Author

Topic: CCminer(SP-MOD) Modded NVIDIA Maxwell / Pascal kernels. - page 580. (Read 2347641 times)

legendary
Activity: 1400
Merit: 1000
I understand the difference very well.
In any TRUE test if you can't run the environments in the EXACT same settings the the results are not valid.

The cards are in the EXACT same settings when you start the miner. But the opensource will trottle and downclock itself automaticly. This is a test of the quality of the kernel at a given powerlimit.

OK then send me your kernel and I will down clock your kernel to the open source to compare the speed. That way it is speed for speed.

I have Maxwell and Pascal and none of my clocks have down clocked on their own. My cards have been the opposite. Boost has kicked in and my 1070's are running at 1840-1850 and never went to 1600.

I just want to see a fair comparison over a period of time.
sp_
legendary
Activity: 2954
Merit: 1087
Team Black developer
1080 g1 2050 core -i 27 405 mh, I think this is stable. 400 mh is stable though with -i 26. 1080 doesnt scale well with overclock. windows 8.1.  -i 28 has nonce problems

Mining on a low difficulty can sometimes give Nonce already sendt error messages, but they should dissapear when the difficulty rise. But if you get Nonce does not validate on the cpu, then you should use an another setting.
legendary
Activity: 1176
Merit: 1015
Guys, why is it so hard to understand?
sp_
legendary
Activity: 2954
Merit: 1087
Team Black developer
I understand the difference very well.
In any TRUE test if you can't run the environments in the EXACT same settings the the results are not valid.

The cards are in the EXACT same settings when you start the miner. But the opensource will trottle and downclock itself automaticly. This is a test of the quality of the kernel at a given powerlimit.
legendary
Activity: 1400
Merit: 1000
NO it is not.
You are comparing apples to oranges with one card with a higher clock. The card on the right needs downclocked to match the one o the left.

Lets say we get the exact same car. You drive 75 MPH and I drive 90 MPH (your 20%), I'll bet you any amount of money in the world I will beat you in a race.

See how it works out when you don't do a TRUE test apples to apples.

He is running with a reduced TDP(70%). 115Watt instead of 150 watt. Do you understand the difference? The NVIDIA boost clock is unable to run at full boost with the opensource@only 115 watt because it is out of power, with the SP-mod the card is running @ full boost clock.

With sp-mod #3 and the Gtx 970 (28nm maxwell) hashing at 157MHASH @ only 115 watt he is close to the perfomance of the pascal chip(16nm) using the opensource kernel when you calculate Hash/watt

I understand the difference very well.

In any TRUE test if you can't run the environments in the EXACT same settings the the results are not valid.
sp_
legendary
Activity: 2954
Merit: 1087
Team Black developer
NO it is not.
You are comparing apples to oranges with one card with a higher clock. The card on the right needs downclocked to match the one o the left.

Lets say we get the exact same car. You drive 75 MPH and I drive 90 MPH (your 20%), I'll bet you any amount of money in the world I will beat you in a race.

See how it works out when you don't do a TRUE test apples to apples.

He is running with a reduced TDP(70%). 115Watt instead of 150 watt. Do you understand the difference? The NVIDIA boost clock is unable to run at full boost with the opensource@only 115 watt because it is out of power, with the SP-mod the card is running @ full boost clock.

With sp-mod #3 and the Gtx 970 (28nm maxwell) hashing at 157MHASH @ only 115 watt he is close to the perfomance of the pascal chip(16nm) using the opensource kernel when you calculate Hash/watt
legendary
Activity: 1400
Merit: 1000
1.8.1 vs sp#3 70% tdp, strix 970 vs strix 970, intensity 25, windows 7x64 368.81



This is a confirmed +20% increase

Did you measure the watt on the sp-mod#1 compared to the opensource?  (70% TDP is around 105Watt? )
Can you try -i 26 or -i 27 or higher if it doesn't crash or trottle.

158MHASH / 105 watt is making 1,5047MHASH per watt. (gtx 970 standard clocks)(70% TDP)
255MHASH / 150 watt is making 1,7MHASH per watt (gtx 1070 standard clocks)

NO it is not.

You are comparing apples to oranges with one card with a higher clock. The card on the right needs downclocked to match the one o the left.

Lets say we get the exact same car. You drive 75 MPH and I drive 90 MPH (your 20%), I'll bet you any amount of money in the world I will beat you in a race.

See how it works out when you don't do a TRUE test apples to apples.
member
Activity: 82
Merit: 10
1080 g1 2050 core -i 27 405 mh, I think this is stable. 400 mh is stable though with -i 26. 1080 doesnt scale well with overclock. windows 8.1.  -i 28 has nonce problems
sp_
legendary
Activity: 2954
Merit: 1087
Team Black developer
2 cards 300 mhash
6 cards 294mhash

But this can be solved. Red antantti last posts. The 1x 8pin cable is not able to deliver enough power, and with 6 cards oc drawing power from the mainboard you need to lower the tdp of all the cards.



sr. member
Activity: 954
Merit: 250


i concur, i need to see a 6 x 1070 rig in action with your mod sp, to decide if it is worth it to get the miner

So do I.
legendary
Activity: 3248
Merit: 1070
i sell my GTX970, and i buy GTX980 . it is a great card .

buy pascal or you'll regret it.
when profit goes down, pascal can still mine and maxwell will shutdown.

yeah lol, buying a 980 now it's not a thing anyone should do, pascal is much better in everything

how much with just 2000 core? i see you reached a crazy core, not so common with all kind of 1070

The point is that the sp-mod is stable on higher clocks than the opensource miner. Rednow just confirmed it with 322MHASH on the gtx 1070. (-i 27)
The highest he got with the opensource miner was 290MHASH.

(+11%)

stability with just one card not always means with 6-gpu rig.
that's why it would be nice to see the hashrate with default clock or even with little OC.

i concur, i need to see a 6 x 1070 rig in action with your mod sp, to decide if it is worth it to get the miner
legendary
Activity: 1176
Merit: 1015
With -pstate=0 on the opensource version it don't throttle my GTX 1080 runs at 2088/1387,8 24h ... on Windows 10 with the latest driver so I don't get the assumption the Opensource version would throttle..i can't let this statement pass ...since it is not the fact.

If you read his post, he has reduced the tdp limit of the card to 70% (105Watt). The sp-mod deliver 20% higher hashrate with the same watt as the opensource.

158MHASH / 105 watt is making 1,5047MHASH per watt. (gtx 970 standard clocks sp-mod #3)(70% TDP)
255MHASH / 150 watt is making 1,7MHASH per watt (gtx 1070 standard clocks)

It is a bit more than 105, 110-120 watts maybe. Those cards have 100% tdp at 163 watts.

Point is that opensource can do max 163MH but it needs 196 (tdp  120%) watts for that. Sp #3 only needs 120-130 watts (tdp 75%).

Sure you can run it at 196 watts with sp-mod too, it does 190-200MH then. I haven't tested absolute max with it, pcie cable with 2x970 gets too warm for my taste.
legendary
Activity: 2716
Merit: 1094
Black Belt Developer
i sell my GTX970, and i buy GTX980 . it is a great card .

buy pascal or you'll regret it.
when profit goes down, pascal can still mine and maxwell will shutdown.
sr. member
Activity: 430
Merit: 250
i sell my GTX970, and i buy GTX980 . it is a great card .
sp_
legendary
Activity: 2954
Merit: 1087
Team Black developer
Gallax HOF GTX 970 standard clocks sp-mod #3


member
Activity: 83
Merit: 10
how much with just 2000 core? i see you reached a crazy core, not so common with all kind of 1070

The point is that the sp-mod is stable on higher clocks than the opensource miner. Rednow just confirmed it with 322MHASH on the gtx 1070. (-i 27)
The highest he got with the opensource miner was 290MHASH.

(+11%)

stability with just one card not always means with 6-gpu rig.
that's why it would be nice to see the hashrate with default clock or even with little OC.
sp_
legendary
Activity: 2954
Merit: 1087
Team Black developer
how much with just 2000 core? i see you reached a crazy core, not so common with all kind of 1070

The point is that the sp-mod is stable on higher clocks than the opensource miner. Rednow just confirmed it with 322MHASH on the gtx 1070. (-i 27)
The highest he got with the opensource miner was 290MHASH.

(+11%)
sp_
legendary
Activity: 2954
Merit: 1087
Team Black developer
With -pstate=0 on the opensource version it don't throttle my GTX 1080 runs at 2088/1387,8 24h ... on Windows 10 with the latest driver so I don't get the assumption the Opensource version would throttle..i can't let this statement pass ...since it is not the fact.

If you read his post, he has reduced the tdp limit of the card to 70% (105Watt). The sp-mod deliver 20% higher hashrate with the same watt as the opensource.

158MHASH / 105 watt is making 1,5047MHASH per watt. (gtx 970 standard clocks sp-mod #3)(70% TDP)
255MHASH / 150 watt is making 1,7MHASH per watt (gtx 1070 standard clocks)
legendary
Activity: 3248
Merit: 1070
gtx1070 319mhs lbry
Damn slow on my poor gtx750.
with what version this hash? with the sp version?
Take a closer look to console window name on the screenshot. ))
I've got 322mhs with -i 27 intensity later ))
it does not work on win 10, above -i 25 it say illegal memory access

This is not the opensource miner.

so your is working with win 10? you said early that it does not work

it does not work on win 10, above -i 25 it say illegal memory access

The opensource miner cannot go above -i 25. The sp-mod #3 can go to -i 29 stable on the gtx 1070
I think I need one more Palit 1070 ))

how much with just 2000 core? i see you reached a crazy core, not so common with all kind of 1070
sr. member
Activity: 574
Merit: 250
Fighting mob law and inquisition in this forum
The clockspeed on the maxwell/pascal are dynamic. With the opensource the cards trottle and downclock themself. I have seen this on many models. Lowering the intensity might help.

Same card, same config. +20% with the sp-mod #3.
With -pstate=0 on the opensource version it don't throttle that much my GTX 1080 runs at 2088/1387,8 24h ... on Windows 10 with the latest driver so I don't get the assumption the Opensource version would throttle..i can't let this statement pass ...since it is not the fact. The max boost clock is 2130MHz with my Zotac AMP!.
Jump to: