Woah epic thread here  I just skimmed through it since 90% is way over my head, so I might have missed it, but just wondering if any more efficient mem straps have been discovered for H5GC2H24BFR, H5GQ2H24AFR, or EDW2032BBBG that are stable in the ~1500MHz range for 290 cards. I know for the latter two The Stilt already posted modified straps for those so I'm more interested in the BFR straps...but I am curious about the other two too if Stilt's timings have since been improved upon. Currently the stock 1250 straps for the BFR are 77 71 33 20 00 00 00 00 8C C5 58 35 80 55 10 11 2C 94 18 05 00 48 C5 00 22 FF 1C 08 6C 0F 14 20 5A 89 00 A0 00 00 01 20 12 0D 24 29 51 24 2D 15 I went ahead and changed it with a stock 1125 strap for the same ICs I found somewhere 77 71 33 20 00 00 00 00 6B BD 57 2F 70 55 0F 10 28 92 97 04 00 48 C4 00 22 EE 1C 08 64 0D 14 20 4A 89 00 A0 00 00 01 20 10 0C 20 24 49 21 29 14 All that did was decrease the max clocks so it just gets the same performance at lower clocks. Anyway, again, just wondering if this info has already been posted and I just missed it. Keep up the great work! Is there a general donation address us plebs can send to, that is shared among everyone involved putting the time in for this? |
|
|
|
Current version of the wallet is 1.3.4 |
|
|
|
The best way to get more hash rate out of 980s is to sell them for RX 480s  |
|
|
|
|
If it wasn't for cryptohunter this thread would not only be not bloated with nothing, there wouldn't be anything, except a random post every few weeks to a month because there is no developers or community members posting here. Speaking of bloat...do you really have to quote the whole conversation every time you reply? On a related note...I'd much rather see this sort of constructive conversation instead of a bunch of hype train pump talk like the blocknet thread.
|
|
|
|
|
@Wolf0
I take it you use customized RAM timings? I don't think anyone else is getting those speeds just by copying stock 1500 or 1625 or whatever straps into higher ones.
EDIT: LOL nice XMR address btw.
|
|
|
|
I recommend having core and memory at same volts (because they will each use the higher of the two anyways).
Umm....you sure about that? That is weird as fuck if it's true. Haven't used Polaris myself but any other card they are two separate voltages. For Hawaii...mem voltage above 1.05V can be dangerous and yet is probably too low for the core under load even at stock clocks. Meanwhile the core can easily do 1.25V and not harm anything. Changed the wording above to make it clearer. What I meant was, the lowest stable clocks that you found for memory should be applied to the core as well, and to from there to undervolt the core you will need to introduce an offset as directly changing mvs below that wont work. Also, Polaris is an efficient architecture so it uses much lower core values than Hawaai/Tahiti so stock clocks are stable at 1000mv on almost every chip, while memory if pushed to 1900+ will need the full 1000mv that it defaults to, so in most cases the higher mvs are on the memory not the core. Unfortunately the higher mvs that you key in for memory are also used by the core (ignoring what you specified it to run at) unless you introduce an offset. I think you missed the point. I know Polaris is more efficient. I was just using Hawaii as an example to illustrate that there is VDDC, and there is VDDCI or AUX voltage, which are two different things. It is dangerous to set VDDCI higher than 1050mV or so, whereas the core (VDDC) can be safely higher....so to say to just set them equal because it will use the higher of the two means if he is already running at, say, 1150mV+ on the core, and then he sets his AUX voltage to 1150mV also, he can fry his card. Also, it seems like you are saying that neither AUX nor VDDC can be set at a value lower than 1000mV. That may be the case for AUX, but I'm pretty sure it's possible to set VDDC to whatever you want in DPM 7, as long as you lower the other DPM states to a value the same or even lower. If DPM 6 is set to 1000 and you try to set DPM 7 to 950, sure, it probably won't take...but as long as you lower DPM 6, 5, etc. to 950 or less as well it should work. EDIT: I should add that AUX/VDDCI is not actually the memory voltage, but rather the memory controller voltage, which is actually on the die itself. AFAIK the memory voltage is about 1.5V and is not configurable even with a BIOS edit. You would have to do a hard mod to change that which is a bit extreme and would require some serious knowledge. The effects of changing AUX voltage are small compared to a change in VDDC so even if you could get it working below 1000mV it wouldn't save you much power in comparison to lowering the core. I'm pretty sure the 480 hash rate is bottlenecked by its memory performance anyway, so it would be best to try and get the core voltage down because even if you had to downclock the core a bit, it shouldn't affect the hash rate too much compared to downclocking the memory. |
|
|
|
I recommend having core and memory at same volts (because they will each use the higher of the two anyways).
Umm....you sure about that? That is weird as fuck if it's true. Haven't used Polaris myself but any other card they are two separate voltages. For Hawaii...mem voltage above 1.05V can be dangerous and yet is probably too low for the core under load even at stock clocks. Meanwhile the core can easily do 1.25V and not harm anything. @Swiped I recommend you try the overclock.net forums for answers. There are lots of helpful and knowledgeable people there that will gladly supply accurate info. http://www.overclock.net/t/1604567/polaris-bios-editing-rx480-rx470-rx460/0_100http://www.overclock.net/t/1605802/official-radeon-rx480-470-460-owners-club/0_100If you have found the lowest stable voltages for the clocks you want and you're already running modded mem timings, then there's probably not a whole lot you can do except downclocking either the core (if you are bottlenecked by memory performance) or memory (if you are bottlenecked by the card's compute performance). If you can downclock one of those parameters without significantly lowering your hashrate then you can probably find a lower stable voltage for your new lower clocks and thus save power. |
|
|
|
Anyone using this miner with 290s or 390s on 15.12 drivers? I tried it with 16.9.2 but still get 31MH per card...no different than when I used Genoil's miner like a year ago. Weren't there some efficiency improvements from recoding using GCN assembler?
i found no difference on my 390 with 15.12 or 16.11.x drivers. 31 mhs @ 1100/1500 sapphire nitro I have the same results. I was running 290s and 390s in Windows 7 64 with 15.12 drivers using Claymore, until I tried sgminer-gm in EthOS (which is basically a custom Ubuntu Linux). I get a higher hash-rate with sgminer-gm, using slightly less power and no fee.I can also mine ETC,ZEC, ZCL or XMR with EthOS but obviously ETH is much more profitable right now. I have had it for many months and I switched to Windows/Claymore for a while but Windows just seems to always have problems. I switched back to Linux for stability. All my 290s and 1 390 are at slightly above 31mh/s except for one which has elpida RAM which is doing 30. The 290s are at core 1100 mem 1125, the 390 is at core 1120 mem 1500 and it often does 32mh/s. If you aren't dual mining and you can use Linux I think sgminer-gm is a no-brainer. I still haven't tried linux yet, but I think I am about to soon, because I just bought all the rest of the parts I need to build a new rig. I am going to try linux on this rig and it will be dedicated to mining/folding and the new one will be my normal day-to-day stuff/gaming rig. |
|
|
|
Anyone using this miner with 290s or 390s on 15.12 drivers? I tried it with 16.9.2 but still get 31MH per card...no different than when I used Genoil's miner like a year ago. Weren't there some efficiency improvements from recoding using GCN assembler?
i found no difference on my 390 with 15.12 or 16.11.x drivers. 31 mhs @ 1100/1500 sapphire nitro Ok cool. Thx!  31mh per card? i thought 290 were down to sub 20 now.
my 280x get about 15
Well when I ran it last night I noticed if I used the new intensity values (-ethi) of 8, 7, 6, etc. they would only run at 26MH. But, when I tried my old setting of -ethi -49152 I got my old hashrate of 31MH again. Not sure if it will work for you but it's worth a shot. It could just be that the DAG is too large compared to your total VRAM. Also I was mining ETC and I think my VRAM was only loaded to 1.6GB. |
|
|
|
|
Anyone using this miner with 290s or 390s on 15.12 drivers? I tried it with 16.9.2 but still get 31MH per card...no different than when I used Genoil's miner like a year ago. Weren't there some efficiency improvements from recoding using GCN assembler?
|
|
|
|
Hey guys, would appreciate an honest answer here please.
I was about to buy a few RX 470 8GB cards to mine either zcash or ETH but have been hearing about these new cards about to be released....Will there be a massive difference in speed from the 470/480s compared to these new cards?
I dont want to build a rig that gets me 1800 - 2100H/s then get blown away by a new miner with new cards that gets double + speeds..
I remember claymore mentioning that he cant do much because of the following..
Memory bandwidth:
RX480: 256 GB/s
If these new cards have the same bandwidth then does that mean no real big speed improvement?
Thanks.
Would like to know more about it too, maybe Claymore can explain this to us. Claymore please, we need your attention here! I cannot comment unreleased cards. Changes in hardware can improve or kill speed on such complex PoW as ZEC, there is no way to know until real tests. The new Vega cards are supposed to have half the memory bus width but a higher speed (2048-bit @ 1600MHz effective vs. 4096-bit @ 1000MHz effective) than Fury X, but with the same 4096 compute cores (although probably at a higher clock and lower power draw with perhaps some architectural improvements too). I believe Claymore said that the 4096-bit bus width of Fury is too wide to be taken advantage of, but the 512-bit bus of Hawaii can be fully taken advantage of...so worst case scenario (Fury is bottlenecked by its compute performance) it should perform at least a little better than Fury X. Best-case (Fury is bottlenecked by its memory) I would assume as much as 60% faster than Fury just considering the memory frequency increase alone but this is just speculation on my part. |
|
|
|
|
Well those people would be wrong no doubt.
|
|
|
|
|
Whelp GTX 1080 Ti with same number of cores as Titan XP is out for $699 USD. I guess all the talk of 1070 vs 1080 price/perf and ROI is all moot now.
|
|
|
|
|
I knew I should have bought those nvidia cards back in the fall...
Great news about Stanford wanting to integrate!
|
|
|
|
|
That should depend only on clocks and not the model (unless there was throttling due to poor cooler design). I think he said he was getting ~600K PPD with "mild" overclocks.
|
|
|
|
|
@QuintLeo
Is there any particular 1070 model that you recommend for longevity / price / cooling, etc.? I was thinking the EVGA ACX 3.0 08G-P4-5171. I know they had an issue with missing thermal pads but they'll send a free kit to fix that and honour their warranty. Seems to be one of the cheapest and also its EVGA with dual fans, so it seems like the one to go with but I wanted to get your opinion first.
|
|
|
|
My figures weren't $400 and $600 - someone else came up with those, I just agreed that they were reasonable for comparison.
I never stated any figures of my own.
Right, I was talking to ComputerGenie when I said that, because he said they were MY figures. - The pricing in that comment is using is his figures, not mine
Actually those were your figures... ... $400 seems quite a bit on the high side for the "additional base system" to me, but that's likely due to the fact I don't build riser systems.
Right, like I say I was being conservative with all of my numbers and erring in favor of the 1080s in order to do a comparison in the worst case scenario for the 1070s...and to not have to endure having everything I said being nitpicked to death and implications of me being mentally challenged. (To no avail concerning that latter part) It IS interesting that the $4000 point was chosen for the comparison though, since it seems to be THE most favorable point for the 1080 - and the 1080 STILL ends up losing past the first month or two.
That's why I chose it  (And presumably why Mr. Genie chose it to counter an earlier post of mine) Now, I'm sure you're still stuck on the "more cards" end of your thought so, remember this...
Cards 7 & 8 will need another full set of hardware to run.
|
|
|
|
Perhaps reply to this topic in the curecoin forums: https://www.curecoin.net/forum/viewtopic.php?f=31&t=3898It hasn't been updated, but I think those forums are followed more closely by Cygnus and Vorksholk than this thread is, which is understandable with the amount of irrelevant posts one has to read through to get caught up if you've been away from this thread for a while (I'm clearly guilty of adding to this number. See above ). Also it was posted by Cygnus so he may get a notification in his e-mail or something if you reply to it. I've chatted with whoever runs the CureCoin account on Facebook and they seem to respond pretty quickly too. I don't think it's either of the devs but I'm sure they could get in touch with them if you ask. |
|
|
|
|
Show Posts
