a little bit more. as this card has a11 line routed to top memory chips, these cards can be memory expanded to 32mb two ways, as four chips 2m x 32 or two chips 4m x 32. moreover, 64mb configuration is possible with small 1.5cm length rework lines at back side. not tested 64mb yet, but very much looks like as i said
correct memory configuration in bios required of course
if we talk about most common 16mb a1 card and bios 1.05, than we have to change addres of LFB to e0000008 (pci 14h). that's it. if we want this card with this bios run faster than 143 mhz, we should fine tune draminit1 regiter. also, as this bios has ibm sub ven, sub sys id, good idea to change it to 121a 0005
finaly tested my a1 daytona. with some kind of register tweaking, found absolutely no problem with available games and tests. tested under win98 and have to say this card gives no shit what driver to use, it work as good as ordinary v4 pci card. used amigamerlin and raziel64 latest drivers and all was just fine. all quakes, hl, carma2, ut, 3dm01, no issues at all. test system is piii 1000 on bx board, but, as mentioned before, i'm very much sure that it will run fine on any system where v3-v4-v5 run good
i know how the v56k made. and in theory, very much possible to connect two v55 cards. but in practice, it's impossible due to very limited length and impedance equal wires should be used to connect vsa chips to synchonize them. in v56 all chips are daisy chained by special wirings
no surprise in price. some people expected to see about $100-150 price level, it's absolutely unrealistic. old chips are not cheap and their cost never be less. and chips in qfp package are more expensive in assembly rather than bga.
the old gets older and the young get stronger
blog about my projects started:
https://zxc64.wordpress.com/
have made rough calculations of assembly cost of this v1 replica at my domestic assembly house, it will be about $100 per single unit. just for work they want for soldering for small batches. plus cost of pcb plus cost of components (one tmu more expensive than vsa chip) plus logistics plus a little bit compensation for hundreds hours of my time. so i see no perspective in production.
by showing him a rough estimation about quantity
correct. i have to estimate estimated estimation. anyway, have strong doubts about quantity more than 50-60 units. feel free to express your opinions about product pricing
heh. i thought about some insane colors. like white )
the main problem with v56k is poor layout, there some traces must be equal lenght. aalchemy has good design by default. although i suspect that the sli connection on aalchemy is differs from v56k. or even there may be no sli connections at all. anyway inline placement for vsa chips is not a good idea. there should be symmetrical layout for optimum signalling. also may exist some impedance quirks
i understand, your guys, desire to see reasonable prices on this v1. this could be achieved by automatic assembly. post your thoughts about market demands for this product
sorry, didn't read all above, but widescreen support for dvi should be added through timing editing in inf driver file. very often existing timings not correct. for example, right timings for 1600x1200 resolutions have to be added manually, at least for win98 drivers. sil164 transmitter able to do everything, so our point is to make clear how to feed it correct data
as i thought, amd. there is some difference in memory aligment seized by card on intel and amd systems
these 2 cards sold on ebay last autumn are same. you don't indicate what system did you use for testing
Könnte doch auch jede Voodoo5 mit SLI nichts mehr anfangen, weil sie
dann auf 66MHz PCI limitiert sind. Wenn es auch Kostengründen sinnvoll
wäre, da eine PCI-E Platine billigerer als eine PCI oder AGP ist, dann
ok.
two bridges in case of v56k, pcie to pci and pci-pci, so no reason for price cut. vise versa, more complex layout
updated faq with recent questions about third tmu, v56k, cad tool
uploaded some pictures to observe. think, would be great to place these at start of this topic, so people will be able to see at the beginning what about all this mess:
[Blockierte Grafik: https://naviburg.ru/tmp/adv_sst1.jpg]
[Blockierte Grafik: https://naviburg.ru/tmp/adv_sst2.jpg]
[Blockierte Grafik: https://naviburg.ru/tmp/adv_sst3.jpg]
can't get this card to work on bx system. tried these drivers http://www.thedodgegarage.com/3dfx/downloads/vsa100_drivers.zip
after booting to win98 i see only artefacts on desktop and shortly after system freezes
looks like amd system needed
hello guys
appreciate your interest on my project
short faq:
q: it's your designed boads on youtube video?
a: surely yes, and even more, i have 3 tmu pcb design and based on it 2 tmu pcb project and have even working prototypes for 2 tmu board
q: why shown boards look different?
a: because pcbs mentioned above intendend to use 6 layers pcbs. and boards you may see at youtube designed to use 4 layers at scratch
q: what about overclocking v1 sli?
a: no, hardware limit, mentioned 20 years ago in q3d drivers. but single board will do overclock as usual
q: any possibility in memory expansion?
a: sure, but for tmus only, so max 10mb per board (4mb per tmu, 2 mb for fbi). on my board this possibility released by using two ras memory modules (1mb per chip). if going to 4mb per tmu with 0.5mb memory chips, there should be utilizing 6 layers design as you can see on any voodoo 2 board.
q: why only 2mb for fbi?
a: because of hardware limitation. on v1 only 2mb for fbi in sli mode supported
q: max resolution for boards in sli?
a: 800x600 for v1 boards. because frame buffer memory amout is doubled, so you will get 4mb fb ram in sli system
q: is it important to have 4mb per tmu?
a: no, any early dos/windows glide game using only 2mb. 4mb texture memory only makes sense on later demanding games, where v5 is better choise
q: any support for sli in dos games
a: i'd like to say yes. sli support exist in any glide driver by default, but to make sure benchmarks or other tests should be done (not done yet)
q: how you can prove you words about v1 limitations:
a: here you ought to go (it's glide2x source, all the quirks are well commented): https://sourcecodebrowser.com/glide/2002.04.…7c03549bfe.html
q: how your boards perform compare to known v1 sli configurations?
a: i have 100sb card and couple of 50sb cards. so i can claim that the perfomance is same
q: any special drivers required?
a: no, support for sli and 2 tmus and 4mb per tmu should be in any existing driver. i'm using diamond monster driver at this moment. canopus driver should be avoided due to it behaviour to detect custom chip (that only canopus boards have) on board
q: how about third tmu on v1 or v2 boards?
a: i have strong doubt concerning drivers support. looks like massive driver development required to take some profit from third tmu. what we know about card of such configuration? only primary image cards equipped with 3 tmus. and most likely they had own custom software to support all these. anyway, competent programmer should take a serious look at source codes to tell us something.
q: is any hope to v2 based projects?
a: i'm able to do pcb design, but not able to do even prototypes assembly. look at this superfine pitch at fbi chip. hate to solder it. considerable investements req'd to make automated assembly. and no warranty that all will work from first batch, so more expenses, etc, etc
q: any upcoming projects?
a: hell, yeah
q: what cad tool do you utilizing?
a: don't like to talk about it, cause did not pay for this tool. it name starts with e and ends with le. not best choise, wish to migrate to allegro or altium
q: can you make a v56k?
a: sure, give me funds and 1.5 years
q: are you serious about v56k?
of course, this card is no more than 2 v55s on own pci bus (with some quirks). i have v55 schematic and have a chance to explore v56k so the rest is technical question
q: how much?
a: eur10k for direct expenses. professional oscilloscope, reflow owen, prototypes pcb ordering, aalchemy card in any state (have to explore advanced design comparing to not the best v56k layout) and few things more
q: how to buy your v1?
a: will place pair of these cards on ebay auction in near future to evaluate interest. in urgent cases contact me in pm
Vielen Dank MadYoshi!
Anthony, sure that would be very interesting. Do you have any surprising experiences to share when it comes to using the Daytona, for example the drivers? Do you use the same one?
yep, the same one, from one source. will put my hands on this card in near future and if there will be interesting experience share it for sure
i have plans to increase memory on my daytona to 32mb. if you wish, i could post my results in benchmarks when it'll be done, and you'll update you article
inquire chinese suppliers. afaik they have SM81C256K16CJ-8
yep, sooner or later it will be
these are prototypes, minor reworks required, though fully functional. unfortunately i don't have enough original samtec connectors to build another couple of these. if you'd be able to buy such type of connectors (i'll tell where) and send it to me, we could make a deal. pm, if interested
tested fully equipped sb100 in q1 and q2, in q1 sb100 scores 97-98 fps. single v2 scores 107. in q2 66 and 90 fps respectively. piii-1000 mhz used. resolution 640x480
