" FPM RAM. If Alliance wouldn't offer chips marked as "-60", the manufacturers of those products either need to get a change approved, or they buy chips from the competition that are still marked "-60". Getting a change approved for a legacy product is a tedious task no manufacturer would like to do, so not offering "-60" chips is lost customer potential.

Would you trust it for all datecodes, or just the datecodes from after that spec. sheet was produced? In my IC photo there are datecodes from 41-2000 to 50-2003. The date on that spec sheet is 11 April 2001. The memory with the white printing is from 2000, while the brown printing from 2002-2003.

pentiumspeed wrote on 2023-11-02, 18:47:

Chips are not consistent (font and color of prints), indicative of chinese sellers. Where did you get these from, have the link?

Cheers,

They were from eBay and shipped from China. The white font units look authentic, as do the units with 0350 and 0311 datecodes. The unit with 0220 datecode has different font. I have been testing with the white font units with 0041 & 0048 datecodes.

pshipkov wrote on 2023-11-02, 14:06:

To your last question - yes.
This is the most interesting mode.

Forcing 1024x768x16bit will require a modification to the BIOS. The chances of it working are slim, so not sure if mkarcher is still interested in providing this hack. I assume we'd need VCLK set to the max? 130 MHz.

CoffeeOne wrote on 2023-11-02, 20:35:

I understand. But my Wintune 2.0 values were below 2MPixel/s for 800x600 24bit c @ 56Hz. So I assume it is worse at 1024x768 16b […]
Show full quote

I understand.
But my Wintune 2.0 values were below 2MPixel/s for 800x600 24bit c @ 56Hz. So I assume it is worse at 1024x768 16bit c @60Hz.
But I did not increase the clk yet.
And I probably should search for Wintune 98, that sounds appropriate for Windows 98 😁

How much worse would 1024x768x16bit be compared to 800x600x24? I could probably live with 5 MP/s w/80 MHz MCLK since I intended to use the Voodoo2 for Windows 3D.

COffeeOne, did you run into the issue whereby shutting down Windows 95 when using 800x600x24 caused a black screen? Maybe it is only an issue in w95? I suspect it is related to using the ARK Tools. Looks like this:
w95_shutdown_blankout_with_24bit.jpg

I also decided to run the tests with the 60 ns Siemens FPM at 80 MHz. 60 ns is not sufficient. This was what it looked like with 60 ns:
MCLK_80mhz_with_60ns_FPM_1.JPG
MCLK_80mhz_with_60ns_FPM_2.jpg
MCLK_80mhz_with_60ns_FPM_3.JPG

The best I could run with 60 ns was 70 MHz MCLK.

CoffeeOne wrote on 2023-11-02, 19:52:

At what speed does your board run? 180 or 200MHz? 21.1 quake and 972 tics doom seems to be incredible fast.

Am5x86 at 180 MHz.

The pixel defects disappears if clocked lower or timings changed for slower settings?

Anyway, I don't trust chinese sellers anymore for complex parts. Add to this, received a wrong blower from the aliexpress last year. Their seller's picture and text was for EBM papst blower. Nor anywhere did seller indicates otherwise. Only thing is last few pictures was what I received but even seller didn't stated this was the one either.
Even amazon was the quality issues also. last month, via amazon, received 16 10k ohm joysticks for specific microsoft controller for xbox one/ series line. All the resistances were all over the map only few unrelated pairs matched. I must require closely matched set for each joystick to prevent drift after repair.
6 years ago, I ordered differential amplifier and were fakes as it was from chinese seller. Leads breaks with gentle bent once. Even very light loads failed to perform and this was recreated from a test diagram in the datasheet by official original maker for this part from USA.
I also had fake 256K x 16 bit vram chips from chinese seller, the alloy metal was so bad that solder would not wet on it even with sanding and flux, markings were badly done, and painted over too.

Back to your issue with few pixel defect, seen this before.
Had a video card back in the day had bad cell area that was addressed from off screen data into viewable area generated thick static bar when I scrolled down long benchmarks in firefox. This was on a ET4000/w32p PCI card. Replaced all chips from donor motherboard fixed completely, that was around 2004.

Cheers,

@Pentiumspeed,
The defects I am showing at the bottom of page 6 are with the Siemens 60 ns DRAM at 80 MHz MCLK. The Alliance branded "50 ns" DRAM I received from eBay worked fine at 80 MHz MCLK. The Siemens branded 60 ns worked fine up to 70 MHz MCLK.

feipoa wrote on 2023-11-03, 23:32:

@Pentiumspeed,
The defects I am showing at the bottom of page 6 are with the Siemens 60 ns DRAM at 80 MHz MCLK. The Alliance branded "50 ns" DRAM I received from eBay worked fine at 80 MHz MCLK. The Siemens branded 60 ns worked fine up to 70 MHz MCLK.

nice. So I have already 50ns capable FPM RAM as second megabyte on my card, as I seem to have the exact same alliance branded type.
So most likely I have to replace the first megabyte.
By the way 80MHz is an impressive overclock, for your initial clock of 55MHz, 80MHz means +45%. For an initial value of 60MHz, 80MHz still is +33% increase.

feipoa wrote on 2023-09-30, 12:24:

I'm starting to get back into testing this ARK1000VL card. […]
Show full quote

I'm starting to get back into testing this ARK1000VL card.

I replaced the RAMDAC with a 110 MHz piece and swapped the CLK GEN with an ARK1294A. Then booted to w95 and tried 1280x1024x256c. The system is still trying to run interlaced mode. So I booted to NT4, which lets me specify that I want 60 Hz, but the screen still looks to be interlaced mode.

ARK_with_110MHz_RAMDAC_and_ARK_PLL.JPG

Next, I altered the 2themax's VGA BIOS per mkarcher's analysis, namely,

offset 4391h, turn 74h into EBh
offset 7FFFh, turn D8 into 61h

Programmed this to a W27E257-12 EEPROM, installed into VLB card, then ran the above test again.

The screen still jitters at 1280x1024x256c at 60 Hz, but it jitters in a different manner compared to how the non-edited BIOS was jittering.

The file:
1280x1024x256c_i43hz.zip contains a video showing the expected interlaced mode jitter on an unsupported monitor. This is also what 60 Hz looks like with the non-edited BIOS.

With the edited bios, 1280x1024x256c_60hz.zip shows how this jitter has changed quite a bit. Unfortunately, the display is still not as it should be, but the editing of the BIOS changed something.
1280x1024x256c_i43hz.zip
1280x1024x256c_60hz.zip

Do you remember where you got the 50ns 20pin DRAMs from?
It seems easier to buy a graphics card with such chips and remove them.
I found:
Mosel V53C104HP45 (with a "fast page mode read or write cycle time" of 25ns, which I believe is very good)
and NPN 514256P 45ns (this is the type you have, but 45ns).

EDIT: According to the datasheet you attached: The NPN 50ns has a "fast page read or write cycle time" of 33ns, the 45ns type has 30ns at this value.
So the Mosel seems better 😀

I bought the NN514256P from utsource.net. I checked my email, but I didn't see a specific vendor name from utsource. I paid $1.93/each for new pieces, which was the cheapest I could find at the time I was looking. A lot of those price offerings on utsource are fake or teasers.

CoffeeOne wrote on 2023-11-04, 15:10:

Mosel V53C104HP45 (with a "fast page mode read or write cycle time" of 25ns, which I believe is very good)

25ns cycle time means 40MHz cycle frequency, which means 80MHz MCLK is in spec, assuming all other timing requirements are also met at 80MHz.

CoffeeOne wrote on 2023-11-04, 15:10:

EDIT: According to the datasheet you attached: The NPN 50ns has a "fast page read or write cycle time" of 33ns, the 45ns type has 30ns at this value.
So the Mosel seems better 😀

This is 60MHz and 66MHz max in-spec MCLK.

feipoa wrote on 2023-11-03, 00:52:

Would you trust it for all datecodes, or just the datecodes from after that spec. sheet was produced? In my IC photo there are datecodes from 41-2000 to 50-2003. The date on that spec sheet is 11 April 2001. The memory with the white printing is from 2000, while the brown printing from 2002-2003.

I would hope that this applies to the whole series, the AS4C256K16F0.

feipoa wrote on 2023-11-03, 00:52:

Forcing 1024x768x16bit will require a modification to the BIOS. The chances of it working are slim, so not sure if mkarcher is still interested in providing this hack. I assume we'd need VCLK set to the max? 130 MHz.

Yeah, 1024x768x64k at the 60 Hz timing used for 1024x768x256 requires 130MHz VCLK. We might get down to 120MHz by reducing blanking (works on LCDs, but likely not on CRTs). At 130MHz, a MCLK strictly higher than 65MHz is required. At 120MHz, a MCLK strictly higher than 60MHz is required. As you confirmed 80MHz MCLK as operational on your card, preparing a patch starts to make sense.

mkarcher wrote on 2023-10-29, 18:46:

OK, on your card, memory clock is configured using JW1, JW2 and JW3 according to this table (C = closed, . = open): […]
Show full quote

CoffeeOne wrote on 2023-10-29, 17:54:

@mkarcher
I posted already an "ultra high resolution" picture here:
Re: 3 (+3 more) retro battle stations

OK, on your card, memory clock is configured using JW1, JW2 and JW3 according to this table (C = closed, . = open):

Copy code to clipboard

11 2 3
2. . . 50MHz
3. . C 80 MHz
4. C . 60 MHz (current configuration)
5. C C 70 MHz
6C . . 40 MHz
7C . C 65 MHz
8C C . 45 MHz
9C C C 55 MHz (feipoa's starting configuration)

It seems JW1/JW2/JW3 can be replaced by a 3x2 pin header to make memory clock jumperable. The only interesting alternate configuration is 65MHz which should improve performance at 56Hz and enable proper display at 60Hz (at an equally bad performance as you have at 56Hz now). 70MHz would provide a sensible amount of margin regarding memory bandwidth, but as the NEC µPD424256C-70 provides a page mode cycle time of 45ns, which seems to allow a maximum memory clock of 44MHz, getting stable operation at 65MHz requires a lot of luck, and 70 feels just way too high.

Thanks a lot for the table!
I have now a jumper block on JW1,2,3.
You were correct. 65MHz seems to work (but I only tested dos apps like Doom, Quake and Wolf3D for now).
Please note my new high score for Wolf 3D (previous high was 153,8) 😁

But at 70MHz nothing works, already in the boot screen there is mostly garbage visible.

So I will try next 8 times 45ns DRAM with a page mode cycle time of 25ns, 80MHz MCLK should be easy then.
Then I will make Windows tests.

mkarcher wrote on 2023-11-04, 23:24:

25ns cycle time means 40MHz cycle frequency, which means 80MHz MCLK is in spec, assuming all other timing requirements are also […]
Show full quote

CoffeeOne wrote on 2023-11-04, 15:10:

Mosel V53C104HP45 (with a "fast page mode read or write cycle time" of 25ns, which I believe is very good)

25ns cycle time means 40MHz cycle frequency, which means 80MHz MCLK is in spec, assuming all other timing requirements are also met at 80MHz.

CoffeeOne wrote on 2023-11-04, 15:10:

EDIT: According to the datasheet you attached: The NPN 50ns has a "fast page read or write cycle time" of 33ns, the 45ns type has 30ns at this value.
So the Mosel seems better 😀

This is 60MHz and 66MHz max in-spec MCLK.

I have an update. I used a donor card to get faster 256kx4 chips. The card now looks like that:

So OK, I just checked under dos that the card still works => everything was like before.
So I switched directly to 80MHz.
Still everythings seems to be OK, same picture quality as before.

Speedsys Video Throughput is now very high, jumped from +26000 kB/s (65 MHz) to +38000 kB/s.
Dos values are the same, a tiny increase in Doom, small increase in Wolf3D, now new record there.

But now to the bad side: No improvements in displaying the modes which are only available with 2 MB.

I tried first 800x600 24bit with 56Hz, same picture (distortions every 2 seconds ,when moving the mouse than more). The mode got much faster though: Wintune 2.0 is now ~ 3300kPixel/s before it was below 2000kPixel/s.
800x600 24 bit 60Hz similar as before not usable at all.

I tested then 1280x1024 256c @60Hz, this can be done without needing the Ark Tools:

So switching to 1280x1024, I have a good picture for one second.

Afterwards the colours change: Sometime into the blue, more often into the yellow. Then the pictures stays like that:

So what is the problem?
1) Card damaged during the soldering?
2) Second megabyte too slow, too bad?
3) The Ark 1491 is not really a 110MHz type and needs to be exchanged?

The most easy point to change for me is point 2. I can get Mosel 45ns 256kx16 ram chips easily. Even more easy would be using Edo chips, can it work? I have V53C16258HK35 chips.

EDIT / UPDATE: I tried the 35ns EDO RAMs, there are distortions also already in 1024x768 mode which is fine with the Fast Page mode ones.
Also tried more combinations of the 2 Rams for the second megabyte, because I have 4 of those in total. Behaviour does not get better: 1024x768 is fine, but colours change randomly in runtime in the 1280x1024 60Hz mode.
So I don't think it is point 2. Also I cannot find anymore the Mosel 45ns 256kx16 chips , ......

800 800 24 seems close. Can this be lcd incompatibility of sort ?

Try running MCLK at 65 or 70 MHz.

You may need a 110 MHz RAMDAC.

There's a Polish seller on eBay selling 35 ns ISSI branded SOJ FPM chips. I bought a few from him which arrived yesterday. I will test them out when my 45 ns DIP Mosel chips arrive.

feipoa wrote on 2023-11-18, 17:18:

Try running MCLK at 65 or 70 MHz.

You may need a 110 MHz RAMDAC.

There's a Polish seller on eBay selling 35 ns ISSI branded SOJ FPM chips. I bought a few from him which arrived yesterday. I will test them out when my 45 ns DIP Mosel chips arrive.

70MHz did not change something, I already tested it. 1280x1024 @60Hz gives bad colours. 800x600 24 bit @60Hz is more or less completely broken.

So do you think it is the second megabyte or the RAMDAC?

EDIT: But thanks for the hint, I bought such RAMs, too.

CoffeeOne wrote on 2023-11-18, 12:24:

So switching to 1280x1024, I have a good picture for one second. PXL_20231118_114446008.jpg
Afterwards the colours change: Sometime into the blue, more often into the yellow. Then the pictures stays like that: 1280second.jpg

This sounds like something stops working at that frequency when it gets warm. As the color changes on the whole screen, ...

CoffeeOne wrote on 2023-11-18, 12:24:

So what is the problem? 1) Card damaged during the soldering? 2) Second megabyte too slow, too bad? 3) The Ark 1491 is not reall […]
Show full quote

So what is the problem?
1) Card damaged during the soldering?
2) Second megabyte too slow, too bad?
3) The Ark 1491 is not really a 110MHz type and needs to be exchanged?

... I think #2 can be eliminated. Uniform bad colors in 256 color mode scream "RAMDAC issue", so the most likely reason is #3. If you have a small heatsink at hand, and the ARK1491 get hot (>50°C), you might try adding a heatsink. As 800*600 / 24bit at 60 Hz uses 120MHz pixel transfer clock (exceeding the 110MHz RAMDAC specs), the complete failure at that mode might indicate that there is absolutely no margin at 110MHz any more, again hinting towards issue #3. As your card still works perfectly at lower clocks, I don't think it is damaged.

CoffeeOne wrote on 2023-11-18, 12:24:

So I don't think it is point 2. Also I cannot find anymore the Mosel 45ns 256kx16 chips , ......

Which agrees with my expectation. Nevertheless, it was worth to test that hypothesis.

I ran a quick test with the 35 ns SOJ-40 FPM, IC41C16257-35k at
800x600x16M
1280x1024x256c.

I had no issue. Temperature of the AT&T 110 MHz rated RAMDAC is only 32 C, with a room temperate of 16 C.

>So switching to 1280x1024, I have a good picture for one second. PXL_20231118_114446008.jpg
>Afterwards the colours change: Sometime into the blue, more often into the yellow.

To make sure colors glitch inside DAC you would need to run another application, something fullscreen 256c that reprograms DAC. Then you can play with cooling DAC (compressed air) while running/quitting (or alt tabbing back-forth) to keep refreshing colormap.

feipoa wrote on 2023-11-19, 01:57:

I ran a quick test with the 35 ns SOJ-40 FPM, IC41C16257-35k at 800x600x16M 1280x1024x256c. […]
Show full quote

I ran a quick test with the 35 ns SOJ-40 FPM, IC41C16257-35k at
800x600x16M
1280x1024x256c.

I had no issue. Temperature of the AT&T 110 MHz rated RAMDAC is only 32 C, with a room temperate of 16 C.
IC41C16257-35K.JPG

The Ark1491 RAMDAC is cold to the touch, my room temperature is similar ~ 16°C, maybe less. It is in the basement, no heating there. It is cold outside, now 4°C.
Running 1280x1024 256c, the colors change a lot (on the whole screen), in the end everything is nearly unreadable.
But after switching back to 1024x768 256c, everything is fine again.
Would you be so kind and test 1280x1024 with the Ark1491 RAMDAC if it works (longer than 10 seconds)?

EDIT:
About the 800x600 24 bit mode. It is diffult to quit (go back to normal) mode. When I change back to 800x600 16bit in the Ark tools, it wants a reboot. In Reboot it hangs, then I have to power off. Afterwards it comes back again in 800x600 24bit..... Sometimes I need to start in safe mode, then confirm 640x480 16 colours, reboot. Then I can switch back to 800x600 16bit for example again reboot. Then I can use the Win98SE again. Pain in the ....
So for testing a 2MB mode 1280x1024 is much nicer.

One more question:

Would that be a compatible (and better!) RAMDAC for my ARK1000VL card?

CoffeeOne wrote on 2023-11-19, 18:42:

One more question:

Would that be a compatible (and better!) RAMDAC for my ARK1000VL card?

That chip is definitely better (up to 135MHz data clock, up to 110MHz pixel clock), and it seems mostly compatible. If you are experienced in soldering SOJ chips, a swap can make sense, but due to a lack of complete datasheet for the STG1700, I can't vouch for it. I found a STG1703 data sheet, and it seems SGS Thomson clearly intends these chips to be compatible with BIOSes targetting AT&T 20C490-like chips.

CoffeeOne, the easiest way to go from 800x600x24 back to 800x600x16 is to use the Windows Display properties page, this way you don't need to restart. The problem is that sometimes, it wants to default to interlaced mode, for which you need an LCD which will at least attempt to display the flashing desktop screen. With your eyes angry at you, then switch to 60 Hz. However, since 1280x1024x256c is fixed at 60 Hz, try switching to this mode using Windows Display Properties.

I also get hanging at shutdown when I try to shutdown the system at 800x600x24, and only on that mode. I've run into this issue with other cards when not using Windows drivers with Windows Display Properties.

I have the STG1702J on my Mach64 card. I'm attaching the pinouts for that chip, which are slightly different compared to the ATT20C490. Are they compatible?

I will try the ARK1491 RAMDAC at 1280x1024x256c within the next few days.