One reasonable way to do this would be to take some action involving this ETH address you've been including in epools_example.txt for the past few years?

0x008c26f3a2Ca8bdC11e5891e0278c9436B6F5d1E

Perhaps announce that you'll be moving a specific amount of ETH at a specific time, and then follow through?
It seems to have been the same address going back to 2018, so should serve as solid proof that you control both this Bitcointalk account and that particular ETH account.

The one aspect of AMD Mem Tweaker that seems ideally suited to include is the new REF adjustment (aka rxboost in claymore).

It's just a single value so not too complex for the end user, it can offer substantial performance boosts, and unlike the memory timing adjustments it's not specific to any particular memory type.
It also acts independently of BIOS timing mods.

I seem to have found a conflict with HWInfo, might be worth mentioning to save others hassle troubleshooting.

It seems if HWinfo has been opened in sensors mode it prevents AMDMemTweak from applying any timing changes until the system is rebooted.  
Even after quitting HWinfo any changes in the timing sections still will not apply, they just revert back to the previous value when I press Apply.
Changes to the Clock/Voltage/Fan Control sections still work, just not the timings or REF adjustments.

Any chance of adding a GPU Memory Errors counter to the tool somewhere so I can ditch HWinfo?
Or do you know of a tool that will show memory errors and does not conflict?

Can confirm, low RAM rigs seem to have issues unless given TONS of virtual memory.
I had issues with some 5 GPU rigs with 4GB RAM and 16GB virtual memory, bumped the virtual memory up to 24GB and everything's fine.

I noticed GPU-Z reports a VRM efficiency of just ~70% for my newest cards (MSI 580), all while my MSI RX 470 report a VRM efficiency of ~90%
it's not something I've paid attention to before, but it seems the reported VRM efficiency drops when you undervolt.

It's possible this is just calculations being skewed by the undervolt and not a real efficiency drop, I've not tested at-the-wall impact.

I too could not get official driver to work with Win7, but the unofficial robinh00d driver works.

Afterburner beta #16 gave me clockspeed and voltage control for RX 470 and 480 using the robinh00d drivers
(have not tried with 570 yet)

Voltage changes are not applied immediately, takes effect when mining is paused/restarted.
Also it seems that changes to clocks end up resetting the effective voltage back to default, so make the voltage changes the final thing you apply.

First rig I've attempted to update isn't working, Claymore crashes at

start building OpenCL program for GPU 0...

Seems to be a mix of the app force quitting, or the entire system crashing & rebooting.

Rig is running Win 7 with 4x 470 and 1x 380, with latest DDU used before driver install and pixel patcher + 6GPU mod used afterwards.
Same result using either ETH 9.8 (ETH-only mode) or ZEC 12.6

All cards are at default clocks, voltage, timings. Tried each card individually using -di flags, same result.

Neat, apparently these now come in 6 slot, 8 slot and 12 slot versions.

TB250-BTC is 6 slot
TB250-BTC+ is 8 slot
TB250-BTC Pro is 12 slot

Although, it seems they also call the 6 slot TB250-BTC the Pro in some contexts (the 6 slot ones I have say PRO on the box in giant letters) so be careful

Some of the newer forks of PBE seem to resolve the garbling issue on cards with BIOSes with timings for multiple RAM types.
https://github.com/jaschaknack/PolarisBiosEditor

as for wattage, low core clocks & low voltages are the way to go, although I've not gotten around to applying at a BIOS level myself.

4 GPUs is usually rather easy to get working, the BIOS changes you mention are usually only needed for 5-7 GPU builds with skylake/kabylake. (perhaps try with optimized defaults?)
I would start with a single GPU on riser in the main PCI-E x16 slot, verify it works, second card to the other full slot, then add the small ones.

MSI's forum support considers mining an illicit activity, they won't provide any support. (and their answers are often outright false)

Here's what your motherboard's manual says:

2x PCIe 3.0 x16 slots (supports x16/x4 mode)*
4x PCIe 3.0 x1 slots**
* The PCI_E4 slot will be unavailable when an M.2 SSD module has been
installed in the M.2_2 slot.
** The PCI_E2 slot will be unavailable when an expansion card has been
installed in the PCI_E5 slot.
** The PCI_E3 slot will be unavailable when an expansion card has been
installed in the PCI_E6 slot.

So, the PCI-E x1 slots use shared lanes and you can only use two in total, and if you use a m2 SSD it disables the second full length slot.
Getting four cards running should be possible using both full-length PCI-E x16 slots and then any two of the x1 slots.

EVGA G2? Yep, the sockets are the same other than labels.
Seems to be common for EVGA PSUs, I've noticed it with G1, G2, GS and PS series. (all of which conveniently use the same modular molex cable pinout)

So long as you use a quality extension cord of sufficient wire gauge you should be fine, just try to minimize excess length as the longer the run the thicker gauge you will need.
Or you could go with longer cables direct to the PSU (C13 connector), I got some nice thick 10ft and 15ft ones from monoprice when I was picking up PDU cabling.

Correct. My 750w G2 are setup similar to your PSU2 above, CPU to 2x 6pin splitter and using 1 molex string per riser.
(I had spare EVGA G2 molex cables, and on the PSU side PERIF/SATA sockets are wired the same)

Okay, there's a terminology problem here.

When we say 6-pin or 8-pin, we mean the 6/8 pin PCI-E power cables. EVGA G2 750w has four.


Basically, you could use a 8pin PCI-E splitter and connect it to both the RX GPU and direct to the riser's 6 pin socket.
Alternately, it's possible to get (or make) cables that convert the PSU's second CPU EPS12V 8pin connector into a pair of 6/8pin PCI-E and use them to power a pair of risers.
Example:
http://www.ebay.com/itm/CPU-8Pin-to-Graphics-Video-Card-Double-PCI-E-PCIe-Power-Supply-Splitter-Cable-/272720358725

Or.... for the risers with the direct molex connection, you could obtain more EVGA G2 compatible molex cables, as for an EVGA G2 all the SATA/PERF sockets on the back of PSU are wired identical/interchangable. (EVGA G2, G1 and GS all use the same pinout for the molex cable)

There were rumors about similar cards a week or two back, this is the first concrete sign I've seen.
Appears to be legit, googling the model number 11256-31-10G finds lots of foreign-language sites with listings mentioning mining + 1x DVI

I'll take five?

That advice was 100% correct for Tahiti cards.

Mining an ETH-based altcoin (MusicCoin, Ubiq, EXP) that was launched more recently means the DAG size is smaller, and thus postpones the TLB thrashing issue that caused 7970/280 on ETH to slow from ~26mh to ~14mh over the past year and a half.

ZEC is also a good option.

Cost on ebay seems to range from $15 to 25 per card.

How often depends on a mix of luck, and how hard you push the fans, running at 80-100% can cut the lifetime considerably.
Some will fail within the first year or so, others will last 2-3 years, some may be running just fine +5 years down the road.

My remaining 7950/280x are nearly all on their second set of fans (a couple sapphires are on their third set).

this 100% depends on the cards.
If these are RX 470/480/570/580 then that's WAY too much draw for a single strand of molex (which has only a single +12v wire)

For safety, I generally use a single strand of Molex per GPU.

That's why I included every possible shred of info about where the BIOS came from
If someone's flashing this onto anything other than a MSI RX 480 Armor 4G OC model #912-V341-028 with Hynix memory they are gonna have a bad time.
MSI has a habit of producing multiple hardware revisions within a single Model/SKU, hence the box code # which identifies the specific variant.

Nice guide, very straightforward.