GPU OpenCL miners using 25% cpu

[shivansps]

Hi, i have two miners right now, one running on a 6870, and another one running on a 6850... each miner is using 25% of cpu, witch gives me a steady 50% cpu usage.

using guiminer with podlbm... both of podlbm process are using 25% cpu. This is normal?

[1714817106]

1714817106

[1714817106]

1714817106

[1714817106]

1714817106

[1714817106]

1714817106

[geebus]

Hi, i have two miners right now, one running on a 6870, and another one running on a 6850... each miner is using 25% of cpu, witch gives me a steady 50% cpu usage.

using guiminer with podlbm... both of podlbm process are using 25% cpu. This is normal?

You have a 4-core CPU, and each one is taking a full core.

The reason for that is that the loops running in poclbm that check for new results or long polling triggers have absolutely no delay between iterations, so it runs as fast as possible, taking 100% of the core. The only reason it doesn't take more is because it isn't written to use multiple cores.

Were it to sleep 1 second per iteration, it would drop to < 1% total CPU usage on a higher end quad core.

It's been a problem with poclbm for quite some time.

[qed]

Hi, i have two miners right now, one running on a 6870, and another one running on a 6850... each miner is using 25% of cpu, witch gives me a steady 50% cpu usage.

using guiminer with podlbm... both of podlbm process are using 25% cpu. This is normal?

You have a 4-core CPU, and each one is taking a full core.

The reason for that is that the loops running in poclbm that check for new results or long polling triggers have absolutely no delay between iterations, so it runs as fast as possible, taking 100% of the core. The only reason it doesn't take more is because it isn't written to use multiple cores.

Were it to sleep 1 second per iteration, it would drop to < 1% total CPU usage on a higher end quad core.

It's been a problem with poclbm for quite some time.

I don't think that's the reason. The problem has alwats been there for HD 6XXX card even before the introduction of long polling and it is present only if you have more than 1 GPU plugged. Under linux the CPU utilization is less than 1%. HD 5XXX cards are not presenting this problem.

[Beremat]

Hi, i have two miners right now, one running on a 6870, and another one running on a 6850... each miner is using 25% of cpu, witch gives me a steady 50% cpu usage.

using guiminer with podlbm... both of podlbm process are using 25% cpu. This is normal?

You have a 4-core CPU, and each one is taking a full core.

The reason for that is that the loops running in poclbm that check for new results or long polling triggers have absolutely no delay between iterations, so it runs as fast as possible, taking 100% of the core. The only reason it doesn't take more is because it isn't written to use multiple cores.

Were it to sleep 1 second per iteration, it would drop to < 1% total CPU usage on a higher end quad core.

It's been a problem with poclbm for quite some time.

I don't think that's the reason. The problem has alwats been there for HD 6XXX card even before the introduction of long polling and it is present only if you have more than 1 GPU plugged. Under linux the CPU utilization is less than 1%. HD 5XXX cards are not presenting this problem.

Uh, yes they are. My 5870 is pulling 100% of one core at all times. I set the affinity of all poclbm clients running to the same CPU core so that they all share a core and save some power.

[qed]

Hi, i have two miners right now, one running on a 6870, and another one running on a 6850... each miner is using 25% of cpu, witch gives me a steady 50% cpu usage.

using guiminer with podlbm... both of podlbm process are using 25% cpu. This is normal?

You have a 4-core CPU, and each one is taking a full core.

The reason for that is that the loops running in poclbm that check for new results or long polling triggers have absolutely no delay between iterations, so it runs as fast as possible, taking 100% of the core. The only reason it doesn't take more is because it isn't written to use multiple cores.

Were it to sleep 1 second per iteration, it would drop to < 1% total CPU usage on a higher end quad core.

It's been a problem with poclbm for quite some time.

I don't think that's the reason. The problem has alwats been there for HD 6XXX card even before the introduction of long polling and it is present only if you have more than 1 GPU plugged. Under linux the CPU utilization is less than 1%. HD 5XXX cards are not presenting this problem.

Uh, yes they are. My 5870 is pulling 100% of one core at all times. I set the affinity of all poclbm clients running to the same CPU core so that they all share a core and save some power.

My 5870 was not, and once again if there is only 1 GPU plugged why the CPU usage is less than 1%?

[shivansps]

I was using 1 HD6870 card... giving me about 250mhash/s and 0-1% cpu...when i plugged in a new HD6850, i now getting 450mhash/s, but at 50% cpu... each miner is using 25% and is static... it never moves of 25%.

[jedi95]

This is a problem with the ATI GPU drivers with more than 1 card installed. You will see the same results regardless of the miner you use. Setting the affinity so all the miners share a single core helps overall CPU usage and doesn't reduce hashrate.