Low CPU hashrate (i7-7700k - cryptonight) - can't figure out why!

[benjim123]

Hey guys,

Been GPU + CPU mining with nicehash for a few months as well as running my AMD GPUs mining eth directly elsewhere.

Recently I upgraded my PC; got a 1080Ti and an i7-7700k. I've seen on the profitability calculator and elsewhere on benchmark tables that the i7-7700k (at around 4.7GHz is capable of 360 H/s for cryptonight (using XMR-Stak-CPU).

However, mine is overclocked to 5GHz core and 4.8GHz cache and I can never get it above ~280H/s...

I've tried everything I could think of. Read and tried all the tips here but no luck at all.
https://github.com/nicehash/NiceHashMiner/wiki/Notes-and-hints-on-CPU-mining

(locked pages, testing different combinations of threads etc).

Have even tried other cpu-miner programs externally without nicehash - all with similar results. I'm stumped!

Would really appreciate some help with this.

Thanks

[1524737162]

1524737162

[1524737162]

1524737162

[1524737162]

1524737162

[philipma1957]

Hey guys,

Been GPU + CPU mining with nicehash for a few months as well as running my AMD GPUs mining eth directly elsewhere.

Recently I upgraded my PC; got a 1080Ti and an i7-7700k. I've seen on the profitability calculator and elsewhere on benchmark tables that the i7-7700k (at around 4.7GHz is capable of 360 H/s for cryptonight (using XMR-Stak-CPU).

However, mine is overclocked to 5GHz core and 4.8GHz cache and I can never get it above ~280H/s...

I've tried everything I could think of. Read and tried all the tips here but no luck at all.
https://github.com/nicehash/NiceHashMiner/wiki/Notes-and-hints-on-CPU-mining

(locked pages, testing different combinations of threads etc).

Have even tried other cpu-miner programs externally without nicehash - all with similar results. I'm stumped!

Would really appreciate some help with this.

Thanks

I have never seen that cpu do 360H  for xmr

did you see a screen shot of it doing that?

Sometimes benchmark tables are not accurate.

[naykos]

With that cpu you should have 4 threads enabled and set affinity for cores 0, 2, 4 and 6.
edit: or was it cores 1, 3, 5 and 7? check both.

[Smontiel]

imho you should've updated to ryzen

[philipma1957]

imho you should've updated to ryzen

I did not want to make him feel bad and tell him my  1800x  gets 580 h on a 3.7 clock

[garytheasshole]

imho you should've updated to ryzen

I did not want to make him feel bad and tell him my  1800x  gets 580 h on a 3.7 clock

My 1500x which is like 33% of the price of the 1080x gets 420 on default clock.
The i6700k I have on default clocks  it can do is 240 h/s I think that's the best it can do with current miners.
Try compiling the miner with gcc 7 on new kernel like fedora 26 or ubuntu 17, enable hugepages do your overclocking from bios and see if you can squeeze moar power. Try out https://github.com/xmrig/xmrig, it gets me slightly more hashes than xmr-stark

[alucard20724]

imho you should've updated to ryzen

I did not want to make him feel bad and tell him my  1800x  gets 580 h on a 3.7 clock

I think he's going by the "NiceHash Calculator" which we all know is 100% accurate 

at stock speeds i get 280 h on my 7700k.

He should try Trupov ccminer xmr... i think it's vers 1.3.1 or something.   I would love for claymore to update his, but he hasn't done that in a year and the gpu miners are priority right now.

[joblo]

4 threads with default affinity should be optimum for that CPU. It might be worthwhile to undo all the tweaks.
Sometimes they can hurt performance if not properly tuned.

Since the poor performance is seen by several miners you should consider a system issue. Check for other processes
running. Check CPU usage, you should have 4 cores at 100% and 4 virtually idle.

Do some performance testing to see if other programs also have reduced performance.

Post your complete specs.

That's all I got.

[antantti]

Out of all systems I have tried for some reason sandy bridge based seem to give the best hash/ thread/ clock speeds, don't know why.

For example, 2600k@4.2GHz (kaby lake default):



Also make sure all threads are running full speed, easy tool for intel+windows:

https://downloadcenter.intel.com/download/24075/Intel-Extreme-Tuning-Utility-Intel-XTU-

[gigi2006]

Can You explain more
Extrem utility Tool?
What can make Whit that?

Out of all systems I have tried for some reason sandy bridge based seem to give the best hash/ thread/ clock speeds, don't know why.

For example, 2600k@4.2GHz (kaby lake default):



Also make sure all threads are running full speed, easy tool for intel+windows:

https://downloadcenter.intel.com/download/24075/Intel-Extreme-Tuning-Utility-Intel-XTU-

[antantti]

Can You explain more
Extrem utility Tool?
What can make Whit that?

You can change clocks + volts on the fly, faster than boot to bios. Makes finding perfects settings a lot faster.

If you don't trust that link just google it.

[gigi2006]

The link is not the problem, I see that this is a legitimate link.

My problem is that I didn't manage to push my CPU. The system remains hung and restarts immediately.

Is there a manual for that somewhere?

Can You explain more
Extrem utility Tool?
What can make Whit that?

You can change clocks + volts on the fly, faster than boot to bios. Makes finding perfects settings a lot faster.

If you don't trust that link just google it.

[antantti]

Is there a manual for that somewhere?

You need to ask Intel, it's their software+hardware.

[RinceWind84]

I am getting 390 hash/s on i7-3770k@4.9Ghz 4-threads HT-off and 458 hash/s on i7-3930k@4.7Ghz 6 threads HT-off

Using xmrig...

[wacko]

I am getting 390 hash/s on i7-3770k@4.9Ghz 4-threads HT-off

That's a lot for 3770k. Do you know by any chance what's the hashrate with HT on? Just curious how much of a difference (if any) hyper-threading makes.

[RinceWind84]

I am getting 390 hash/s on i7-3770k@4.9Ghz 4-threads HT-off

That's a lot for 3770k. Do you know by any chance what's the hashrate with HT on? Just curious how much of a difference (if any) hyper-threading makes.

Yes. around 380 hash/s
I turn HT off mainly to get a more stable OC. And since I can only run 1 CryptoNight thread pr. 2Mb L3-cache I would still only be running 4 threads with HT-on.