realized I actually had the box and driver disk for the VLB I/O card in the closet. It's labeled as a VL Pro card from CMD and the marketing on the box claims it's the "world's fastest controller for IDE". anyways installed the DOS drivers and my Speedsys HDD score went from 3700ish to over 8000, nice!
Nice video ! Regarding your soldering thing for the turbo switch, if you are afraid to solder (which I understand since I myself have 2 left hands as we say in French), simply use a pair of Arduino cables : their connectors will fit perfectly on both ends (panel and motherboard). So basically you replace the original cable with these. I use this trick without any issue.
In general 486 boards, due to lacking tag support in their chipset and caching technique, the cacheable region is directly tied to cache size as the tag only provides a fixed 7/8 bits to the data adress(stored in the "tag" sram). Take the log(base 2) of the cache size and add 7 bits for WB / 8 for WT and you get the log of the cacheable mem size. 256KiB cache -> 2^18 bytes 2^(18+8) = 64MiB cacheable WT 2^(18+7) = 32MiB cacheable WB 1024KiB cache -> 2^20 bytes 2^(20+8) = 256MiB cacheable WT 2^(20+7) = 128MiB cacheable WB etc.
I used to love messing around with write back cache... everyone loved me at work when I got some cheep COAS cache and installed it on computers.. it was like getting an entirely new computer
I think that optical drive is a lite-on model, an LTD-163D or something like that. I used the exact same model for my 486 build because I wanted something that could read data DVDs just in case I ever had any need, but that still had analog CD playback and CD player functionality with the dual buttons. Being able to just pop a music CD in the drive and hit play while I'm at a DOS prompt sounds like a novelty, but it's actually pretty fun. I didn't sand off the DVD logo like you did though. I personally have an Alaris Cougar build with the IBM Blue Lightning CPU. People argue over whether it's a " real " 486 or just a hot rodded 386, for most intents and purposes I consider it a slow 486 but I can see the argument, and oftentimes when I feel like poking fun at it I call it a 386 on stilts. Originally mine ran a 25 megahertz bus and a three times multiplier, but with some help from a friend I modified my bios to run a 2x multiplier instead so I could overclock the system bus to 33 megahertz and have it reliably function. I lost a bit of ALU speed but gained in every other category, from disk speed to memory speed to video speed and finally to FPU speed, since it uses an external 387 compatible FPU. For that, I'm using an IIT 4C87-DLC which benchmarks pretty close to a Cyrix FasMath but was less than a third of the cost of one on eBay. For the three things that I do that can even use an FPU, it's plenty good enough. Performance-wise, I would say that it's a rough equal to a 486DX2-50. It has better IO performance but worse CPU / FPU performance so it evens out. I'm using a Cirrus Logic GD5428 1MB VLB where I used to use a Trio64V+ VLB, The trio wasn't as compatible as I wanted it to be and honestly for the speed grade I'm looking at the Cirrus card is plenty quick. Interestingly, the boards IBM paired these CPUs with have excellent IO, with nice 16550A UARTs and VLB IDE and such onboard. It also has a real socket 2 so when I really feel like it I can use my 5X86 on an interposer. These days, I've made my peace with what it can and can't do, and have other machines for what it can't, so I don't feel an immense urge to crank as much out of it as I can like I used to. Another interesting part about the IBM blue lightning CPU, is the cache controller. It has 16K L1 but it can only cache 16MB of memory, anything after that point is only cached by the L2. My theory about that stems from the fact that the blue lightning has some heritage in the older 486SLC design where they gave it L1 cache but being that the CPU had a 24-bit address bus, being based on the 386 SX, they didn't bother giving it the ability to cache more than 16MB and the blue lightning just copied that wholesale. It's fine, 16MB is enough, but it was annoying to find that Windows 95 was even slower with 32 than 16 given how 9x fills memory top down, as my initial impulse was to fit 32MB too.
I've always wanted a Blue Lightning setup though not for any particular reason except I like unique builds so the idea that its kind of is a suped up 386 appeals to me. Can you give examples of compatability issues with the trio based VLB card? I've never had any issues with S3 based cards as far as compatability goes and I always read as well as tell viewers in my videos. that compatability is a strong point of S3 video cards though I've mostly only used PCI implementations. I have a Trio based VLB card myself but I've never gotten it to work.
@@AncientElectronics in particular the Trio64V+ was never meant to exist on VLB, the fact that it works is just a hold over from the older Trio64 design. There's only about a year's worth of drivers that will work with the thing on Windows 9X, from early 1996 to January 1997, And those drivers have issues with graphical corruption both in general and in screensavers in particular. The S3 driver in both NT 3.51 and 4.0 behave in different ways -- The NT3 driver will install and mostly work but it will exhibit major frame buffer corruption issues, and the NT4 driver simply won't install at all. To the best of my knowledge, there isn't a single Windows 3.x driver that will work with it at all. The S3 trio package for 3.x will install but refuse to initialize. ~1999 Linux that otherwise easily supports the Trio line doesn't know what to make of it and will give you a working VGA text console but freeze when initializing X with both the S3 specific server and the generic SVGA server. Like, it freezes the entire machine solid, stops responding to ping requests, that kind of thing. Even with an updated VBIOS, it supports VBE 1.2 at best, there is a TSR available for DOS that brings that to 2.0, and UniVBE can bring it to 3.0, but I remember going out of my way to burn an EPROM with a copy of a newer release of its BIOS only to be disappointed that it didn't fix anything. Their PCI cards are rock solid, 100% fine. Their more common VLB cards are also, I'm assuming, fine. Mine in particular is just an abomination of God that shouldn't have existed. Ironically though, in DOS, it's darn near perfect - - I haven't found a single DOS game that I haven't been able to prod into working, it doesn't exhibit any SVGA bugs in the commander keen games that I can see, And it's fast as hell. Easily the fastest VLB card I've ever seen, though to be fair I've never personally seen a tseng labs or ark1000pvl card. As a bonus, I never had any extra issues out of it that stemmed from running it at a 40 megahertz bus, like on my old 486 build. It wasn't a prototype or anything, STB manufactured and marketed these cards in 1995 as the PowerGraph64V+ VL, it just doesn't have very good support in the form of drivers at all. I'm told that there's no major difference between S3 drivers targeting PCI and VLB except for where in the memory map it's looking for the devices, but I don't have the wherewithal to attempt to rewrite any of them.
I love your little "90s style review tech hobbist show" I plan to start a channel myself eventually doing the same thing with retro And modern semi-budget hardware builds and going for a sorta retro late 90s-early 2000s tech tv show (Along the lines of "The Screen Savers", But with my own "twist".), dont care about monitisation though I wouldnt complain about light monitisation ither, mainly I just want to do what i know and have been doing since around 1995 sort of as hobby, and plan to use (mostly) my own money.
Whoa, I thought that M912 looked familiar. I looked at a picture of an old computer parts display I had in my room years ago and that board is the centerpiece. Unfortunately I didn't have the parts back then to build it up.
L1 cache made a huge amount of difference on 486. It's tiny, but it is *very* close to the CPU, has low latency and ~3 times the bandwidth of DRAM. The L2 cache was on the motherboard and running at FSB speed. Since the FSB is 33 MHz it is only running at 33 MHz and not the full 66. It's not *that* hugely faster than DRAM in either bandwidth or latency; sustained reads are maybe only 30% higher bandwidth than DRAM. On am386DX-40 systems it was also quite common to have some motherboard cache; it helped more there than it did your 486 for two reasons. The cache is running full CPU speed (40 MHz) and the CPU doesn't have any L1 at all. This motherboard cache could give you easily 10%-15% increased performance. Even so, 1 kB of on-die L1 cache helped a lot more than 128 kB motherboard cache (there were some 386 upgrade chips with 1 kB L1 on-die).
Ack! Watching this vid again and I realized I still haven't built in my new old stock case of this type. I have a couple of good 486 boards ready to go but need to get off of my butt and put them together.
With 15ns cache you really should be able to use the fastest cache timings even on a 40MHz bus. All of my boards accept the fastest cache and RAM timings with 15ns cache and 60ns RAM. The mem throughput increases dramatically with these timings, especially at 40MHz bus. My first overclock was an Intel DX2-66 to 80 in 1994, and my first upgrade were 15ns cache chips in the same year. I was able to play Wing Commander III and enjoy stutter free cutscenes from CDROM with this machine ;-)
LGR has a lot of subs, im sure even if he did it full time he couldn't keep up reading and replying to them all. Thank you though, I try to do the best I can reading every comment and replying if I have something to add.
Did a 486 DX2 WB (P24D-S, SX955) build two weeks ago. Couldn't get the motherboard to recognize the CPU type correctly en let me set the L1 to WB instead of WT. Jumpers where in correct positions (P24D), new EPROM with the latest BIOS version, fiddling with the BIOS settings, no success. Turns out I had to set the jumpers on the mainboard for a regular 486 for the CPU to be recognized as a WB version, get the correct CPUID code and the WB performance increase. Go figure. Mainboard was a Chicony CH-498B. PS: I'd like to do a DX2-80 build some day. The CPU MHz difference won't make that much of a difference but the bus speed increase from 33 to 40 will do wonders for memory speed/bandwith, VLB video card performance etc.
i walked away from my pc when this playing and came back into my room at this point 1:00:16 and thought you somehow made some old dos game sync the guys mouth with yours or something XD and it cracked me up so much to think of some old man from the year 1 in a village going on about retro computer stuff hahaha
Your problem with your ISA slots is really common. I had a lot of those boards some 10 years ago or so, and almost all of them had that issue. But it's just not an issue only with VLB, even on 16-bit ISA slots the cards never seem to go all the way down in the slot. I found that concerning too at the time, I tried everything I could possibly think of (even using pliers to create a slight bend in the I/O brackets) and nothing worked. They are just kind of '' conceived '' that way for some weird odd reason. But none of them (from my experience) had a problem with contacts in the slot. It's just a wack design lol
These standard baby AT pizza boxes are so nice. My 486 setup is a lot jankier than yours, an old HP Vectra VL2 with all sorts of proprietary stuff. It's nice and compact with 4 ISA slots and a built in VLB Cirrus Logic card, but the PSU is weird and I doubt I'll ever find L2 cache for it. Still, it was cheap and runs the stuff I need it for, paired with a DX33 it's just right for Ultima 7 and disabling L1 gives good compatibility with even older stuff. Doom and newer games are happier on my Pentium II machine anyway.
I always liked the Vectras, I have a socket 7 Vectra I use as a "fast DOS" PC. It also has a proprietary PSU. It's as long as the case and also acts as a cooling fan for the CPU.
@@AncientElectronics That's how mine is laid out as well, thankfully 486 doesn't run very hot and no bulging caps so far. Also, a real MPU-401 and multiple midi modules, that's some lovely stuff you have. I raise an SB16, an Ultrasound Classic and an LAPC-I. I feel like I've spent my fair share of retro hardware luck with that last one, haven't really bought anything since.
Unfortunately I wasnt able to make a direct comparison on this board. When I would switch to write-through in the BIOS it didn't seem to have any effect. The CPU still detected as running in write-back and benchmarks were the same. Maybe there's a utility program out there that can force it.
I have a similar board to this and when I upgraded the Cache to 1MB i was having issues with benchmarks not detecting it properly. I updated the bios to the latest version i could find online and everything then worked fine. was detected in benchmarks and got better performance on the benchmarks i had run beforehand.
I think the latest BIOS is dated 12/01/1995 and mine seems to be 07/02/1995 so pretty close but I'll see if I can get it upgraded and report the results. thanks!
Apparently, there is an entire thread over at the Vogons site on this board and its funky BIOS versions I must have missed when putting this build together. I guess someone has patched a later BIOS to fix a lot of the bugs on this board such as the cache issues and detection of later AMD dx5 CPUs. wish I had a BIOS flasher to make and test them.
@@AncientElectronics Also the "RetroWeb" site has the M912 (several versions) with different bios files. If we lived closer I'd happily flash that for you.
I have that same motherboard with a P24D. In my testing the WB option actually causes WORSE performance in most benchmarks. I've noticed this on every motherboard I've tested with, except maybe the FinALi chipset boards (not enough testing to know for sure). With WT L1 cache and 512kb L2 cache and a fast enough video card you can break 30fps in Doom. I'm using a S3 Trio64 VLB to get that result. BIOS timings have to be as fast as possible as well.
I have a new old stock case just like that one. Have one very simar that I have my main 486 system in. Just haven't decided what I want to put in the new case.
Its a common DVD drive, there are millions of them out there and also anyone can pretty easily figure that out regardless. Second I'll likly be the user until the thing dies.
You dont have the 32bit mode enabled on the ide. If you want to take advantage of the IDE transfer speeds on that VESA controller you have to enable 32bit mode. Just saying.
Realized this after publishing the video. I installed the drivers for the vlb card and now performance is much improved. I belive I pinned a comment about this.
486 CPU's can address up to 4GB memory or 3.5GB when the system also has PCI slots (PCI takes up 512MB address space). As far as I know the biggest 72 pin Simms are 128MB and the biggest 30 pin Simms 16 MB, so 320MB max for this board. If you would ever use that on a system like this is probably a big no :)
I'm fairly confident this system will not detect more than 64MB of RAM regardless of what the CPU can address. all official specs indicate 64MB max though this does not always mean anything as sometimes larger amounts became available after the board was produced. I could certainly be wrong but I believe chipset and BIOS can also be limiting factors but please, if anyone knows i'm wrong please correct me. I wish I had a 128MB stick of 72-pin memory to test here.
@@AncientElectronics The 4GB is because it is a 32-bit CPU. Technically a 386 CPU can address 4GB as well. There just isn't a mainboard with enough Simm slots to make that happen :D For the 486 is this review 32MB is more then you ever need for the games from that era.
@@AncientElectronics I guess it would be useless of me to speculate on what i don't have, but it would a combination of bios, chipset and OS that set the limit of the amount of RAM that can be addressed, i won't be surprised if the NS rating of the sticks play a part in it as well. Can DOS 6.22 even address more than 64mb anyway?
@@Viczarratt DOS can address up to 4 gigs with XMS 3.0. Usually the true limit is a combination of available memory slots and whatever density of memory your BIOS will accept. What's extra funny is that Windows 98SE (or ME) will get all sulky if you try to give it more than 1GB.
any idea why the bios though it was a P24D-S (overdrive)? my main 486 has a intel dx2 66 mhz, 32 mb, 256k L2, awe64 gold, Cirus logic VLB, IDE VLB i have a second board ASUS PVI-486SP3 but i can the L2 cache to run right yet. Might have bad trace.
No idea, probably the BIOS just misidentifiying it. Apparently the BIOS on these boards have a few quirks and only the very latest versions correctly identify and even POST with chips like a AMD dx4. I. Currently working on finding a later BIOS then mine but I do not own a flasher.
I decided to check again, P24D-S isn't an overdrive, it is indeed a write-back enhanced 66MHz DX2, the S designates the chip has power management features. The 486 overdrives would be designated P4T, P23T, P24T or
Not sure I'd say always. There seem to be a number of people that believe the more filtered sound of the SB pro 2 sounds "better" to them, especially on older titles that were developed with the older SB cards in mind. Part of it also has the age of the games being played in mind. A number of games also do not take advantage of 16-bit sound anyways and 8-bit digital sound tends to have more hisses and pops when played on an SB-16 (at least the early SB-16 cards, I tend to really like the sound on the Vibra based cards). The SB-16 also is NOT 100% compatible with games that support only the Sound Blaster Pro and not the SB16 in digital sounds, as it will only play digital sound in mono. Admittedly this is a small number of games and note this only affects digital sound and not FM sound through the OPL3 chip. In the end they are both excellent choices as sound cards with minor pluses and minuses between them.
There were so many bad SB16 variants that you really have to wade through a lot of chaff to get a decent example and were still balancing out which bugs you wanted to deal with, hanging/phantom note MIDI issues, noisy DAC's, distortion/ringing with Vibra chipsets, etc... Even with "good" variants, watch out for older/noisy DAC's being implemented on otherwise decent models like the CT2230/90 since Creative was building their cards as quickly and cheaply as possible using whatever components they had on-hand. While I own a bin of old Creative Labs cards for nostalgia's sake and resale, I find the Media Vision Pro Audio Spectrum 16 cards are a lot quieter due to much higher build quality, while the Aztech cards and more modern Yamaha/ESS/Crystal single chip PnP cards are more affordable for similar or superior levels of performance including working MIDI interfaces for waveblaster or external devices.
The board is made by PC Chips, so I wouldn't put too much faith in it. (Amptron is one of the names PC Chips used after people started catching on to their fraud and shoddy quality)
realized I actually had the box and driver disk for the VLB I/O card in the closet. It's labeled as a VL Pro card from CMD and the marketing on the box claims it's the "world's fastest controller for IDE". anyways installed the DOS drivers and my Speedsys HDD score went from 3700ish to over 8000, nice!
Nice video ! Regarding your soldering thing for the turbo switch, if you are afraid to solder (which I understand since I myself have 2 left hands as we say in French), simply use a pair of Arduino cables : their connectors will fit perfectly on both ends (panel and motherboard). So basically you replace the original cable with these. I use this trick without any issue.
In general 486 boards, due to lacking tag support in their chipset and caching technique, the cacheable region is directly tied to cache size as the tag only provides a fixed 7/8 bits to the data adress(stored in the "tag" sram).
Take the log(base 2) of the cache size and add 7 bits for WB / 8 for WT and you get the log of the cacheable mem size.
256KiB cache -> 2^18 bytes
2^(18+8) = 64MiB cacheable WT
2^(18+7) = 32MiB cacheable WB
1024KiB cache -> 2^20 bytes
2^(20+8) = 256MiB cacheable WT
2^(20+7) = 128MiB cacheable WB
etc.
You have a pc museum 😂
WOW!486 DX2
The 486DX2 Logo is beautiful
I used to love messing around with write back cache... everyone loved me at work when I got some cheep COAS cache and installed it on computers.. it was like getting an entirely new computer
Honestly don't wanting to sodder the reset button is a mood and I can dig it.
The SiliconDrives are pretty good. I bought a huge lot of 2GB ones on a lot on ebay a while back. They haven't given me any trouble yet.
You shouldn't call your computer pathetic; lots of people would love to have a computer like that.
Really? Come on, I know for a fact you guys have jokes in Canada too.
@@AncientElectronics Sure we do. I think it's a nice computer, though.
I think that optical drive is a lite-on model, an LTD-163D or something like that. I used the exact same model for my 486 build because I wanted something that could read data DVDs just in case I ever had any need, but that still had analog CD playback and CD player functionality with the dual buttons. Being able to just pop a music CD in the drive and hit play while I'm at a DOS prompt sounds like a novelty, but it's actually pretty fun. I didn't sand off the DVD logo like you did though.
I personally have an Alaris Cougar build with the IBM Blue Lightning CPU. People argue over whether it's a " real " 486 or just a hot rodded 386, for most intents and purposes I consider it a slow 486 but I can see the argument, and oftentimes when I feel like poking fun at it I call it a 386 on stilts. Originally mine ran a 25 megahertz bus and a three times multiplier, but with some help from a friend I modified my bios to run a 2x multiplier instead so I could overclock the system bus to 33 megahertz and have it reliably function. I lost a bit of ALU speed but gained in every other category, from disk speed to memory speed to video speed and finally to FPU speed, since it uses an external 387 compatible FPU. For that, I'm using an IIT 4C87-DLC which benchmarks pretty close to a Cyrix FasMath but was less than a third of the cost of one on eBay. For the three things that I do that can even use an FPU, it's plenty good enough.
Performance-wise, I would say that it's a rough equal to a 486DX2-50. It has better IO performance but worse CPU / FPU performance so it evens out. I'm using a Cirrus Logic GD5428 1MB VLB where I used to use a Trio64V+ VLB, The trio wasn't as compatible as I wanted it to be and honestly for the speed grade I'm looking at the Cirrus card is plenty quick. Interestingly, the boards IBM paired these CPUs with have excellent IO, with nice 16550A UARTs and VLB IDE and such onboard. It also has a real socket 2 so when I really feel like it I can use my 5X86 on an interposer. These days, I've made my peace with what it can and can't do, and have other machines for what it can't, so I don't feel an immense urge to crank as much out of it as I can like I used to.
Another interesting part about the IBM blue lightning CPU, is the cache controller. It has 16K L1 but it can only cache 16MB of memory, anything after that point is only cached by the L2. My theory about that stems from the fact that the blue lightning has some heritage in the older 486SLC design where they gave it L1 cache but being that the CPU had a 24-bit address bus, being based on the 386 SX, they didn't bother giving it the ability to cache more than 16MB and the blue lightning just copied that wholesale. It's fine, 16MB is enough, but it was annoying to find that Windows 95 was even slower with 32 than 16 given how 9x fills memory top down, as my initial impulse was to fit 32MB too.
I've always wanted a Blue Lightning setup though not for any particular reason except I like unique builds so the idea that its kind of is a suped up 386 appeals to me. Can you give examples of compatability issues with the trio based VLB card? I've never had any issues with S3 based cards as far as compatability goes and I always read as well as tell viewers in my videos. that compatability is a strong point of S3 video cards though I've mostly only used PCI implementations. I have a Trio based VLB card myself but I've never gotten it to work.
@@AncientElectronics in particular the Trio64V+ was never meant to exist on VLB, the fact that it works is just a hold over from the older Trio64 design. There's only about a year's worth of drivers that will work with the thing on Windows 9X, from early 1996 to January 1997, And those drivers have issues with graphical corruption both in general and in screensavers in particular. The S3 driver in both NT 3.51 and 4.0 behave in different ways -- The NT3 driver will install and mostly work but it will exhibit major frame buffer corruption issues, and the NT4 driver simply won't install at all. To the best of my knowledge, there isn't a single Windows 3.x driver that will work with it at all. The S3 trio package for 3.x will install but refuse to initialize.
~1999 Linux that otherwise easily supports the Trio line doesn't know what to make of it and will give you a working VGA text console but freeze when initializing X with both the S3 specific server and the generic SVGA server. Like, it freezes the entire machine solid, stops responding to ping requests, that kind of thing.
Even with an updated VBIOS, it supports VBE 1.2 at best, there is a TSR available for DOS that brings that to 2.0, and UniVBE can bring it to 3.0, but I remember going out of my way to burn an EPROM with a copy of a newer release of its BIOS only to be disappointed that it didn't fix anything.
Their PCI cards are rock solid, 100% fine. Their more common VLB cards are also, I'm assuming, fine. Mine in particular is just an abomination of God that shouldn't have existed. Ironically though, in DOS, it's darn near perfect - - I haven't found a single DOS game that I haven't been able to prod into working, it doesn't exhibit any SVGA bugs in the commander keen games that I can see, And it's fast as hell. Easily the fastest VLB card I've ever seen, though to be fair I've never personally seen a tseng labs or ark1000pvl card. As a bonus, I never had any extra issues out of it that stemmed from running it at a 40 megahertz bus, like on my old 486 build.
It wasn't a prototype or anything, STB manufactured and marketed these cards in 1995 as the PowerGraph64V+ VL, it just doesn't have very good support in the form of drivers at all. I'm told that there's no major difference between S3 drivers targeting PCI and VLB except for where in the memory map it's looking for the devices, but I don't have the wherewithal to attempt to rewrite any of them.
I love your little "90s style review tech hobbist show" I plan to start a channel myself eventually doing the same thing with retro And modern semi-budget hardware builds and going for a sorta retro late 90s-early 2000s tech tv show (Along the lines of "The Screen Savers", But with my own "twist".), dont care about monitisation though I wouldnt complain about light monitisation ither, mainly I just want to do what i know and have been doing since around 1995 sort of as hobby, and plan to use (mostly) my own money.
I wish you luck and hope you have fun doing it. I'd like to check it out once you start it up
Whoa, I thought that M912 looked familiar. I looked at a picture of an old computer parts display I had in my room years ago and that board is the centerpiece. Unfortunately I didn't have the parts back then to build it up.
L1 cache made a huge amount of difference on 486. It's tiny, but it is *very* close to the CPU, has low latency and ~3 times the bandwidth of DRAM.
The L2 cache was on the motherboard and running at FSB speed. Since the FSB is 33 MHz it is only running at 33 MHz and not the full 66. It's not *that* hugely faster than DRAM in either bandwidth or latency; sustained reads are maybe only 30% higher bandwidth than DRAM.
On am386DX-40 systems it was also quite common to have some motherboard cache; it helped more there than it did your 486 for two reasons. The cache is running full CPU speed (40 MHz) and the CPU doesn't have any L1 at all. This motherboard cache could give you easily 10%-15% increased performance. Even so, 1 kB of on-die L1 cache helped a lot more than 128 kB motherboard cache (there were some 386 upgrade chips with 1 kB L1 on-die).
Ack! Watching this vid again and I realized I still haven't built in my new old stock case of this type. I have a couple of good 486 boards ready to go but need to get off of my butt and put them together.
With 15ns cache you really should be able to use the fastest cache timings even on a 40MHz bus. All of my boards accept the fastest cache and RAM timings with 15ns cache and 60ns RAM. The mem throughput increases dramatically with these timings, especially at 40MHz bus. My first overclock was an Intel DX2-66 to 80 in 1994, and my first upgrade were 15ns cache chips in the same year. I was able to play Wing Commander III and enjoy stutter free cutscenes from CDROM with this machine ;-)
maybe add some heat shrink on the reset button
I prefer you to LGR. You listen to your viewers, and I like your presentation.
LGR has a lot of subs, im sure even if he did it full time he couldn't keep up reading and replying to them all. Thank you though, I try to do the best I can reading every comment and replying if I have something to add.
@@AncientElectronics I prefer a channel where I can have more engagement with someone. LGR doesn't even consider requests.
@@CovenantAgentLazarus He never listened to a suggestion of mine.
@@CovenantAgentLazarus I especially want him to review one of the 'Xtreme' series of games, like 'Xtreme Rockclimbing' (which he already did).
Did a 486 DX2 WB (P24D-S, SX955) build two weeks ago. Couldn't get the motherboard to recognize the CPU type correctly en let me set the L1 to WB instead of WT. Jumpers where in correct positions (P24D), new EPROM with the latest BIOS version, fiddling with the BIOS settings, no success. Turns out I had to set the jumpers on the mainboard for a regular 486 for the CPU to be recognized as a WB version, get the correct CPUID code and the WB performance increase. Go figure. Mainboard was a Chicony CH-498B.
PS: I'd like to do a DX2-80 build some day. The CPU MHz difference won't make that much of a difference but the bus speed increase from 33 to 40 will do wonders for memory speed/bandwith, VLB video card performance etc.
i walked away from my pc when this playing and came back into my room at this point 1:00:16 and thought you somehow made some old dos game sync the guys mouth with yours or something XD and it cracked me up so much to think of some old man from the year 1 in a village going on about retro computer stuff hahaha
Your problem with your ISA slots is really common.
I had a lot of those boards some 10 years ago or so, and almost all of them had that issue. But it's just not an issue only with VLB, even on 16-bit ISA slots the cards never seem to go all the way down in the slot. I found that concerning too at the time, I tried everything I could possibly think of (even using pliers to create a slight bend in the I/O brackets) and nothing worked. They are just kind of '' conceived '' that way for some weird odd reason. But none of them (from my experience) had a problem with contacts in the slot. It's just a wack design lol
These standard baby AT pizza boxes are so nice. My 486 setup is a lot jankier than yours, an old HP Vectra VL2 with all sorts of proprietary stuff. It's nice and compact with 4 ISA slots and a built in VLB Cirrus Logic card, but the PSU is weird and I doubt I'll ever find L2 cache for it. Still, it was cheap and runs the stuff I need it for, paired with a DX33 it's just right for Ultima 7 and disabling L1 gives good compatibility with even older stuff. Doom and newer games are happier on my Pentium II machine anyway.
I always liked the Vectras, I have a socket 7 Vectra I use as a "fast DOS" PC. It also has a proprietary PSU. It's as long as the case and also acts as a cooling fan for the CPU.
@@AncientElectronics That's how mine is laid out as well, thankfully 486 doesn't run very hot and no bulging caps so far.
Also, a real MPU-401 and multiple midi modules, that's some lovely stuff you have. I raise an SB16, an Ultrasound Classic and an LAPC-I. I feel like I've spent my fair share of retro hardware luck with that last one, haven't really bought anything since.
I will be nice to see difference between regular and write back 486 -66
Unfortunately I wasnt able to make a direct comparison on this board. When I would switch to write-through in the BIOS it didn't seem to have any effect. The CPU still detected as running in write-back and benchmarks were the same. Maybe there's a utility program out there that can force it.
I have a similar board to this and when I upgraded the Cache to 1MB i was having issues with benchmarks not detecting it properly. I updated the bios to the latest version i could find online and everything then worked fine. was detected in benchmarks and got better performance on the benchmarks i had run beforehand.
I think the latest BIOS is dated 12/01/1995 and mine seems to be 07/02/1995 so pretty close but I'll see if I can get it upgraded and report the results. thanks!
Same issue with my M915 ( The one that says "Secondray" on it ), 512k was the most mine would work properly with.
Apparently, there is an entire thread over at the Vogons site on this board and its funky BIOS versions I must have missed when putting this build together. I guess someone has patched a later BIOS to fix a lot of the bugs on this board such as the cache issues and detection of later AMD dx5 CPUs. wish I had a BIOS flasher to make and test them.
@@AncientElectronics Also the "RetroWeb" site has the M912 (several versions) with different bios files. If we lived closer I'd happily flash that for you.
I have that same motherboard with a P24D. In my testing the WB option actually causes WORSE performance in most benchmarks. I've noticed this on every motherboard I've tested with, except maybe the FinALi chipset boards (not enough testing to know for sure). With WT L1 cache and 512kb L2 cache and a fast enough video card you can break 30fps in Doom. I'm using a S3 Trio64 VLB to get that result. BIOS timings have to be as fast as possible as well.
I have a new old stock case just like that one. Have one very simar that I have my main 486 system in. Just haven't decided what I want to put in the new case.
It would have been a better idea to leave the logo on that drive in case it ends up in someone else's hands and they need to know more about it.
Its a common DVD drive, there are millions of them out there and also anyone can pretty easily figure that out regardless. Second I'll likly be the user until the thing dies.
There was no stereo in Duke Nukem setup sound test
Nice
You dont have the 32bit mode enabled on the ide. If you want to take advantage of the IDE transfer speeds on that VESA controller you have to enable 32bit mode. Just saying.
Realized this after publishing the video. I installed the drivers for the vlb card and now performance is much improved. I belive I pinned a comment about this.
jeez, dude -- get on with it, fer chrissakes.
486 CPU's can address up to 4GB memory or 3.5GB when the system also has PCI slots (PCI takes up 512MB address space). As far as I know the biggest 72 pin Simms are 128MB and the biggest 30 pin Simms 16 MB, so 320MB max for this board. If you would ever use that on a system like this is probably a big no :)
I'm fairly confident this system will not detect more than 64MB of RAM regardless of what the CPU can address. all official specs indicate 64MB max though this does not always mean anything as sometimes larger amounts became available after the board was produced. I could certainly be wrong but I believe chipset and BIOS can also be limiting factors but please, if anyone knows i'm wrong please correct me. I wish I had a 128MB stick of 72-pin memory to test here.
@@AncientElectronics The 4GB is because it is a 32-bit CPU. Technically a 386 CPU can address 4GB as well. There just isn't a mainboard with enough Simm slots to make that happen :D For the 486 is this review 32MB is more then you ever need for the games from that era.
@@AncientElectronics I guess it would be useless of me to speculate on what i don't have, but it would a combination of bios, chipset and OS that set the limit of the amount of RAM that can be addressed, i won't be surprised if the NS rating of the sticks play a part in it as well. Can DOS 6.22 even address more than 64mb anyway?
@@Viczarratt DOS can address up to 4 gigs with XMS 3.0. Usually the true limit is a combination of available memory slots and whatever density of memory your BIOS will accept.
What's extra funny is that Windows 98SE (or ME) will get all sulky if you try to give it more than 1GB.
From my understanding the issues with Win98 start after 512MB though there are patches that correct the issues.
any idea why the bios though it was a P24D-S (overdrive)?
my main 486 has a intel dx2 66 mhz, 32 mb, 256k L2, awe64 gold, Cirus logic VLB, IDE VLB
i have a second board ASUS PVI-486SP3 but i can the L2 cache to run right yet. Might have bad trace.
No idea, probably the BIOS just misidentifiying it. Apparently the BIOS on these boards have a few quirks and only the very latest versions correctly identify and even POST with chips like a AMD dx4. I. Currently working on finding a later BIOS then mine but I do not own a flasher.
I decided to check again, P24D-S isn't an overdrive, it is indeed a write-back enhanced 66MHz DX2, the S designates the chip has power management features. The 486 overdrives would be designated P4T, P23T, P24T or
@@AncientElectronics ok thanks for figuring it out and letting me know.
"D" as in "Duh... Write back! Duh!!!"
Not too bad, but a SB16 is always the best choice.
Not sure I'd say always. There seem to be a number of people that believe the more filtered sound of the SB pro 2 sounds "better" to them, especially on older titles that were developed with the older SB cards in mind. Part of it also has the age of the games being played in mind. A number of games also do not take advantage of 16-bit sound anyways and 8-bit digital sound tends to have more hisses and pops when played on an SB-16 (at least the early SB-16 cards, I tend to really like the sound on the Vibra based cards). The SB-16 also is NOT 100% compatible with games that support only the Sound Blaster Pro and not the SB16 in digital sounds, as it will only play digital sound in mono. Admittedly this is a small number of games and note this only affects digital sound and not FM sound through the OPL3 chip. In the end they are both excellent choices as sound cards with minor pluses and minuses between them.
There were so many bad SB16 variants that you really have to wade through a lot of chaff to get a decent example and were still balancing out which bugs you wanted to deal with, hanging/phantom note MIDI issues, noisy DAC's, distortion/ringing with Vibra chipsets, etc... Even with "good" variants, watch out for older/noisy DAC's being implemented on otherwise decent models like the CT2230/90 since Creative was building their cards as quickly and cheaply as possible using whatever components they had on-hand.
While I own a bin of old Creative Labs cards for nostalgia's sake and resale, I find the Media Vision Pro Audio Spectrum 16 cards are a lot quieter due to much higher build quality, while the Aztech cards and more modern Yamaha/ESS/Crystal single chip PnP cards are more affordable for similar or superior levels of performance including working MIDI interfaces for waveblaster or external devices.
Ive never realized how irrelevant L2 cache is on i486.
The board is made by PC Chips, so I wouldn't put too much faith in it. (Amptron is one of the names PC Chips used after people started catching on to their fraud and shoddy quality)