VOGONS


Reply 20 of 49, by retro games 100

User metadata
Rank l33t
Rank
l33t
sliderider wrote:
Old Thrashbarg wrote:

You can't up the voltage of the POD since it runs on 5v already.

Externally, yes, though there is a way of modifying the onboard voltage regulator to change the voltage going to the chip itself. I'll be damned if I can remember the details of it, though, and I also recall it making little/no difference in the overclocking ability.

One other random tidbit about those things... IIRC they automatically switch to a 1X multiplier if you remove the built-in fan. (You'd want to attach an externally-powered fan in its place though.) That might provide some possibilities for testing a 66mhz bus speed.

Ummm...there are no socket 3 motherboards with a 66mhz bus setting. 😉

I am testing a Biostar 486 mobo with a working 66 MHz bus setting. Previously, I got a 100% overclock on an SX-33 to get it to run at 33->66 MHz, but I've just this very moment got the P75 chip running at a 3x multi = 198 MHz. And it works! I'll post more after I tweak the BIOS timings..

Edit: At the end of page 3 on this thread, I explain that the 66 MHz bus speed option does not work well for the P75 chip. It's OK for an SX-33, but either the P75 or the 60ns EDO RAM, or possibly both of these components can't cope with this bus speed.

Last edited by retro games 100 on 2011-02-18, 11:15. Edited 1 time in total.

Reply 21 of 49, by udam_u

User metadata
Rank Member
Rank
Member

Ummm...there are no 486 motherboards with a 66mhz bus setting.
Wink

You are wrong. (: It seems that all 486 motherboard based on UMC 8881 have got undocumented 66MHz bus option.

Reply 22 of 49, by Tetrium

User metadata
Rank l33t++
Rank
l33t++
sliderider wrote:
Old Thrashbarg wrote:

You can't up the voltage of the POD since it runs on 5v already.

Externally, yes, though there is a way of modifying the onboard voltage regulator to change the voltage going to the chip itself. I'll be damned if I can remember the details of it, though, and I also recall it making little/no difference in the overclocking ability.

One other random tidbit about those things... IIRC they automatically switch to a 1X multiplier if you remove the built-in fan. (You'd want to attach an externally-powered fan in its place though.) That might provide some possibilities for testing a 66mhz bus speed.

Ummm...there are no 486 motherboards with a 66mhz bus setting. 😉

Yes there are! 😁
Test:486 undocumented jumper settings

Reply 23 of 49, by Yushatak

User metadata
Rank Member
Rank
Member

I just found out that a POD can run on a Socket 1 box, the added pins are just for added voltage and it gets plenty from the 5v socket as it is - perfectly stable. Anyway I wanted to use this "PCPBench" I see referenced all over VOGONS and I can't find a copy.

Also, I find in Quake it gives a pretty significant boost to performance vs. an Am5x86-133, 4FPS gain (and on a 486 box 4FPS is a big deal, because it's struggling with Quake).

My board can only do 25 or 33Mhz, is there any way to get the POD up to 100Mhz by some other trick, perhaps change the internal multiplier somehow?

The real bottleneck in my box is that it's stuck with an onboard GD5420, which is OK for most things, but a bit scant for Quake. If it were the 1MB variety it might do better, but it's just 512K VRAM.

I'm about to check out Duke3D, see how much better it runs now that there's a POD in there.

Concerning your overclock via 66Mhz FSB... if the problem is the RAM, you can easily change the timings - there's a program called "HyperRAM" - it's purpose is to free up a few CPU cycles by raising the refresh of the RAM slightly - you can raise them very far up if you want, and that might help it cope. I've tried the default 65, 120, and 240, all of them stable on my 60 RAM.

Edit: Just enough of a difference gained from that FPU to take previously very laggy scenes in Duke3D and make them smooth. Much more significant difference than seen in Quake.

Reply 24 of 49, by sliderider

User metadata
Rank l33t++
Rank
l33t++

I don't see how they could have made a 66mhz fsb 486 board when they couldn't even get a 50mhz board to run stable without giving off more radiation than Chernobyl and there were never any 486 class chips that supported it and you'd never be able to plug a PCI card into it because it wouldn't tolerate the overclock.

Reply 25 of 49, by Tetrium

User metadata
Rank l33t++
Rank
l33t++
sliderider wrote:

I don't see how they could have made a 66mhz fsb 486 board when they couldn't even get a 50mhz board to run stable without giving off more radiation than Chernobyl and there were never any 486 class chips that supported it and you'd never be able to plug a PCI card into it because it wouldn't tolerate the overclock.

And that is what the "magic jumper" is for! 😉

Firstly, yes, you're right that when 486 went 50Mhz fsb, they had many troubles. But the boards I tested were made years later, after the original Pentium Socket 4 got released which also ran on 66Mhz fsb.

The magic jumper, as we call it, is a setting (either by jumper, in the BIOS or automatically set) that can halve the PCI bus relatively to the fsb. This way you can run a cpu at a much higher fsb without needing to overclock the PCI bus at all!
Another way to look at it (though technically incorrect) is that the magic jumper doubles the cpu's multiplier (compared to the PCI bus).

So if all relevant components on the board (cache chips, memory and cpu) can handle the overclock, there should be no stopping you! 😉
This is the only realistic way to reach 200Mhz on any Socket 3 chip.

Reply 26 of 49, by retro games 100

User metadata
Rank l33t
Rank
l33t

The host clock & PCI clock divider setting can be set to 1/2 on some boards, but on others such as the Biostar MB-8433UUD-A, it can be set to either 1/2 or 2/3. (It can also be set to 1:1.) This flexible 2/3 option is good, because it allows the PCI clock to be "gently overclocked", if required.

Also, Tetrium, you say that the only realistic way to reach 200Mhz is via the use of this divider option. That's not true, because on the Biostar mobo, I got a CPU clock speed of 200 MHz by using a 50 MHz bus speed, in conjunction with a 4x multiplier setting, and I specifically set this BIOS PCI divider option to 1:1, because I knew that the Virge and Viper V330 video cards could cope with this PCI speed.

Reply 27 of 49, by Tetrium

User metadata
Rank l33t++
Rank
l33t++
retro games 100 wrote:

The host clock & PCI clock divider setting can be set to 1/2 on some boards, but on others such as the Biostar MB-8433UUD-A, it can be set to either 1/2 or 2/3. (It can also be set to 1:1.) This flexible 2/3 option is good, because it allows the PCI clock to be "gently overclocked", if required.

Also, Tetrium, you say that the only realistic way to reach 200Mhz is via the use of this divider option. That's not true, because on the Biostar mobo, I got a CPU clock speed of 200 MHz by using a 50 MHz bus speed, in conjunction with a 4x multiplier setting, and I specifically set this BIOS PCI divider option to 1:1, because I knew that the Virge and Viper V330 video cards could cope with this PCI speed.

Ah, well you are right about that, but in your case it's a must to have hardware that can cope with this extremely overclocked PCI bus. While the V330 may be able to cope, harddrives most likely won't, and probably neither will soundcards, NIC's etc.
I suppose you were using a flash card as a drive? That should fix it I guess, I don't have such a card.

Reply 28 of 49, by retro games 100

User metadata
Rank l33t
Rank
l33t

Yes, that's a good point. I am using a compact flash card. I think using a CF card is ideal, for a "slow" 486 system. (Not so for Windows 98, because of its slow writes.) I haven't tried a PCI sound card, but for a 486 system, I wouldn't want to! As for a NIC, unfortunately I don't own one, and have never tested one. I expect it would have problems @ 50 MHz.

Reply 29 of 49, by 5u3

User metadata
Rank Oldbie
Rank
Oldbie

Cheap PCI network cards often have problems with high bus speeds. I couldn't get Realtek cards (8029, 8139) to work reliably at 40 MHz.

Reply 30 of 49, by feipoa

User metadata
Rank l33t++
Rank
l33t++

@swaaye

I saw these results when I was playing with a Voodoo3 in my PCI 486 board comparing the Cyrix, AMD and POD chips.

What 486 motherboard were you able to get a Voodoo3 working in? What FSB and cpu(s) were used? I didn't even get a boot screen with a Voodoo Banshee in an M919 (40Mhz FSB, Cyrix 5x86-120).

Reply 31 of 49, by swaaye

User metadata
Rank l33t++
Rank
l33t++

I have a MSI MS-4144. It's a SiS 496 PCI board. It use it with a 33 or 40 MHz FSB. The Voodoo3 just worked, but I had to use old drivers or it would blue screen probably because the later drivers require Pentium instructions (the POD stopped the blue screens). I used it with the POD, Cx5x86 and Am5x86.

I had Jedi Knight running on it and used that to compare the CPU FPU capabilities against the Voodoo3. POD was definitely the fastest.

Reply 32 of 49, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Huh, it is probably chipset related then. The POD83 didn't make any of my non-working PCI graphics cards work. They all work in other PIII-based systems though.

Anyone with an Intel chipset 486-PCI motherboard try a Matrox G450?

Reply 33 of 49, by Tetrium

User metadata
Rank l33t++
Rank
l33t++
swaaye wrote:

I have a MSI MS-4144. It's a SiS 496 PCI board. It use it with a 33 or 40 MHz FSB. The Voodoo3 just worked, but I had to use old drivers or it would blue screen probably because the later drivers require Pentium instructions (the POD stopped the blue screens). I used it with the POD, Cx5x86 and Am5x86.

I had Jedi Knight running on it and used that to compare the CPU FPU capabilities against the Voodoo3. POD was definitely the fastest.

This might be interesting info though.

Anyone actually ever tried to test "modern" PCI cards in 486 PCI boards just to check if it worked?
It might be chipset, but it might also be how the PCI bus was implemented on any particular board.
I could even imagine 2 boards, exactly the same except slight different revision number, in which one works with the more modern PCI cards and the other one doesn't.

Edit:
And 🤣 @ feipoa's avatar 😜

Whats missing in your collections?
My retro rigs (old topic)
Interesting Vogons threads (links to Vogonswiki)
Report spammers here!

Reply 34 of 49, by Yushatak

User metadata
Rank Member
Rank
Member
swaaye wrote:

I never had any luck with overvolting 486s.... Some boards have 4v setting but it didn't really help. 5v would probably blow these 350nm chips up.

I unintentionally ran several 3.3v chips at 5v by putting them in a Socket 1 board without realizing the discrepancy at one point. They ran for months at a time with no issue - if the board was capable of anything higher than 33Mhz FSB I could have gotten a nice OC out of them I imagine.

Reply 36 of 49, by DonutKing

User metadata
Rank Oldbie
Rank
Oldbie

I unintentionally ran several 3.3v chips at 5v by putting them in a Socket 1 board without realizing the discrepancy at one point. They ran for months at a time with no issue - if the board was capable of anything higher than 33Mhz FSB I could have gotten a nice OC out of them I imagine.

I'm trying this now with an Am486DX4-100 in a 5v only board. For some reason my board won't work with any chip with 16kb L1 cache so the DX4 overdrives aren't an option. The CPU I'm using is running fine, barely gets warm with a HSF on it, however overclocking the bus to 40MHz means that I had to loosen the RAM/cache timings which actually hindered performance more than than just running at 33MHz using tight timings.

If you are squeamish, don't prod the beach rubble.

Reply 37 of 49, by swaaye

User metadata
Rank l33t++
Rank
l33t++
Yushatak wrote:

I unintentionally ran several 3.3v chips at 5v by putting them in a Socket 1 board without realizing the discrepancy at one point. They ran for months at a time with no issue - if the board was capable of anything higher than 33Mhz FSB I could have gotten a nice OC out of them I imagine.

Cool. They must have been running pretty warm though.

Reply 38 of 49, by feipoa

User metadata
Rank l33t++
Rank
l33t++

@Tetrium
Funny you should ask that, I recently started the following post which coincides with your question, but didn't get much feedback on 'modern' graphics cards.
Fastest PCI graphics card in a 486

I have tried these but was unable to get an image from the monitor. These have been,
1. nVidia Riva TNT2 16MB
2. Matrox G450 32MB (these may only work with Intel MB chipsets)
3. nVidia GeForce6200 512MB
4. 3dfx Voodoo Banshee (released between the Voodoo2 and Voodoo3 w/2D acceleration)

The best PCI-based video card I've been able to use stably on the M919 and Biostar MB-8433UUD has been a Matrox G200 w/16MB RAM. The display is crystal clear my 1280x1024 LCD testbed. The Voodoo3 that was mentioned by swaaye on a 486 came as a surprise to me. I doubt there would be any performance boost in 2D graphics on a 486 though as compared to the Matrox G200.

The reason for the avatar is to enourage anyone with this processor to come forward. I've been looking for one on and off for more than a decade. If I cannot get one to put into use myself, I'd at least like to see some test results with proven stability. If anyone else out there has one, please PM me -- if not for sale, I'll at least offer to test it FOC to the highest degree of professionalism.

Reply 39 of 49, by feipoa

User metadata
Rank l33t++
Rank
l33t++

A general notice to overclockers, continued use in this state will eventually lead to pre-mature component failure. This is not limited to CPU overclocking, but also to RAM and cache. Northbridge, southbridge, and even superIO ckts have been known to fail prematurely when pushed beyond their limits.

I've had motherboards were the northbridge would fail slowly over time, or immediately. Once such example is where the memory controller stopped being able to address its maximum amount of RAM -- i.e. after a year of use, it could not longer address 256MB of ram, but could only address 192MB, and a few months later only 128MB of RAM. I've seen similar cases with cache ram. In all such cases, the cache and RAM timings was beyond the recommendation.

The moderate increase in temperature due to cpu overclocking encourages the growth of impurities in the semiconductor's crystaline structure, in effect causing shorts. The more dislocations your chip has from the onset, the quicker it will fail. All semiconductor devices have crystaline impurities to some extent, and in theory, they all will eventually fail depending on the yield limits imposed at the time of growth.

Just thought I'd mention this to anyone who was considering overclocking or over-voltaging a rare cpu or motherboard, or especially a Cyrix 5x86-133 -- don't!