Claymore, can you please add a counter for "lowdifficultyshare" error?
It is hard to check by scrolling, too much text after a while, so just like we have invalid shares count, it could be very useful to have lowdifficultyshar error count.

Thanks!

The bios is completely custom made with custom timings, didn't use the common trick with the straps which only get you up to 10-15% more speed.

Confirmed cards that have 4gb Samsung memory: Sapphire Nitro+ RX470 and RX480. These also come sometimes with Hynix (especially 470s), so make sure to check before contacting me. Latest batch as far as I know comes only with Samsung.

Mining speed: +25-30% more than stock on monero (up to 835h/s on RX470, up to 850h/s on RX480).

Didn't test on 8gb Samsung memory, but it should be compatible (timings are similarly made between 4 and 8gb on stock bios). Still, I don't guarantee the same speeds.

PM me only if properly interested, and only if you have 40 rigs or more, or the price might be a bit steep for you (fixed, one-time price for the bios file set at 0.9 BTC). This investment can be recovered in ~2 months with ~40 rigs, ~1 month with ~80 rigs, ~15 days with ~160 rigs, so on and so forth). As a bonus, any update I make to improve the bios will be given to the clients who purchased, free of charge.

I will only send the bioses if at least 2 clients are in the final stage of acquisition (sent payment). Also the sale will be limited to 5 clients max to not affect the difficulty and make the ROI of this investment harder.

I will post proof in a few hours.

Dwarf is really going haywire, not showing stats for a few hours now, and not paying...

I have 16.11.2/4 and all these fan speed issues people are talking about aren't happening.

Finally, some down-to-earth users around here...

Well, again, the fault belongs to the manufacturer, so Claymore can bear no blame.
And you not wanting to tinker with your systems to make it work, is your own choice. It is not something that everybody has/wants to do.
If you have not realised yet, you are in the bitcoin farming business, where if you don't stay competitive by all means, you don't win that much. Complaining only doesn't get you anything.

One more thing about these power spikes...

The fact that it isn't the miner's fault, doesn't mean that Claymore can't do something about it. He can change the code to be more progressive on that part where power usually spikes. This will mean that he will do what the manufacturers didn't, place a more proper limit on the use of hardware. But I repeat, it has nothing to do with the miner itself as software, it is well done, with no obvious glitches. This change, if done by Claymore, might also result in lower speeds, so we might get like a v8+ version, where speed is only a bit higher, but not as high as v9. So careful what you ask, you can either work to make your systems stable and/or change equipment with more quality ones, or just ask for this modification, but expect consequences.

And to understand why the power spike happens, it appears when gpu is fully utilized, and ram is being purged and reloaded for the next job (not "share" like I previously said, it would happen too often). This is something that games don't do, and if some do, they don't do it as fast as this miner. So, he isn't using some magic trick to go over the limit, it is just the combined usage of both gpu core and mem controller being used at their highest point.

If I was able to get 5x470 oced working on a 750w psu at 1065h/s (claymore v9), which previously at stock settings would just shut down, then you can too.

Read my post to understand, you clearly don't: https://bitcointalk.org/index.php?topic=1670733.msg17127240#msg17127240

Any of the "Over" stuff. Don't know that much to be able to tell exactly.

The Zalman is fine with 4 cards. Your power is actually quite balanced between the 2 psu's.

If you read back, you'll see that it was I who discovered the spikes...

Back on track, it is not a miner issue that the power spikes, the software can only control hardware within hardware specs. Only software that has the ability to over/under clock and over/under volt can fuck with hardware, any other software without these functions won't be able to do anything.

If a regular software can push hardware beyond its limits, it is because of the manufacturer, he didn't properly test his equipment before releasing, and didn't foresee this kind of usage. So, now we shall know which PSU are really quality-built, also which video cards are quality-built (some spike more than others, even though they are the same brand and model; this might also have something to do with ASIC quality).

I managed to get a system with 5x470 running oc on a 750w power supply,  on -i 8. Its been running nonstop for 24h,  so dont say that v9 has stability issues. If you know how to control the spikes,  you will have a stable system,  even on a low power psu like mine.

To all having PSU issues/reboot/freeze/shutdown, I suggest keeping the same settings as in Clay V8, but modifying power limit in AB to 80% for each card, see if it happens. If it's stable, also check if hashing speed is lower. I am not at home to try this.

Just follow the main claymore thread and you will know why.

We all know that already, but the idea of using gpu-z is to monitor the actual spikes, not what values the spikes have. That's why I didn't provide any numbers besides psu rating when giving my example in a previous post.

If your cards are stock, then it might be that the protection kicks in for too much power usage on a single pciex cable. Did you connect each card to an individual pciex cable?

A regular wall meter cannot tell because the spikes are really short, but enough to trigger the PSU's protection systems. So only with a high update rate wall meter you can tell.
Are your cards overclocked/overvolted?

How much power for the PSU? The fact that it's platinum means very little.

Careful with this version, while people report no major power increase, if you ever watched the gpu power consumption in gpu-z, you will see that each card has a power spike when it finishes a share (or something in the same ballpark). If you take the avg power consumption, it might look ok for the PSU, but those spikes, if they are done at the same time on most or all GPUs, even a relatively good PSU will make the system freeze or even shutdown. Happened on 2 systems of mine after switching to V9. I have pretty poor power supplies, a seasonic 750w (it can get to 830-840w without shuting down) and a segotep 800w (at 50 degrees, 900w at 25 degrees). The first one has 5x 470s, and the 2nd one has 2x470 and a 290x and a 390. I undervolt the shit out of every card so that systems are stable 24/7. Now with v9 I had to undervolt more and also underclock (at least on the core) to make systems not freeze (I didn't get any bsods or restarts, or artifacts, only screen turned black and no signal from card, but systems kept running, so cards are not the issue).
So careful with what PSU you choose, in my opinion, a 1000w is the minimum for a 5x470/480 if you plan to leave stock or overclock a bit. For 6 cards you will need 1150-1200w psu.
The other option is to undervolt, like I did, and you could also lower power limit in AB for each card, maybe that helps with power spikes (I will experiment with that).

In conclusion, while on avg power consumption the eth miner is higher, those power spikes on zcash miner are overloading our psu's if done simultaneously on more cards or all. My sistems were completely stable at the same settings used in clay v8 on eth miner, so not the avg power consumption is the issue, clearly.

There is even a separate thread created by someone having these issues, and some others are joining in saying they have the same issue, but didn't bother into actually checking why it happens. It is not the miner's fault!

I've gotten two b150a-x1 to work with 5 cards, you need to set pcie to gen2 for all slots and alocate 3.5gb resources to vgas in bios. Also, disable sound from bios, else you might get no resources available.