Whats better? More rigs with less gpus or less rigs with more gpus

[minfet]

Hello everyone,

I'm new here and I'm confused what is better

It is now 2 month while i'm in all this mining thing

I had no problems getting gpu's but I had problems getting normal psu like 1200W or more (had to wait 2-3 month)
All I was able to get was 850W psu all gold.

So.. For now I have 4 rigs of 4 gpu and 1 rig of 7 gpu

All 4 rigs runs on one 850W psu, electricity draw is 600-650W from wall, they works fine, no errors, no hash rates down to 0 or something. All stable all good.
Another one rig have one 750W psu and 850W psu connected with dual psu cable. This one works well too.

Before all this setup I had less rigs with more gpu, one rig had 12 gpu connected to ASROCK H110 PRO BTC+ and had 3 power supplies of 850W each, but it got crashed like 5 times a day, Issues was various, sometimes OS stucks (ethos), sometimes hash rates drops down to 0 on random gpu, sometimes risers (but after changing risers to new ones there were same problems)

Another rig had 2x750W psu connected with dual psu cable and 6 gpu and there was same issues as in rig with 12 gpus, hash rates drops down to 0 on random gpu and etc..

So I decided to split all cards to be powered with one psu. Now one mobo and 4 gpus are powered by one 850W psu.

I have total 680 mh/s rate mining ETH, In nanopool they find 440-520 shares / hour. Or 0.06 eth/ day.

As I have 2 motherboards with 12+ pci-e slots I'm able to assembly 2 rigs of 12 gpus each. But I was wondering could I get better results of founding shares If I have 2 rigs instead of 5 with same quantity of gpus?

And should I use add2psu connectors instead of dual psu cable using 3 psu of 850W each powering 12 cards. Or there is better way to get one good 1800W mining edition power supply?

[AAile]

You more or less answered your own question. When it comes to hash power it doesn't matter at all whether you have divide your cards to more rigs.
It's usually easier to set up a 4 gpu rig, less issues, they somehow solve themselves, cheaper power supplies and a better choice. The downside is more CPUs, motherboards, RAMs, hard drives etc.  - it can add up! On the other side, testing anything on a 13 gpu is expensive, anything goes wrong with a 13 gpu rig - you lose a lot of money. If I would start all over now, I would get 8 gpu motherboards. It seems to be a good number where you still have enough gpus there to save money on hardware, you still can get a decent power supply solution and you still don't get to many other issues.

As of power supply, I use all kinds: regular, "mining" and server. I've managed to make everything work. If I would start over, I'd get only server supplies for gpus with some regular pu for the motherboard, cpu and hard drive. Mining ones are a new thing, my guess is they are not properly tested and just an extension of the regular pus. Server ones are cheaper, quieter, have all kinds of protections in them, and best of all, they are built for 24/7 operations by serious companies.

I don't think there is any difference worth worrying about the add2psu and other solutions. I've used all of them, now I prefer anything with less electronics, I just find something to connect the pins (properly!).

[farkest]

Well, when I was playing with mining in the beginning with one GPU (#1), I get used to overclock it with +X core +Y mem.

When I got my second GPU (#2), applied the same +X core +Y mem. The miner was mining for a few seconds before the GPU #2 crash (0 hashrate or crash the miner).
I tought "Ok, this GPU can't overclock like the other" and decrease a bit the overclock of the GPU #2 and stop happen. But then noticed that it was happen to the GPU #1 some minutes after the miner start and had to decrease the overclock of GPU #1 also and never happened again

My conclusion: One GPU alone is more stable with their overclock than working with others. That may causing your mining rigs with more GPU to crash or get some 0 hashrates, maybe you have to decrease the overclock of yours GPUs when using more GPUs per rig.

The BUT: I can't say with it happen only to [1 GPU -> multiple GPUs] or [1 GPU  > 2 GPU > 3 GPU > 4 GPU > ...] since I only tested with 1 -> 2 GPUs.

[AAile]

My conclusion: One GPU alone is more stable with their overclock than working with others. That may causing your mining rigs with more GPU to crash or get some 0 hashrates, maybe you have to decrease the overclock of yours GPUs when using more GPUs per rig.

That's a very wrong conclusion. GPU's a very independent. You probably had a problem with correct power supplying to GPUs.

[minfet]

You more or less answered your own question. When it comes to hash power it doesn't matter at all whether you have divide your cards to more rigs.
It's usually easier to set up a 4 gpu rig, less issues, they somehow solve themselves, cheaper power supplies and a better choice. The downside is more CPUs, motherboards, RAMs, hard drives etc.  - it can add up! On the other side, testing anything on a 13 gpu is expensive, anything goes wrong with a 13 gpu rig - you lose a lot of money. If I would start all over now, I would get 8 gpu motherboards. It seems to be a good number where you still have enough gpus there to save money on hardware, you still can get a decent power supply solution and you still don't get to many other issues.

As of power supply, I use all kinds: regular, "mining" and server. I've managed to make everything work. If I would start over, I'd get only server supplies for gpus with some regular pu for the motherboard, cpu and hard drive. Mining ones are a new thing, my guess is they are not properly tested and just an extension of the regular pus. Server ones are cheaper, quieter, have all kinds of protections in them, and best of all, they are built for 24/7 operations by serious companies.

I don't think there is any difference worth worrying about the add2psu and other solutions. I've used all of them, now I prefer anything with less electronics, I just find something to connect the pins (properly!).

Thank you for your answer. And thanks for advice about using server power supply. So I should leave everything as is, splitted 4 cards per rig if there are no effect to finding shares
Now I have:
3x ASROCK H81 PRO BTC R2.0 with 6 slots (4 gpu each)
1x ASRock H110 PRO BTC+ with 13 slots (7 gpu)
1x Biostar TB250 BTC PRO ver 6.0 with 12 slots (4 gpu)

I bought everything while there were in stock. And I was planning to buy more GPU, but those are now double priced.
I bought RX570 8GB Sapphire NITRO+ DUAL for 246 euros per piece, now they are 449 euros per piece. For RX580 8GB Sapphire PULSE DUAL 279 euros now they are 499 euros. For RX580 4GB Sapphire NITRO+ DUAL I paid 222 euros now they are 449 euros ) and prices are VAT excluded... Crazy prices now..

[AAile]

Yes, crazy prices now, it's just not the time to buy. In a couple of weeks, or a month at most there must be another supply of cards (or even new models) so I expect there will be more value for the money.

If you undervolt the cards (and you should!) there is still plenty of space in that 850w power supply. You can add a 5-th card for sure even if you do dual mining. Well, you don't say the models but it's safe to assume anyway.

[treanski]

i have only 12 cards amd rigs, if you know what you are doing stability will be no problem...my rigs run stable 24/7. you wont get better results with smaller rigs...like i said if you know what you do it will be fine
vega only rig i would recommend only 8 gpus, that you can use blockchaindriver (if you go for cryptonight otherwise it doesnt matteR)
if you make rigs with many gpus you dont have to pay the basic powercosts multiple times(cpu, hdd, risers...)

some important things
-never connect risers with sata or molex to sata adapter, it will cause problems sooner or later
-max 2 risers / molex or 6pin cable

[toptek]

Whats better? More rigs with less gpus or less rigs with more gpus

either or .... it's more what you can afford and how much room you have .  with the price of cards were they are it might be what you can afford for some , my self i go with either 4 to 5 five card rigs but one or two big ones works as good .

[BennyT]

Neither is better. The only thing that matters is that if rigs have less gpu’s then you’re buying more motherboards and having to monitor/maintain more systems.

For me it’s way easier to throw 9 GPU’s on each system. I just ordered one of those pciE splitter cards and I’ll see if running 12 GPU’s per system is stable.

[philipma1957]

Easy six and eight slot riser free rigs running prebuilt waterblocks are the best rigs.

If I could do magic I would have onda six slot with three Evga 1070 ti hybrids and three Evga 1080 ti hybrids

Using 1200 watt plat Corsair psus.

Mobo was only 99 with a psu a ram stick was 30.

You can set the 3 1070ti to 105 watts and the 3 1080 ti to 195 watts

Use 900 watts and does 3600 sols

Small foot print quiet you can fit 2 on a four foot shelf

Which means 8 on a full rack easy actually 10 will work

That is 36000 sols for zec in a small space