He says in the video that the one he used in his previous video wasn't stable at 200Mhz so he had to use a different one.
I really need to see if I have a board that will do 66Mhz fsb.
The current board I am using now, the ASUS PVI-486AP4 maxes out at 40Mhz fsb because the chipset divides the input signal by 2. It can already qualify as the slowest 486 since I can get the CPU to run at a reported 3.90Mhz with some jumper trickery and board mods but it would be really cool to test it out with a clock doubling circuit so I could feed the chipset faster clockspeeds for it to halve. That would potentially give me up to an 80Mhz fsb, but not sure if it would work as I am unsure if the PCI bus has any sort of divider above a certain speed. There are no settings in the BIOS for it.
I have watched the video now. I enjoy his enthusiasm.
It would be interesting if he ran the results with the PCI bus using a 1/2 multiplier, thus at 33 MHz, so we could see how much the clock doubled PCI bus is increasing the results. I would also like to know if this setup can achieve the same stability using more of a conventional socket 370 heatsink/fan. In this way, one might actually be able to build a cased system out it. Lastly, it would be neat to see how much performance loss there is when running the Am5x86 at 50x4 rather than 66x3.
I have my IBM 5x86c-133/2x fully setup in a case and dozens of software installed, but I don't think I have ever tried running the PCI bus at the full 66 MHz. With the PCI bus at 33 MHz, I get 28.5 fps in GLQuake, so I can only imagine what the result would be like at 66 MHz.
If I recall, the MB-8433UUD has an 83 MHz FSB jumper, so perhaps CPUGalaxy can shoot for 250 MHz in his next video! I know I was able to run the Am5x86 at 166 MHz using the 83 MHz setting, but there were some mild screen artifacts. Better yet would be to find a motherboard with a 75 MHz FSB, or swap out the PLL. This should yield an Am5x86 at 225 MHz.
Plan your life wisely, you'll be dead before you know it.
I have watched the video now. I enjoy his enthusiasm.
It would be interesting if he ran the results with the PCI bus using a 1/2 multiplier, thus at 33 MHz, so we could see how much the clock doubled PCI bus is increasing the results. I would also like to know if this setup can achieve the same stability using more of a conventional socket 370 heatsink/fan. In this way, one might actually be able to build a cased system out it. Lastly, it would be neat to see how much performance loss there is when running the Am5x86 at 50x4 rather than 66x3.
I have my IBM 5x86c-133/2x fully setup in a case and dozens of software installed, but I don't think I have ever tried running the PCI bus at the full 66 MHz. With the PCI bus at 33 MHz, I get 28.5 fps in GLQuake, so I can only imagine what the result would be like at 66 MHz.
If I recall, the MB-8433UUD has an 83 MHz FSB jumper, so perhaps CPUGalaxy can shoot for 250 MHz in his next video! I know I was able to run the Am5x86 at 166 MHz using the 83 MHz setting, but there were some mild screen artifacts. Better yet would be to find a motherboard with a 75 MHz FSB, or swap out the PLL. This should yield an Am5x86 at 225 MHz.
Please try the Cyrix with 66Mhz PCI! It might beat the AMD.
Do you remember which video card you were able to boot with the 8433UUD 83MHz setting?
I was using the 1/2 multiplier, so the PCI bus was running at 41.67 MHz. So most video cards will work.
My IBM 5x86c-133/2x system has a Voodoo2 in it normally. Will Voodoo2's run at 66 MHz? I have a vague recollection that they won't, so I would likely need to pull it for 66 MHz PCI tests.
Plan your life wisely, you'll be dead before you know it.
I was using the 1/2 multiplier, so the PCI bus was running at 41.67 MHz. So most video cards will work.
My IBM 5x86c-133/2x system has a Voodoo2 in it normally. Will Voodoo2's run at 66 MHz? I have a vague recollection that they won't, so I would likely need to pull it for 66 MHz PCI tests.
Only one way to find out, fire it up and bring balance to the force. Restore honor to the house of Cyrix.
I fired it up, but the system would hang on AWARD BIOS Extension something... Where it would normally list your ISA cards before POST. So there must be some component not happy with the 66 MHz. Also, on this board, the ISA bus runs at a divisor of the PCI clock and there isn't a fixed 7.x MHz option for ISA like there is for the keyboard controller. The largest divisor is 4, meaning the ISA bus would be running at 16.67 MHz, which is asking too much. So to really test this, I'd have to clear all the hardware out of the system and disable ISA. Something like this is more suitable for benchtop work, but I won't have workbench space for some time.
Plan your life wisely, you'll be dead before you know it.
I fired it up, but the system would hang on AWARD BIOS Extension something... Where it would normally list your ISA cards before POST. So there must be some component not happy with the 66 MHz. Also, on this board, the ISA bus runs at a divisor of the PCI clock and there isn't a fixed 7.x MHz option for ISA like there is for the keyboard controller. The largest divisor is 4, meaning the ISA bus would be running at 16.67 MHz, which is asking too much. So to really test this, I'd have to clear all the hardware out of the system and disable ISA. Something like this is more suitable for benchtop work, but I won't have workbench space for some time.
Was worth a shot, hopefully you’ll find some time soon. Slightly off topic but what kinda of case and power supply do you use for this setup? Thinking you found an AT case that has a ps/2 hole, or do you use the ps/2 header and bracket?
I had my Gigabyte GA-486 AM/S up and running and used a peltier element with a capable version of the 5x86. It's unfortunately not 100% stable which I can maybe change by hopping onto another board with a variable voltage regulator.
Nonetheless a nice experiment.
I had my Gigabyte GA-486 AM/S up and running and used a peltier element with a capable version of the 5x86. It's unfortunately not 100% stable which I can maybe change by hopping onto another board with a variable voltage regulator.
Nonetheless a nice experiment.
Hell yes! Awesome work!
Any pics of the setup? Was this with Phil’s Dosbench for the quake timedemo?
Any recommendations on where to get a Peltier cooler for socket 3?
I had my Gigabyte GA-486 AM/S up and running and used a peltier element with a capable version of the 5x86. It's unfortunately not 100% stable which I can maybe change by hopping onto another board with a variable voltage regulator.
Nonetheless a nice experiment.
Hell yes! Awesome work!
Any pics of the setup? Was this with Phil’s Dosbench for the quake timedemo?
Any recommendations on where to get a Peltier cooler for socket 3?
Yes, it was from Phil's menu.
You can get the peltier element for little money on amazon. Be aware that, naturally, water droplets build around it, so prolonged use is dangerous as in fluids sinking into the socket!
I have changed the setup to a LS486E, so unfortunately I can't post pics of the rig running with the GA-486AM/S. The LS486E has much lower PCI performance so I will commence tests at 3x66 instead of 4x50. I will need some custom made srams which will arrive in a few weeks... until then some screens from the old setup.
The cpu needs juice in the 4.2 to 4.4 region [(4.2, 4.4) exclusive] to work properly - minimum being 3.98V for a boot screen.
Once it goes up to 4.4 it does not start up anymore.
I lost the score of 3DBench2 - which was over 122fps. I'll redo scores once I got the new srams.
I had my Gigabyte GA-486 AM/S up and running and used a peltier element with a capable version of the 5x86. It's unfortunately not 100% stable which I can maybe change by hopping onto another board with a variable voltage regulator.
Nonetheless a nice experiment.
Hell yes! Awesome work!
Any pics of the setup? Was this with Phil’s Dosbench for the quake timedemo?
Any recommendations on where to get a Peltier cooler for socket 3?
Yes, it was from Phil's menu.
You can get the peltier element for little money on amazon. Be aware that, naturally, water droplets build around it, so prolonged use is dangerous as in fluids sinking into the socket!
I have changed the setup to a LS486E, so unfortunately I can't post pics of the rig running with the GA-486AM/S. The LS486E has much lower PCI performance so I will commence tests at 3x66 instead of 4x50. I will need some custom made srams which will arrive in a few weeks... until then some screens from the old setup.
The cpu needs juice in the 4.2 to 4.4 region [(4.2, 4.4) exclusive] to work properly - minimum being 3.98V for a boot screen.
Once it goes up to 4.4 it does not start up anymore.
I lost the score of 3DBench2 - which was over 122fps. I'll redo scores once I got the new srams.
Awesome scores and thanks for the info! Could you share your jumper settings?
If my memory doesn't deceive me, the GA-486 AM/S has a through hole resistor for the voltage set resistor and can easily be swapped out with a trim pot to give you variable voltage. I did this on mine.
Can you elaborate some on these "custom made SRAMs"? Are they in a SOJ to DIP adapter? What speed and brand are they? You can still get SOJ SRAMs new in the 10 ns flavour from Cypress.
Plan your life wisely, you'll be dead before you know it.
@Chadti99 - it will take me a few days since I have to take care of some boards and the GA-486AM/S needs some testing plus bios re/flash.
@feipoa - I would be interested in modding it as I have a 4015 DC/DC down converter lying around - any instructions for the mod?
These srams are soj ones soldered onto adapter boards. Luckily Tiido(thanks again!) did it for me and they seem to be working great. Unfortunately 10ns seems not enough to get the system cache stable at the best timings. I have a possible solution for this but can only deliver results in a week or so. There's authentic ISSI 12ns parts among the modules and they do not work stable at all at 60mhz...
Made my own little 486 'hot rod' box. Not trying to get too crazy here, just throwing together something that can put that 40MHz FSB to good use.
Even so it's much faster than any of my previous 486 builds. As a stock CPU the AMD DX4-120 is really not to be underestimated.
Motherboard: Shuttle HOT-433 /w UMC PCI chipset
CPU: AMD Am486DX4-120 SV8B @ 120MHz, 8Kb write back L1
RAM: 16MB FPM (for now), 1WS read (yuck)
L2 Cache: 512KB asynchronous @ 15ns, 0WS
HDD: Conner CFA-540A 540MB IDE, 4500rpm, 256K cache (rather uncommon in a Conner drive)
Video: Diamond Stealth PCI (S3 Trio64), 2MB
Sound: ESS 1868 sound card
NIC: 3Com 16-bit ISA network card
OS: IBM PC DOS 7.00 /w Y2K patches, Windows For Workgroups 3.11
The low DRAM result is due to me having to run with one read waitstate on the DRAM. Without the waitstate the memory throughput jumps to 110-120MB/s, but then HIMEM.SYS begins complaining about the A20 line. (The /MACHINE=11 switch bypasses the error, but then Doom crashes in a spectacular manner.)
Supposedly the board (Shuttle HOT-433) supports EDO DRAM proper, but it doesn't seem to extend beyond electrical/signalling compliance as there was no measurable improvement going from FPM to EDO on it.
A lot of people use Phil's Benchmark Pack. it should pop up within the first dozen search results on Duckduckgo. I personally don't use it, but use benchmarks that have been amalgamated into Phil's pack.
Plan your life wisely, you'll be dead before you know it.
Few benchmarks that i can add from myself: A strictly non-PCI 486.
CPU: AMD 5x86-75
RAM: HP 2x32MB 60ns 72pin DIMM
GPU: Cirrus Logic CL-GD5429-86QC-B VLB 2MB
Motherboard: Soyo 025P2 / QDI V4S471/G (benchmarked both)
L2 Cache: Soyo - 256kB (15ns); QDI - 1024kB (mix of 10ns and 12ns - probably fake timings)
Motherboard's BIOS: Modified to support L2 cache writeback properly and enabled/made visible all options in bios (on both boards)
Timings: Everything maxed out (including GPU with mclk tool), except QDI @ 3x50, when i had to loosen RAM, L2 cache and GPU memory timings.
I used Phil's benchmark set.
There might be some fine-tuning still to be done. It seems some benchmarks are modifying some settings on the PC, as the order of running benchmarks affects the result. I've run the benchmarks all in the same order as on the picture, but just as an observation, when i run "Doom max" on QDI @3x50 as a first thing, i get around 1250 ticks instead of 1291 ticks. So still some mysteries to solve.
"640K ought to be enough for anybody." - And i intend to get every last bit out of it even after loading every damn driver!
Interesting. Perhaps run various instances of two benchmarks - one benchmark prior to DOOM per reboot - to determine which benchmark is altering some VGA setting, thus reducing the frame rate.
Plan your life wisely, you'll be dead before you know it.
Hey, cannot find any direct comparison of DX4-75 and DX2-66. Anyone with 486 DX4-100 mhz can try and clock it to 75mhz(25x3) and 66mhz (33x2) and compare them in various benchmarks? Or at least one or two benchmarks. Would like to see, which one is faster and where (in what application).
If anyone knows link to such benchmark (or make a new benchmark, preferably on same CPU (DX4-100 either AMD or Intel)), I would be glad.
(edit: On fast VGA, so it is not slowed by VGA, preferably PCI based socket 3 motherboard).
Last edited by Stiletto on 2021-11-29, 21:57. Edited 1 time in total.
Hey, cannot find any direct comparsion of DX4-75 and DX2-66. Anyone with 486 DX4-100 mhz can try and clock it to 75mhz(25x3) and 66mhz (33x2) and compare them in various benchmarks?
On a case by case
Which was faster related more to the cache, video and motherboard configuration
No cache + faster FPM 75mhz dx4 will do better than the dx2
Decent cache + decent VLB video hardware Dx2 wins
I had a 486dx4-75mhz Compaq laptop, performance was below a dx2-50 VLB desktop, terrible video and no cache killed the system.
