Litecoin (and any other Scrypt-based coin) has been UNPROFITABLE to mine for GPU-based rigs for ballpark 3 years now.
ASIC took them over and are now in what is arguable the 3'd generation for ASIC units (Innosilicon A4).
2'nd gen ASIC (KNC Titan, Innosilicon A2, Alcheminer and it's private-label varients like the MAT) are still profitable though if you have fairly low cost electric.


 For a GPU based rig, depending on the specific GPUs, the "basket" of ETH / ZEC / XMR / varients on those coins is going to be the way to make any money at this time.


 Cloud mining is generally a loss for the user - the company has to charge you more than THEY make doing the actual mining or THEY lose money after all.

 I've actually made money on mining contracts through NiceHash, but the profits at BEST are very small, and you've got to be VERY carefull to watch out for all the fees involved and watch market pricing VERY closely and even then sometimes there is no room to make anything.

 Up side for me on the contracts I've done - the contract itself might be barely better than break-even, but I've got enough hardware hashpower pointed at Nicehash that I help increase my profitability on THAT enough by putting a "floor" under the price I get paid for THAT work that I end up making more THERE.

Don't waste time trying to mine Bitcoin (or any other coin using the SHA256 algorythm) on a computer.

 Even a high-end machine is SO FAR BEHIND on efficiency compared to the ASIC available as to be definitve money-losers.


 Trying to mine on a laptop is a very good way to destroy that laptop quickly, in almost all cases - Laptops are NOT designed to operate at high load factors 24/7 as mining imposes on machines.


 Using the Laptop as a USB host controller for something like  theAntminer U2 or Gridseed "Orb" / "Blade" type units where the actual MINING is done on the ASIC is a different story - that puts a very low load on the computer running the control software.

Latency matters, but not as much as you might think.

 2 years experience mining on a Exede sat connection (typical 700 ms round-trip ping, rarely as low as 660, most of that the double-hop out through geosynch) only kicked my rejects up into the 1-2% range instead of the sub-1% range I've had since my move last summer an aquisition of my first *GOOD* internet connection in about 6 years.

 ANY Relive (16.12.x) version to date has been very poor for mining - they don't get along with modded BIOS at all, and they impact hashrates NEGATIVELY by 5-10% in my testing on them.

 I recommend 16.10.1 (most recent WQHL version that isn't Relive), but 16.9.2 also worked well in my testing, and I suspect any 16.9 16.10 or 16.11 version will have competative results.

 15.12 is still a viable "go-to" version for pre-RX series cards but I've not found there to be any difference between that version and 16.10.1 on hashrates for my various R9 2xx and HD7xxx series cards.
 Definitely worth going to 16.10.1 if you ever plan to upgrade from older cards to the RX 4xxx series.

 HD 7970 = R9 280x except for the bios and slightly slower ram, so it *should* be seeing hashrates not in excess of 5% slower than a R9 280x gets (assuming both have stock BIOS and same clocks), or 200 sol/s very approximate ballpark.

 Unless you're overclocking the CPU for some reason (that CPU is SERIOUS overkill even stock for any mining rig), that should be a ton more power supply than it needs.

Hmmm, too bad I sold off my C64 ages ago.

 They DID have a couple of hard drive options.....

 (5MB on Burst isn't gonna mine even a penny worth a month though, IF you could get a miner to work on that low end of a machine).





 BTW - I've got quite a few "10 year old appx." pieces of hardware in use in my rigs - motherboards, CPUs, RAM, hard drives - but nothing near that old on GPUs, though some of the high end AMD HD5xxx and HD 6xxx series GPUs that are 7 or more years old would still be viable for mining with (very inefficiently but they would WORK).
 Some of the hard drives are probably closer to 15 years old (in design if not the specific HD).
 It's kinda funny in a way, watching you complain about "old tech" while at least one of my GTX 1070 cards pulls 380 sol/s on a Sempron 2800+ with 1 GB of DDR (400 I think but might be 333) and a 60 GB EIDE HD based system where I *KNOW* the MB CPU and RAM date from 2007 (Hard drive on that system at the moment has a date of manufacture of 2005, but it wasn't a new design at the time).

 Do keep in mind that the OP never specified ANY details about their system - some folks think anything more than "last year's model" is an "Old Computer".

 For that matter, the HD 7xxx GCN cards are about 5 years old (AMD added support for them in the Febuary or April 2012 Catalyst version) and THOSE do in fact mine quite effectively on some current coins - the R9 2xx and R9 3xx series mostly used the SAME GPU CHIPS as the HD 7xxx series, with no more than a BIOS upgrade and a small improvement in memory speed.

 One should handle a SP20 if you set the volts in the web interface fairly low - can probably get it into the 1.2-1.3 TH ballpark for mining speed.
 Again, definitely set the "power limit per connector" option - I'd go no higher than 180 per connector just to have a bit of leeway.
 

 Pair of them should handle a SP20 comfortably even if you push it as hard as possible, with some fair leeway for better efficiency on the PS side.

Point.

 Been a while but I remember that issue when I was one of the first ADSL customers for a particular ISP and also remember it being an issue at times on the ISDN connection I had before that (2 bonded 64k down, control channel at 16k also used as up).

 However, OP did specify that their download speed was crazy-low at times for a 70 Mbps (max) connection - I suspect that is their issue, not the load put on the connection by mining.

 There is one reason why I moved last summer to The Land Of Very Very Cheap Electric.

 Ballpark 28, some folks have reported up to 33 with cards with "good" memory.
 1070 are a LOT more effective on ZEC since it's proving to be quite a bit less "memory bound" than ETH is.

I wonder if any of these would be worthwhile for PoC (HD mining) type stuff?

 How much drive space do you get to use?

 Up side - it's not high CPU usage, so it probably wouldn't be a "ban" issue.

The only time I've had to worry about this, I used a spare HDMI -> VGA converter like those pictured in previous posts - did NOT need to use resistors.

 As a general rule though, it's not an issue - as I generally run mining machines on LINUX which DOES NOT HAVE THIS LAME ISSUE.

A full bitcoin node uses a MAXIMUM of 1 Megabyte every 10 minutes to download a block.

 INSIGNIFICANT.

 I was managing that *and* a multi-machine farm on a Virgin Mobile 3G cell connection for quite a while - connection was LUCKY to see 400k download speed in wee hours of the morning, more commonly 150-200k - my rejected ratio was a bit high (1-2% was the norm) but otherwise it worked fine.


 Cable is NOTORIOUS for being oversubscribed - my current cable connection ALWAYS gets very slow on Sunday afternoons and somewhat slower the entire weekend, as folks on the same "loop" tie up lots of bandwidth watching TV. THAT, OP, is probably the issue - mining bandwidth usage is VERY small.


 Luckily, cable internet for me is a backup connection, not my primary "fiber to the interface for the electric meter remote reading box, 100 Megabit Ethernet to ME" connection.

I'm not sure if that old of machine can even do AltCoin mining - if it doesn't have a PCI-E video card connection (most machines of that vintage would be AGP) it's best to just forget about mining on it and set it up as a media center machine or some such.

You can run it from 2 PS - make sure you connect each one to the SAME hash board, don't mix-and-match on a single hashboard.
 I don't recommend it as it's an added layer of management complexity that doesn't help anything.

 It needs 4 x PCI-E 6-pin connections AND a beefy +12 volt output from the power supply - most lower-end PS do NOT supply their entire rated output on +12VDC, CHECK THE SPECS FIRST!
 2 of the Corsair 550w units you posted would work - to a point.
 They're only rated 504 Watts on the +12VDC rail, so you're not going to be able to push your SP20 very hard at all.
 It'll do more than a S5 though - those only used about 600 watts at full rated speed.
 Set the "power limit" on each connector to 250 watts OR LESS in the SP20E web interface and it should work OK without overloading the power supplies, though I'd recommend 200 watts limit on each connector to give yourself some leeway, especially on lower-end PS like those.


 *DO NOT TRY TO RUN A SP20 FROM A SINGLE 450 WATT POWER SUPPPLY*

 Even at the LOWEST throttled down setting you CAN set one to, it will eat quite a bit more power than that.


 You don't control the mining frequency directly - you set the voltage to the hash chips in the web-based interface and then the software auto-optimises the frequency vs. max temp and current draw.
 Give it 5-10 minutes to "settle in" before you look at the hashrate.

 They got quite a bit more efficient as you dropped the voltage (down to the point they just stopped running), at the cost of hashrate.
 I forget what the specific set points were (I used to have a chart I'd made up for mine, but it's long gone with the SP20E I had when I sold it), but I rememeber one setpoint giving almost identical hashrate and power draw to the Antminer S5 - lower setpoints slower but MORE efficient, higher setpoins faster but LESS efficient.

 1.7 might have been achieveable - on a VERY COLD ROOM installation at the max voltage setpoint.
 Most I ever saw out of mine was 1.55 or so but that was in a room that was commonly 75-90 F.

 I ran mine off of either Seasonic X1250 or EVGA 1300 G2 for a power supply - if you're going to go lower clocks for efficiency, I'd consider the smaller options in those series rather than 2x small PS for better reliability and ease of management.

 Given the hot environment and my inherent very conservative attitude towards 24/7 PS reliability and efficiency, my actual setup had one side of the SP20E and an antire Antminer S5 running from one PS, the other side of the SP20 and another entire Antminer S5 from a second PS - which kept both PS at well under 80% of max capasity and close to their optimal point on the efficiency curve.
 (Yep, do as I say not as I did - I KNOW what a pain the multi-PS setup became to manage from personal experience).

You can mine BURST with less than 1TB - you just don't make squat.

 If the machine supports SATA 3Gb/s or newer, it probably would be viable as a BURST miner.
 It MIGHT have issues supporting drives larger than 4TB if it's not fairly current (UEFI support needed for Windows, I believe LINUX can work around that limit without UEFI if you keep the boot partition small enough and at the start of the drive).

 If it only supports the older EIDE interface, don't bother - those drives never came larger than 1TB, machines that old had issues running anything over 500GB (except under LINUX) and the drives over 500GB were rare and expen$$$$ive, and ANY EIDE drive is very old and probably on it's last legs (or is a refurb with unknown lifespan left).


 BTW - we're ALL shooting in the dark here, since OP has never posted ANY SORT OF SPECS to the "old computers" in question.

 Not for everyone. I haven't seen it come up as "most profitable" for anything I run *except* my GTX 750 Ti cards yet (those are in XP gaming machines that CAN'T run anything else AFAIK), though it's been getting semi-close on my RX 470s (though whattomine hashrate figures need to be taken with LARGE quantities of salt at times).

 On the other hand, the pool I've been running the 750 Ti cards on seems to be not accepting a TON of hashrate - any recommendations for a RELIABLE pool, preferably with a low "payout to exchange" minimum?

 Why not just run the NATIVE Windows version?

 My RX 470 Sapphire ref cards like -i 8

 I'm quite certain the 480 should like it even better.