I agree, just checked one open rig. 17 days, 118mh (TBM) at the pool side avg., 330W (I have not done much tuning on this card).

Hi,

I have forgotten what is the logic of --xintensity -1 ?

Does TBM try to find best possible intensity setting for card & pool combination?

I assume that it starts at some default setting and then tries to increase intensity (more MHASH) until it gets too many rejected shares and lowers the intensity.

And finally it arrives at some kind of sweet spot (according to code logic)?

If so, does it show the intensity it *thinks* is the best?

Regarding RTXA2000 6GB vs. RTXA4000 16GB, it is more of a question what is available.

RTXA2000 ~40MHs/70W, ~1.75W/MHs, no extra power needed & have not tried reducing power yet.
RTXA4000 ~58MHs/115W, ~1.98W/MHs (philipma1957 settings)

RTXA4000 costs 2X, so that also needs to be considered.

But as I said, I could get RTXA2000, RTXA4000 is MIA currently and I doubt I will get more of those RTXA2000s until next year.

And these cards will not be used for mining for long, they will go to actual workstation PCs :-)

@philipma1957

I just got 5 pieces of HP NVIDIA RTX A2000 6 GB GDDR6 Blower (MFG part number 340L0AA) and 2 AMD W6600 Pro cards.

Running HIVEOS, TRM (AMD), TBM (NVidia). So far no luck with OC with W6600 (I assume 30MHs would be max with these?)

They are exactly the same as PNY "retail" versions. I have one PNY and 3 HPs in our heating via mining prototype. I started this project just for fun, further development depends how looming ETH POS affects financials.

https://ibb.co/RBWMzwk

https://ibb.co/qmtPrnp

https://ibb.co/TTw58x7

https://ibb.co/c2nZhK8

RTX A2000 card,

These are very easy, no messing with power limits, volts, etc.

12 X these = 480MH = 840W

(I have only one but maybe with those monster cases)

Just put memory as high it goes and enjoy, 70W is the limit anyway :-)

(HIVEOS & TBM 1.21)

01:00.0
RTX A2000 5940 MB · NVIDIA
Samsung GDDR6 · 94.06.2C.00.01 · PL 10 W, 70 W, 70 W
41.62 MH 67° 65% 68 W 65 0 2400

TBM 1.11, Nicehash & RTX 3080

I am pretty pleased with -xxintensity 177 setting (I used 7 on version 1.05)

TBM reports average 98.44MH/s (Nicehash reports 108.88 MH/s as accepted speed?)

(1042/8/0), 226w

core -175MHz
memory +1000MHz
PL 75%

I just like simplicity of Nicehash even if it means less profit.

And TBM gives 3-5MHs more (with this setup) than other miners, great job :-)

HP RTX 3080 OEM

--xintensity 7 works best for me (no stales)

(I assume Nicehash does not pay for stales.)

(15411/23/0), 0.15% rejects

Hashrate 91-93 MH/s (GPU is in original OMEN case)

Settings in Afterburner:

Core -175Mhz (~ 1446Mhz)
Memory +1000Mhz (=10241Mhz)
PL 75% (~239W)

A couple of suggestions:

- cumulative average MH/s for current minig session
- cost / 24H would be nice addition

Hi,

I am testing now with RTX 3080 (Nicehash & --xintensity -1).

Where can one see the dynamic xintensity calculated value (so I can report it  )?

It seems that with default values no log file is generated.

I did order one, maybe if I am lucky, I will be getting it from Santa Claus (no confirmed ETA)

I was looking at NVIDIA RTX A2000, price seems to be about 45% of RTX A4000 from one of our suppliers.

(That is if they could deliver anything  )

It has interesting specs (a cut down/slower clocked/slower memory RTX 3060) and maybe no LHR tricks?

If one looks at stock settings, RTX A2000 performance could be about 64% of RTX A4000 at 45% price & 50% power.

Maybe RTX A2000 memory can be tweaked and still be at 75W board power.

(Info is from Nividia & Techpowerup)

RTX A2000

GPU memory 6 GB GDDR6 (memory is running at 1500 MHz, 12 Gbps effective)
Memory interface 192-bit
Memory bandwidth 288 GB/s
CUDA Cores 3,328
Power consumption Total board power: 70 W

RTX A4000

GPU memory 16 GB GDDR6 (memory is running at 1750 MHz, 14 Gbps effective)
Memory interface 256-bit
Memory bandwidth 448 GB/s
CUDA Cores 6,144
Power consumption Total board power: 140 W

I have one of those 30L Omens (RTX3080 / I7-10700K). It is a very nice looking case.

Currently NBMiner running on windows is getting about 89M @225W (card power, have not checked total from the wall).
Card is running fans at 100% (afterburner, core clock -175, memory +1000, PL 76). I am going to try hiveos when I have the time for it.

Thanks,

I was testing my 2 pieces in windows and had no way lower voltages yet.
I have several combos (mb+cpu+memory+ssd+psu)  lying around which can take two cards so I did not use hiveos yet.

I will try hiveos on these combos (with USB stick). I think even with extra overhead / card (mb+cpu+memory takes about 35W) it makes sense as only investment needed is cards.

Hi,

I can also report high % of stales / rejected (>10%) with RX 6600XT with Phoenix 5.7b

With NBMiner 39.0 only about 0.05% stales/rejected.

OS:     Windows 10 Version 2009, 10.0.19043
Driver: non-whql-radeon-software-adrenalin-2020-21.8.1-win10-64bit-aug10

I have been testing two RX 6600XTs since yesterday.

With minimal tweaking (basically only GPU 1305Mhz & MEM 2250MHz, no voltage tweaks) my result so far:

87.5W for ~ 32MHs (total card power on the wall)

More details in this post:

https://bitcointalk.org/index.php?topic=5354335.msg57701633#msg57701633

Hi,

I got lucky (in a sense), I noticed the cards on local (Finnish) online store (jimms.fi).

Price was 450€ (VAT 0%) so not exactly @MSRP but not too bad either :-)

It seems almost all RX6600XT cards are gone for now at that store also :-(

AMD does not actually "lie" about the wattage. The power usage reported is only for the *GPU*. VRAM + other stuff is on top of that.

Nvidia gives (at least tries to) power usage for total board if I have understood correctly.

I have two NITRO+ AMD Radeon™ RX 6600 XT cards (11309-01-20G), testing now with NBMiner. After 2 hours no rejected/invalid shares so far.

Phoenix Miner 5.6d - Release build works also but gets lot of stale/invalid shares (for example during 11h46min 4314/460/155)

Driver 27.20.22023.1004 (non-whql-radeon-software-adrenalin-2020-21.8.1-win10-64bit-aug10)
W10P version 21H1 (19043.1165)

[17:23:14] INFO - ================ [nbminer v39.0] Summary 2021-08-15 17:23:14 ================
[17:23:14] INFO - |ID|Device|Hashrate|Accept|Reject|Inv|Powr|Temp|Fan|CClk|GMClk|MUtl|Eff/Watt|
[17:23:14] INFO - | 0|6600xt| 32.37 M|   386|     0|  0|  64|  63| 33|1311| 2238|  45| 505.8 K|
[17:23:14] INFO - | 1|6600xt| 32.38 M|   385|     0|  0|  62|  51| 33|1308| 2238|  45| 522.2 K|
[17:23:14] INFO - |------------------+------+------+---+----+---------------------------------|
[17:23:14] INFO - |    Total: 64.75 M|   771|     0|  0| 126| Uptime:  0D 02:12:01   CPU:  7% |

This two card rig has total power from the wall about 210W. (85+ PSU, 230V)

Rig without cards consumes about 35W on idle desktop (Windows 10 Pro).

So total power per card is (210-35)/2 = 87.5W for ~ 32MHs on the wall with this PSU (Chieftec CTG-750C)

(GPU power is 64W & 62W, 64W card is getting hot air from the 62W card backplate, Gigabyte mATX board GA-A75M-UD2H)

(maybe a bit lower than 87.5W as CPU usage is a bit higher during mining vs. idle desktop)

Settings during this test so far with Afterburner:

GPU: 1330MHz
MEM: 2250MHz

No other tweaks tested yet.

I am testing currently one new server with 3950X and is there a better way to test stability :-)

Nicehash running on Asrockrack X470D4U2-2T motherboard with 2X8GB DDR4 @3200 (KVR32N22S8/8, jedec timings 22-22 something, no tuning/oc, will replace these with 32GB ECC dimms when I get them). OS is Windows Server 2019 (yes, W10 drivers work fine)

I have set TPP to 88W in bios which translates to ~65W for CPU (checked with HWINFO)

I am using "only" 15 cores (--threads=30). I like to leave one real core free for other stuff. Helps with remote control etc.

3950X @TPP 88W (65W CPU), 30T, CPU temp ~60c on open bench (cooler Arctic Freezer 7XCO, will replace it with NH-U12S when I get it, Freezer is a temp solution)

11427~12762 H/s (761-850H/s per core)

I personally use commercial Teamviewer but I've seen this also in use:

https://anydesk.com/en

There is a free version for personal use. I have not checked too much if there are limitations on that tier.

I believe Anydesk was started by former Teamviewer people around 2014.

I have 5 x RX580 8GB, total at the wall around 610W (230V power). 28.XXmhs / GPU.

Not saying that this a great result, just hopefully a helpful data point for somebody.

I did try to go up to 30mhs, but lost stability (and patience).
I am mining mostly to stay current with this crazy virtual coin thingy and it is fun too :-)

https://i.imgur.com/IIbV0d6.jpg
https://i.imgur.com/lj3Czz6.jpg

RIG:
W10 Pro + Nicehash running Phoenix
ASROCK H110 PRO BTC+
i3-6100 & stock fan
5 x RX580 (Powercolor Red Dragon/2 x Sapphire Pulse/2 x Sapphire Pulse BE)
8GB RAM
120GB SSD M.2
PSU1 750W (80+)
PSU2 550W (80+)

Phoenix launch parameters (I found these here in Phoenix miner thread)
-mt 1 -gt 64 -rmode 1 -cvddc 820 -mvddc 825 -cclock 1050 -mclock 2075