Author

Topic: strange behavior overclocking 1070's (Read 137 times)

full member
Activity: 258
Merit: 104
April 12, 2021, 02:59:04 PM
#9
One of my rigs has 8 1070's that have been running nicely for almost 4 years.  But something I have never understood is that overclocking the memory makes no difference to the hashrate. But increasing the core clock does.  It's precisely the opposite of what is supposed to happen.  My other rigs which have 1070ti's don't exhibit this behavior.

My 1070's are tuned to use around 110 watts hashing at around 28mh/s which seems pretty optimum but I only achieve this by leaving the mem clock alone and bumping up the core clock.  All my rigs hardware are identical excepting the 1070ti's on the other rigs, which I have running at around 110 watts and 32.3mh/s by increasing the mem clock.

I haven't bothered asking this before because I doubt very much there is an answer.  But has anyone experienced this and know what causes it?

This has to do with the TLB bug on 1070s/1080s/1080ti's. Increase the core and power helps increase hashrate because of this bug. 1070ti is not affected because it was released later in the year with the "fix"

https://forums.developer.nvidia.com/t/blockchain-drivers/110847

"Here is the reply. Please note that I likely will not be able to respond to any follow-up questions about this. Therefore, if you ask a question, and I do not respond, that is the reason. I’m not permitted to share any more details than what is presented here:

We’ve confirmed that the performance drop is due to the size of the DAG exceeding the total on-chip TLB capacity on the Pascal GPU. As a result, there is an increased number of TLB misses, which affects performance. Because the TLB is a fixed capacity hardware resource, and the ETH algorithm accesses the DAG randomly, we don’t believe there are any software optimizations that could reduce the TLB miss rate.

In the Volta generation, TLB coverage was increased by 4x, and large DAG sizes (up to ~8GB, which won’t be reached for many years) will still fit in the on-chip TLB. So these newer GPUs (Volta and beyond) will show much less performance sensitivity due to DAG size."

Thank you!  Finally, an answer that explains whats causing the issue.  Even though there isn't anything to be done to fix it, I feel a lot better knowing the why.
member
Activity: 1201
Merit: 26
April 12, 2021, 01:10:57 PM
#8
I thought many people allready know 1000series dag problem.in 2018-2019 people complained about 0.5-1mhs decrease.
member
Activity: 1558
Merit: 69
April 12, 2021, 11:21:14 AM
#7
One of my rigs has 8 1070's that have been running nicely for almost 4 years.  But something I have never understood is that overclocking the memory makes no difference to the hashrate. But increasing the core clock does.  It's precisely the opposite of what is supposed to happen.  My other rigs which have 1070ti's don't exhibit this behavior.

My 1070's are tuned to use around 110 watts hashing at around 28mh/s which seems pretty optimum but I only achieve this by leaving the mem clock alone and bumping up the core clock.  All my rigs hardware are identical excepting the 1070ti's on the other rigs, which I have running at around 110 watts and 32.3mh/s by increasing the mem clock.

I haven't bothered asking this before because I doubt very much there is an answer.  But has anyone experienced this and know what causes it?

This has to do with the TLB bug on 1070s/1080s/1080ti's. Increase the core and power helps increase hashrate because of this bug. 1070ti is not affected because it was released later in the year with the "fix"

https://forums.developer.nvidia.com/t/blockchain-drivers/110847

"Here is the reply. Please note that I likely will not be able to respond to any follow-up questions about this. Therefore, if you ask a question, and I do not respond, that is the reason. I’m not permitted to share any more details than what is presented here:

We’ve confirmed that the performance drop is due to the size of the DAG exceeding the total on-chip TLB capacity on the Pascal GPU. As a result, there is an increased number of TLB misses, which affects performance. Because the TLB is a fixed capacity hardware resource, and the ETH algorithm accesses the DAG randomly, we don’t believe there are any software optimizations that could reduce the TLB miss rate.

In the Volta generation, TLB coverage was increased by 4x, and large DAG sizes (up to ~8GB, which won’t be reached for many years) will still fit in the on-chip TLB. So these newer GPUs (Volta and beyond) will show much less performance sensitivity due to DAG size."

Thanks, the only right answer in this thread.
member
Activity: 420
Merit: 13
$CYBERCASH METAVERSE
April 12, 2021, 11:00:06 AM
#6
Cards behavior are base on their silicon, just because they have the same name doesn't mean they will behave the same way, this is why it's better to spend more time on GPUS before making up your mind that you are the optimum settings, if you can drop the memory type your 1070 has then you can have different settings from people who still use this card for mining including myself

Yeah... I dont think memory type matters in this case as adjusting mem clock makes zero difference.  Only the core clock seems to affect the hashrate.  Higher core clock = higher hashrate.

But for what it's worth the 8 cards are all GeForce® GTX 1070 G1 Gaming 8G (rev. 1.0)

I think it's more likely to be a bios setting or something in windows 10.  But damned if I know what.
Memory type matters because we have samsung, hynix, micron, the reason why you are asked to drop your memory type is because people who have same card as yours can easily share their own settings, stop acting like you understand your card better than anybody, my gtx1660 super gives me 30mh no matter what I tried and after several months I later find out that with a particular settings i can achieve 32mh
member
Activity: 246
Merit: 24
April 12, 2021, 10:45:18 AM
#5
One of my rigs has 8 1070's that have been running nicely for almost 4 years.  But something I have never understood is that overclocking the memory makes no difference to the hashrate. But increasing the core clock does.  It's precisely the opposite of what is supposed to happen.  My other rigs which have 1070ti's don't exhibit this behavior.

My 1070's are tuned to use around 110 watts hashing at around 28mh/s which seems pretty optimum but I only achieve this by leaving the mem clock alone and bumping up the core clock.  All my rigs hardware are identical excepting the 1070ti's on the other rigs, which I have running at around 110 watts and 32.3mh/s by increasing the mem clock.

I haven't bothered asking this before because I doubt very much there is an answer.  But has anyone experienced this and know what causes it?

This has to do with the TLB bug on 1070s/1080s/1080ti's. Increase the core and power helps increase hashrate because of this bug. 1070ti is not affected because it was released later in the year with the "fix"

https://forums.developer.nvidia.com/t/blockchain-drivers/110847

"Here is the reply. Please note that I likely will not be able to respond to any follow-up questions about this. Therefore, if you ask a question, and I do not respond, that is the reason. I’m not permitted to share any more details than what is presented here:

We’ve confirmed that the performance drop is due to the size of the DAG exceeding the total on-chip TLB capacity on the Pascal GPU. As a result, there is an increased number of TLB misses, which affects performance. Because the TLB is a fixed capacity hardware resource, and the ETH algorithm accesses the DAG randomly, we don’t believe there are any software optimizations that could reduce the TLB miss rate.

In the Volta generation, TLB coverage was increased by 4x, and large DAG sizes (up to ~8GB, which won’t be reached for many years) will still fit in the on-chip TLB. So these newer GPUs (Volta and beyond) will show much less performance sensitivity due to DAG size."
member
Activity: 1201
Merit: 26
April 12, 2021, 09:49:06 AM
#4
There is many types of gtx 1070.
For example i have asus expedition 1070 (8pin) and msi gaminx 1070 (6+8pins) asus needs core clock and it gets hot. But gaminx without core clock ok and default clock is higher than asus thats why asus needs more push. They same memory both micron. and memory can be overclocked on msi far away than asus.Any way i sold all 3 asus away 2 years ago Smiley
full member
Activity: 258
Merit: 104
April 12, 2021, 07:53:12 AM
#3
Cards behavior are base on their silicon, just because they have the same name doesn't mean they will behave the same way, this is why it's better to spend more time on GPUS before making up your mind that you are the optimum settings, if you can drop the memory type your 1070 has then you can have different settings from people who still use this card for mining including myself

Yeah... I dont think memory type matters in this case as adjusting mem clock makes zero difference.  Only the core clock seems to affect the hashrate.  Higher core clock = higher hashrate.

But for what it's worth the 8 cards are all GeForce® GTX 1070 G1 Gaming 8G (rev. 1.0)

I think it's more likely to be a bios setting or something in windows 10.  But damned if I know what.
full member
Activity: 1148
Merit: 116
April 12, 2021, 02:17:33 AM
#2
Cards behavior are base on their silicon, just because they have the same name doesn't mean they will behave the same way, this is why it's better to spend more time on GPUS before making up your mind that you are the optimum settings, if you can drop the memory type your 1070 has then you can have different settings from people who still use this card for mining including myself
full member
Activity: 258
Merit: 104
April 11, 2021, 09:27:38 PM
#1
One of my rigs has 8 1070's that have been running nicely for almost 4 years.  But something I have never understood is that overclocking the memory makes no difference to the hashrate. But increasing the core clock does.  It's precisely the opposite of what is supposed to happen.  My other rigs which have 1070ti's don't exhibit this behavior.

My 1070's are tuned to use around 110 watts hashing at around 28mh/s which seems pretty optimum but I only achieve this by leaving the mem clock alone and bumping up the core clock.  All my rigs hardware are identical excepting the 1070ti's on the other rigs, which I have running at around 110 watts and 32.3mh/s by increasing the mem clock.

I haven't bothered asking this before because I doubt very much there is an answer.  But has anyone experienced this and know what causes it?
Jump to: