Bitcoin Forum

https://imagescdn.tweaktown.com/content/5/8/5857_03_amd_radeon_r9_290_4gb_reference_video_card_review.jpg


Due to the huge success of my Tahiti GPU appreciation thread, I figured I would make one for the higher end GPUs from the same era; Hawaii.

The Hawaii GPU chipset was released with the R9 290 and R9 290X which were released in Sept 2013. These GPUs were mostly ordered for Litecoin mining (Scrypt) to profit off the Crypto boom of 2013 when BTC peaked at $1100 USD.

However the profitability was short lived because once the R9 290 were delivered to customers the mining profitability took a huge turn to the down side, so many many miners were actually sitting at a huge loss if they ordered this GPU too late in the game.

In 2013/2014, it was mostly used to mine SCRYPT coins such as Litecoin/Dogecoin. If I recall the speed was around 800KH-900KH/s depending on clock rate.

In 2015, it was mostly used to run Darkcoin X11 algo.

In 2016-2018 it spent most of its time with the Dagger Hashimoto ETH mining algo. Basically it didn't suffer the early Thrashing bug so under stock clocks (947/1250mhz) it achieved a speed of 26.5MH/s and could easily break 31MH/s if noise/power consumption wasn't an issue.

However it looks like the Hawaii's are suffering the same DAG thrashing as did the Tahiti GPU and will lead to their eventual shutdown.

Currently under stock clocks all I can achieve is around 23.5MH/s and mining ETC which is about 7 DAGs ahead, it hashes at 22.5MH/s and there is no fix in sight.

Putting these numbers into What-To-Mine under 10 cent KwH power results in a profit of NOTHING.

Unfortunately the Hawaii GPUs are under the 28nm technology and are basically power hogs. The best undervolt I can only achieve around 170 Watts or so mining at 23.5MH/s. This is horrible since an RX GPU can achieve around 28MH/s using only 130 Watts or so.

I don't hold too many of these GPUs so I haven't had much time to investigate what speeds it generates with XMR V8. But under XMR V7 it only achieved 700H/s so most likely with V8 it will be much slower.

So lets discuss what coin to mine next with this once powerful GPU.

What to mine next?  I mined a lot of Ethereum and ZCash with these cards, mostly R9-390's.  I have sold probably 30 of these and more for sale now.

I don't have power going to a single R9-class card at this time.

I have a few XFX R9 390. I mined that with ETH over the years. If I want to mine XMR V8, it is better to modify the memory timing? Any source of memory timing for XFX cards?

Mined LTC Scrypt back in the day with these had 4 of these monsters, in heat, power and hash! Sold off in 2014 before the whole Mt Gox incident and crash due to me moving out of WA to CA. Got back into mining in summer of 2017 and starting looking for these bad boys again... but people wanted insane prices 290's were going for $200-300 then... sure as heck was a mining craze.

These cards are still good if you happen to be in a country with low electricity, I can think of Washington DC with only 0.02 cent/kilowatt these cards can still be good enough especially if you have the R9 390 card or the Radeon 295x2.

I think these cards are being sold for very cheap in marketplaces like Ebay or Amazon right now.

I still mine cryptonight based coins with my reference 290x and a few 290s.  On v7 the 290x's get 840 h/s and the 290's got about 790 I want to say.  On v8 the 290x get about 790 and the 290 get 745.  I haven't checked wattage since v8 came out but it should still be profitable I think.

My nitro 390's get about 790 on v8.

Lately I've been using my test machine with a non-reference 290x to optimize wildrig ( https://bitcointalk.org/index.php?topic=5023676.0 ) within RainbowMiner in hopes finding a diamond in the rough. I've only done about 3 algos but cnv8 still seems to be the most profitable.  I think the 290x has tons of room to be optimized but the developer doesn't have a hawaii card.  Maybe he'll take a donation card. ;)

The only rig I ever had I built with my own hands and used 5 R9 290 cards as the "money makers" :D. During cold months (through October to March) those monsters kept my bedroom comfortably warm and cozy :). Made them silent by scrapping the reference cooler and DIY watercooled them. Good times! :)

Mined ETH and Cryptonight coins

Great thread, but I think you should add fijis with hawaiis there: fury, fury x and nano cards since they are all from the same era.

anyways, my 390 and 390x get 800 and a little above 800 H/s on the v7, on the v8 they perform slightly worse - around 780 H/s all stock clocks.

On the other hand, my fijis (have them all 3) on v7 get the same hashrate as 390, but on v8 even worse then hawaii, 650 H/s at best, obviously they're running less hot and consuming less power than hawaiis,
but I don't know if I'm doing something wrong here.
How is it possible that fijis have worse mining performance than hawaii cards in v7 and v8 when fijis are clearly more powerful than hawaiis?

Does somebody here have a better experience with fijis than me in v7/v8 mining?

I have had good luck with srbminer on hawaii, what miner are you using?

Bought 10 R9 390s for $260 each (after selling copies of Civ 6 I think it was for $40) and mined Monero when everybody else was doing X11

Sold Monero on it's first run up for BTC - made more money than I would have made had I mined X11 because Monero difficulty was so low because it was the unpopular coin

Mined ETH with R9 390s for 2 years at insane 31MH/s rates. Sure they were hot as hell sucking 250+ watts and required I buy Gold/Plat PSU in the 1000+ range but they paid themselves off pulling in over 10 ETH a day.

Sold the R9s when people went full retard and spent more on the cards then they could have realistically mined in a reasonable timeframe. 8 of my USED R9 390s sold for over $500 after fees.  I took that money and bought more crypto with it.

So other than wasting my time and associated electricity costs I was basically able to get free crypto by not trying to min/max the market. I probably could have done better if I had a time machine or mined 0 day coins, but it was passive income mostly.

I have said it before and I'll say it again - only mine a coin if you expect it to fail - in this case I expected Monero to fail and was pleasantly rewarded by excess crypto because I mined as insanely lower difficulty because WhatToMine sent all the "idiots" to the wrong. If you really think a coin is going to take off just buy it as soon as you can.

Most miners "chase" coins only after they hit big in the profits. I had a similar experience with my R7 370 4GB's cards, i got them for a great deal but had many issues getting 12 GPUs to ran stably, only finding out later it was the SATA to PCI-E risers, sold them off on ebay and actually made a profit from them after mining. Same with the R9 290X Lighting cards I bought. Those MSI 290X Lighting had beast coolers and were massive. I achieved a max hashrate of 36mh/s (not stable with a ton of power) but settled around 32-33Mh/s.

Thanks for reporting your V8 speeds. Putting these numbers into WTM results in basically identical profitability as ETH mining. However I am assuming XMR V8 will consume much less power than ETH. Each week the speeds decline, currently under stock clocks I only get 23.2Mhs on DAG 221.

Really surprised we can't achieve 1000Hs with these powerful GPUs.

Regarding the Fuji and Nano GPUs, I didn't combine them because the chipset is completely different and i don't think there are much Fuji GPU out there. They were a flagship model which was very expensive and sold for a high price and they produced them in limited quantities.

They should definitely be able to perform better but unfortunately no one seems to be optimizing for our hardware, at least not publicly. 
In my experience yes cn based coins don't use as much power as eth.  My non-reference card gets about 21 mh/s on eth so the reference probably get about 23 as well.
For cn based coins use these calculators  www.cryptunit.com and https://cryptoisme.com/  They will give you prices for the other cn coins based on v8 speeds (or maybe v7?)
If you want to go further you can use srbminer and set it up with moneroocean to mine the most profitable cn coin and get paid in xmr.
Or go 1 more step and set up RainbowMiner, not a tough learning curve, and hope we get lucky with one of our cards doing well on one of the lesser known algos.
Here are some speeds on those algos from WildRig Multi

hmq1725 : 2.9 mh/s
x22i : 2.7 mh/s
x16r : 5 mh/s estimated, due to the random order

For 'exotic' algos, you can use this calculator, https://crypt0.zone/calculator but still need to know the speeds.  RainbowMiner will auto benchmark these but with non-optimized settings.

I've played around with the different algos but it seems the profitability is more or less the same. For some reason the Hawaii chipsets are all badly optimized and perform very poorly with other algos.

For example, the RX 470 is a very low budget GPU and it can almost get 30MH/s depending on the memory type. Its cost/hash factor is top notch compared to the Nvidia counterparts.

I've tried playing around by overclocking the memory and trying to up the core clock as high as I can get but I either get a small improvement in speed, or a huge power gain, or it just causes the system to crash.

SRBminer 1.6.8

after darkcoin they were actively used for groestlcoin mining. (at least by me)

I have a few Fiji (R9 nano) cards. I run them at 900/370MHz on the CNV8 using JCE miners. The hash rate is about 600H/s. Not so profitable.

I’m using R9 290 with modified BIOS for the lowest power consumption. They give about 23 mh/s on Ethash consuming up to 200 watts. Now it gives about $ 4 per month and additional heating useful in winter time. The time of these cards is coming to an end.

I had 8 of these and had to RMA most of them due to attempts at repasting and also just GPU death in general; I have one 290X Tri-X Sapphire still holding strong with the middle fan broken (cheap Tri-X fans eventually melt the center pin) but yeah it's a loss-leader it basically voids the other 5 1070s on the same rig in terms of profitability.

JCE CNv8 you can get 900 on with 1120/1420

For some reason, I've never had any luck with temps with repastes on reference type of coolers.

With the aftermarket coolers, I usually could achieve a better temperature if the card is a few years old but with the reference models, repasting doesn't do anything.

I am assuming that they use a better-longer-durable quality of thermal paste than the aftermarket manufacteurs or that simply the reference way of cooling is bad in general and without upgrading to an aftermarket cooler, the temps will always run high.

Thanks for the reference ;)
could you share your GPU config file content, i don't have this same card to bench myself, so i could add your config as a reference in my doc :)

I've also had mostly bad experience with 290s. Had only 5 or 6 of them in total, and 2 or 3 went bad within a few months (all bought new). Haven't ever had such high failure rates with any other GPU series.

I think 900 is very fast. I can get only 600 when the core frequency is 820MHz. That is equivalent to 820 at 1120MHz.

What were you mining with them? If you bought new, I am assuming it was late 2013 and you were mining Scrypt which was litecoin.

From what I recall, the R9 290 was a complete power-hog with Scrypt and if you ran with stock voltage they pulled 350-400Watts per GPU easily and due to poor heat control, they easily burnt up.

Not sure what coins I was mining back then, probably scrypt (vtc and doges mostly, iirc) and other algos that made sense at that time, but never mined with stock voltage on any gpu and definitely not 290s - always undervolting. Maybe just got unlucky.

Also earlier there were not so simple risers as we have now. I burnt one R9 290x after incorrect connection of additional power to the riser. It cost a lot at the beginning of 2014.

Yes back then we had to use the ribbon risers.

I remember people had issues with the powered risers due to the +12V line being shared with the motherboards and the PSU +12V rail.

Basically you had to make sure the +12V lines were cut and weren't connected to the motherboard because if they were then you got a feedback loop because the motherboard +12V wasn't exactly the same as the +12V from the PSU, even if you used the same PSU. And it was a quick way to destroy a GPU, motherboard and a PSU.

These GPUs are still profitable mining Blake2b at 12c/kWh and Ethash too but the hashrate on these late DAGs is around 22Mh/s