Bitcoin Forum
December 07, 2016, 10:37:42 AM *
News: To be able to use the next phase of the beta forum software, please ensure that your email address is correct/functional.
 
   Home   Help Search Donate Login Register  
Pages: [1] 2 »  All
  Print  
Author Topic: Unstable hash rate with Radeon 5970  (Read 4526 times)
JayPhizzt
Newbie
*
Offline Offline

Activity: 22


View Profile
July 11, 2012, 02:32:39 PM
 #1

I have a Radeon 5970 in my server that I'm using for mining, but I'm having some issues with it. At first I had problems getting both cores to work, but I seem to have fixed that now. The problem now though is that the hash rate is fluctuating a lot between about 170-350Mhash/s on each core. Shouldn't it be more stable than that? When I try mining with only one core its a lot more stable and stays at around 340-350Mhash/s.
The GPU's are at stock-speed(725MHz) and I haven't done any clocking other than underclocking the memory to 500MHz(wont go lower than that in either CCC or MSI Afterburner).
Should also mention that I'm using WinXP 32bit, Catalyst 11.10 and Stream SDK 2.1. And I'm using GUIminer with phoenix to mine.

Also, I'm currently using these flags:

-k phatk VECTORS BFI_INT FASTLOOP=FALSE AGGRESSION=7 worksize=128

For some strange reason I get the highest hash rate when I set AGGRESSION to 7, any other number and hash rate drops.

Anyone have any idea of what could be the problem?
1481107062
Hero Member
*
Offline Offline

Posts: 1481107062

View Profile Personal Message (Offline)

Ignore
1481107062
Reply with quote  #2

1481107062
Report to moderator
"There should not be any signed int. If you've found a signed int somewhere, please tell me (within the next 25 years please) and I'll change it to unsigned int." -- Satoshi
Advertised sites are not endorsed by the Bitcoin Forum. They may be unsafe, untrustworthy, or illegal in your jurisdiction. Advertise here.
Ram
Newbie
*
Offline Offline

Activity: 13


View Profile
July 11, 2012, 06:53:11 PM
 #2

Check you powersupply.

But what you are explaining sound more like a temperature problem. Use GPUZ to check what your VRM's temps are... Make sure they dont go over 90c and you should see "throttling" stop.
JayPhizzt
Newbie
*
Offline Offline

Activity: 22


View Profile
July 11, 2012, 08:13:11 PM
 #3

Check you powersupply.

But what you are explaining sound more like a temperature problem. Use GPUZ to check what your VRM's temps are... Make sure they dont go over 90c and you should see "throttling" stop.

I don't think my power supply is the problem. Its a 650W 80+ gold FSP power supply which should be more than sufficient.
And I have checked all temps with GPU-Z and nothing is above 90C.
zvs
Legendary
*
Offline Offline

Activity: 1386



View Profile WWW
July 12, 2012, 12:20:13 AM
 #4

I have a Radeon 5970 in my server that I'm using for mining, but I'm having some issues with it. At first I had problems getting both cores to work, but I seem to have fixed that now. The problem now though is that the hash rate is fluctuating a lot between about 170-350Mhash/s on each core. Shouldn't it be more stable than that? When I try mining with only one core its a lot more stable and stays at around 340-350Mhash/s.
The GPU's are at stock-speed(725MHz) and I haven't done any clocking other than underclocking the memory to 500MHz(wont go lower than that in either CCC or MSI Afterburner).
Should also mention that I'm using WinXP 32bit, Catalyst 11.10 and Stream SDK 2.1. And I'm using GUIminer with phoenix to mine.

Also, I'm currently using these flags:

-k phatk VECTORS BFI_INT FASTLOOP=FALSE AGGRESSION=7 worksize=128

For some strange reason I get the highest hash rate when I set AGGRESSION to 7, any other number and hash rate drops.

Anyone have any idea of what could be the problem?
i get the highest hash rate at 6 and 7 aggression too, but i don't believe it, so I use 9 instead

it shouldn't fluctuate that much though.  @ 6 or 7 i might see jumps from 360 to 390 or something like that.  did you try checking temps?  it could be throttling itself

Dacentec, best deals for US dedicated servers. They regularly restock $20-$25 Opterons with 8-16GB RAM & 2x1-2TB HDD's (ofc, usually lots of other good stuff to choose from).  I did a Serverbear benchmark of one of my $20/mo Opteron (June last year), it's here.  Have had about a half dozen different servers with Dacentec, & none have failed to sustain at least 40MB/s (burst higher). My favorite is a 12-month rent-to-own ZT Systems 2XL5520 16GB 2x2TB SATA for $40/month (got lucky with the 'off-brand', haven't seen a RTO 2xL5520 for under $50/mo since -- at least for monthly contracts).  wholesaleinternet.com has some ancient 2-core intel CPUs @ $10/mo sometimes (I got an Intel Core 2 6300 @ 1.86GHz, with a 250GB HDD with 46000 hours on it, LOL. $20 @ Dacentec is much better, if you can grab one). joesdatacenter.com (same location as Wholesale Internet) also occasionally has specials (or if you don't want to wait, it has an AMD Opteron 170 @ $16/mo).
ssateneth
Legendary
*
Offline Offline

Activity: 1288



View Profile
July 12, 2012, 09:09:45 AM
 #5

You shouldnt get more than 340 mhash on a 5970 core @ 725 MHz, even with perfect ram speeds and miner/kernel options. If you're spiking 350-370, definately something awry going on there. Is your GPU activity maintaining 99+%? Doing nothing on the computer running on the miner? I'm running 5830, 5870, 5870, 5970, and 5970 and I get sudden dips of GPU activity (40-70%) if I do ANYTHING on it, and it kind of cascades between the 7 GPU's for about 30 seconds before stabilizing and getting back to 99.5+%

JayPhizzt
Newbie
*
Offline Offline

Activity: 22


View Profile
July 12, 2012, 11:21:56 AM
 #6

You shouldnt get more than 340 mhash on a 5970 core @ 725 MHz, even with perfect ram speeds and miner/kernel options. If you're spiking 350-370, definately something awry going on there. Is your GPU activity maintaining 99+%? Doing nothing on the computer running on the miner? I'm running 5830, 5870, 5870, 5970, and 5970 and I get sudden dips of GPU activity (40-70%) if I do ANYTHING on it, and it kind of cascades between the 7 GPU's for about 30 seconds before stabilizing and getting back to 99.5+%

My GPU's maintain about 60-75% activity. And yes, its a dedicated mining-server at the momemt, so I'm not doing anything else on it.
Tried only running one core again and then the activity goes up to around 96-98%. Strange..
crazyates
Legendary
*
Offline Offline

Activity: 938



View Profile
July 13, 2012, 01:56:51 AM
 #7

You shouldnt get more than 340 mhash on a 5970 core @ 725 MHz, even with perfect ram speeds and miner/kernel options. If you're spiking 350-370, definately something awry going on there. Is your GPU activity maintaining 99+%? Doing nothing on the computer running on the miner? I'm running 5830, 5870, 5870, 5970, and 5970 and I get sudden dips of GPU activity (40-70%) if I do ANYTHING on it, and it kind of cascades between the 7 GPU's for about 30 seconds before stabilizing and getting back to 99.5+%

My GPU's maintain about 60-75% activity. And yes, its a dedicated mining-server at the momemt, so I'm not doing anything else on it.
Tried only running one core again and then the activity goes up to around 96-98%. Strange..

What driver/SDK? Back with the old 100% bug if I ran both my cards on the same CPU, their hash rates would suffer.

Tips? 1crazy8pMqgwJ7tX7ZPZmyPwFbc6xZKM9
Previous Trade History - Sale Thread
JayPhizzt
Newbie
*
Offline Offline

Activity: 22


View Profile
July 13, 2012, 08:00:34 AM
 #8

You shouldnt get more than 340 mhash on a 5970 core @ 725 MHz, even with perfect ram speeds and miner/kernel options. If you're spiking 350-370, definately something awry going on there. Is your GPU activity maintaining 99+%? Doing nothing on the computer running on the miner? I'm running 5830, 5870, 5870, 5970, and 5970 and I get sudden dips of GPU activity (40-70%) if I do ANYTHING on it, and it kind of cascades between the 7 GPU's for about 30 seconds before stabilizing and getting back to 99.5+%

My GPU's maintain about 60-75% activity. And yes, its a dedicated mining-server at the momemt, so I'm not doing anything else on it.
Tried only running one core again and then the activity goes up to around 96-98%. Strange..

What driver/SDK? Back with the old 100% bug if I ran both my cards on the same CPU, their hash rates would suffer.

As I wrote in the main post, 11.10 and SDK 2.1. When it comes to the CPU it maintains about a 50% load.
ssateneth
Legendary
*
Offline Offline

Activity: 1288



View Profile
July 13, 2012, 08:57:01 AM
 #9

You shouldnt get more than 340 mhash on a 5970 core @ 725 MHz, even with perfect ram speeds and miner/kernel options. If you're spiking 350-370, definately something awry going on there. Is your GPU activity maintaining 99+%? Doing nothing on the computer running on the miner? I'm running 5830, 5870, 5870, 5970, and 5970 and I get sudden dips of GPU activity (40-70%) if I do ANYTHING on it, and it kind of cascades between the 7 GPU's for about 30 seconds before stabilizing and getting back to 99.5+%

My GPU's maintain about 60-75% activity. And yes, its a dedicated mining-server at the momemt, so I'm not doing anything else on it.
Tried only running one core again and then the activity goes up to around 96-98%. Strange..

What driver/SDK? Back with the old 100% bug if I ran both my cards on the same CPU, their hash rates would suffer.

As I wrote in the main post, 11.10 and SDK 2.1. When it comes to the CPU it maintains about a 50% load.

You shouldnt even get that much load either. I get 1% CPU on a 7 GPU rig, single core 2GHz. I'm going to bet you have a dual core CPU. You still have CPU bug. 11.12 and 12.1 don't exhibit the CPU bug with 2.1 SDK. Unfortunately, I tried 2.5 SDK a little while back on 11.12 and I had CPU bug. Will take some futzing around to get a 2.5 SDK combo working..

JayPhizzt
Newbie
*
Offline Offline

Activity: 22


View Profile
July 13, 2012, 10:42:24 AM
 #10

You shouldnt get more than 340 mhash on a 5970 core @ 725 MHz, even with perfect ram speeds and miner/kernel options. If you're spiking 350-370, definately something awry going on there. Is your GPU activity maintaining 99+%? Doing nothing on the computer running on the miner? I'm running 5830, 5870, 5870, 5970, and 5970 and I get sudden dips of GPU activity (40-70%) if I do ANYTHING on it, and it kind of cascades between the 7 GPU's for about 30 seconds before stabilizing and getting back to 99.5+%

My GPU's maintain about 60-75% activity. And yes, its a dedicated mining-server at the momemt, so I'm not doing anything else on it.
Tried only running one core again and then the activity goes up to around 96-98%. Strange..

What driver/SDK? Back with the old 100% bug if I ran both my cards on the same CPU, their hash rates would suffer.

As I wrote in the main post, 11.10 and SDK 2.1. When it comes to the CPU it maintains about a 50% load.

You shouldnt even get that much load either. I get 1% CPU on a 7 GPU rig, single core 2GHz. I'm going to bet you have a dual core CPU. You still have CPU bug. 11.12 and 12.1 don't exhibit the CPU bug with 2.1 SDK. Unfortunately, I tried 2.5 SDK a little while back on 11.12 and I had CPU bug. Will take some futzing around to get a 2.5 SDK combo working..

Then I'm gonna give 11.12 and/or 12.1 with SDK 2.1 a try. Thanks for the tip. And yes, I do have in fact have a dual core CPU(C2D).
crazyates
Legendary
*
Offline Offline

Activity: 938



View Profile
July 13, 2012, 01:52:32 PM
 #11

You shouldnt get more than 340 mhash on a 5970 core @ 725 MHz, even with perfect ram speeds and miner/kernel options. If you're spiking 350-370, definately something awry going on there. Is your GPU activity maintaining 99+%? Doing nothing on the computer running on the miner? I'm running 5830, 5870, 5870, 5970, and 5970 and I get sudden dips of GPU activity (40-70%) if I do ANYTHING on it, and it kind of cascades between the 7 GPU's for about 30 seconds before stabilizing and getting back to 99.5+%

My GPU's maintain about 60-75% activity. And yes, its a dedicated mining-server at the momemt, so I'm not doing anything else on it.
Tried only running one core again and then the activity goes up to around 96-98%. Strange..

What driver/SDK? Back with the old 100% bug if I ran both my cards on the same CPU, their hash rates would suffer.

As I wrote in the main post, 11.10 and SDK 2.1. When it comes to the CPU it maintains about a 50% load.

Ah I see that now. Must have glossed right over it. Tongue

Like ssateneth said, 50% of a dual core CPU is 100% on one core, and if you run both instances of Phoenix on the same core, the hash rates suffer. Upgrading to 11.12 will fix your problem.

Tips? 1crazy8pMqgwJ7tX7ZPZmyPwFbc6xZKM9
Previous Trade History - Sale Thread
JayPhizzt
Newbie
*
Offline Offline

Activity: 22


View Profile
July 13, 2012, 02:44:54 PM
 #12

I've tried both 12.1 and 11.12 with SDK 2.1 now and it didn't work... Still 50% load on the CPU and still fluctuating hash rates.. I'm really getting tired of this **** and I'm about to give up... And before you ask, yes I used Driver Sweeper to uninstall the drivers.
Maybe I should try a different miner? Highly doubt that'll work either though..
crazyates
Legendary
*
Offline Offline

Activity: 938



View Profile
July 13, 2012, 03:36:44 PM
 #13

I've tried both 12.1 and 11.12 with SDK 2.1 now and it didn't work... Still 50% load on the CPU and still fluctuating hash rates.. I'm really getting tired of this **** and I'm about to give up... And before you ask, yes I used Driver Sweeper to uninstall the drivers.
Maybe I should try a different miner? Highly doubt that'll work either though..

Try 12.2 with SDK 2.5. I always used 2.5 on my 5970s with no issues.

I'm all for switching to CGMiner, but this sounds more like a kernel issue.

Tips? 1crazy8pMqgwJ7tX7ZPZmyPwFbc6xZKM9
Previous Trade History - Sale Thread
JayPhizzt
Newbie
*
Offline Offline

Activity: 22


View Profile
July 13, 2012, 05:35:47 PM
 #14

I've tried both 12.1 and 11.12 with SDK 2.1 now and it didn't work... Still 50% load on the CPU and still fluctuating hash rates.. I'm really getting tired of this **** and I'm about to give up... And before you ask, yes I used Driver Sweeper to uninstall the drivers.
Maybe I should try a different miner? Highly doubt that'll work either though..

Try 12.2 with SDK 2.5. I always used 2.5 on my 5970s with no issues.

I'm all for switching to CGMiner, but this sounds more like a kernel issue.

As far as I can tell 2.5 isn't available for WinXP. 2.3 seems to be the newest version for XP.
crazyates
Legendary
*
Offline Offline

Activity: 938



View Profile
July 13, 2012, 06:10:36 PM
 #15

I've tried both 12.1 and 11.12 with SDK 2.1 now and it didn't work... Still 50% load on the CPU and still fluctuating hash rates.. I'm really getting tired of this **** and I'm about to give up... And before you ask, yes I used Driver Sweeper to uninstall the drivers.
Maybe I should try a different miner? Highly doubt that'll work either though..

Try 12.2 with SDK 2.5. I always used 2.5 on my 5970s with no issues.

I'm all for switching to CGMiner, but this sounds more like a kernel issue.

As far as I can tell 2.5 isn't available for WinXP. 2.3 seems to be the newest version for XP.

Oh man, I am full of fail advice today.

Tips? 1crazy8pMqgwJ7tX7ZPZmyPwFbc6xZKM9
Previous Trade History - Sale Thread
Aseras
Hero Member
*****
Offline Offline

Activity: 658


View Profile
July 15, 2012, 03:37:26 PM
 #16

Just use 12.1. It's the last ocl driver for xp. Sdk 2.1 or use the one with 12.1. CPU bug should be gone.
zvs
Legendary
*
Offline Offline

Activity: 1386



View Profile WWW
July 15, 2012, 03:47:45 PM
 #17

yeah, 12.1 should be fine.

i have 3 5870's in a system w/ phoenix & aggression rating 9, it's about 1-2% per instance (this based on my dual core e6850 that's been underclocked to like 1600mhz and 1 core disabled)

Dacentec, best deals for US dedicated servers. They regularly restock $20-$25 Opterons with 8-16GB RAM & 2x1-2TB HDD's (ofc, usually lots of other good stuff to choose from).  I did a Serverbear benchmark of one of my $20/mo Opteron (June last year), it's here.  Have had about a half dozen different servers with Dacentec, & none have failed to sustain at least 40MB/s (burst higher). My favorite is a 12-month rent-to-own ZT Systems 2XL5520 16GB 2x2TB SATA for $40/month (got lucky with the 'off-brand', haven't seen a RTO 2xL5520 for under $50/mo since -- at least for monthly contracts).  wholesaleinternet.com has some ancient 2-core intel CPUs @ $10/mo sometimes (I got an Intel Core 2 6300 @ 1.86GHz, with a 250GB HDD with 46000 hours on it, LOL. $20 @ Dacentec is much better, if you can grab one). joesdatacenter.com (same location as Wholesale Internet) also occasionally has specials (or if you don't want to wait, it has an AMD Opteron 170 @ $16/mo).
JayPhizzt
Newbie
*
Offline Offline

Activity: 22


View Profile
July 15, 2012, 04:47:23 PM
 #18

Just use 12.1. It's the last ocl driver for xp. Sdk 2.1 or use the one with 12.1. CPU bug should be gone.

I have tried 12.1 and it doesn't help.. It doesn't seem to matter what drivers and SDK I use, both the CPU-bug and fluctuating hash rate remain.. And I've tried many different drivers and SDK's now..

Does anyone have any other idea of what could be causing my problems?
Aseras
Hero Member
*****
Offline Offline

Activity: 658


View Profile
July 15, 2012, 11:53:47 PM
 #19

Uninstall whatever sdk you have. Do a search for opencl.dll and delete them all. Then Reinstall.
MrTeal
Legendary
*
Offline Offline

Activity: 1246


View Profile
July 16, 2012, 12:27:06 AM
 #20

Drop your aggression down. I have setups where 9 will jump to 100% on a core, but 8 is fine. I even have one where 9 is fine on all the GPUs, except the one driving the display which needs to be 8 or lower otherwise a core gets pegged.
Pages: [1] 2 »  All
  Print  
 
Jump to:  

Sponsored by , a Bitcoin-accepting VPN.
Powered by MySQL Powered by PHP Powered by SMF 1.1.19 | SMF © 2006-2009, Simple Machines Valid XHTML 1.0! Valid CSS!