Ozzuneoj wrote on 2025-12-15, 03:43:

Straightened a bunch of pins on some "scrap" K6 processors I picked up recently and got to test a few out. Started with a couple […]
Show full quote

Straightened a bunch of pins on some "scrap" K6 processors I picked up recently and got to test a few out. Started with a couple of K6-2 500AFX, which are working great. Then I moved to a K6-2+ 550ACZ (2.0v) and a K6-III+ 500ANZ (1.8v), both of which seem to work fine at 600Mhz at 2.0v. It's so easy to overclock those chips to 600Mhz, and they run so ridiculously cool... Seriously. I have some cheap little heatsink+fan combos that I have rewired to run at 5v rather than 12v so they are very quiet. Even with this, I ran some programs on the K6-III+ at 600Mhz 2.0v, immediately shut the system down, pulled off the heatsink and the CPU heatspreader was cold to the touch, as was the underside between the pins.

Makes me want to overclock it to the moon... though I know they tend to hit their limit well before 650Mhz. I also tried running the K6 III+ at stock 1.8v at 550Mhz and it was stable while playing a game and doing some stuff in Windows 98. That seems really good. 😀

Also, the Soyo SY-5EMA+ v1.1 I have seems to top out at 6x100Mhz so I will need to try them in a different board to push them any further than that. Not that it is really necessary of course...

I think one of these would make an excellent mid level test system for parts since I will from time to time run into cards that need a slower system or just don't like running in a PII/PIII. Being able to flick a couple switches to go from 600Mhz to 200Mhz (100x2) or 166Mhz (83.3 x 2) is handy. Come to think of it, I wonder if the K6-III+ has the same issues running the NV1 that the PII\PIII have? The card actually runs games slower on Intel chips newer than a Pentium MMX I believe.

Now that you mentioned the 5EMA topping at that speed... I need to check my newly bought Tekram P5M3-A+ to see how high it goes. It claims 133FSB, and uses 686A southbridge surprisingly.

Tested 7 sticks of DDR1 1GB:

- 5 x Corsair: 4 100% Ok, one has issues: one chip has oxidized leads and looks like it took a blow.
- 1 x Adata: 100% Ok
- 1 x Wise: 100% OK

If I mix Corsair, Wise and Adata I get random errors... Old compatibility issues 🤣

Locutus wrote on 2025-12-16, 10:00:

Winter: when the sun disappears and the tanning bed steps in 😉

The attachment EPROM-eraser_LR.jpeg is no longer available

I used to erase EPROMs with sunlight. In a tropical country, it took between one week and one month to erase, depending on the EPROM and the weather (cloud cover).
Erasing more than 30 EPROMs of different types (24 of them for an "SSD") took me a month.
Then I got hold of a discarded UV toothbrush sanitizer, and now I erase them in about 25-30 minutes, 3 or 4 at a time.

Alien Legacy, 1994, floppy version.

I made a VGM pack for VGMRips.net here: https://vgmrips.net/forum/viewtopic.php?t=6329 it's submitted but it takes a while to get approved, if at all

I also went ahead and played it using VGMslap on my 486 with a SB16, and slapped that on YouTube. My little contribution for #DOScember:

https://www.youtube.com/watch?v=FfjENbxmWF0

If you are deeply familiar with the game, I'd love to attach a name to each of the unnamed tracks. I haven't gotten far into the game to immediately recognize those.

So my brother has an arcade machine at work that, well, doesn't work.

I took a look at it. It basically freezes whenever you launch a game. After I came around with an USB keyboard, we were able to check the BIOS. It's one of those off-the-shelf DIY arcade modules from china. Shady windows 2000 + emulator + hardware on top of it.

Well it turns out, the cooler mounting is broken so that Pentium 4 2.4GHz is getting no cooling at all. And it doesn't like that. Luckily I have another S478 board in the lab somewhere, where I will just steal the cooler bracket. With forced cooling, everything works fine.

While I'm at it I'll probably create an image of that 40GB spinning rust seagate drive, before it leaves us for good.
The board is quite interesting. It somewhat looks way too modern in some parts for S478 and appears to be stripped down purposefully. So perhaps a custom jobby.

Not my normal field of work, but hey, it's a fun exercise.

Received a Sapphire X800GTO AGP (R430), only to find out it has a Rialto chip.

Spent a while to source a heatsink to paste onto that little bugger and now it's running sweet.

Now, to test whether it's gonna be happy with me old KG7-RAID and its pimped Win2k Pro install on an adapted 74GB SATA WD Raptor. My main concern is whether it'll be kind to the already installed Catalyst 6.5 driver (I previously used that MB with a X800XT from @Socket3, which was native AGP) and if I don't have to do any other hijinks to get it running.

PcBytes wrote on 2025-12-17, 19:23:

Received a Sapphire X800GTO AGP (R430), only to find out it has a Rialto chip.

Spent a while to source a heatsink to paste onto that little bugger and now it's running sweet.

Where does that urge from everybody to cool the rialto bridge come from?
It had no cooler per default and with a proper airflow usually lives just happy without one.

I was testing some S3 Trio64V+ cards yesterday and started to notice something very confusing that now has my mind boggled. I'm hoping someone else has experienced this as well or has a possible answer.

I experienced this on two totally different systems (440BX + 850Mhz PIII; MVP3 + K6-III+ 550). All cards are run at 800x600x32bit and have been run on a CRT as well as an LCD that can do 75Hz.

Basically, what I'm noticing is that there is some very inconsistent performance with regard to desktop usage in Windows 98SE. Particularly when dragging a window around (with contents shown while dragging). On some cards I can flick the window around as fast as I can and it just moves perfectly. On other cards I get a big trail behind the window and sometimes the icons disappear briefly as they are uncovered by the window. The behavior also changes after rebooting sometimes, though some cards are clearly affected by it more than others. On some cards even just moving the mouse cursor around is clearly faster and more responsive than on others, and this is at either 60Hz or 75Hz. I am using PS2Rate to set my PS/2 Port to 200Hz to get smoother mouse movement, so it is very obvious when one card is slower than another.

Swapping BIOS chips from one card to another does not cause the behavior to change so it is something about the card itself that is causing this.

I thought it might be a memory thing, but I'm a bit confused about that too since there are cards on both sides of the issue that have all EDO chips (I have checked the datesheets for those too). After reading the Trio64V+ datasheet it seems that there are three possible memory configurations for these cards. It says: Supports standard fast page mode and EDO DRAMs (60 MHz) and 1-cycle EDO DRAMs (50 MHz)

I'll be honest, I never even knew about this faster single-cycle mode of EDO. What exactly enables or disables this feature? It seems like some cards have it and some do not.

Confusingly, some of my cards that are faster have a BIOS chip marked with -60 and others -50. I would have thought this was indicating the memory speed applied by the BIOS... which normally would make the -60 better, but if the single-cycle EDO cards are 50Mhz, then maybe those are better.

TL;DR:
Sorry for the confusing post... this is just a mind-bending thing to discover. If someone would have asked me, I'd have said that every Trio64V+ would give the same solid compatibility, decent performance and be totally boring otherwise... but I have a couple cards here that feel more like using a Voodoo 3 or Matrox card on the desktop, where as the others are noticeably sluggish. If anyone has any explanation for what is happening here, or something specific I can check (differences between cards, etc.) please let me know. If it seems like there's a deeper rabbit hole to explore here, then I will make a dedicated thread for this where I'll post pictures of all of the cards and try to document which ones are faster.

Also, is there any kind of Windows 9x desktop graphical benchmark I can use to try to quantify this difference? Something that will just run at whatever resolution and color depth I have selected and give me a performance number that might reflect desktop smoothness or responsiveness? That'd be way better than me just flicking the mouse around and going "huh". 😅

Do you have a desktop wallpaper set on some systems but not others? I noticed with the VESA drivers that some operations are thousands of times slower when drawing over the desktop wallpaper than over another window or a blank desktop. The WELCOME.EXE fade-in is especially bad, drawing over the "Inside Your Computer" wallpaper it takes a couple of minutes while over a solid color it's done in less than a second.

You could try the CrystalMark Retro benchmark, it includes a section of 2D GDI tests. I think PCMark 2002 also has a 2D GDI test.

onethirdxcubed wrote on 2025-12-17, 23:57:

Do you have a desktop wallpaper set on some systems but not others? I noticed with the VESA drivers that some operations are thousands of times slower when drawing over the desktop wallpaper than over another window or a blank desktop. The WELCOME.EXE fade-in is especially bad, drawing over the "Inside Your Computer" wallpaper it takes a couple of minutes while over a solid color it's done in less than a second.

You could try the CrystalMark Retro benchmark, it includes a section of 2D GDI tests. I think PCMark 2002 also has a 2D GDI test.

No, the desktop is always the default blank desktop with the same few icons on it. And, I can swap cards back and forth on the same system and some are clearly much faster than others.

I will try the benchmarks you mentioned to see if they show any measurable differences. Thanks!

One thing I noticed is that it almost feels like vsync is off on the cards that are faster... but I'm not sure how that would be possible with the exact same drivers (they don't even reinstall... they just work after swapping cards) on the same system, and even with the same BIOS chip swapped from one card to another.

And the X800GTO is happily running. The bridge chip really was getting at uncomfortable levels of heat, but yeah to each his own.

I will change the platform it runs on tho - I think the XP-M 2400+ on there is quite a bottleneck for it.

The Dell Pentium 3 ATI strong to run Half Life 1 arrived.

Ozzuneoj wrote on 2025-12-17, 21:12:

I was testing some S3 Trio64V+ cards yesterday and started to notice something very confusing that now has my mind boggled. I'm […]
Show full quote

I was testing some S3 Trio64V+ cards yesterday and started to notice something very confusing that now has my mind boggled. I'm hoping someone else has experienced this as well or has a possible answer.

I experienced this on two totally different systems (440BX + 850Mhz PIII; MVP3 + K6-III+ 550). All cards are run at 800x600x32bit and have been run on a CRT as well as an LCD that can do 75Hz.

Basically, what I'm noticing is that there is some very inconsistent performance with regard to desktop usage in Windows 98SE. Particularly when dragging a window around (with contents shown while dragging). On some cards I can flick the window around as fast as I can and it just moves perfectly. On other cards I get a big trail behind the window and sometimes the icons disappear briefly as they are uncovered by the window. The behavior also changes after rebooting sometimes, though some cards are clearly affected by it more than others. On some cards even just moving the mouse cursor around is clearly faster and more responsive than on others, and this is at either 60Hz or 75Hz. I am using PS2Rate to set my PS/2 Port to 200Hz to get smoother mouse movement, so it is very obvious when one card is slower than another.

Swapping BIOS chips from one card to another does not cause the behavior to change so it is something about the card itself that is causing this.

I thought it might be a memory thing, but I'm a bit confused about that too since there are cards on both sides of the issue that have all EDO chips (I have checked the datesheets for those too). After reading the Trio64V+ datasheet it seems that there are three possible memory configurations for these cards. It says: Supports standard fast page mode and EDO DRAMs (60 MHz) and 1-cycle EDO DRAMs (50 MHz)

I'll be honest, I never even knew about this faster single-cycle mode of EDO. What exactly enables or disables this feature? It seems like some cards have it and some do not.

Confusingly, some of my cards that are faster have a BIOS chip marked with -60 and others -50. I would have thought this was indicating the memory speed applied by the BIOS... which normally would make the -60 better, but if the single-cycle EDO cards are 50Mhz, then maybe those are better.

TL;DR:
Sorry for the confusing post... this is just a mind-bending thing to discover. If someone would have asked me, I'd have said that every Trio64V+ would give the same solid compatibility, decent performance and be totally boring otherwise... but I have a couple cards here that feel more like using a Voodoo 3 or Matrox card on the desktop, where as the others are noticeably sluggish. If anyone has any explanation for what is happening here, or something specific I can check (differences between cards, etc.) please let me know. If it seems like there's a deeper rabbit hole to explore here, then I will make a dedicated thread for this where I'll post pictures of all of the cards and try to document which ones are faster.

Also, is there any kind of Windows 9x desktop graphical benchmark I can use to try to quantify this difference? Something that will just run at whatever resolution and color depth I have selected and give me a performance number that might reflect desktop smoothness or responsiveness? That'd be way better than me just flicking the mouse around and going "huh". 😅

It deserves a separate thread. How much VRAM does each S3 Trio have? Do they each have the same device ID - which drivers is Win98 choosing for each device ID?

Ozzuneoj wrote on Yesterday, 09:54:

No, the desktop is always the default blank desktop with the same few icons on it. And, I can swap cards back and forth on the same system and some are clearly much faster than others.

I will try the benchmarks you mentioned to see if they show any measurable differences. Thanks!

One thing I noticed is that it almost feels like vsync is off on the cards that are faster... but I'm not sure how that would be possible with the exact same drivers (they don't even reinstall... they just work after swapping cards) on the same system, and even with the same BIOS chip swapped from one card to another.

You could check the PCI revision ID on the GPUs, the driver inf supports revisions from 0x40 to 0x5F of the S3 86C7675 chip but it doesn't set any Registry keys for different behavior. (It does add some more video modes for the Trio64V2/DX which is the 86C775.) You could also check the manufacture date code. There's a full version of the S3 Virge driver in the Windows 98 DDK if you can find that somewhere and that's probably identical in the 2D part.

onethirdxcubed wrote on Yesterday, 21:36:

You could check the PCI revision ID on the GPUs, the driver inf supports revisions from 0x40 to 0x5F of the S3 86C7675 chip but it doesn't set any Registry keys for different behavior. (It does add some more video modes for the Trio64V2/DX which is the 86C775.) You could also check the manufacture date code. There's a full version of the S3 Virge driver in the Windows 98 DDK if you can find that somewhere and that's probably identical in the 2D part.

MattRocks wrote on Yesterday, 21:28:

It deserves a separate thread. How much VRAM does each S3 Trio have? Do they each have the same device ID - which drivers is Win98 choosing for each device ID?

Thanks for the input guys. I ended up making a thread for it, which you can check out here.

Re: Some Trio64V+ Cards significantly faster in Windows 98 than others?

Through some datasheet sleuthing Weedeewee discovered that one of the two cards I posted about was indeed running in 2-cycle EDO mode, while the other was running in 1-cycle EDO mode, which is significantly faster. Normally, the 1-cycle mode cards are supposed to run the RAM at 50Mhz, but the fastest card I have is also running the RAM at 60Mhz in 1-cycle mode. So it is a pretty potent combo and is exceptionally fast for a Trio64V+. There are a couple of pins on the Trio64V+ chip that will either have resistors on them to enable 1-cycle, or will have no resistors to leave the chip in 2-cycle. The card's BIOS will determine the memory clock speed, which is why they are often marked with -60 or -50. Confusingly, they tend to have RAM chips marked as 60ns or 50ns, but this is completely unrelated to the 60Mhz or 50Mhz setting applied by the BIOS. It is worth mentioning that it seems like the cards cannot be made to run at 50Mhz in 2-cycle mode, or at the very least the performance doesn't drop any further than 60Mhz 2-cycle mode.

As crazy as it sounds, in the Wintune97 video benchmark the 1-cycle 60Mhz card I have scores 24MP (whatever that means) while the 2-cycle 60Mhz cards score 7.8MP... so at least on this setup in this situation, at this resolution (800x600x32bit at 85Hz) the 1-cycle card is a full 3 TIMES faster than any 2-cycle card. It's a bit premature to make too many assumptions, but it seems like all that is needed for this performance gain is to add a couple resistors, assuming the RAM chips are fast enough to handle 1-cycle mode (I'm not sure exactly how fast they need to be). The chips on my fastest card are marked as 45ns, which is apparently fast enough for 1-cycle at 60Mhz.

Anyway, there is more to read in the thread there too so I'd recommend checking it out.

All in all, a really interesting discovery. These cards are super popular but I have never heard any discussion about this feature that can take a card from being barely adequate to super fast.

Finally found a Socket A motherboard with ISA slot at my local recycling center.

A Biostar M7VKD.

Being a 2000's board all capacitors around the VRMs were bloated and leaking, so I replaced them all.

After that the board works perfectly, tested with a few regular Athlons and to my surprise it also works and recognizes Athlon XP Palominos.

With thoroughbreads and Bartons it displays "Unknown CPU" at boot splash screen, but works perfectly as well.

FSB 133Mhz works great, since this board has a KT133A.

Now I will try to mod one of my Athlon XPs into a Athlon XP-M, will be a test to my old eyes.