Beiträge von osckhar

    Hello,

    Anthony, Correct. My modded Daytona IBM 64Mb DDR SGRAM works perfect using 4 banks with 16Mb each chip with a total 64Mb, and yes, you need to connect rear side A11 pin memory to chip daytona. If you want to use only 2 front chips with 16Mb each chip for total 32Mb, you dont need to do any rework.

    I did not modify directly register 'on fly' if not I did all the work via STRAPPING, difficult work since without Daytona Databook, you must guess what are the correct points but with patience and checking PCI18 IO data you can start to check bits in dramInit0.

    As you declarated before of original IBM does not work correctly since membase PCI14 is set up with only 64Mb and for perfect function it needs to be 128Mb.

    I still working on IBM Daytona since my main idea is to mod the daytona IBM with 128Mb- I have in mind some crazy idea for doing it - komponent style - , but not sure if it will work though. Just waiting for new Qtsop memory.

    Anyway, a Daytona with 64Mb is not bad option. You can play it 1280x1020 2xFSAA. About O/C, with new memory 4ns I am awaiting I hope to reach more than 220MHz- actually I use memory of 5ns and it reached 190-195MHz estable. Daytona chip O/C as the hell, just you need to find correct memory. I own an original Daytona 4-2 4800 agp 32Mb with not heatsink and it works with 0 problem over 220MHz and it has not any VMod or similar and it uses standar dramInit0 and dramInit1 register used on daytona bios 1.06 or 1.05c.

    Regards,

    Oscar.

    Hello,

    Seems Bernd's 55000 will reach 210MHz. At the moment 200MHz stable looping demo with not VMOD.

    ketfri8002, I did some progress and added more modd's, AGP4x support, VSA100 3rd rev, panel control with jumper cap for setting different VSA100 voltage 2.8/3.0/3.1xVSA100 and new m9dded heatsink built by me.

    [Blockierte Grafik: http://i.imgur.com/5tFEZjd.jpg]

    [Blockierte Grafik: http://i.imgur.com/5YyS9C1.jpg]

    Yupe, card looks amazing. : )

    SK1, I dont know but when I test modded card with not heatsink using 3RD VSA100 rev. with not heatsink and only a12cm big fan blowing air onto the chips card handles 185MHz clock speed.

    Regards,
    Oscar.

    Hello at all,

    Tweakstone, thanks!!!
    @stivy, no problem. We can talk about tech details here. There are 2 important points, the theory and practice. Theory says a perfect BGA reworks has a curve with a maximum point of heat of arround 220ºC when we are talking about leaded balls but since I dont like stress the PCB with high temperature and I use a profile where max temperature reach not more than 200ºC enough for soldering and desoldering any BGA IC. Later is very important you know very well your BGA MACHINE rework- it means a lot of hours of practice and checking different values for using the best profile you needs.

    Yes, high temperatura while you are doing a reflow at a BGA chip can damage it since it can joint different pads. So be ware with reflows since it needs more heat.

    Edited 26/03/2027 - I forgot to say I used new and original VSA100 since is not a worth to reball the original chip because the shorts can damage the chips and for reworks I prefer to use new ones. Anyway I have reballed used VSA100 and Avenger or even Rampage chip for my prototypes with not problem. Yes, renalling is an option too.

    Kaitou, Here I am for fixing your V5 but not always I will be to able to fix them. ; )

    Regards,
    Oscar.

    Thanks @GAT for the info. : )

    I started again to work on 3400 A3 and reemplaced MASTER VSA100 CHIP.

    TAKING OUT MASTER VSA100 CHIP
    [Blockierte Grafik: http://i66.tinypic.com/2a4ppia.jpg]

    CLEANING BGA PADS
    [Blockierte Grafik: http://i67.tinypic.com/rthsm0.jpg]
    [Blockierte Grafik: http://i64.tinypic.com/xp93pc.jpg]

    STARTING PROCESS BGA REWORK
    [Blockierte Grafik: http://i63.tinypic.com/2ebfsw4.jpg]

    PERFECT ENDED BGA REWORK
    [Blockierte Grafik: http://i64.tinypic.com/2115q9d.jpg]

    BOOTING UP 3400 A3
    [Blockierte Grafik: http://i66.tinypic.com/11ieno2.jpg]

    QUAKE 3 FAST MODE WITH NOT LINES ERRORS.
    [Blockierte Grafik: http://i65.tinypic.com/684j29.jpg]

    Very nice feeling when later a BGA REWORK you see again the card is booting up without problems - each BGA rework is a risk - since I dont have here stuff for checking if process is correct and if there are not balls in short or some ball is not soldered correctly - I would need a X-Ray machine - and when I run a Quake 3 in fast mode or 2xFSAA and lines vanish- not words here. Amazing feeling of 'peace'. 8)

    Another 3Dfx V5-6000 fixed. :respekt:

    Hello guys,

    Later a lot of hours testing component by component checking smd IC components controll 2D engine and all worked perfect seems I found the problem.

    As you can see in the pictures there are 2 pads in short on VSA100. GND+MA_A6 and it can cause the lines on memory. Such pad make reference to Memory Address Bus:

    [i]"This multiplexed bus supplies the address to the SGRAM/
    SDRAMs providing MD[63:0].
    The exact connections of the MA, MBA, and MCS lines to the SGRAM/SDRAM inputs
    depends on the memory configuration. See Chapter 5

    [Blockierte Grafik: http://i65.tinypic.com/vsc35y.jpg]

    Is it time for [b]BGA rework??? 8)

    ps. Why picture is showed as link and not as picture? In the editor message picture is showed but once posted it is showed as address... ?(

    Regards,
    Oscar.

    Agent500,

    Thank you so much for posting images from 3400 N.62. I have been looking them without luck. I also lost the clue once card is sold on ebay back 2003 year. I contacted with auction winner and he told me " I won the auction but never paid for the card." I still trying to catch it.

    I had really problems tracking all 3700 A but finally I did- the only card missing is 3400 N.62 - Idont know who own it and where is the card...

    Though I catch N.58 & N.63.

    [Blockierte Grafik: https://dl.dropboxusercontent.com/u/4383540/3dfx/3dfx_museum/V5-6000/6k_N58_N6.jpg]

    Regards,
    Oscar.

    Hi at all,

    This topic is very familiar at me. We tried re-build a V5-6000 even we alredy had a company here on Spain for doing the engineering reverse but later some talks with Nvidia's lawyers we decided stop the project. I prefer to sleep quietly and money needed is very important without nvidia's agree... it is a risk.

    What I got in clear is, build a batch of 10 samples is not the problem. They can't to check it but this topic can be a boom and if people starts to be interested in such cards not sure how Nvidia law can atc.

    About VSA-100 chips. Be ware with USA / ASIAN market. I found a lot of them comes re-balled and the most funny. They comes with lead free balls. ROHS Normative back year 2000? LoL
    Go for 355-0026-320 date code 0034-0036. Since my experience they are the best. I got them for arround 20Euros per chip shipment at home. Possibly a big batch can low the price.

    Very good luck in this crazy project. Any kind of help here I am.

    Regards,
    Oscar.

    Zitat

    I think your card died for something else, maybe was already malfunctioning ... or something went wrong during the trip. The weird performances of which you are speaking seems to be a clue in that sense. Moreover, the A1 you owned for a while was clearly a 'prototype'. It had the NFR stickers that proved that was an ES card, so we have no info about the way that card worked or not. These A2's instead, dunno if all the A2's, but these which came out from eBay lately aren't ES cards. They don't have NFR stickers but have even "warranty" stickers instead.

    I was wrong. I was checking my notes and I saw my dead card was an A2 and not A1 and it never had NFR stickers. I dont know if previous owner removed them. Rolo01 has another A2, we got them from same source and he not had luck with his card. It only worked ok 3.3V At least It is what I remember. Though sloomy owns an A1 and seems he handled to run the card correctly with an Intel motherboard 1.5V.

    Here my dead V5 agp4x A2 card.
    https://dl.dropboxusercontent.com/u/4383540/V5_A…4x_3Dfx_FAN.jpg
    https://dl.dropboxusercontent.com/u/4383540/V5_AGP4x/v5_4x_BACK.jpg

    [Blockierte Grafik: https://dl.dropboxusercontent.com/u/4383540/V5_AGP4x/V5_AGP4x_POWERSTRIP.jpg]

    Zitat

    Actually, to check if the card is wired to accept 1.5V signal you should expect 0 on the display of your multimeter, not 1.


    Right, it must be 0 and not 1.

    Hoping what Dominik tell us later he can do hard test with the card.

    Regards,
    Oscar.

    Hi Oscar, the card runs fine and very cool, no instability issues at all. It runs with 1.5v (meassured this on pin B40). The 3DMark99 MAX scores are normal and the cpu score also normal for an A64 system. Only 3DMark2000 scores are too low... mainly because of way too low CPU scores, which I think is a driver bug or a 3DMark related problem.


    Hi Dominik,

    Oky, please test PINA2 and let me know. Perhaps I am wrong.

    Regards,
    Oscar.

    Be ware with V5 AGP4x. I am sure card is not fully compatible AGP 1.5V. With my ABIT IS7 I handled to get it working but it never run ok. All the scores in test bench was very low and more when FSAA was enabled. Finally card dead for NOT right voltage- this is my opinion. Well, my card was an A1, but I think A2 will behaviour in same manner.

    There are others cards looks with an universal AGP4x as PROHET 4500 card but it only work 3.3V. There is a manner to check it via Multimeter. Here it worked fine but not sure if it is a definitive and real manner to test it. Check PINA2 to GROUND with a multimeter set 0hom. 0 mean 3.3V and 1 means 1.5V compatible.

    Good luck,
    Oscar.

    Main problem is HW engineers worked on rampage project do not remember since it does 13 years ago. I am requesting info about those test jumpers.

    Yes, once we have more info Gary will update his Rampage page section.

    Regards,
    Oscar.