Відео

I am not really sure on what is your problem there. To update your bios you need a flash program (award flash, to do it while the PC runs), or a programmer (to do it offline by removing and reprogramming the chip). The video and the tools here are about modding the bios files to add modify some options, not related to programming or updating the bios. Latest bios probably available retro web, but first check that version you have.

@@atheatos its the very first family computer so i doubt it was never done. Could this bios modding tool assist in updating to its 1997 release according to rw? Or do you think awardFlash is enough without needing this modding tool?

Moding tools are not for flashing. As said, these are used to edit bios files (add maybe options) before flashing. AwardFalsh is enough for update, but again check first what version you have. The date is stated on boot.

@@atheatos Thank you so much for the clarification. I needed this advice for quite some time as i had to put my socket 7 project on hold because of this. 😅

The Cyrix 486DLC has 1KB of Cache. It is just that some tools do not detect this small amount of cache. But the cache is there and has a significant difference when you have it enabled. It is the same with 1KB cache clock double CPUs. 8KB ones are detected by the tools.

I have no ability to do LN2. I had tried with a Peltier and had no real effect there. Also it looks like 55MHz is not only the limit for the CPU but for other parts of the system. I think in general voltage is more important there. The maximum I tried was 5.4v, 5.5v. It could be that even 6v is safe but in general I do not want to risk it as this goes to all the components not only the CPU.

The thing is, If I want to play Doom I for sure do not do this on my 386. And then with normal doom is better / easier for others to compare with my results. However, all this is interesting so I will try it at some point.

This is not possible with AMIBCP. You need a tool called "Bios Patcher" (rom.by).

@@atheatos I think Bios Patcher works for Award bios, not AMI

You are right. I confused AMIBCP and MODBIN. Either way... I do not think AMIBCP can do it and I do not know something to patch AMI Bios for this.

I think I also remember that there was performance hit when with more that 256MB of ram. Now normally with a system like that, 256MB of ram is already more than enough (most people will use it with win win98, max winXP). So I just used 256MB and did my job. Unfortunately I never investigated in depth the cacheability topic. I might check in the future. For the tags... this has nothing to do with Asus. According to the chipset datasheet 1MB cache is only possible with the external tag installed. Now excluding all the mods with 10bit external tag as this means 2 external tag chips that nearly no M/B has, there are two options: 1) Only external tag 8bit that limits cacheability to 128MB to all cache sizes. 2) Both tags enabled that should be 4GB for 1MB cache. Combined tag is 12bit. As 128MB vs 256MB, I did not notice any performance difference, the M/B is probably in the 2) case. Also looking at the WPCRs. register 40 (val 53) and 41 (val 08), both tags are marked as enabled on the 1MB board. So either the 4GB cacheability claim is not actually true or there is some other reason that cuts performance above 256MB of ram. I need to check more, last time I worked on this board was 2 years ago. This is just what I remember.

@@atheatos thank you. I don't believe 4GB cacheability claim is actually true in the 1MB cache type as there's a performance hit when with more that 256MB of ram. In the other hand the 512kb cache one improves performance with 512MB even 768MB ram. I've ever seen some programs which can detect the cacheable range, I don't know if that's true. I'll be happy if you investigate about that.

There is the cache check tool in DOS. It is true, but the more Ram you have the longer it takes to run. It will take a few hours probably with 512MB :/ I will only know more if I do tests again with this M/B. But in the end 256MB is fine for me.

​@@atheatos It would be great, if you compared the 1MB cache with 512KB one with 512MB Ram. I´ve already done it, it seems to be about 12% faster the 512KB cache version than the 1MB one doing Sifsoft Sandra 2007 SP1 Memory - Cache test and 7% faster the 512KB cache version doing 3Dmark2001 test. I would be very grateful with you if you do the double check about that. Best wishes.

I will include a section about cacheability on my next video about this board.

Yeah there is still some room for improvement there. One or the other way I have now the 486SXL (cyrix 2x40, 8KB cache). I also have many VGAs that are probably faster like the WD90C31 (I just had some incompatibility problems there). IBM BL is a different story all together. It is undocumented and to my understanding not 100% compatible with these M/B. However I have a motherboard with it so I can maybe run some benches and compare.

No, these are are drop-in replacements for 386DX. However enabling the cache without problems depends on the M/B circuits. And a bios that is aware of CPUs like that could save you from having to use the Cyrix tool.

The CPU is fine. For sure it supports clock doubling. The problem most likely here is an incompatibility of this clock doubling feature with the motherboard (with the M/B chipset or the bios). Others tried on the same M/B failed the same way, while this works on other M/Bs.

Very hard to tell, very hard to reverse to this level. I do not have equipment to try something like this :/ My guess is that this is not possible. The PLL there is probably fixed at 2.5x. Then this is just a safety thing that disables the PLL and so it is clocked directly from the input 1x.

This is a 64KB Bios. More or less all motherboards like that use a 27C512 Chip. Brand, window or not does not matter.

@@atheatos That's exactly what I tried, I programmed an M27C512, used both of the Roms at Retro Web but neither worked, the post card just shows 4 dashes. Popped the Good chip in and it posted. None of the Chip types I tried with the TL866 can read the good Bios chip ( I was hoping to dump the Bios). I rewatched your video about Bios Modding and I'm wondering if this good bios chip is a 'one time programmable' perhaps that's why it's impossible to find online. The M27C512 I'm using is from a parts board (I've never seen it actually functioning in a running machine) so I'm going to try programming it for another machine and see if it will work in that, if not I'll try another 27C512. Really appreciate your reply Good Sir. I'll be sure to credit your assistance in my next video concerning this. Thank you

If you program and verify your chip with your programmer, it should be fine. So my guess is that the roms you downloaded are just incompatible with your board. Unfortunately many OEM of this era used custom ordered one time programable chips with custom prints for their motherboards. So it is impossible to tell the real model / brand of these chips. However for sure these are one of the many 27c512 listed in Xgpro tool. Just try all of this and try read your chip. If you get an invalid ID massage you can still use the option to ignore the ID and try read it. Or You can see the ID and go online and check it. There is a site "EPROM Device ID List" with a bit list of IDs and eeprom models. Then an other approach is maybe to use some dos utility to dump your bios into a file. Search for "dosdays BIOS Utilities".

@@atheatos Fantastic, I will try all the above until I get that rom dumped and transferred. I read on a forum there is a potential to damage the Bios chip by reading it with the wrong Chip selected, but you see a lot of half-baked opinions regarding these things. So far after trying 6 or more types my chip is still functioning. You think there is any truth to the programmer damaging a Bios chip with a Read?

As you predicted, Yes indeed my M27C512 Chip is fully functional in my Fortiva 5000 machine. On to the next step. Thanks again.

Nice :)

True, this (offset 41) should be changed from 0... to 4... This sets the 6th bit, and enables "Fast DRAM Read". This increases performance, but is only stable when the M/B cache is disabled. The model or cache size does not matter, as long as it is the same chipset. That is for all these opts. Happy tuning ^^.

@@atheatos Thank you! Those Ali AGP 1.9 drivers made my 3d Mark 99 score go up 1000 points! Almost 4000 total now. 😁 I also did a couple of those tweaks from wpcredit which helped mem bandwidth & latency quite nicely 👌 I haven't done any of the command line stuff in my autoexec yet. Finally got over 2000 in 3d mark 2000 👍 Currently running a gigabyte ga 5ax rev 5.2 with a K6III+ @ 600 (120x5@2.2v) And a voodoo 3500. I have a p5a 512kb 1.06 as well sitting in storage. Going to do the 1k ohm mod on it and tinker to see how it compares to my ga-5ax next week. Question: I have the following agp gpus available... Which would be fastest in the ali chipset with these two boards I have... And which would you run: Riva TNT2 Ultra GeForce 2 MX400 GeForce 3 Ti500 Radeon 9700 Pro Or keep the voodoo3 3500?

Nice results there :) Fastest overall and best option is the "GeForce 3 Ti500" combined with some old driver (7.76??). Especially this is true in benchmarks. Big difference from the Voodoo. Now some games with the Glide API might run faster on the Voodoo, so hard to tell. Personally I do not have stable configurations and PC cases/ full pcs etc. So normally I would had just tested all for fun, make maybe and video then put away.

Probably true. The clock doubler not working is probably the same thing too. I have a few M/Bs with the RC2016A5 chipset now. I might try there too. I have also some RC2016A5 chips I might try swapping chips.

I added a link to a Vogons thread with what you are looking for. I might make a packet with everything for these Cyrix CPUs in the future. I just do not have these files not on my windows machine right now.

Cyrix / TI ones did not license from Intel, but to my understanding these are their own 486 core just for 386 socket (they never had a 386). IBM is always confused with these but it is actually totally different. They licensed the 486 core directly from Intel. The pin out is mostly compatible with 386sx, but not 100% as there are extra pins regarding the cache (unfortunately undocumented). So unless the datasheet is found... or someone reverses a M/B with an IBM chip... modding one into a 386sx M/B with caches working is impossible.

Thank you for the comment :) More videos to come at some point soon.

Unfortunately, not much has changed here. I just too busy with my day job and other simpler projects right now. I also want to improve this further. This will take some time so please be patience. Check also my second video on this.

I have never seen a board with an 486sxlc2 all ready soldered in. You can take any 386sx M/B and try a mod maybe. As for my mod / little adapter boards, I only sell these to very few people that have exactly my Model of M/B and the skills to solder. Regarding making an 486sxlc2 M/B with local SRAM chips for ram, yeah this is on my scope. It will be nice for sure. The main issue is to find a suitable chipset with a datasheet and no intergraded memory controller. Then the problem is the Bios. I can find solutions to all that but it will take time to design this.

I'd like to get your board and 486sxlc2 to challenge my soldering skills /*I have ALI M1217*/ Can I ?@@atheatos

But 486sxlc2 with local SRAM will not give enough boost, as 133MHz ARM or 166Mhz ESP. Amiga users made PIstorm, PC users also could... Is it possible to ignore the m/b bios /*there are few open-source bioses for QEMU/Bosh*/and DRAM controller and use only CPU pins to access ISA bus?@@atheatos

So far I did not had the time to research this in depth but it looks like the main problem is the intergraded DMA + memory controller. If you put local Sram for RAM, then DMA transfers will still go to the memory controller. So I think this needs a barebone chipset like from the 286 era that do not intergrade a memory controller, Using the socket to just access the busses will not work (probably). Regarding performance difference it is very hard to tell, all 16bit bus 386, 486 chips are bottlenecked massively by the low bus thought put. So this could be probably performing up to a normal 486SX at same speed with Dram. I am just looking making my own mini boards in general, and this is one of the options. I need to check more the bios thing also, it is a bit outside my skills.

I also bought about it... so: #0 how 486SLC cache works then with DMA? There must be a solution to make SRAM on PIstorm board too #1 even if we have problems with #0 all RAM above 16Mb /*fast-ram*/ is not accessible by ISA DMA could be made SRAM without a problem, and 16MB DRAM on mother-board /*slow-ram*/ - similar like in AMIGA.@@atheatos

Thank you! I checked you video on this topic also a few days ago. It is nice you are also checking older platforms now.

@@atheatos I've started doing a bit of research and Oh My there are so many variants and versions to explore. Very interesting but also confusing! Some seem to be Cyrix CPUs in disguise and other interesting nuggets of details.

I have now also the 486SXL2 (8KB L1 version of the 486DLC). Managed to run this also at 55MHz and 80MHz+ (X2 mul). The performance is very 486 like... ok with the limitation of the ISA VGA. I do not know when I will have the time to make this video. Quite busy these days :/

@@atheatos Amazing! The clock doubled ones are really hard to find it seems.

yeah it is nearly impossible to find the 486sxl2 132p (386 socket) version, I do not have this either. But there is also a 168p (486 socket) one. It is way easier to find this one and I have made an adapter to 132p for this. I also have now the rare 486Drx2 (486DLC with x2 mul), not tested this one yet.

XD

wow 450MHz very nice! I have a few of these, most do not go above easily. My best is this on the video.

@atheatos I got the freqency with FSB 110 (interpreted as 112) * 4 on my Asus P5A with 512kb Cache. V-Core is 2,4V. Test with 2.2V currently not tested but comming soon. More Power is not possible. My pci latency timer is working with 128. That is the only the different to your P5A. 128MB PC133 CL2, Voodoo3 3000 @180, A short sumery of my CPU performens. 3DMark 99 CPU: 2927 PCMark2002: 773

Very nice!

Nice suggestion, I will check.

Not sure, probably not possible. The 386/486 usually have something like 27c512 ROM chips. Check your programmer support list.

Totally true.

Thank you for the information. I will have to check a few things again then. For OP1 I am 100% sure it connects to the FLUSH# input of the CPU. It is also marked as an option for Cyrix CPU on the M/B PCB. Without that for sure the internal cache do not work correctly. I do not remember if I checked the A20 line to be connected. In any case, with my setup all was fine except the clock doubling. I have also more M/Bs to check now, we will see.

I do not have one, probably not any time soon. These are too expensive and rare, also in the end while faster these have very limited support, so I do not have a priority to get one now.

Unfortunately I do not have a Weitek, very expensive and rare. CPU Galaxy has done this test with a Rapidcad + Weitek. I do not have a Rapidcad, same reason.

I would try, not sure as I have never run these, probably ok. My system is stable.

yeah, a very strange problem... I do not know probably incompatibility with the chipset, I will have to check on other motherboards too.

I agree all these comparisons could be very interesting and I have more than enough systems to do this! The only problem is that to make it fair I would have to optimize the other systems too and check a-lot of stuff. On top of this I am very busy these days and I have no time to make videos :/ But I will come back to this at some point.

Interesting... I am not Finish, I am Greek ^^. Regarding DRAM / SRAM, I have made something like that actually (simm-30s with SRAM). I just have no time to make videos now :/ Long story short unless you replace the memory controller there is no benefit as refresh can not be completely disabled. Only a custom M/B chipset and benefit from something like that. I might try this also in the future.

I have so many new motherboards myself now too. Just absolutely no time for anything right now :/ On top of this, I have accumulated so much hardware I do not know what I have and where. I am trying to organize now. I have a few of IBM SLCs. These are for sure totally deferent. Based on original 486 core by intel. The biggest problem for me is the lack of documentation / pinout info. I will come back to videos at some point, I have so much to do and show to the people. :/

@@atheatos Very nice work you did here indeed with those little interposer boards 👌 Yeah, there's plenty of documentation on TI/Cyrix chips but IBM has always been very secretive about their designs and didn't share a lot because they didn't need to. You know: Those TI chips were meant for clone PC builds so they had a lot of incentive to share the way to implement them because it was directly linked to their sales. IBM was on the other side of the spectrum though. They were sick of the vast amount of clones eating up their market share and they very much wanted to dominate the PC industry again like they did with the 5150, 5160 and 5170 (their very succesful 1st gen/old school PCs) so they locked down every bit of low level technical detail and wanted to produce their own CPUs.. Even their technical reference manuals were very limited in terms of design decisions.. They really didn't want anyone to steal their designs again and hence they were very protective, which is of course understandible but which is also why I have great respect for other companies like Compaq who blew the clone market wide open with their crusade for open system standards 👍👍 It's the guys from Compaq who initiated IBM's demise in the desktop PC market but that was inevitable in the end as they opened the doors ten years before by designing their first PC by using of the shelf parts.. The irony 😄

I think I got mine up to like 680 (or something weird because of the clock multiplier). Mine ran like total shit though and i wound up underclocking to 500MHz just to keep it stable under win2k. I still have the board, would be fun to see what it could do with a proper modern CPU cooler.

55MHz is more or less the limit here, both for my M/B and 386 CPUs. It might be possible to go higher on another board + with LN on the CPU. However I do not think 100MHz or anything near this is possible.

Well, we are actually talking quite often as we are basically live in the same city in Austria... But ok... I do not think this could ever happen. I think he told me he had to bin over 50 CPUs to find this 200MHz one. So based also on my tests.. it is safe to say that these are very rare. Also my M/B might be modded as hell, but in the end his M/B is better as it supports EDO ram that is a bit faster. Borrowing his CPU is improbable, but we might cooperate in the future. Maybe I mod his M/B :p

@@atheatos yes, volt mod and pll mod. His graphics card should also handle it. How would it stack up to against some of the recent records posted on Vogons?

As for right now we are probably a bit behind compared to what is posted on Vogons. We might have a chance to catch up. I also have now more M/Bs and mods to test. However the real problem is what we are now just super busy with our day jobs. So yeah...

I am aware of week/year codes on chips... it is funny how I missed this point on the video. Now 9507 is the latest one the and it is not the fastest. In total we have: 9507AK1 (The one I used on most on my videos) 9419A3D (The new to my collection and a bit faster, small difference) 9345A3D (Same as the one above) 9250PA (The CPU galaxy one, the slowest with significant difference?) So the week/ year it is partially helpful here. The most interesting stuff is the maybe the code after that AK1, A3D, PA. Yeah normally this is the plant location or something, but it is interesting that only the ones with A3D performed the same. If I ever get more of this I will check again.

There is indeed a planed follow up video to this, but this is now a victim of a long hiatus. The video is actually half made and stored on my disk for over a year now :/ Just super busy with my day work and my personal life after the pandemic :/ Very sad.