pshipkov wrote on 2022-09-07, 00:58:

if you have eeprom programming device it will be trivial to read the BIOS, but it does not sound like you have one. […]
Show full quote

if you have eeprom programming device it will be trivial to read the BIOS, but it does not sound like you have one.

small utilities like NSSI or checkit v3 can read the video card bios and write it to file on disk.

---

i didnt look at 4x50MHz for the M919 motherboard. 3x66 is a no go which demoralized me for further testing.

i thought the PCI bus divider happens past 50MHz, no ?

Yes I have an XGecu Pro, don’t I have to tell it the exact chip I’m using? Just want to be extra careful, the steps I take, before I damage the rom on accident.

Honestly not sure when the divider is forced on. Was thinking it was anything above 33?

Hello pshipkov and Chadti99,

I've been following your guys' work for some time now and it is quite impressive. Talk about taking things to the next level!

I've been meaning to contribute here, but my highest performance 486 boards at the moment consist of the APC Predator 747 w/ Chips4041 chipset, which has given me the fastest scores I've seen in VLB benchmarks but has not quite proven easy to work with, being an industrial board, and the Nice SUPER EISA, which, while excellent, is older and finnicky. I've also got an APC Predator I Plus with the Symphony Haydn II chipset, like pshipkov's excellent DTK 386 board, and it has great performance but is, too, unfortunately limited by its status as an industrial board in ways that impede performance (no L2 > 256k, etc.).

Some day I'd like to find a VL/I-486SV2GX4 and join in on the fun with y'all! I've really taken to this board after your guys' work with it!

@pshipkov
Sorry to butt in on the conversation, but I just wanted to offer a BIOS for the MB-1212C that came with my board. I've attached it to this post for your convenience. It's a Chips & Technologies BIOS, seemingly OEM, and it is much better than the AMI BIOS otherwise available for the board. I have been able to run my board at 16MHz, and cannot figure out why yours was unable to do so. Perhaps this BIOS can offer some assistance? I'd be curious to hear your thoughts.

@Chadti99
I saw your VLB thread and have to say that I am quite impressed with your work in the area! These benchmarks here are interesting; I have a Paradise Barbados64 S3 Vision864 card and a Miro20SD S3 Vision868 card; having briefly benched both, I can say that the Miro is slower. I find it interesting that your Vision868 is slower than your Trio64; I wonder if memory clocks were more conservative? As far as I understood, the Trio64 is basically an 864 with all components integrated for cost-savings, and the 868 is an 864 without CGA/EGA backwards compatibility and with some basic video scaling capabilities. It seems that this may be something worth looking into, as far as to what differentiates the Trio64 and S3 Vision864/868. Have you messed with memory clocks at all with MCLK to see if that changes things on the Vision868?

Sincerely,
Will

@chadti99
since you are reading from the eeprom there is no chance to damage the microcodes.
usually these chips are model m27c512 by ST, AMD, or Intel.

i have a memory for the divider after 50mhz.
which is confirmed by your recent test.

@WJG6260
Good to hear you enjoy the "casual facts finding" soap opera going on here.
Feel free to contribute episodes to it as you see. It is a free-style thread anyway.
On a related note - your youtube soap opera is great too.

These two motherboards look real tight.
Being industrial grade certainly means OC-limited, but if you benchmarked them please share the results and any interesting details.
I have a soft spot for these Symphony based assemblies. Most of them are not great, but the few that "click" are in a class of their own.

Going to check your MB-1212C BIOS tonight. Mobo happens to be on the bench so time is right.
One more question - did you have problems with CFs or HDDs through standard IDE adapters on this board ?
POST may not complete often.

About your Vision cards - are you talking about DOS graphics, or Windows accelerated GUI, or both ?

I think i saw the other day on ebay some Asus VLI motherboards, but may be wrong, and with their prices these days is more like an investment than hobby shopping.

Definitely! This thread has been most interesting and elucidating, to say the least. I have learned quite a bit from the experiments done here, and really appreciate the work that you all have been putting in. Thank you for having me here!

As for YouTube, I really appreciate it. I haven’t been on there in a while due to some health issues, but I plan to get back sometime soon. In the meanwhile, I wholly intend to stay active tinkering and learning, and it seems like this thread is certainly the right place to learn quite a few neat things.

Regarding the boards, I have the Chips one laying around; I still need to extract the other Symphony one from its case because I plan on swapping to the SuperEISA temporarily, at least. The Symphony boards really are interesting; I find it odd that the later 491 chipset from them is slower, as shown by your tests with the Young Microsystems boards. This particular Symphony board is quite well-rounded and very fast. I am unhappy with industrial BIOS on there though, as it does not work well—or at all—with 486DLC/SXL chips, despite 486DLC being silkscreened on the board as a jumper option. It is most troubling. The MR-BIOS does not work there either, unfortunately. I’ve tried everything up to hardware modification, but I am confident that it is cache coherency issues. Cyrix.exe and other associated utilities don’t help much.

The Chips one is a real beast. It has a full VLB 2.0 implementation. Of course, no card can utilize such, but the onboard EIDE controller should also conform to VLB 2.0 and work at 50MHz. Unfortunately I cannot find any drivers for it and it seems none were provided. Given that this is the only board to ever sport this chipset, at least that has so far been discovered, I’d say the odds are against me there. The SpeedSys memory stats are surprisingly unimpressive, but graphics performance is unparalleled. Other benchmarks like CacheChk agree that this board is special as well. Unfortunately the board only accepts 256k of L2 cache, and I am working to determine if this can be bypassed as 10 DIP-32 sockets are already onboard.

I’d be interested to hear your thoughts regarding the MB-1212C BIOS. I haven’t pulled out the board in a while, but I do have some notes in my files mentioning IDE pickiness. I recall using a CF card w/ an ISA caching controller that has its own BIOS (and ironically equipped with a Harris 286-20) to handle the work.

Regarding the S3 Visions, I am talking mainly DOS. I am planning on setting up one machine to bench all of my VLB cards with a 160MHz Am5x86 as a “standard” benchmark at 40MHz, and ditto at 133MHz for a 33Mhz FSB.

Truth be told my Windows benchmarks are lacking. I usually use WinBench 4.0 and WinTach v1.0 for my testing. I can’t decide if Windows 95 or Windows 3.1 is more appropriate; I am leaning toward the latter. Any advice on that front? I’d love to hear your thoughts.

I saw those boards too. Seems like eBay pricing is worse by the day, unfortunately. I have been waiting for one of those, as well as a Mach64 VRAM ISA. Maybe by some dumb luck I’ll find them 🤣

You said your SuperEISA was a bit hard to work with. Can you explain?

pshipkov, the m919 definately starts employing the 1/2*PCI divisor at 40 Mhz, not 50 Mhz.

Anonymous Coward wrote on 2022-09-07, 06:10:

You said your SuperEISA was a bit hard to work with. Can you explain?

Absolutely!

I was quite fortunate and initially had two SuperEISAs. I traded the more commonplace one that’s like yours with a ZIF socket and removable/replaceable clock gen, and kept the older version, which uses a LIF socket and lacks Pentium OverDrive support. I don’t mind that, nor do I mind the fact that the board uses oscillator crystals; that’s fine by me, since it’s infinitely flexible.

My main gripe with the SuperEISA is that it just lacks a BIOS newer than v1.21, and the 06/06/92 AMI core is cranky sometimes; its DX4 support is nonexistent (although they work fine, so that’s good), and its Cyrix 5x86 support is also limited, as I’ve both seen firsthand and witnessed via your posts. I need to get ahold of some ACARD adapters to switch my equipment over to SCSI, but using my EISA IDE controller with the stock BIOS for now has been annoying and necessitates the XT-IDE BIOS. I can get over that, because IDE is not ideal compared to SCSI.

It’s a beastly fast board and really full of performance and customization options, but some things about it have never made sense to me: namely, the board can’t take 64k x 8 SRAMs for 512k L2, but it seems fine with 128k x 8 ones? I am not complaining, but more or less confused. I also wish that 16MB 30-pin SIMMs were more cheaply and easily found, but I digress.

I suppose the key limiting factor (or benefit?) of the SuperEISA is its one VLB slot, which doesn’t work 100% with some of my VLB cards. I know for a fact that my Cardex Challenger ET4000w32p is not a good match for it, and my S3 Trio32 is hit-and-miss. I would ideally like in a 486 the ability to test a couple of VLB cards, and I really would like EISA. Something like your Tyan S1437 would be great, but it’s hard to find another EISA/VLB board these days. I wouldn’t mind even testing something with the later OPTi 691/696 chipset, as that might support L1 WB, from what it seems. That’s another feature it’s hard not to want.

Overall, I’d say I really like the board but I guess I want more expandability out of it? It’s definitely the most badass 486 board I’ve tested yet though.

@feipoa
Checked my previous posts about M919 and i say there 40MHz as well. So you are probably right. Just wonder why Chad is not seeing perf diff when switching from 20 to 50 MHz upon BOOT - worth checking that.

---
---
---

@WJG6260

286
---
The AWARD BIOS for MB-1212C provided an important hint for the IDE issues that we have seen with that motherboard.
That BIOS allows user to enter up to 1024 cylinders, while the AMI BIOS supports much higher number in UI, but the hardware seems to be actually limited to 1024 only.
I use 2Gb Transcend CF card which translates to 3884 cylinders and that causes problems.
Switched to small volume mechanical HDD and things were fine.
Otherwise XT-IDE adapter is the way to go for storage devices with bigger size.
Thanks for bringing it up.
(updated the original post above with this info)

It looks like the assembly here didn't age as well as yours. It refuses to go past the default 12MHz.
Will be good if you can try go the 16MHz you are currently running yours at - a soft request.

486 Symphony
------------
The situation with early/late Symphony chipsets is similar to VLSI:
VLSI-100 - ok at best.
VLSI-200 - the monster.
VLSI-300 - the clunker.

I also noticed that some of the Symphony based boards don't fully support the advertised set of capabilities.
Usually smaller electronic manufacturing companies picked that chipset, so maybe they didn't have the resources to complete all specifications - especially the less mainstream ones like DLC processors.

486 Chips
--------
Yeah, SpeedSys memory metrics are useless - they mean exactly nothing about the qualities of a given system. This can be extended to all synthetic tests really. : )
But i am even more curious now about that motherboard.
Over time i limited my search criteria for 486 boards by excluding industrial ones and mostly excluding 256Kb L2 cache ones. So that Predator guy went under the radar.
If you find the time - please share some perf numbers for it. You probably noticed we use here 3 set of tests:
DOS interactive graphics - mostly Phil's benchmark suite + Wolf3D
Windows accelerated GUI - mostly WinTune2 (draws bunch of primitives on screen, probably not very accurate, but is quick, and also Windows GUI perf kind of does not matter much for 99% of the cards)
2-3 more heavy computation tests - some 3D rendering apps that are very sensitive to system instabilities.

---

My Windoz matrix (with WinTune2):
For 286, 386 and 486 systems running at 160MHz i use Win3.1.
For late 486 systems running at 180/200 MHz - Win95.
For Pentium 2/3 - WinNT4.

---

S3 Vision cards, especially the higher versions are rare, so not a lot of structured info about them.
Anything you can share will be a plus.

Is the symphony 461 the VLSI 200 analog here?

@pshipkov

Re: 286
—————
Interesting developments indeed. It’s a shame that your board seems stuck at 12MHz. I’ll dig mine out and see where I can go with an overclock to 16Mhz, but certainly it seems like the AWARD BIOS is the better option for starting. The AMI BIOS limitations do not surprise. The 6/6/92 HI-COLOR core seems a bit more robust; a shame no such BIOS exists for this board.

Re: Symphony 486
———————————
Agreed on all fronts. I am troubled by the lack of DLC support. It frustrates to see the silkscreen referencing the DLC, and yet the board failing to POST with it. I do think your DTK board is of better quality, as far as implementation goes. The Symphony chipset’s VLB implementation is a bit odd too. I have noticed some oddities in utilities like X-VESA, whereby anything over 16-bit reads and writes occur at the same speed. This is troubling, as the cards tested (ET4000/W32p, S3 864, S3 Trio64) are all 32-bit host bus cards and should not have this limitation like the GD-542x cards. Perhaps there is more to this, as UNIVBE overcomes this, at least somewhat, and increases performance across the range. Maybe X-VESA is a faulty benchmark?

@ maxtherabbit - I believe so, as the VLSI 200 is quite performant, just as is the Symphony Haydn II, and both of their successors were not great.

Re: Chips
—————
Synthetic tests certainly don’t tell the story, but there is undoubtedly some usefulness in being able to compare numbers. It is unfortunate that SpeedSys is quirky as it is.

The F4041 data sheet indicates support for 1MB of L2. I do not understand why the manufacturer would cripple it like so, but I have not tried 128k x 8 SRAMs, only 64k x 8s. I will have to try the 128k x 8 chips and see. I have attached the data sheet, if you’re curious at all.

I will absolutely do so! Which Wolf3D benchmark do you use? I can use WinTune2 and see how that board performs. I’ll take some notes as I have been amassing a number of VLB cards and would be cirrus to see how they compare.

Regarding the rendering tests, are there any standard renders built into the software, or do you run your own routines?

———
I think I may eventually run benchmarks in Windows 3.1 and Windows 95, but for now will work on 3.1 so we can all compare our work.

Re: Vision 86x cards
————————————
I can get to work and standardize my findings against a Trio64.

Interestingly, my Diamond Stealth 64 DRAM T is identified as a Vision864 by SpeedSys (as aforementioned, not reliable, but still interesting) on the Chips board. I am uncertain why; it has BIOS v2.02. This may lend to the theory/understanding that the Trio32/64 is just an integrated 864 and Trio64V+ is an integrated 868.

If anyone here has any of the few VLB Trio64V+ cards, does/do theirs compare favorably against the Trio64 VLB or are they pretty much the same numbers-wise? I am curious because my 868 seems slower (slightly) than my 864, but I suspect memory frequency/timings is at play here since both cards have SIEMENS HYB51417BJ-60 chips (possibly EDO, as the pinouts match other EDO chips).

Does the Trio64 have a 32-bit datapath to the RAMDAC? The Vision86x cards I’ve tested are 16-bit and, perhaps, that might lend to whether there’s slight difference in Windows GUI and GDI performance?

Sounds good about the MB-1212C.

Didn't spent much time on tweaking VESA modes with the usual utils myself.
Like many others, i mess around with that stuff every once in a while, but never had a case where it is actually a factor.
Maybe some specific DOS games or apps that require external intervention ? I don't know. Do you ?
My experience was more towards handling them programatically.

Datasheet of the chipset clearly indicates support for 512Kb and 1024Kb, but is not clear if the PCB implements it.
Yet it sports DIP-32 sockets which is suspicious. I only skimmed through the document. Need to better comprehend it really.

Testing
1. You can download this archive: https://www.petershipkov.com/temp/retro_pc_im … /test_suite.zip and copy it to C:\
It contains Phil's benchmark tools. Bunch of other benchmarking stuff. Wolf3D test is located at c:/mrk/bench/3d/wolf3d. Run test.bat there. We capture readings from second or third run.
2. There is 3D Studio R3 for DOS in c:\a\3dsr3. Start it with 3ds.exe. Load the "chevy" asset. Press CTRL+R. Then press Enter. Let it think about the situation. Compute time is displayed at the bottom/middle of the screen in seconds and hours:minutes:seconds.
3. Windows 3.1 is located at c:\win. Default video driver is VGA for compatibility.
4. WinTune2 is located at c:\mrk\wt2. Start it. Run it. It will complete shortly after that. Click on details. Click on "video" - it will display the KPixels per second number there.
5. Start LightWave3D's Layout application. Load space\blade.lws. Press "render" at the bottom-left of the screen. Set both start and end frame to 50. Press "render". Let it think for a while - can be long. Compute time is displayed in the same dialog upon completion. This test is very sensitive.

About these early S3 chipsets - there is information scattered around but never saw a full matrix in one place, so always have to go search online about this stuff only to forget it shortly after. :}

@ph4nt0m

Swapped the L2 cache buffers on one of the LuckyStar motherboards with the 0-wait state ones from TI.
No apparent difference in stability/performance at 3x60 and 3x66.
Doesn't allow for no-PCI-bus divider at 3x66, which was expected since the issues are unrelated to cache and memory but the PCI bus.

I am under the impression that the difference w/ respect to the VESA modes must be something of a fluke. I cannot for the life of me understand these APC boards sometimes; I constantly wonder if it's an implementation quirk with respect to the chipset or whether it's just the chipset itself, but there's something about the VLB on that board that, while fast, is just funky. I will have to test further and report back.

Only in Quake have I ever seen any increase in scores with VESA modes; usually loading the S3VBE on S3 86x and Trio cards seems to give me an extra 0.1 or so FPS. It is variable, but overall I cannot say if it's a factor either. I do not know of any DOS applications that tend to need UNVIBE/VESA drivers, but there must be a few. I have a strange ISA card based on the IIT AGX014 which lacks a VESA implementation in the BIOS and it is not happy with certain video modes, including non-VESA ones, until I load a TSR. Strange stuff indeed, but I admit readily that my knowledge is very limited in this area. Programmatic has been my experience as well.

I will try 1024kb of cache later this evening and report back. I recently ordered some "10ns" cache from China and most of the chips tested good. It works for certain at the fastest timings on the SuperEISA.

Thank you for the suite! This is awesome. I will make certain to put this to good work. I am going to put the Chips board on the bench for further inspection in a bit.

Agreed regarding the S3 chipsets. It is a pain to find much; there are some weird things about them, going back to even the 801/805; e.g., I have an S3 805i-based ISA card, but the 801 is the model that was typically seen on ISA. There are discrepancies into the width of the host bus connection on the 80x series; is it 16-bit, or 32-bit? Depends who we ask. It is very fun uncovering and exploring this information, that's for sure! I cannot wait to report back and add to the findings you all have here so far! 😀

IIT AGX014 ISA VGA you say ?
Been wondering about this video card for a while now. Was even thinking to obtain one from somewhere. But from your notes it sounds like waste of money.
Apart from the missing VESA modes in the BIOS do you have any impressions from its performance characteristics in the context of interactive DOS graphics ?

Yes, Q1 can benefit slightly from VESA modes. There is a console command that somebody pointed at in another thread. It establishes some VESA mode that bumps perf tiny bit.

Funky assemblies like some PC-Chips stuff or maybe these APC boards can be a great pass-time toys.
Let's see what comes-out from the benchmarks ...

A while ago I ran some benchmarks on the AGX014 on my Symphony board with a DX2-66. I normalized the results to a GD5426 VLB, as the WD Paradise Accelerator Pro box amusingly says "faster than some VL-BUS accelerators." I was skeptical, to say the least.

A few erratum of note:

• This card only has 1MB of RAM installed at the moment. I actually ordered a second MB of 80ns dual ported ZIP VRAM, and I am not sure as to whether the memory bus is halved or not.
• This card is definitely faster than an average ISA card in Windows; the NCR77C22E isn't far behind, however. In DOS, it is nothing special.
• I do not have the Windows 3.1 drivers for this card. It is not cross-compatible with the Hercules Graphite divers; while using the same chip, the Graphite drivers appear vendor-locked, hence my use of Win95.

The AGX015 might be faster, but my guess is marginal improvement on ISA. VLB is probably where this card shines, but by the standards of faster, later VLB cards, it's certainly old and outdated. AGX016 is PCI-only as far as I am aware, and it only has NT drivers. For a card based on the XGA, it's pretty impressive.

I've been wondering if VESA modes in Quake are "cheating" for performance, as far as benchmarks go. Do you all use them at all?

Agreed. I am tinkering with your benchmark setup now on the SuperEISA since it's on my desk. It seems the LightWave files are missing, but I can reinstall as I have the disc. Actually, I am currently running the 3dsmax benchmark for grins. I will be grabbing the Chips board and beginning work soon. Will keep you posted.

Looks like I have some catching up to do.

Give this a spin on your Trio64 VLB cards.

@Chadti99
Thanks for the BIOS.
Going to spin it tomorrow.

@WJG6260
Good info.
You saved me $50 (or even more $ ) by providing the details for the IIT card. Looks like a decent ISA Windows GUI accelerator. Lacking in DOS graphics.
Also, the details for NCR 77C22E are very welcome too. Been wondering about that one as well. Looks like a good silicon - on par with ET4000AX in DOS, but 2x better in Windows GUI. Anything else you can share about it ?
I don't know if you have the capability to test ISA/VLB video cards on a 4x40MHz system, for conformity. If not - not a big deal at all.

Edit: the more i look at the numbers of the NCR card the more i like them.
We know that the fastest ISA video cards are based on Cirrus Logic GD-5434 with 2Mb RAM.
They put in the pocket ET4000AX and W32i/p.
Eyeball-refitting the known ratio between et4000ax and gd-5434 and rescaling that with the shown by you delta between et4000ax and ncr - it seems that NCR is close to 5434, in DOS graphics at least.

There is a mutual agreement around here to present results with default Phil Benchmark settings only. So it is apples to apples.

About the archive i provided. If something is not ticking properly - let's fix it. I thought this is a decent mirror, but maybe something got corrupted. Will verify tomorrow and reupload.
Btw, do i sense the offline 3D graphics force over there ? (noticed inserts like "i have the LW3D installation CD. Running 3dsmax.") : )

@Chadti99
Awesome! I want to see how this differs from the Diamond BIOS v2.02.
I’ve seen a couple of those #9s out there. Very nice cards. How’s the image quality? They’re always top of the game in PCI-land.

@pshipkov
Glad to hear! I am considering an IIT AGX015 for testing, but the VLB version is quite hard to come by and the ISA version is usually capped at 2MB of RAM rather than 6MB max for the VLB version.
Agreed, it is lacking in DOS. But, I would like to add that it is impressive for a 16-bit core. In Windows it is very efficient, even with 1MB of RAM.
I think it’s worth retesting when the 1MB of ZIPs comes in. Will keep you posted.

The NCR 77C22E is excellent, but quirky.
I have a Boca SVGAN1, which is capped at a meager 1MB of RAM. The card requires a TSR for color in DOS otherwise it operates in monochrome in text mode. I am not certain as to why, but it does work in color in DOS graphics and in Windows. It will automatically switch to color in graphical situations. There are 2MB models out there, and a few VLB ones. The VLB versions are unfortunately crippled with a RAMDAC (or perhaps driver?) that only allows 256 colors. The card is very fast out of the box and very easy to work with. It is stable at up to ~12MHz ISA. I would rate it on par with the ET4000AX, and in many ways prefer it for Windows. It seems to be a beefy design, especially for something dating to 1990!

I am really trying to join you all and test ISA/VLB at 4x40MHz.
That leads me to an update on the Chips board. Rudimentary tests revealed two things: the board does not work with >256k of L2 without modification (perhaps this can be done via bodges?) and the board is somewhat crippled by the BIOS.
At 40MHz, I cannot tighten timings with the default -15ns cache. I plan on replacing the set with 10ns cache and tightening timings for further testing.
In the meanwhile, I have ordered a couple of 40MHz oscillators to bring out the real beast: the Nice SuperEISA.
The SuperEISA is real fast in all senses of the word; it only has one VLB slot, but works well with pretty much any card. For an old design it is impressive, with 1024k WB L2, 256MB max fully-cacheable RAM, and memory bandwidth and throughput in the range of the SiS471 boards.

The NCR is interesting. I will retest it on the SuperEISA once I get first-pass numbers finished for the Chips board. I have completed tests with an ARK1000VL (Bali32) and S3 864 (Bahamas64). I plan to run one more set of tests with the 868. To give a hint at how it seems there is a bottleneck, no matter the card, Windows graphics scores and 3DSMAX scores are not changing. This is troubling. There is potential in this board, but BIOS handicaps are seemingly present.

Apples to apples sounds good to me! I prefer good comparisons for science’s sake 😁

Re: the download.
It could be on my end as well. I noticed the LW3D files did not seem to download, and am not sure why. The icons are there in Windows and all else works great.

Offline 3D graphics is something I have always wanted to explore further. I will admit I have always been intimidated, but something about 3D rendering, even on these older machines, is beyond whimsical and magical. It is a foreign language and brilliant art to itself that’s beyond fascinating! I will have to keep messing with this software, even beyond the tests! Seeing that Chevy render is too cool! 😀

@chadti99

That N9 BIOS for the S3 Trio64 VLB video card is cool.
On the Asus VLI-486SV2GX4 with POD100 CPU i am seeing these differences:

Diamond V2.01 BIOS:
Wolf3D: 135 fps
PC Player: 28.8 fps
Doom: 60.03 fps (1244 ticks)
Quake 1: 26.6
WinTune 2: 13300 KPixels/second

N9 BIOS:
Wolf3D: 137.3 fps [+]
PC Player: 29 fps [+]
Doom: 59.95 fps (1254 ticks) [-]
Quake 1: 26.6 fps [=]
WinTune 2: 13400 KPixels/second [+]

As you can see - same in Quake 1, tiny bit worse in Doom, better in Wolf3D, PC Player and WinTune.
Also, the Diamond BIOS keeps the card dark for much longer during POST while the N9 one turns on the lights right away.
Will keep the N9 BIOS in the card.

---

@WJG6260

NCR quirks are a turn-off despite the seemingly good DOS perf.
Still, it will be good to see where things are on a 4x40 basis.

Are you saying that the IIT video card has 6Mb VLB version or that in theory supports up to 6Mb of video memory ?
Never saw 6Mb VLB card so far. Or at least cannot remember right now.

Tested the archive with the retro 3d rendering stuff.
Worked well. Not sure what was the problem you hit with it but reuploaded with fresh tested files.
Hope it works better this time.

Yes, i find retro graphics processing fun and everything.