Topic: 400MH/s+ on a 5850? (Read 5043 times)

I'd worry more about the VRM temps for 5xxx cards as long as you are using some form of stock cooling.

Pics or it didn't happen?

I've been able to hit 394 Mhash with 930/300 @ 1.137v.  GPU temps in at 72 (I rigged an old Twin Turbo on it with dual 120mm fans).  Phoenix phatk with the usual settings and worksize 256 / aggression 13.  Also that 3% performance improvement edit (also disabled most of Windows 7 performance visuals). Only thing that prevents me going higher is VRM temps at 97-105c.  I know they are rated for higher than that but I don't feel comfortable with testing my luck.

seeing as i'm cranking 320 out of my 5830(1025mhz shader) that sounds witin reason depending on your cooling you may be able to get a bit more out of it even

No steps need to be taken.  The card will clock itself down if it goes over 90c, all the way down to 300mhz core if it has to, which is enough to keep it cool without the fan running even under full load (ask me how I know).

Try SDK 2.1, its a bit faster than 2.4.

Based on what I've read in different forums of the BOINC community, the 5800 series have max temps of about 105 degrees celsius. Of course, steps must be taken to cool the card a lot better if the temps goes over 90 degrees.

I get a very stable 390 MHash at 950/300 MHz under Linux with the newest poclbm. 79C at 64% fan speed, default (1.088v) voltage.

i get 400 out of my msi 5850 twin frozr II with 1005 core, 340 mem @ 1.3 mV and temps stay below 80°C.
BUT, my phoenix dies after around 3 hours on gui miner, and my backup opencl miner kicks in (has -f20 as backup) and the opencl backup can only get around 385 =/
any tips how to make phoenix more stable? the opencl one can run days without problems.

I'm getting 395 mhash/sec at 990MHz core 330MHz mem using phoenix and phatk on linux on a few of my 5850s

Thats part of my reason for being involved. I really want to see bitcoin as a currency (or commodity, depending on who you talk to) flourish. If all it takes to put people off being invovled is that its no longer profitable to mine, because some people with deep pockets (and perhaps with an interest in seeing the ideal disappear) flood the system - then whats the point? If i were a large economy and felt threatened by bitcoin I would do my best to make sure people found no reason to mine through lack of profit. But if miners are here without the drive of profit, then the 'flood the market' move is, to say the least, stale. The ideal stays alive and presses on.

I know a lot of people out there who made this ideal get off the ground invested a lot of money through mining to get it going. I cant thank them enough. I'm sure they will make more money in the years to come, but I hope the ideal of what bitcoin is about stays with this forever - long after some people give up mining because they no longer make a few bucks a month.

I get 400 MH/s at that setting. it's a bit lower because i ran it at a lower aggression.

Can't compete with that

Then again, it does make the network more secure. If everyone just quits mining the network becomes susceptible to forging attacks

Well, i was wondering if its just a quirk of GUIminer reporting. Personally I prefer to run linux too, btu for some reason I could not get the ATI drivers to play nicely under 2.6.36 so i made a windows install to mine. If i run one miner thread with my clock settings I get 386MH/sec. If i run multiple threads, it goes higher, but again im not sure if this is an accurate report, or a bug in guiminer.

However, as you are using linux with a pretty intense overclock this proves an interesting oppertunity. Can you try running multiple miner threads with different CPU affinity? (install the 'util-linux' package if its not already).

Why not just run three different miner instances and use the 'taskset' command to specify CPU affinity, then see if you get the same results as me. If you dont get similar results than what I see, then what I report is very likely a quirk of guiminer in reporting the MH rate for multiple miners (perhaps refresh rate from the actual miners it runs?).

If you do get the same result as what I see, then draw your own conclusions

Start three phoenix miners. Set instance 1 to agression 13, instance 2 and 3 to agression 8. Leave other switches the same.

Then set CPU affinity using taskset. get the PID of each of the three miner processes and set accordingly. for example, if you have 2 CPU cores and your miner PIDs are 1234 (agression 13), 1258 and 1299 (agression 8 ) respectively:

sudo taskset -c 1-2 1234
sudo taskset -c 1 1258
sudo taskset -c 2 1299

Then look at all three miner processes and compute the total hashrate. If you see a hashrate greater than your single miner instance of 394, then we are on to something. If you see a total of 394, then what I am seeing is just a quirk in the way guiminer reports hashrates

Its as simple as that. My personal feeling is this is a guiminer issue in the way it refreshes the hashrates form the console (usleep 1000000 anyone?), but id love to be proven wrong.



Different strokes for different folks. Personally im just an old skool geek who loves fucking around with stuff. I love squeezing every last drop out of things that I do. Look at the bitcoin total hashrate since may - do you really think all of that exponential increase is down to people looking for money? Sure some of it is the big boys with their farms, the pools and then people trying to make a quick buck. But, theres a lot of us out there that dont worry about money and just like new technologies, understding them inside out and dabbling in them.

I guess above all, put simply, Bitcoin has a certain ideal to it that attracts me.

Im based in the UK, so with current difficulty and what my system draws in terms of power, im not going to make a profit on BTC, especially if difficulty keeps going up. But does that put me off mining? Nope. This shit is fun, is a great ideal in terms of what its offering and on top of that its a challenge to get some cheap piece of hardware to do more than its designed to. Hell, I probably make more in a day with my work than I could even with a 50GH/sec rig. You do the math?

Theres millions just like me out there, and the more that hear about this the more you can expect to see being involved. Ill still be doing this if the difficulty hits 500 miilion. Im not in it for money

:\

Do people even mine to get btc/profit in the long term
or just drive up the difficulty for fun & to see a number go higher on their screen

I wonder how you get so much higher hashrate on lower clocks than me:


With these flags under linux and 2.4 SDK: -k phatk DEVICE=0 VECTORS AGGRESSION=11 WORKSIZE=256 FASTLOOP=false BFI_INT

LOL it was only £100. pocket change.

Its lasted 2 weeks already so im happy. I was more interested in seeing what it can do and what all this mining business is about.

That being said, 960/300 @ 81oC isnt that bad. ive seen 5850s running much hotter and happy with it - one of my firends runs his at 90 to keep the noise down and that thing hasnt broke yet (6 months in). 58xx seem to be cards which can perform nicely.

But back to the point of my question - is this hashrate reporting a bug/quirk of guiminer?

I think that card isn't going to last very long. Good luck.

So GUIminer is saying I have broke through 400MH/sec on my 5850.

What do you guys think? Is this just a bug/quirk in the way guiminer updates the hashrate when multiple miners are running, or have I found a way to squeeze out an average of 400 MH/sec from a 5850 without a massively insane overclock? It runs stable if i dont try to move anything on screen

My GPU settings are
-5870 Bios
-960 clock
-300 RAM
-1.162 Vc

I set up three phoenix instances. The first instance I set to

-k phatk DEVICE=0 VECTORS BFI_INT WORKSIZE=256 AGGRESSION=13  FASTLOOP=false
that worker was set with CPU affinity to use both CPU cores

Second instance:
-k phatk DEVICE=0 VECTORS BFI_INT WORKSIZE=256 AGGRESSION=8  FASTLOOP=false
that worker was set with CPU affinity to use CPU core #1

Third instance:
-k phatk DEVICE=0 VECTORS BFI_INT WORKSIZE=256 AGGRESSION=8  FASTLOOP=false
that worker was set with CPU affinity to use CPU core #2

When running, the MH/sec goes as low as 386 and as high as 412, but seems to average out at about 402 to 404

A screen grab of guiminer in operation, along with GPU-Z and Afterburner settings is here

http://www.b3tards.com/u/f6d22a29e775a1426924/wtf.jpg

This might seem like a cock length post, but i promise its not. I just want to know if this is a true hashrate or if GUIMiner has some kind of weirdness in calculating hashrates when multiple miners are running and hoped someone here would know.