Radeon HD 58XX Series users get in here!

[frankiebits]

So i just booted up a used 5830...

800/1000 (Stock Speeds)
SDK 2.3
Catalyst 11.2
POCLBM -f 60 -v
65 C
200 mhash/s

Now I have a few ideas on how to pull some more out but before I start screwing around I thought maybe someone could share their settings and experience with this particular series of cards and have any advice for me....

lol I did everything to squeeze 80 mhash/s out of my 4850 and to boot this guy up and immediately see 200 mhash felt pretty good.

If anyone is wondering I paid $150 for the card , I am going to sell the 4850 on Ebay for somewhere between $40-$60 (still have original packaging and everything) , so basically I spent $100 , not bad right?

[1715052913]

1715052913

[1715052913]

1715052913

[1715052913]

1715052913

[Beremat]

5870 here. I have it clocked at 980 core, 300 mem for mining. Temps are around the same as non-overclocked if you set the mem clock that low. I get around 340-380mhash with this. Using flags -v -w128, AMD APP 2.2, catalysm 10.somethingorother.

[frankiebits]

5870 here. I have it clocked at 980 core, 300 mem for mining. Temps are around the same as non-overclocked if you set the mem clock that low. I get around 340-380mhash with this. Using flags -v -w128, AMD APP 2.2, catalysm 10.somethingorother.

I tried to lower the the mem to 400 and it froze up.

Ok I changed everything except app version and driver version to exactly the same as you and I am getting 255 mhash/s

You think changing driver and app version will push me up anymore?

[frankiebits]

I dropped down to 10.7 and SDK 2.2 and I got an increase of only 2-5 mhash/s , not really what I was hoping for, going to try SDK 2.1 and if no increase I will stick with my original config. If you can check what version of driver you have I would appreciate it , thanks.

Update: Ok now I uninstalled 2.2 and installed 2.1 and actually see a decent increase, I am up to 270 mhash/s , so sdk 2.1 seems the best, but driver wise there are so many I can be here forever trying them all, anyone have an idea? I am using 10.7 right now.

Thanks in advance, if anyone could help I really appreciate it. 

[snedie]

How do you change the SDK?

[dishwara]

I dropped down to 10.7 and SDK 2.2 and I got an increase of only 2-5 mhash/s , not really what I was hoping for, going to try SDK 2.1 and if no increase I will stick with my original config. If you can check what version of driver you have I would appreciate it , thanks.

Update: Ok now I uninstalled 2.2 and installed 2.1 and actually see a decent increase, I am up to 270 mhash/s , so sdk 2.1 seems the best, but driver wise there are so many I can be here forever trying them all, anyone have an idea? I am using 10.7 right now.

Thanks in advance, if anyone could help I really appreciate it. 

Try 10.12 or 10.11, Most says for 5000 series this gives max hash/s

How do you change the SDK?

You have to uninstall SDK through add/remove pgms if in windows until 2.2
ATI stream sdk 2.x
& also delete the folder in C:\ATI

If 2.3 then you can use driver installer itself, by running it & selecting uninstall & can remove what u don't want & can then again run 11.4 driver to install 2.4

[Beremat]

Looks like I was wrong with what I posted about actually, I'm running catalyst 11.2.

Yes, lowering your mem clock too low will freeze the card. I think it depends on how lucky you are. I can go as low as 250mhz, but any lower will start artifacting even the windows desktop and 225 will freeze it.

[BOARBEAR]

im using cat 11.4 beta with steam sdk 2.4

clock 870/1250
getting 325000 khash/s

you can find the driver here http://support.amd.com/us/kbarticles/Pages/Catalyst114earlypreview.aspx

[frankiebits]

How do you change the SDK?

I always install the SDK separately from the drivers... then in CCleane you can uninstall them.


You can also switch sdk's without uninstalling the drivers (doesn't even need a reboot) if you had installed them completely separately.

[frankiebits]

Looks like I was wrong with what I posted about actually, I'm running catalyst 11.2.

Yes, lowering your mem clock too low will freeze the card. I think it depends on how lucky you are. I can go as low as 250mhz, but any lower will start artifacting even the windows desktop and 225 will freeze it.

I got it sorted out, thanks, I'll keep it at 300 for now.

[frankiebits]

im using cat 11.4 beta with steam sdk 2.4

clock 870/1250
getting 325000 khash/s

you can find the driver here http://support.amd.com/us/kbarticles/Pages/Catalyst114earlypreview.aspx

Hmm just lost 10 mhash/s using these newer versions. Man this is frustrating.
So far the best results I have got at my current clocks (1000/300) is 275 mhash/s with SDK2.1 and 10.7 Drivers...


What version of card do you have?

[dishwara]

im using cat 11.4 beta with steam sdk 2.4

clock 870/1250
getting 325000 khash/s

you can find the driver here http://support.amd.com/us/kbarticles/Pages/Catalyst114earlypreview.aspx

Hmm just lost 10 mhash/s using these newer versions. Man this is frustrating.
So far the best results I have got at my current clocks (1000/300) is 275 mhash/s with SDK2.1 and 10.7 Drivers...


What version of card do you have?

For 5000 series card sdk below 2.3 only works good.

The best seems 10.12 or 10.11 with 2.1   as said before.
The 11.4 & 2.4 preview akka BETA works well with 6000 series.
http://bitcointalk.org/index.php?topic=4493.0

Lowering memory below 300 Mhz actually decreases hash/s & freezes windows.
Also, it seems there is some kind of formula there between memory frequency & core frequency, it seems in the form of core clk =3(3.??) x mem clk which gives the Maximum hash/s
I get Maximum 280 Mhash/s at 1038 core clk & 360-370 mem clk.
Seems it differ for everyone. Please try increasing mem clk every 5-10 mhz while mining in back & see the change in hash/s & Please also post it, if u can able to get any max hash/s at any particular frequency while keep the core clk at max with out freezing.

lowering mem clk reduces TEMP up to 10C to me.

you can increase the core clk up to <1040 with out freezing. even 1041 freezes after some time.
i got freezing at 1047 after 1-2 hours mining & not touching pc

[frankiebits]

im using cat 11.4 beta with steam sdk 2.4

clock 870/1250
getting 325000 khash/s

you can find the driver here http://support.amd.com/us/kbarticles/Pages/Catalyst114earlypreview.aspx

Hmm just lost 10 mhash/s using these newer versions. Man this is frustrating.
So far the best results I have got at my current clocks (1000/300) is 275 mhash/s with SDK2.1 and 10.7 Drivers...


What version of card do you have?

For 5000 series card sdk below 2.3 only works good.

The best seems 10.12 or 10.11 with 2.1   as said before.
The 11.4 & 2.4 preview akka BETA works well with 6000 series.
http://bitcointalk.org/index.php?topic=4493.0

Lowering memory below 300 Mhz actually decreases hash/s & freezes windows.
Also, it seems there is some kind of formula there between memory frequency & core frequency, it seems in the form of core clk =3(3.??) x mem clk which gives the Maximum hash/s
I get Maximum 280 Mhash/s at 1038 core clk & 360-370 mem clk.
Seems it differ for everyone. Please try increasing mem clk every 5-10 mhz while mining in back & see the change in hash/s & Please also post it, if u can able to get any max hash/s at any particular frequency while keep the core clk at max with out freezing.

lowering mem clk reduces TEMP up to 10C to me.

you can increase the core clk up to <1040 with out freezing. even 1041 freezes after some time.
i got freezing at 1047 after 1-2 hours mining & not touching pc

Thanks I will play around and report back.

[frankiebits]

Here is where I am at right now

Card: Sapphire 5830

Driver : 10.11
Stream SDK: 2.1
Clocks: 1025/300
HashSpeed: 285 mhash/s

These setting beat out (driver/stream) 11.2/2.3 and and 11.4/2.4

For now I am going to stick with these settings , they seem stable , temps are good.
If anyone wants to try and see if they can something better with drivers 10.8-10.9 it might be worth a shot.

Hope they help, if they do throw me a coin or two 

[frankiebits]

Hmm I can't get it stable. Well I can get it stable when only mining , but it I am mining for a while and come to use the computer it freezes up.

This never happened with my previous card, but then again I can't control the voltage on this card, something I didn't know until I received it....

Anyone know if it is possible to control the voltages on the Sapphire 5830, I only have 2 options...

Well otherwise I found the 10.7-10.10 drivers with stream sdk 2.1 to be the fastest.

I tried to see if the freezing up was due to the software versions but it happened on all of them.....


What to do , what to do....

Edit: Sticking with 10.7 and SDK 2.1 , got it stable (where i can still use PC @ decent speed without freezing) at 990/300 at 275 mhash/s...

Wish I would have known I couldn't change the voltage, the card to go with if your getting a 5830 is the newer Asus version...

[BOARBEAR]

try to use a tool called OCCT to test your clock speed
for your card choose shader complexity 8 and check for error

[frankiebits]

try to use a tool called OCCT to test your clock speed
for your card choose shader complexity 8 and check for error

Thanks, it kills me I can't control the voltage though. I could push so much more out of this card if I could. I am going to stick with it for a month or two , hopefully be able to sell it for almost what I paid for it and get either the Asus 5830 or something else, but from what I see the best bang for the buck mining wise is still a 5830 , i see the Asus ones going for around $120 on ebay and I bet you can squeeze alteast 320 mhash/s out of it.


On here  http://en.wikipedia.org/wiki/Ati_comparison you can see why the 58xx series kicks ass for mining, 2154 million transistors compared to the 68xx series 1700. I assume for gaming they are better but certainly not for mining.

This also caught my eye and made me get a 5830 even though I am not sure what it means    but only a few cards have it, does anyone know what "Double-precision FP" is  and if it effects mining speeds?

[urizane]

DP FP: Double precision floating point
This won't affect mining at all.  Hash calculations are done with integer operations and some compare operations.  That said, a 6870 can compete with a 5870 in mining, but only if you overclock the crap out of the 6870.  Even moderate overclocks on a 5870 can pull it ahead.  As far as mining performance per unit price, the 5830 does look like the best option.  Mining performance per watt is still going to be the 5970 with the 5850 and 6870 really close behind.
Also, AMD's primary impetus behind creating the 68xx cards was to reduce wattage but keep performance nearly the high as the 58xx cards.  Basically, consider Barts the Cypress redesign.  A 5850 and a 6870 have roughly the same TDP rating, yet the 6870 performs better in games than the 5850.

[Elanzer]

The best bang for the buck mining is the HD 5870.

A few places around here sell it for about $185-$200 CAD new, and it will have 43% better performance per clock than a HD 5830 because it has 43% more shader units (1600 vs 1120 on the 5830).

At only 900mhz it puts out about 330,000 khash/s, I've also seen 390,000 khash/s at 1030mhz (although with my reference cooler, I can only keep it at about 930mhz unless I want a hairdryer in my room). This is all using SDK 2.3 on Windows with the 11.2 driver. I tried SDK 2.1 as well as older drivers and found no significant difference at all (less than ~5m increase/decrease). I also didn't notice anything significant upgrading to SDK 2.4.

I wouldn't agree that the 6870 is a better miner than the 5850 either. The 58xx and 68xx/69xx series all top out at roughly the same core frequencies, most will cap out anywhere from 950-1000mhz for mining. HD 6870 is already 900mhz and doesn't have much headroom, HD 5850 is 775mhz and has much more headroom for overclocking. That all being said, HD 5850 already performs better for mining when both are at stock frequency.

My HD 5850 at ~920mhz sees roughly 300,000 khash/s. Comparing my HD 6850 at the same frequency of 920mhz.. it only sees about 215,000 khash/s.

[urizane]

The best bang for the buck mining is the HD 5870.

A few places around here sell it for about $185-$200 CAD new, and it will have 43% better performance per clock than a HD 5830 because it has 43% more shader units (1600 vs 1120 on the 5830).

Well, at stock clocks the 5870 has a 51.7% theoretical advantage over the 5830, but at the price frankiebits saw it for ($120) versus $185, the 5870 has a 54.2% higher price.  OK, so the 5830 wins in price to mining performance by only a tiny amount at those prices.

At only 900mhz it puts out about 330,000 khash/s, I've also seen 390,000 khash/s at 1030mhz (although with my reference cooler, I can only keep it at about 930mhz unless I want a hairdryer in my room). This is all using SDK 2.3 on Windows with the 11.2 driver. I tried SDK 2.1 as well as older drivers and found no significant difference at all (less than ~5m increase/decrease). I also didn't notice anything significant upgrading to SDK 2.4.

I wouldn't agree that the 6870 is a better miner than the 5850 either. The 58xx and 68xx/69xx series all top out at roughly the same core frequencies, most will cap out anywhere from 950-1000mhz for mining. HD 6870 is already 900mhz and doesn't have much headroom, HD 5850 is 775mhz and has much more headroom for overclocking. That all being said, HD 5850 already performs better for mining when both are at stock frequency.

My HD 5850 at ~920mhz sees roughly 300,000 khash/s. Comparing my HD 6850 at the same frequency of 920mhz.. it only sees about 215,000 khash/s.

The relevant comparison should not be a 5850 at 920MHz.  Also, in my post, a 5850 was not compared to a 6850.  Those two cards aren't the same wattage.  While the 5850 will most definately overclock better than a 6870, both cards at stock clocks will produce mining results that are similar.  However, the 6870 will perform better in games.  That's how I have it worded in my previous post.  Also, in my post I point out that even a moderate overclock on a 5870 will pull it ahead of the 6870.  From the way I've ordered the sentences, it can be inferred that the 6870 at any level of overclock can't compete with the 5870 at even a moderate overclock.  I'm just restating that to clarify.

I will agree, though, that a 5870 will give you the best performance for the given clock speed.

EDIT: Looking back, I did not emphasize the fact that the 5850 and 6870 have similar mining performance.  Oops.