Quote from: jeremyismer on May 26, 2018, 08:32:12 PM

Quote from: iSuX on May 26, 2018, 06:32:42 AM

TBH, I would not waste time on MB, system ram, or even PSU. (in terms if variables, there is nothing to tune, and little to gain, (assuming you have a stable rig right now)
The key point with PSU is, they are a big portion of your asset cost, so ensuring they run at optimal efficiency vs number of GPU is a quick and easy calculation, that's it, there is nothing else to tune there, no more variables to consider, (other than maintenance and mtbf)

Ok, I'm going to through my MB question. I started mining with whatever I had lying around. In my case, I have an AMD PHenom II X2 545 CPU with DDR2 ram and . I've got a PCIe x1 splitter to run out extra cards.

I've got 7 cards in total, split across two machines.

However, I would ideally run these on a single machine, so I have less CPU/RAM/HDD power overhead. But, the problem I have found is I can't run more than 5 cards on this machine, or the other one (AMD A4-4000, with DDR3 ram). The 6th card would detect but give that error-33 (32?). I swapped many cards around, tried cables, risers, rechecked mb bios settings, but I started to get the feeling that these old CPUs just don't have enough PCI lanes to run this many cards, or else there is a similar MB limit. So I ended up giving up on a single machine setup, and have the two running.

I looked at a mining motherboard, etc. but the saving in power overhead doesn't outweigh the cost of hardware for over a year. Which still makes these "free" motherboards (they were just lying around home) still appear attractive.

So, in summary, do these old boards/CPUs have a GPU limit?

(P.S. sorry to hijack the claymore thread for these kind of questions, but most people here actually seem to contribute constructively to questions!)

Good morning jeremyismer

Thanks for adding additional detail.

Yeah, that's a tough one, I don't want to speculate on your specific hardware, as I have no experience with those MBs, but...

I did exactly the same thing as you initially, in that I had Dell Optiplex 7010 kicking about, and used that as my prototype rig. As it only had 3 PCIe slots, I used a PCIe 4 port expansion card, but that rig never had more than 5 GPU running off it, BUT it was REALLY unstable, I had a lot of issues every time I added a card, or changed anything, or even powered it down.

But I learned the importance of stability and uptime, and when building my first rig from scratch, opted for the ASUS Prime 270-a, and MAN was that like a breath of fresh air, it just worked... Until I added the 5th GPU, then it would not boot, or if it did, I would have only 4, or sometimes 3 or 2 GPU detected.

Now on that board, you need to do some tweaks in the BIOS to get beyond the 4 GPU, various youtube videos are out there about that, (guess what, they vary in detail, sigh), but essentially, the only critical change you need to make is, (on the Prime 270-a), called enable above 4G decoding.

The definition of Above 4G decoding is to allow the user to enable or disable memory mapped I/O for a 64-bit PCIe device to 4GB or greater address space, because the primary VGA card should always be mapped below 4GB address."

OK, so after that, I used the ASUS Prime 370-A, (same situation as above, (above 4G decoding), gets you up to 13GPU, and on the ASUS B250ME, specific mining mb, you have that feature enabled by default.

Now, as it happened, a lot of back ordered GPU started flooding in, after months and months waiting, and I was able to finish the rigs I built for other people, and even then had some spares kicking about, and again found myself with a spare Dell Precision 3650, so decided if I could use that to employ 5 "spare" GPU.

This was a tough one, 2 PCIe slots, so I used the 4port card in the 16x slot, and a single PCIe in the 1x slot, and no matter what combination, cards, ports, even tried some older AMD and Nvidia cards, but it would not boot with more than 4 GPU.

I played about with that for some weeks, building up to the conclusions, it lacked the ability to cope, whether by hardware design, or BIOS limitations, (something Dell are renowned for, their goal, (understandably) is mass market, stable, reliable, nothing special or pushed too hard, means lower support costs for them, and steady revenue streams for replacements after 3 years warranty expires).
At the same time, having 5 GPU kicking about and no way to use them, I forked out for another Prime 270-A, which I chose over the 370-A, simply because I wanted to pull the i7 6700 CPU and 16MB 2400DDR RAM from the Dell precision, so in my case, that additional rig, was only really costing me the new MB, and time, as always ;-)

Interestingly, I noticed 2 other things,

1. Mining with a HDD sucks, staggering how slow that was to do everything, boot, init, you name it.
2. Mining with an i7, even an older mid range 6700, was a lot faster to init Claymore, than all my other rigs, i3 7100 or i3 8100. Probably that is obvious, but all the same it just goes to illustrate how much code is executing in Claymore during the init, so that was the real surprise I mean here.

So... I'm sorry man, I don't have any real solutions for you, I would have liked to get some definitive answers on those Dells, but time is always divided between things, and looking at the money draining away from not using those 5 GPU, that was the kick in the arse to build another rig.

BTW: One mistake I made early on, was to opt for a celeron CPU, and while that worked ok, it means you have to use one of your mining GPU to drive a monitor, and that in itself brings in a whole raft of issues, mostly when you're debugging driver problems, init issues etc. Now, that was my first from-scratch rig, and I didn't fully appreciate what the loss of intel graphics would mean, so I swapped that out for a i3 7100, and on all rigs since, having intel graphics is a check-box item for me since. It means, no matter what crap MS or AMD or drivers or PSUs through at you, at least you can get a display on it.
The i3 uses about 1W more than the celeron, (on paper), in reality, for mining, once everything is running, I don't see any difference in power consumption from the celeron.

So, in conclusion, you could be flogging a dead horse there by adding the 5th GPU, and weighing up the cost of 4 GPU doing nothing while you work on that, possibly unsolvable issue, is a situation probably already very much playing on your mind.

So... time to go shopping maybe? MB, CPU, RAM? Or by a 2nd hand bare-bones maybe.

Good luck man, lmk how you get on.