Topic: My RX580 Benchmarking and Results: 30.5Mh/s @ 77.5W (+ 1130MH/s Dcr w/ 125W) - page 3. (Read 189842 times)

Well it seems one of my PSUs had a knack and also my mobo suffered (partly melted PSU connector... So I got myself a new PSU (Corsair H1200) and a new MOBO (Asus B250 expert mining) and 4 additional cards (sapphire nitro+ SE).

However the MOBO can only address upto 13 similar cards (mine are all AMD RX580.

I have
PSU 1 (1200W) 4 x RX580 + MOBO and 4 risers + SSD
PSU 2 (1200W) 5 x RX580
PSU 3 (750W) 4 x RX50

Total power draw from the wall (incl monitor) is 1700 watt

Hashing power is around 390 mH/s with modded BIOS and OverdriveNTTool at 1150/875 and 2150/875

win10-64 with AMD drivers Jan 2018 and compute mode by AMD compute switcher. 8GB RAM 228 GB SSD with, 80 GB set temp memory.

Cards are
6 x Sapphire Nitro+ SE RX580 8GB GDDR5 Hynix
4 x Sapphire Nitro+ SE RX580 8GB GDDR5 Micron
4 x MSI Armor OC RX580 8 GB GDDR5 Hynix
All cards are between 29.50 and 30.50 mh/s

Depends on what you want to mine, and how proficient you are w/ strap editing.  Samsung does a little better for cryptonight using the best publicly available timings (a.k.a 'ubermix' -- ~1025H/s) vs the best available Micron timings (~1000H/s).  On the other hand, the best publicly available Micron straps are better for ethash (~33MH/s) vs the best available Samsung ones (~31.5MH/s), though if you can create your own custom straps, you should be able to get to 33MH/s w/ Samsung as well, according to certain people on these forums. 

Other coins/algos generally fall into one of these two camps, or are generally memory agnostic (like LUX/phi.)

So... what would you choose between Samsung and Micron memory? (Sapphire nitro+ 8gb).

Don't bother w/ compute mode until after you've updated the bios on all your cards - you'll want to reinstall drivers once all bios' have been updated.

PBE is for editing the bios (hence the name).  Trixx/Afterburner (as well as Wattman and OverdriveNTool) are for updating clocks/frequencies post install.  I use OverdriveNTool exclusively.

think not, so increase windows virtual memory / set card to compute mode and flip the switch on the card? Open Gpu-z and ATI Winflash and save both files before modding. And modding should be done only with one card (removing the other 5), right? What I don't get, what's the difference between the Polaris Bios EditorSapphire Trixx and MSI Afterburner? Which one is the way to go?

Did you set all cards to compute mode?
https://support.amd.com/PublishingImages/Support/KB/DH-024/DH-024-03.png
Also, why are you testing with this driver specifically? Imo it isn't worth it to download random bios from internets, much more right thing to do is mod bios by yourself (either using PBE 'one-click' or just testing with straps from 1750, then 1500 and so on), doing by one card.

Hi guys, can someone help me out? Im a noobie regarding mining and just started. Just built my rig yesterday and was running a first test to see if everything works. However my hashrates are devastating. Currently all the settings are orginal, so next step is to optimize my system (undervoltage / oc).

Running 6x Sapphire Nitro RX580, 8GB on a ASrock 110 pro. Everything is yet out of the box. System draws currently 1100w.
System: Windows 10 Pro 64bit, 1709
Driver: Crimson 17.11.4 Beta

Did run a "Quick Start" on https://nanopool.org/ for testingt. So far this are my results and funny thing, the two cards with 18mh/s are Micron chips and the other 4 are Hynix . So next goal is to join the 30mh/s club, is this tutorial still the way to achieve it? http://1stminingrig.com/best-windows-setup-configuration-tweaks-for-mining/ --> Tweaking Windows / flash the bios / oc the cards?
https://imgur.com/XEDs6JZ

It doesn't matter what you set mem voltage (mvddc) to, it's always running at 1.6V.  To actually change mem voltage, you would have to muck around in the bios.  The setting, as you point out, acts as a floor for the core voltage (cvddc), though it's actual purpose escapes me.  I can only imagine that a.) it's a leftover of previous models which is actually no longer necessary, or b.) it's meant to be a sort of signal of the minimum core settings necessary to fully utilize the memory bandwidth available at that given clock setting.

Bottom line, just set it somewhere below your P7 core voltage, and everything will be fine.  In fact, I usually set it somewhere around my P4 core voltage, so I can fiddle w/ P7 while the miner is running and not worry about being limited by the memory voltage setting (which can only be adjusted while the miner is offline.)

Btw, I find OverdriveNTool much more straightforward for adjusting clocks/voltages than any of the branded tools.

The clack is when a PSU port line OCP is overloaded and shuts down to protect the circuitry. With the balanced wiring you should be able to run you rig with two PSUs only. How are your risers connected to the PSU? Hint: Connect the risers to the 8-pin ports (as many as you can). For the rest - when you run out of 8 port cables-empty ports on PSU) - connect to the 6-pin ports. Only connect max 2 risers per sata/molex cable.
Identify (User Manual) what are the individual ports max amperage values and make sure you are not chaining too much to the same port or port group.

Background: On my PSU (Corsair AX1500i) there are ten (10) 8-pin port connectors each protected by 40 A OCP and six 6-pin for SATA/peripheral protected all together by 40A OCP. So all 6 ports (6-pin) combined can  only sustain at max level 40 A. This can sustain up to 6 risers. Connecting 7th riser on this same set makes in my case exactly your behavior, which in my case happens in 10 minutes of running the miner, consistently.
When you connect your GPU to the risers and an 8 port power it will draw up to 75 W (about 5-6A). Connecting 10 risers, all to SATA ports would yield 60 A, combined.
In your case I guess you are combining 10 risers to the 6-pin PSU ports via SATA-molex cables and so pulling too much power and the PSU is clacking it self down. This only happens at maximum load when you start mining at highest rate.

The occurrence of power shutdown when an PSU OCP is exceeded is harmful and can burn your components.

If you need more help PM me.
If this solved your problem send your gratitude to ETH: 0x4d45B1B0467e3D02e686936DBE0659e74Bb2Cf52

may be you hit a claymore bug in the "hashrate" increase function  
anyway..on restart it should be ok.

Thats just it, I didn't. The OC settings are the same on all 5 GPUs across the board. 1130 Core, 2200 Mem. Undervolt to 850 on core and mem.
I have since then lowered the mem clock down to 2150, as this seems to be the most stable for this type of GPU.

The hash rate reported by Claymore seems flaky at times. I can started once and it will report 24.5 to 27.5 MH/S, just stopping and restarting the miner will change the hash rate to 27.5 to 29.5.

You have OCed too much. It's not so unusuall with Claymore, but it won't last long.
You more likely need better timings.

I took a photo of this so people wouldn't call me a lier.

All 5 gpu are MSI Armour RX580 8GB

All have exactly the same OC, BIOS mod

The 2 with Micron mem are running ~29.8 MH/S

The rest have the Hinex mem running ~27.8 MH/S

But this 1 freak cpu with Hinex is somehow running at 45.5 MH/S @48C

Alas it wasn't meant to be, it only lasted 10 min before Claymore reset itself and the freak went back to its old self. lol

https://www.dropbox.com/s/xyt79p48vct2tc2/20180219_192738.jpg?dl=0

So I was racking by brains out since I only recently had a setup of Vega cards and sold some of them to switch over to 580's. I was getting a great hash rate from the V64 and V56 so I thought switching over to the 580 would be easy. After two days of getting nothing but 20.X no matter what I did I finally got my Sapphire RX580 to be hashing at 31.9X.

I almost was ready to burn down the world because I couldn't get higher than 20 no matter what I did. There's so much information out there that's out of date and so many different tools and I was going crazy.

In the end, ran DDU, only had 1 card installed, bios mod only the memory strap, and used overdriveNTool to do the rest and when I started up Claymore I'm at 31.XX.

What a relief and I'm happy now that I'm in the 30's club on an RX580.

(BTW, my card had Micron memory... unlucky me but it seems that it doesn't matter what memory really.)

Now to start testing tomorrow on the MSI RX580 cards and see if I can replicate the same hashrate.

Hi guys.

I have bought a Sapphire RX 580 8GB with Hynix memory.
After some tweaking arrived at the mirific 30 Mh but for a small amount of time (2 to 3 hours and then "crash").
I' ve configured my card with a Bios mod, running at 92 Watts, 66º Celsius degrees and 28 Mh stable for more than a week (24/7 work).
Did anyone get better hash values with this amount of watts consumed and temperature?



Share your findings.

Thanks.

i will be getting 8 of these cards next week and I’m excited as heck!

can you try modding the bios via polaris bios edition v1.6.7 and use one-click patching?

Try setting core to 1200 and memory clock to 2200 or 2250 whichever is stable and if could report back, that’d be awesome.

Just saw you use ethOS but if you have access to Win10 to bios mod, that’s super!

Should be allright. I'm running three cards (RX580) on a 650 Platinum without any trouble. The uptake isn't that high.

I received two of 8 Nitro 580 8gb cards this week and one has samsung memory and one has hynix. Used pbe v1.6 one-click patching and set the clocks to 1200/2100/875.

Hynix can only do 30 Mh but the samsung does 30.8 but with a mem clock of 2090, as it’s crashing with 2100 setting.

my settings in wattman reseted while running the miner lol, i guess 2250 is too high (950 default mv). wth

Afterburner app maybe? I so something on YouTube. I didn't try it, this is  just suggestion.