It doesn’t work and we’re not surprised considering the can of worms that comes with RAM addressing. Right off the bat we assume timing problems due to variance in the trace lengths and EM issues. But you have to hand it to [cyandyedeyecandy] for even trying. The self-proclaimed upgrade seeks to readjust how the DIMM works without changing the edge pinout.
The stick shown here is a 512 MB module that, because of the computer using it (unspecified in the post), is only allowing access to 256 MB. The added chips and free-form circuit make up an AND for the chip-select line, and flip-flop for the bank address.
The post is a gorgeous cry for help. We already weighed in from the peanut gallery at the top (seriously, that’s somewhat baseless guessing) so step up to the computer-engineering plate and let us know what needs to be done to make this most-awesome-of-non-working hacks actually work.
Once you’ve figured this out, here’s another one to scratch at your brain with.
It may not work, but that’s a gorgeous bit of airwiring. Give it a brass sheen and it’d look like a Steampunk DIMM.
Doing this with the address lines only works if it is simple RAM like static RAM which has no pipelining or command interfaces. SDRAM requires a whole series of operations to control it and doing this mod means you’re basically sending two interleaved sets of commands to a single set of RAM chips.
I wouldn’t be surprised if the tiny delay introduced by the extra logic is also throwing it off.
What might work better is to put two DIMMs together and multiplex the data lines. This way any updates and row selects will be sent to both modules. Basically making a 64-bit DIMM and selecting either upper or lower 32 bits of it.
FYI: http://en.wikipedia.org/wiki/DIMM
>Another difference is that standard SIMMs have a 32-bit data path, while standard DIMMs have a 64-bit data path.
O M G that soldering! it is beautiful! I want to know this guys equipment and technique. Would be great if he recoded himself soldering this stuff (no doubt thru microscope)
I have a stereo microscope and I can’t solder half that neatly. I too would like to see a video.
From the Reddit thread:
“I thought the soldering was pretty bad really XD… I normally do much better when there are no wires involved: http://i.imgur.com/I117E5G.jpg
I use a 50W Weller station with LT-GW tip, temperature at 300°C, normal 60/40 solder, and rosin-based no-clean flux.
There’s not much of a technique to it. Put a drop of solder on the wire ends and the iron, drown everything in flux, hold the wire in place with tweezers and then hold the iron to the wire for < 2 seconds or so. To remove excess solder, I only use the hollow part of the tip, then brush it off on the sponge."
The submitter is apparently having clock signal problems, as April Fool’s was Wednesday. But better late than never!
Seriously, registered RAM, and such problems? The only problem with over-ranked ECC REG I had was that it never worked in a first place.
Congratulations, for courage and patience. Learning takes a lot of both. Still awesome if it’s a joke :)
Back in old times, when electronics were very expensive, I tried adding more RAM to Commodore 128 D(CR), without understanding how DRAM works. Documentation? Just a book in french, with a *lot* of typos (La Bible du Commodore 128), french dictionary, and determination. C128D(CR) works still, but I released the blue smoke from some 6526s and 8520s when playing with serial communications.
how is it refreshing the dram?
12 address bits, interleaved (via RAS and CAS signals) to form a 24 bit address. 64 bits of data returned (plus 8 parity bits). To expand from 128MB to 512MB you need to add an extra address bit, 13 bits total, and you need to interleave the two new new address bits from the CPU.
Would be far easier to find a DIMM with chips of 1/2 the capacity and 2x as many of them. That often works on computers that *should* support more RAM but don’t.
If there’s a RAM ceiling in the ROM or BIOS then you’re SOL unless you or someone else knows how to hack that. Apple was infamous for artificial RAM limits in some of their mid and low end Macintosh models, and also known for some models like the IIci that could be expanded to a what was then the crazy amount of 128 meg, in a system that at release date was only available with up to 4 meg.
Many PCs are hard limited by their motherboard chipsets. I’ve run into several than can only address 3.25 gig even with a 64bit CPU and OS. I’ve also had one laptop where some arcane trickery had to be involved because it could address *and use* a full 4 gig while running 32 bit Win XP. (It had dedicated VRAM for the GPU.) Dunno how because at most XP is only supposed to be capable of using up to 3.5 gig with 3.25 gig typical. I gave it a Core 2 Duo brain transplant and a Win 7 upgrade.
PAE support ? I think it can be enabled in XP.
Interesting mental challenge… Is it possible with (assumed) SDRAM? Gonna mull it over a bit.
Definitely feasible with older memory technologies… but now with configuration-registers, non-address-related commands, etc. that use the address-lines, it’d be a bit more difficult.
And, agreed, that soldering/wiring job is pretty nifty.