Sapphire r9 280x only getting 20.5 mh

drovdrov Member Posts: 3
Hello, I've seen a lot of people saying that they get 22-25 mh from their r9 280x card. I've tried many configs but I'm stuck at 20.56 maximum.

My current config is like this : ethminer --farm-recheck 200 -G -F http://127.0.0.1:8080/drov --cl-global-work 8192 --cl-local-work 128

using the dwardpool stratum proxy default configuration.

Do you guys do anything different to get higher hashrate ?

Thanks in advance.

Comments

  • adasebadaseb Member Posts: 1,043 ✭✭✭
    What is your clockrate?

    Personally I wouldn't overclock these Sapphire 280X because they got weak VRMs and last month one of mine went on fire.
  • ericeric Member Posts: 76
    is it hynix or elpida?
    on my experience elpida is much lower than hynix
  • blueboxbluebox Member Posts: 181 ✭✭
    edited March 2016
    It's well known that hashrate on AMD has been declining with the increasing DAG file size. At least many in this forum have seen the decline, so the benchmark here likely has merit:
    http://cryptomining-blog.com/7316-the-latest-ethminer-0-9-41-genoil-1-0-4b3-for-windows/
    Notably, the GTX doesn't show any decline. Wondering if it's real or an artifact of benchmarking, but given other's experience with their AMD's it appears real. Maybe @Genoil can help explain better. The DAG file has only grown about 5% since I started, so I can't say for sure but my hashrate hasn't budged on the GTX.

    EDIT: Uh-oh, just said something negative about AMD, better get my firesuit on. :D
  • MrYukonCMrYukonC Member Posts: 627 ✭✭✭
    bluebox said:

    It's well known that hashrate on AMD has been declining with the increasing DAG file size.

    This is an incomplete statement, therefore it's wrong by omission.

    Cards with 2GB - 3GB have seen a slight decrease in hashrate. 4GB AMD cards have not seen any drops. Neither have 4GB NVidia cards. Thus, it's no surprise that your GTX 970 4GB has not dropped in hashrate.

    There is no grand conspiracy here about AMD vs NVidia. It mostly has to do with computer science and computer hardware/engineering.
  • blueboxbluebox Member Posts: 181 ✭✭
    @MrYukonC So let's see what an authority says, I asked @Genoil (thanks, btw):

    "both Nvidia and AMD are affected by growing DAG size, but in different ways. AMD card gradually become slower, while NVidia cards have a threshold after which there is a steep decline in hashrate. For GTX9x0 series using WDDM 1.0 or Linux the threshold is 2GB DAG size, so well after planned PoS switch."

    This means a 3GB GTX 780 hasn't declined since mining began (with a 1GB DAG). Would have been more constructive for cryptomining-blog to compare the 280x with a 780, but it appears the story would be the same.

    Hope this clears things up. (still wearing firesuit though)
  • happytreefriendshappytreefriends Member Posts: 533 ✭✭
    edited March 2016
    Your problem is Dwarfpool. :wink:


    On my only rig populated with a 280 and a 280x, I get 46MH but I'm running at : 1060/1500, undervolted on the 280x to 1075mV. I only use XFX boards with the 'good' memory. :wink: (3GB models each)

    For real though, a 280x will now yield about 20-22MH with stock or slight overclocks.
  • c3p8c3p8 Member Posts: 28
    which pool would you recommend?
  • blueboxbluebox Member Posts: 181 ✭✭
    I wonder how much faster the DAG is growing, now that the global hash rate is so much higher than a month or two ago. Haven't been tracking it, but I'd think it has to be relative. Since it's about 1.38GB now, perhaps another 100-200GB comes in a couple months and they might be on par with each other (3GB 780GTX does 17.5MH no o/c, 280x getting 20 or so now)? Anyone able to extrapolate when the DAG might get to 1.6GB?

    Just an innocent question/observation is all.
  • happytreefriendshappytreefriends Member Posts: 533 ✭✭
    I'd suggest ethpool.org IF you have to use a pool. Otherwise try your luck at solo mining. It's been very good to me .
  • dlehenkydlehenky Member Posts: 2,249 ✭✭✭✭
    @bluebox The DAG growth rate never changes, unless nyoc lowers the target block time. Network hash rate has no effect on DAG growth. The DAG grows 8.3 MB per epoch, period. If the block time gets shorter, it will simply cause the epoch to happen more frequently time-wise (every 30,000 blocks).
  • GenoilGenoil 0xeb9310b185455f863f526dab3d245809f6854b4dMember Posts: 769 ✭✭✭
    @bluebox regarding AMD I don't know exactly the effects of DAG size between different versions of the GCN architecture. For 1.0 there's a definite slowdown, for 1.1 and 1.2 I don't know exactly.

    The dag growth rate is determined by block time, not difficulty. You can look up the DAG size for each epoch (==block/30000) in the ethash design rationale.
  • davethetrousersdavethetrousers Member Posts: 46
    High quality 280X's can get to 23MH at this point. One of my Asus' (die quality ~90%) does, at 1140/1670 clocks, without overvolt. These really do have good memory, I must say.
    So the absolute maximum for 280X might be like 24MH, if you overvolt such a card crazy (not a good idea on high quality dice) and do maximum overclocks.

    Regarding pools, I'm on the really small pool eth.pp.ua for a while now and, thus, can recommend it. To me, a small pool is a compromise of overall mining decentralization and hedging my luck. When I reach my maximum hashrate in the next few days, I might switch to solo, though.
  • blueboxbluebox Member Posts: 181 ✭✭
    @dlehenky @Genoil Ah, thanks — now that you (re)explained it I recall seeing this before. Low-level mechanics of mining code are a mystery to me, but if you ever needed to know about optimizing HPC apps with infiniband just ask. :)

    So (thanks to wikipedia) Oland, Cape Verde, Pitcairn and Tahiti based cards like the 280/280x show decline, though we don't know about 7790/8770, R7 260/260X, R9 290/290X, R9 295X2 etc. cards yet. Good to know.

    Apparently we'll be many more moons before growing the DAG another 200MB. Meanwhile I see people with 280x (gcn1.0) now managing ~20-22MH, 290's (gcn1.1) around 24MH. Will check back in a few months. ;)
  • r928520r928520 Member Posts: 9

    I'd suggest ethpool.org IF you have to use a pool. Otherwise try your luck at solo mining. It's been very good to me .

    What is your MH/s?
  • JBaETHJBaETH Member Posts: 40
    adaseb said:

    What is your clockrate?

    Personally I wouldn't overclock these Sapphire 280X because they got weak VRMs and last month one of mine went on fire.

    I really hope you're kidding on this because I have 4 Sapphire 280x OC'd @ 1175, right next to my bed.
  • happytreefriendshappytreefriends Member Posts: 533 ✭✭
    Sapphires are one of the lowest QL Graphics Card makers, just like EVGA. They've been since starting many years ago. That's why they are pretty cheap. Watch out.

    My solo MH is only 400+, will be back to 1PH soon. (Taking my pool rigs (1/2) to do solo only).
  • adasebadaseb Member Posts: 1,043 ✭✭✭
    JBaETH said:

    adaseb said:

    What is your clockrate?

    Personally I wouldn't overclock these Sapphire 280X because they got weak VRMs and last month one of mine went on fire.

    I really hope you're kidding on this because I have 4 Sapphire 280x OC'd @ 1175, right next to my bed.
    Which 280x do you have? The Dual-X ? then those usually start fires. Search bitcointalk.

    The newer Sapphires don't have this problem or the Toxic ones.
  • davethetrousersdavethetrousers Member Posts: 46

    Sapphires are one of the lowest QL Graphics Card makers

    At a cursory glance, it might seem like it. As some dudes on a forum. we can't know it better than anecdotally, though. I'd really like to find some actual statistical comparison of chip qualities used by the different vendors in their different products.
    adaseb said:

    Which 280x do you have? The Dual-X ? then those usually start fires. Search bitcointalk.

    Well, that seems a bit extreme of a blanket statement. :sweat_smile:
    I do however suppose Sapphire don't buy the highest quality chips for their basic product. The one I once had was like a ~55% die quality, which is quite low. But suggesting they deliberately under-engineer their card such that they set themselves on fire I find a bit extreme.
  • adasebadaseb Member Posts: 1,043 ✭✭✭
    You don't have to believe me but one day I came home rig was turned off with a weird smell. I located the faulty GPU and it was the Sapphire. I looked and there was a huge burn mark near the VRMs. The card is garbage. Only value is the fans since they are still spinning, I can probably sell those on eBay for like $20.

    They aren't meant for mining only gaming. That's why Sapphire started throttling their newer cards because most were burning out on them.

    They might not start a fire but the VRM burns out and something shorts and then the power supply overcurrent protection kicks in and the computer turns off.

    If you have warranty then mine away but the warranty was only 2 years on those cards.
  • JBaETHJBaETH Member Posts: 40
    adaseb said:

    JBaETH said:

    adaseb said:

    What is your clockrate?

    Personally I wouldn't overclock these Sapphire 280X because they got weak VRMs and last month one of mine went on fire.

    I really hope you're kidding on this because I have 4 Sapphire 280x OC'd @ 1175, right next to my bed.
    Which 280x do you have? The Dual-X ? then those usually start fires. Search bitcointalk.

    The newer Sapphires don't have this problem or the Toxic ones.
    Yes 2 of them ard Dual-X, rest being Vapor-X.
    Which one are the new ones?
  • adasebadaseb Member Posts: 1,043 ✭✭✭
    Vapor-X, the vbios makes the card throttle when it gets too hot.

  • adasebadaseb Member Posts: 1,043 ✭✭✭
or to comment.