Topic: What flags should I use for 5830s? (Read 11701 times)

Use
R
T
F
M
and
SEARCH

Remember, a higher MHs/W does not mean more profitable. I suggest you do some testing, and figure out what undervolting and overclocking settings run you the most PROFIT with your given electric costs, and the current price of BTCs.

There was some useful information about improving Mh/w in there. How many pages can you spend on "best flags" for a given GPU anyway? 

A shame this thread went off topic and for so many pages

I think I just discovered your secret on how to double your bitcoins:
https://bitcointalksearch.org/topic/bitcoin-doubler-submit-and-wait-51726

If only I had seen that earlier, Id have been rich!

and if you re-read i corrected my errors before your reply. But im mashed, so thats my excuse

I do it for BTC's, im just gathering them together actually. Looking nice. and electrics paid for so i'm sticking with bob on this one and dont worry, be happy

Current drop aint good though. Still, more BTC's with my other projects. Its win win

Exactly  5850 is a crippled 5870 as I said.  It is the exact same chip just w/ SP turned off due to defects.   The 5850, 5870, and 5970 have the exact same number of transistors (2.154M to be exact). .


100% wrong.  Easily verified.
http://en.wikipedia.org/wiki/Comparison_of_AMD_graphics_processing_units#PCIe_.28HD_5xxx.29

The 5970 (5990 doesn't exist) uses less power per hash.  It is one of the most efficient cards on W/MH basis.  TDP is not a measure of power consumed.  It is a measure of max power that can be displaced without damaging the card or other nearby components.  If you did all your estimating on TDP well it is worthless.

5970 TDP is 300W.  I have 3x5970 w/ MB, CPU, RAM, 2x 120mm fans, and USB drive for ~870W at the wall.  Backup out 12% PSU inefficiency at that is ~800W DC.  If we guesstimate that rest of system uses 50 to 100W that puts the peak load on each 5970s at 230 to 250W.

The whole rig is >2.3GH/s @ 870W (AC at the wall) for ~2.6MH/W.



Who said anything about games.  This is about mining efficiency.  The most hashes for the minimum amount of hardware & electrical cost. X2 cards lets you get more GPU per rig and that means the hardware & electrical cost of the rest of system is ammortized over more MH.


Just because you can't find one doesn't mean they don't exist. 

http://www.ebay.com/csc/i.html?LH_ItemCondition=12&rt=nc&LH_Complete=1&_nkw=HD%205970&_dmpt=PCC_Video_TV_Cards&_fln=1&_trksid=p3286.c0.m283

Average price over last 10 sales was $330 ea.  Thats ~ 210 GBP.  Let me know when you are going to pay that 10 BTC.

Nar, its annoying that renaming, i read it back haha

If you find a real Radeon HD 5990, I'll buy it off you

No such thing exists.  Radeon HD 5970 is the high-end dual-GPU card in the 5xxx model range.  It is that particular line's 6990.

[edit] I see you edited your post already, sorry for calling you out

[edit] Furthermore, just for clarification.  The 5970 is two Cypress cores - the stream processor count would indicate that the GPUs are infact 5870 chips (1600 x 2) and not 5850 cores ( 1440 ).

Also the 6870 is the equivalent upgrade of a 5830, not a 5770 - it has an identical stream processor count, and improved performance in all other respects.  In regards to mining, the 5830 performs better due only to better overclocking ability.

A 5850 is a few cores less than 5870, There is no such thing as a 5950, just 5970, equivilent to the 6990 which are 2x 5870's/6970's respectively. the 5970 may of been 2x5850's due to TDP but i cant remember and THEY DO USE MORE POWER PER HASH. I cant be bothered go find evidence on this. Do it your self. point is few 5870's actually out perform the 5850's as the 5850's seem to clock higher with less cores. It's a close call Dependant on makes and models compared. All comes down to price at end of day, but the X2 cards are just silly, for silly people. The only game i have that needs more than my 6870 (thats equivalent to 5770) is S.T.A.L.K.E.R C.O.P with S.M.R.T.E.R addon, All 3 Absolute addons and Atmosfear 2.3, v3 makes it not load That includes BF3 on ULTRA @1920x1080, Total War 2. Point is for games, theres nothing to push these cards. Crysis 1 seemed to be a power hog, never interested me though. Hate BF3 been EA, grrrrr. Stalker COP is wonderful and its only because Atmosfear on high requires 3GB VGA ram so duel VGA cards is recommended. Then again, lower some other settings. But this is maxing out

Thius is why i see AMD going to CGN straight through on there 8 series cards, 9XXX latest when they die shrink again and the only thing left for mining will be FPGA boards. With a few people working on it already, hopefully prices will drop before end of next year or before the 9 series and we get a nice transition

Good banter with you 2 and kept it polite. Im sure we all agree AMD's renaming scheme is really annoying

Coming from the guy who wrote "I live in uk (expensive) and at the rate now i mine less than 24 hours in the week to cover the electric. Its expensive here too. Very. the other 6+ days is profit." when in reality you pay 2.5x market price for bitcoins in electricity cost alone.

If you have a clue, I wonder what the ones without a clue do. Probably buying 5970s .

What makes you think of 5870 or 5970 would use more electricity per hash? Given 5850 is simply a crippled 5870 it isn't going to get superior performance on a watt per hash basis. 


Most miner's don't buy 5970s @ 700+ pounds.  Not even close.   Hell I didn't pay that much for 3.

Bitcoin 2 is the new revision, Also as better airflow as its more open frame. The rig that is. As i say, they can swap around i believe, i dont notice too much as i dont even look at them. They sit there and work away these days. Im on other projects and until i have spare cash for FPGA boards i dont think ill be doing much. Maybe a 4 card 78XX series but i think thatll be it for VGA mining if AMD take the CGN architecture route. And calling me mad for getting 2 5850's that cost 240£ doing 800m/hash compared to people that have spent 700£+ on single 5990/6990's for less than 800m/hash at extra cost of electric shows that theres plenty more out there that dont have a clue, and i aint one of them

So whats different between left and right machine? Left machine looks like what Id expect for 5850s at those clocks. Right screen looks like higher clocks to me. 1 Ghz would seem right. If not the clock, what else is different?

Man, Stop biting my balls. Our they chocolate salted enough for you? haha Here, Averages around 400 between both give or take (they been running 6 days, We had a power outage in the area) and have gone a month with no problem before. I can only argue what Im getiing

http://imageshack.us/photo/my-images/5/screenshottg.png/



Uploaded with ImageShack.us

And also, one rig as 4 different revisions, which i suspect is causing the lesser speeds on bitcoin 1. But im happy with that, Could lower voltage on a few but i cant be bothered with that. Not as fun as seen what the max is

AND ALSO.... i use a pool that bad/good luck doesnt effect. I get payed for..... wait for it...... MY AVERAGE HASHRATE PER HOUR (or 30 minutes or something). Damn, Dont you hate it when your proven wrong, I could get 410/420 out of them if i refined things down. A friend runs one of these cards at 1.247 all day long. 4 months now. Dont ask me prove that too, That would be a mission getting a picture of his desktop ha.

No offense, i respect both you 2 but get off my sweetbreads Search 'blue waffle' here, I do alright in all 3 charts

https://arsbitcoin.com/statsAll.php 50-70 in all 3 charts

Oh that what I get for not reading whole thread.  Yeah instant rate is utterly useless.  Even average rate over <1 hour is completely useless.  Good "luck" or bad "luck" can vary those by 30% or more. 

DnT, he is not using time averaged effective hashrates. He is just looking at instantaneous hashrates with all the variability that comes with it. You dont get "400MH+" sustained average out of a 5850 @965Mhz. But dont bother telling him, because he knows all that already .

Did you check your load?  If load isn't 100% then while the card may be running at 1000MHz it isn't doing 1000MHz of work.  Thermal and current throttling will kick in and cut the load.  To be apples to apples you need to ensure cards are at equivelent load.  If load remains 99% then you should get a 1:1 relationship between increased clocks & increased hashrate. However if you raise the clock 10% but load falls 10% then obviously you aren't getting any more work done. 

Hashrate = (clock speed)*(effective load)*(# of SP)/(ALU efficiency of miner code)

965/300 gets me 400 m/hash+ per card. when set to 1000/300 (only when i was seeing how far they do go) i got around 415-425m/hash. Iys around .40-45% the hash rate is to core speed, least on 5850's ;P)

Tnx for the info.