• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

X11 configurations for 270X, 280X, 290X, and 7950

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
Lyra2RE results:

Tahiti: 1.0-1.1mhs
Pitcairn: 660-670khs
6870 915/1050: 450khs
7770 1020/1350: 230khs

Seems most stable at xI 64 for me, all cards, though I haven't done a ton of tinkering yet. 128 doesn't seem to give me much benefit, and it's not so stable at the moment. I did swap in the ROTL and ROTR code improvements, haven't done anything since. Haven't really kept up with the development... too busy a week for me.

The 6870 likes one thread, everything else seems better with two.

Edit - the unstable rig was down to a 270 that didn't like running at 1050/1500, 1.15V. It's now running 1000/1500, 1.12V, seems to like that better. I suspect this card's been flaky a while now... it's given me problems on XMR before. I'll have to pull it soon and re-apply the thermal grease. Runs hotter than all the other 270s.

Numbers updated after grabbing the latest Lyra2RE build. Hynix/Elpida seems to make no difference anymore, the Tahitis all run equally fast clock for clock.
 
Last edited:
What are your Lyra2RE settings, OW? I'll probably try to mine it for a bit when it switches algos with my remaining cards.

I should probably figure out what miner I need before then LOL.
 
Xintensity 64 or 128, worksize 64, 2 threads... that's just about it. Not much needs to be in the conf for it to work. Everything can be found in the Vertcoin thread at BCT.

I'm probably already using an outdated version of the custom SGMiner, but it seems to work just dandy on the test pools. Even if I had one of my 270s go sick and torpedo the drivers last night.
 
Only if it proves more profitable than FTC ;)

Roughly 6 hours to go before the switch.
 
Only if it proves more profitable than FTC ;)

Roughly 6 hours to go before the switch.

Good luck guys, hope it's profitable! :cheers:
I think as far as I can see I'll be folding for T32.

I just can't risk putting money into mining now.
I'd wager I would need A LOT of GPU's to make it worth while :(

Maybe one day I'll return to it
 
Good luck guys, hope it's profitable! :cheers:

Thanks - I'm sure I'll need it. Right now I'm barely doing 0.008-0.009 BTC a day on FTC, and that's with 8 cards. Terrible, but better than nothing and it's still all profit for me.

We should probably stop cluttering up Wolf0's thread now.
 
Last edited:
Back somewhat more on topic...

I'm catching up. Starting to see how the Neoscrypt kernel was improved. Not making much progress on getting it running faster, because the compiler sucks out loud sideways, but my brain is braining things in code again. In no hurry with this... as long as only a couple people have improved kernels, I'm not at any real disadvantage. Adding more cards easily makes up for it.

Got Lyra2RE up to 450khs on the HD6870, but that's only because I pulled it out of the dedicated miner it was in and shoved it into a rig of its own. It's not trying to coexist with Tahiti and Cape Verde anymore. Not going to bother trying to optimize Neoscrypt for it. That card's a waste of time with two more 280Xs incoming. Once they do, the farm will be all Tahiti and Pitcairn. The two slowpoke power hogs are going bye-bye.
 
Very cool. Been too involved with the power supply torture stuff to do any more fiddling around with it myself. Maybe I'll get some more time to do that next week or the week after. Been looking into the vector stuff...

I'll worry about Hawaii when I can finally afford one :D
 
Very cool. Been too involved with the power supply torture stuff to do any more fiddling around with it myself. Maybe I'll get some more time to do that next week or the week after. Been looking into the vector stuff...

I'll worry about Hawaii when I can finally afford one :D

270X is no slouch on it either, above 290kh/s.
 
That should keep me motivated...

Found out the hard way one of my 270s just won't do 1500 on the memory, which explains why it'll sometimes hash fine for days on one algo and then go to whitescreen city when I switch to Cryptonight or Lyra2RE. Stupid Elpida. 1485 finally made that card happy.
 
Incidentally, I do have to say that this is reminding me why coders do need to be compensated properly. People think there's nothing to it, like you just sit there and change a couple variables here and there and bam... instant improvement. No, this takes a lot of time. Especially dealing with a buggy compiler like this one.

The amount of time I've already put in to try and catch up is not at all insignificant. Dragging 18 year old knowledge out of my head based on coding C for Pentiums and trying to update it for hardware completely new to me (I mean, come on... you can't code for a GPU the same as a CPU and expect optimal results) has been a major headache. We didn't even have Radeons back then, let alone R9s. I'm getting there... slowly... but it's going to take a while. Even for the guy who graduated from CDI College at the top of the class back in the day ;)

It's a lot like reviewing power supplies. A lot of folks think it's easy and all I do is sit there pushing buttons and writing down results, but they're just not aware it's a full blown full time job that pays less than minimum wage.

I'm hoping you're still getting the odd donation every now and then. I might do another one later, once I rebuild my XMR holdings after that cashout last week.
 
Incidentally, I do have to say that this is reminding me why coders do need to be compensated properly. People think there's nothing to it, like you just sit there and change a couple variables here and there and bam... instant improvement. No, this takes a lot of time. Especially dealing with a buggy compiler like this one.

The amount of time I've already put in to try and catch up is not at all insignificant. Dragging 18 year old knowledge out of my head based on coding C for Pentiums and trying to update it for hardware completely new to me (I mean, come on... you can't code for a GPU the same as a CPU and expect optimal results) has been a major headache. We didn't even have Radeons back then, let alone R9s. I'm getting there... slowly... but it's going to take a while. Even for the guy who graduated from CDI College at the top of the class back in the day ;)

It's a lot like reviewing power supplies. A lot of folks think it's easy and all I do is sit there pushing buttons and writing down results, but they're just not aware it's a full blown full time job that pays less than minimum wage.

I'm hoping you're still getting the odd donation every now and then. I might do another one later, once I rebuild my XMR holdings after that cashout last week.

Don't worry - not releasing this Neoscrypt for a while, if I do release it. I started coding for GPUs in June, I'm just all right at it. I'm above average at CPU coding, I think - I was badass at assembly when I was 13 - 15, but I haven't used that in a long time...
 
No problemo. If I'm not caught up by then, comparing your code then with your code now will probably catch me up real fast :D

Assembly... that was the one part of the programming course I hated second most. Still did fairly well at it, but I was always much better at finding inefficient code in C and streamlining things. Enough to get the attention of the textbook people, apparently. At any rate, I like working in C. Not so much assembly. The Visual Basic part of the course was ok too, but I still liked C better. COBOL was hell on earth, though. I'd rather do assembly.

It's not that COBOL was any harder than assembly, it was just the fact that these guys gave me no choice: "COBOL or FORTRAN, pick one or the other, it's mandatory." I thought they were wasting my time with old stuff I didn't want a career coding in.

And if you ask me, if you're already doing this well at the GPU stuff after only a few months, you're better than just all right at it. Keep up the good work, and so on, and so forth...
 
Last edited:
So what are you guys coding?

The .bin files the miner makes?
I hate code, it just never makes sense - to complex.

I'm a hardware guy anyway :p
 
Nope, the OpenCL source code. I'm mostly just fiddling with the kernels right now, but I have looked at the rest of the SGMiner code a couple times.
 
Two more cards added to the dedicated mining rigs, both 280X Hynix.

Suddenly, the leaked X11 Tahiti bin is no longer working on either of those rigs. Hardware errors galore on all cards, unless I delete the leaked bin and let it build an old bin.

Both of the new cards are on the bottom rig... the two new 280X cards and the three stable Pitcairns. This rig will work without hardware errors if I kill the leaked Tahiti and leave the leaked Pitcairn. Top rig has the two 7950s and one Cape Verde. Same deal there... insane HWs unless I remove the leaked Tahiti bin.

Looks like I need to stop mixing and matching to get this working (all Tahiti on one, all Pitcairn on the other), or the drivers need me to clean wipe and reinstall. Both rigs are fine with Neoscrypt though, so I went back to that for now. The leaked Tahiti bin is still playing nice with all the single card machines.

Edit - got them sorted Pitcairns on one rig, Tahitis on the other. X11 bin is still not playing nice on the Tahitis. 14.6 drivers.

Not gonna worry about this. Back to Neoscrypt.
 
Last edited:
I've got Freya doing well on Neoscrypt - 1 290X, 1 280X, and 3 7950s doing 2210kh/s average while pulling almost exactly 1100W.
 
Back