23 Mh/s for a R9 280x?

farwarefarware Member Posts: 116
edited November 2015 in Mining
UPDATE: See test results below and why you underclock memory, overclock core clock and undervolt VDCC.

Hi guys,

how much are you getting from a single R9 280x?

I have three different models: VaporX, TriX and Asus OC edition - most of them are clocked at 1070 core / 1550 memory. The maximum hash rate per card is 23 MH/s for the VaporX, TriX. The Asus cards are doing 25 MH/s max. I tested that on Win8, Win10 and Ubuntu.



Thanks
FW




Post edited by farware on
«1

Comments

  • farwarefarware Member Posts: 116
    So my conclusion is if you can afford it I would really recommending to buy ASUS cards only because they really get the 25MH/s you'd expect from a R9 280x. They perform the best out of the box.

    Sapphire Tri-Vapor X are ok too but the VaporX (DualX) is not as good (23 MH).

    7950 are pretty cheap and I can get them to 23MH so that's pretty efficient and they dont draw that much power if you undervolt them. But out of the box they draw way too much power.




  • farwarefarware Member Posts: 116
    Ok did some further digging, I can reach about 26 MH/s per card now but the wattage per rig goes up by 50-100 Watt which is way too much for such a small gain.

    My recommendation for 280x: Memory 1300 - Core clock 1000 @ 1.05 VDCC - a lot less energy wasted.
  • MrYukonCMrYukonC Member Posts: 627 ✭✭✭
    All of my R9 280Xs are PowerColors along with one XFX.

    I have them all set to core clocks of 1050 and memory clocks of 1350. Most of them have core voltages set at 1.037 with a few set at around 1.05.

    They all hash at 24.5 - 25.5 MH/s and consume very low wattage.
  • farwarefarware Member Posts: 116
    @MrYukonC After much fiddling and comparing how much energy the rigs use overclocked vs underclocked, I eventually decided to overclock my 280x's to 1600 memory and 1150 core @ 1.084 and they are running quite nicely. The key is the cooling. Temperatures outsides are really going down now where I live so I can overclock and undervolt.

    I am quite happy now with my rigs running at 350 MH/s at only 75 kW per day. So that's quite good. The 280x rig is far more efficient than my 7950 rig..

  • estevaozabotestevaozabot Member Posts: 29
    I have a PowerColor 280x 1500/1030 = 22.6Mh/s..
  • MrYukonCMrYukonC Member Posts: 627 ✭✭✭

    I have a PowerColor 280x 1500/1030 = 22.6Mh/s..

    I would suggest experimenting with dropping the memory clock down to 1350 and see what that does. Like I said, all of my R9 280X PolowColor cards hash faster with the memory clock at 1350 versus the stock 1500.
  • estevaozabotestevaozabot Member Posts: 29
    MrYukonC said:

    I have a PowerColor 280x 1500/1030 = 22.6Mh/s..

    I would suggest experimenting with dropping the memory clock down to 1350 and see what that does. Like I said, all of my R9 280X PolowColor cards hash faster with the memory clock at 1350 versus the stock 1500.
    You changed the Clocks in the BIOS? or by MSI Afterburner?
  • MrYukonCMrYukonC Member Posts: 627 ✭✭✭

    MrYukonC said:

    I have a PowerColor 280x 1500/1030 = 22.6Mh/s..

    I would suggest experimenting with dropping the memory clock down to 1350 and see what that does. Like I said, all of my R9 280X PolowColor cards hash faster with the memory clock at 1350 versus the stock 1500.
    You changed the Clocks in the BIOS? or by MSI Afterburner?
    I change the core voltage in the BIOS using VBE7 and then reflash the cards with the new BIOS.

    I change the memory clock using aticonfig at runtime (I'm running Linux).
  • ethfanethfan Member Posts: 458 ✭✭✭
    MrYukonC said:



    I change the memory clock using aticonfig at runtime (I'm running Linux).

    Tried decreased memory clock as you suggested but the hash rate seems unaffected. Dropping the core clock does reduce the hash rate. Was testing the 7970 which is the same as the 280x.
  • MrYukonCMrYukonC Member Posts: 627 ✭✭✭
    ethfan said:

    MrYukonC said:



    I change the memory clock using aticonfig at runtime (I'm running Linux).

    Tried decreased memory clock as you suggested but the hash rate seems unaffected. Dropping the core clock does reduce the hash rate. Was testing the 7970 which is the same as the 280x.
    You have to issue

    amdconfig --od-enable

    before aticonfig with --od-setclocks will work.

    Also, the syntax for --od-setclocks is pretty crazy. You have to specify which card you are targeting, by using the --adapter=n flag.

    For example, to set the clocks for card #3 (numbering starts at 0), the command would look like the following:

    aticonfig --adapter=2 --od-setclocks=CoreClock,MemClock
  • ethfanethfan Member Posts: 458 ✭✭✭
    Yes it was done right. I did --odsc=0,1450 down to 0,1100 in steps of 50. Each time checking with --odgc. The hash rate was monitored in a separate terminal that was running ethminer (a version that reports hash rate continuously). Remained about the same throughout. Which is why I mentioned the change with altered core clock. Dropping to 950,0 reduced hash rate by about 2Mh/s. 900,0 even more. Which means the commands were getting through to the graphics card in real time!
  • dont12dont12 Member Posts: 60
    I find that if I undervolt, works fine for a few days and then just locks up. Card is running at 98% utilization, but its not actually giving any results. Interestingly if you try to close the terminal window It freezers the whole machine.
  • estevaozabotestevaozabot Member Posts: 29
    dont12 said:

    I find that if I undervolt, works fine for a few days and then just locks up. Card is running at 98% utilization, but its not actually giving any results. Interestingly if you try to close the terminal window It freezers the whole machine.

    I'm having trouble with crashes também..mesmo not making any changes to the board ..
    Two R9 280x cards and win 10
  • estevaozabotestevaozabot Member Posts: 29
    another question, you use the power of the risers?
  • dont12dont12 Member Posts: 60
    Yeah risers powered, had a molex plugged into the board, but mono failed so now no molex in the board. Its a HD 7990 card I have undervalued it to 1.18V runs at 86C if I don't underplot goes to 94C, fast though 52MHS
  • estevaozabotestevaozabot Member Posts: 29
    I am using this extender and am having trouble with crashes ..

    Use it in PCI x16 port motherboard.

    Can you be defective?



  • PansyfaustPansyfaust Member Posts: 29
    Yeah, my combined hashrate for my 7990 and 2 280x's has gone from 100Mhs down to around 96Mhs. No idea why.
  • farwarefarware Member Posts: 116

    I am using this extender and am having trouble with crashes ..

    Use it in PCI x16 port motherboard.

    Can you be defective?



    If you have another 11 bucks buy the powered USB riser (the blue one). This is a powered riser but the USB risers seems to work better.

    However before you do that I would try and see if its properly connected or you somehow damaged the PCIe slot by incorrectly inserting it.
  • HelioxHeliox Member, Moderator Posts: 634 mod
    Hey, i know it's not a 280X but still want to add some info.

    I'm running a 4x Fury X rig that's doing 115Mh combined (no oc, stock clocks) @ exactly 1000watt

    So 115Mhz / 4 = 28.75Mhz per card



    1000 - 50(give or take)watt for mobo = 950 / 4 = 235Watt per card (Roughly)

    OCing a 280x to get 28.75Mhz will give you around 100 watt extra as i read here.

    So what is that? also 250 watt?

    But of course, a FuryX is around 3x the price of a 280X. So as already confirmed by so many others. 280X is the card to go on Ethereum.


    However!

    I just got a Nano card and did some testing.

    At stock clocks, 1000core/500mem. I reach exactly 25Mhz (Windows 10)

    here comes the good part. This @ 220 watt. (Mobo, cpu, mem included)

    It's an i7 processor that eats around 75watt idle.

    So mobo + cpu + mem = around, lets say 90-100watt? Give or take..

    So in other words, 120 watt @ 25Mhz.

    So downvolting this is even more interesting i think!

    Again, downside. 500+ EUR for 1 card :neutral:
  • RDashRDash Member Posts: 7
    edited December 2015


    I get near acurate 25MH/s all time with my R9 280X ;)

    Card:
    i.imgur.com/GQXTvd9.gif
  • dont12dont12 Member Posts: 60
    I am getting 20 MHs on my 380 cards power is 175W vs 250 on the 280Xs but the 280Xs doing 25.5Mhs
    So .11Mhs/W on the 380 versus .1Mhs on the 280Xs Getting a water cooled 390X today gonna be interesting to see what power draw is.
  • peepeedogpeepeedog Member Posts: 32
    RDash said:



    I get near acurate 25MH/s all time with my R9 280X ;)

    Card:
    i.imgur.com/GQXTvd9.gif

    Is this the view for the standard miner? I don't see any hashrates on mine. What command do you run to see the hashrate? the only time i saw the hashrate was when i benched the cards, but after starting to mine, i don't see any hash.

    When i do web3.eth.hashrate on the geth console, i get a number like this that changes all the time:
    2067391578384541400
    > web3.eth.hashrate
    2613119980733287400
    > web3.eth.hashrate
    1168908908985762000
    > web3.eth.hashrate

    thanks in advance!
  • ethfanethfan Member Posts: 458 ✭✭✭
    edited December 2015
  • peepeedogpeepeedog Member Posts: 32
    Thanks everyone.

    Not to hijack this thread but i think they are relevant newbie questions pertaining to hashrates:
    1) I've been mining with 2 x 280X for close to 2 days now with 0 Eth - when people say "you need to mine for 5 days or X days" does this mean continuous mining? Can I reset the miner during this period? Does it even matter if it's continuous or does some tracker/counter reset somewhere and I start from scratch again after say, configuring a new GPU and having to restart it for 5 seconds?

    2) The common info i see on Geth is something like:
    commit new work on block xxxxxxx with 0 txs & 0 uncles. Took 1ms

    So i know what uncles are and i know you do get rewarded for them, but how about txs? Sometimes I see 0 txs, sometimes 17 etc etc.. I assume these are the actual txs for the current block, and I've somehow contributed to securing them via some txs? Do we not get rewarded for these small txs?

    3) When people say "my hash rate is 200 MH/s" - do they just mean they have roughly 10 GPUs with 10 miners running at the same time per Geth console? I'm currently mining with 2 x 280x with a hashrate of around 23 each, so does that mean my total hashrate is 46? (sorry if this question sounds dumb, i'm trying to figure out if there's a collective way to mine i.e. generate one hashrate per RIG, rather than per GPU, and if it even matters)

    thanks all!
    Shaun

  • dont12dont12 Member Posts: 60
    There's no counter, you can reset. If you get an uncle no txs reward. Your hash rate is 46 mhs Doesn't matter whether its one rig or ten.
    By the way if the number of txs gets very high, the gas spent is much higher than the block reward of 5 ether.
    Not at the moment, but when Olympic finished they drove txs as a test, as a miner you ended up with way more ether, than the 5 per block.
  • peepeedogpeepeedog Member Posts: 32
    dont12 said:

    There's no counter, you can reset. If you get an uncle no txs reward. Your hash rate is 46 mhs Doesn't matter whether its one rig or ten.
    By the way if the number of txs gets very high, the gas spent is much higher than the block reward of 5 ether.
    Not at the moment, but when Olympic finished they drove txs as a test, as a miner you ended up with way more ether, than the 5 per block.

    Thank you dont12
  • peepeedogpeepeedog Member Posts: 32
    I've run into a strange error today - had 2 cards mining successfully over 2-3 days, but when i checked in today, black screen. I rebooted the box and reinitialized the GPUs.

    When trying to start the miners, both would start but the second one would always cause a crash on the first guy, throwing a "Bus error (core dumped)." If i try to start the crashed miner, it would start successfully, but crash the one that was already running, effectively only letting me use 1 card.

    I initially thought this was to do with my forced reboot of the box and maybe it had DAG issues, so i rebuilt that but still no go.

    Has anyone seen this behavior?
  • peepeedogpeepeedog Member Posts: 32
    edited December 2015
    Okay it's working now - somehow even tho the DAG% completed (100%), both miners wouldn't work at the same time until the Message "ethminer Full DAG loaded" appears. After starting the second miner it's working now.

    Strange ! Hope this helps someone
  • aminlvaminlv Member Posts: 7
    kinda wondering what do i do wrong!
    my local shows 13.XX mhs hashrate but when i look @ https://ethermine.org/miners
    it shows between 1~2 hashrate!!!

    anyone knows why?!
Sign In or Register to comment.