Are 1070 TI's the next best GPU for mining? 4.7 sols per watt

[wacko]

Set up my EVGA 1070 Ti this afternoon (Micron memory  )and adjusted settings to 80% TDP +200 Core +600 memory.  Ran fine with about 525-530 sol/s for about 45 mins then artifacted on me on and crashed system as my monitor was hooked up to this.  This something I should RMA if it happens again?

You overclocked the crap out of the card and then want to RMA it because it crashes? EVGA won't be happy about that. Seriously though, overclocking is not only ever guaranteed, but with most manufacturers it actually voids your warranty. Hard to enforce since customers never admit that they did any overclocking and there's no sure way to check whether they did, but still it's there most of the time, in the fine print.

[butmuncher]

Set up my EVGA 1070 Ti this afternoon (Micron memory  )and adjusted settings to 80% TDP +200 Core +600 memory.  Ran fine with about 525-530 sol/s for about 45 mins then artifacted on me on and crashed system as my monitor was hooked up to this.  This something I should RMA if it happens again?

You overclocked the crap out of the card and then want to RMA it because it crashes? EVGA won't be happy about that. Seriously though, overclocking is not only ever guaranteed, but with most manufacturers it actually voids your warranty. Hard to enforce since customers never admit that they did any overclocking and there's no sure way to check whether they did, but still it's there most of the time, in the fine print.

If that was a new card it was faulty on arrival if it crashes in game and with mining at stock, its not wacko's fault the card died unless its a unstable overclock then its user error.
Obviously if it does'nt crash in game at stock settings its 100% perfectly ok, if it does crash in games with no oc then it is indeed faulty and will need to be rma'd.

Your oc would'nt of hurt this card in the slightest

[wacko]

If that was a new card it was faulty on arrival if it crashes in game and with mining at stock

Hmm, I've got the impression from his message that he had artifacts on his screen while the card was @ +200/+600. Which is a very hefty overclock at PL80 and not every card can run stable with such an OC. Of course if those issues show up at stock clocks — then RMA it.

[QuintLeo]

Set up my EVGA 1070 Ti this afternoon (Micron memory  )and adjusted settings to 80% TDP +200 Core +600 memory.  Ran fine with about 525-530 sol/s for about 45 mins then artifacted on me on and crashed system as my monitor was hooked up to this.  This something I should RMA if it happens again?

 No - it just means you didn't get lucky in the silicon sweepstakes on your memory or your core - OR you have poor cooling for the card and you managed to overheat it.

 What temp does the card get up to after 5 minutes or so?

 I have no idea what memory type my EVGA SC 1070 ti cards have, as I don't run them on Windows so no GPU-Z to easily tell without disassembly.


 Not sure why "in game" was mentioned in some of the replies, OP didn't say anything about gaming on the card.

[flip4flop]

Set up my EVGA 1070 Ti this afternoon (Micron memory  )and adjusted settings to 80% TDP +200 Core +600 memory.  Ran fine with about 525-530 sol/s for about 45 mins then artifacted on me on and crashed system as my monitor was hooked up to this.  This something I should RMA if it happens again?

 No - it just means you didn't get lucky in the silicon sweepstakes on your memory or your core - OR you have poor cooling for the card and you managed to overheat it.

 What temp does the card get up to after 5 minutes or so?

 I have no idea what memory type my EVGA SC 1070 ti cards have, as I don't run them on Windows so no GPU-Z to easily tell without disassembly.


 Not sure why "in game" was mentioned in some of the replies, OP didn't say anything about gaming on the card.

Micron garbage memory and the temp is fine, never gets above 55C.  I tried lowering OC settings and it still errors out of EWBF.  Its basically a worse 1070.  My 3x 1070s are pushing +200/+600 no issues @ 70 TDP (and they have Micron mem as well).  This 1070 Ti cant hold +50/+200 at 80 TDP for more than a hour maybe two.  Hashrate is equal to or a little less than the 1070s. Disappointing.

Even tried putting all my 1070s on the risers and the 1070 Ti directly on the MB but still same issue.

[wacko]

Micron garbage memory and the temp is fine, never gets above 55C.  I tried lowering OC settings and it still errors out of EWBF.  Its basically a worse 1070.  My 3x 1070s are pushing +200/+600 no issues @ 70 TDP (and they have Micron mem as well).  This 1070 Ti cant hold +50/+200 at 80 TDP for more than a hour maybe two.  Hashrate is equal to or a little less than the 1070s. Disappointing.

Even tried putting all my 1070s on the risers and the 1070 Ti directly on the MB but still same issue.

Did you try overclocking them separately? It's always +xx/+xx @ xx TDP in your posts, so it's hard to be sure what's what. Set the powerlimit to 100, don't touch the core clock and simply increase the memclock to check the actual memory OC limit. Cause +200 max doesn't sound plausible, it should be able to OC more.

[flip4flop]

Micron garbage memory and the temp is fine, never gets above 55C.  I tried lowering OC settings and it still errors out of EWBF.  Its basically a worse 1070.  My 3x 1070s are pushing +200/+600 no issues @ 70 TDP (and they have Micron mem as well).  This 1070 Ti cant hold +50/+200 at 80 TDP for more than a hour maybe two.  Hashrate is equal to or a little less than the 1070s. Disappointing.

Even tried putting all my 1070s on the risers and the 1070 Ti directly on the MB but still same issue.

Did you try overclocking them separately? It's always +xx/+xx @ xx TDP in your posts, so it's hard to be sure what's what. Set the powerlimit to 100, don't touch the core clock and simply increase the memclock to check the actual memory OC limit. Cause +200 max doesn't sound plausible, it should be able to OC more.

All 3 of my MSI 1070 run stable at +200 core +600 Mem at 70 TDP

The EVGA 1070 Ti is my only issue. I will try tonight to leave TDP at 100 and then slowly up the memory to see when it fails.  My concern is the same, it should be able to handle +200 memory with no issue.

[Yutard]

Set up my EVGA 1070 Ti this afternoon (Micron memory  )and adjusted settings to 80% TDP +200 Core +600 memory.  Ran fine with about 525-530 sol/s for about 45 mins then artifacted on me on and crashed system as my monitor was hooked up to this.  This something I should RMA if it happens again?

You overclocked the crap out of the card and then want to RMA it because it crashes? EVGA won't be happy about that. Seriously though, overclocking is not only ever guaranteed, but with most manufacturers it actually voids your warranty. Hard to enforce since customers never admit that they did any overclocking and there's no sure way to check whether they did, but still it's there most of the time, in the fine print.

I agree with that. I do heavy underclocking to save electricity.

[wacko]

All 3 of my MSI 1070 run stable at +200 core +600 Mem at 70 TDP

The EVGA 1070 Ti is my only issue. I will try tonight to leave TDP at 100 and then slowly up the memory to see when it fails.  My concern is the same, it should be able to handle +200 memory with no issue.

TDP (Powerlimit slider in MSI AB) is just one part of the OC equation that people often don't even fully understand. It's the maximum allowed percentage of the "total" power that the card is allowed to burn. The thing is, the "full" TDP differs from card to card, it's a value in the card's BIOS, yet people in the forums often share their set Powerlimits like they're universal between different cards. They're obviously not. Your MSI 1070 cards might be higher-end Gaming (X/Z) cards and their factory TDP might be well above the reference 150W. Could be 180W, or even more. So when you set your powerlimit to 70% with these cards, you're telling them to burn no more than 180*0.7 = 126W. Not that much of a limit for a 1070. But your EVGA 1070 Ti cards might have their TDP set to 180W in BIOS, yet it's a more power-hungry chip, so when you set PL 80 for them you're limiting them to 144W, which might just not be enough for +200/+600 mode that you initially set.

I'm not saying it's the powerlimit setting that is the reason for your troubles in this case (although it could be), just trying to explain that controlling voltage and clocks on Pascal cards by simply moving one powerlimit slider is not very precise and reliable, and any meaningful tuning with this slider at least requires the user to know the actual (100%) TDP for the particular card in Watts.

[QuintLeo]

Micron garbage memory and the temp is fine, never gets above 55C.  I tried lowering OC settings and it still errors out of EWBF.  Its basically a worse 1070.  My 3x 1070s are pushing +200/+600 no issues @ 70 TDP (and they have Micron mem as well).  This 1070 Ti cant hold +50/+200 at 80 TDP for more than a hour maybe two.  Hashrate is equal to or a little less than the 1070s. Disappointing.

Even tried putting all my 1070s on the risers and the 1070 Ti directly on the MB but still same issue.

Did you try overclocking them separately? It's always +xx/+xx @ xx TDP in your posts, so it's hard to be sure what's what. Set the powerlimit to 100, don't touch the core clock and simply increase the memclock to check the actual memory OC limit. Cause +200 max doesn't sound plausible, it should be able to OC more.

All 3 of my MSI 1070 run stable at +200 core +600 Mem at 70 TDP

The EVGA 1070 Ti is my only issue. I will try tonight to leave TDP at 100 and then slowly up the memory to see when it fails.  My concern is the same, it should be able to handle +200 memory with no issue.

 Depends in part on where it sets the STOCK clock of the memory - part of the reason my EVGA 1070 ti SC cards overclock memory so well is that the BASE MEMORY CLOCK is underclocked when they are running a mining program.

 Card models VARY, and even individual CARDS of the SAME model can vary - memory brand seems to make a big difference, and EVGA (like ALL card makers) tends to use whatever memory it can get ahold of to make a batch of cards, even if it's not the SAME brand as what went into a previous batch of the same card model.


 MSI Armor Gaming 1070 card is a VERY GOOD example of the "variation in TDP between models" issue - 240 watt TDP by factory default, where MOST 1070 models are 151 or 180 watts.


 For the record, I've NEVER had a 1070 model that could do "+600" in afterburner settings AT ALL, most were unstable at +550 and at least a couple unstable at +500 - even on models other folks reported being able to STABLE overclock more.
 Like I said, cards VARY.

[Kompik]

Micron garbage memory and the temp is fine, never gets above 55C.  I tried lowering OC settings and it still errors out of EWBF.  Its basically a worse 1070.  My 3x 1070s are pushing +200/+600 no issues @ 70 TDP (and they have Micron mem as well).  This 1070 Ti cant hold +50/+200 at 80 TDP for more than a hour maybe two.  Hashrate is equal to or a little less than the 1070s. Disappointing.

Even tried putting all my 1070s on the risers and the 1070 Ti directly on the MB but still same issue.

Did you try overclocking them separately? It's always +xx/+xx @ xx TDP in your posts, so it's hard to be sure what's what. Set the powerlimit to 100, don't touch the core clock and simply increase the memclock to check the actual memory OC limit. Cause +200 max doesn't sound plausible, it should be able to OC more.

All 3 of my MSI 1070 run stable at +200 core +600 Mem at 70 TDP

The EVGA 1070 Ti is my only issue. I will try tonight to leave TDP at 100 and then slowly up the memory to see when it fails.  My concern is the same, it should be able to handle +200 memory with no issue.

 Depends in part on where it sets the STOCK clock of the memory - part of the reason my EVGA 1070 ti SC cards overclock memory so well is that the BASE MEMORY CLOCK is underclocked when they are running a mining program.

 Card models VARY, and even individual CARDS of the SAME model can vary - memory brand seems to make a big difference, and EVGA (like ALL card makers) tends to use whatever memory it can get ahold of to make a batch of cards, even if it's not the SAME brand as what went into a previous batch of the same card model.


 MSI Armor Gaming 1070 card is a VERY GOOD example of the "variation in TDP between models" issue - 240 watt TDP by factory default, where MOST 1070 models are 151 or 180 watts.


 For the record, I've NEVER had a 1070 model that could do "+600" in afterburner settings AT ALL, most were unstable at +550 and at least a couple unstable at +500 - even on models other folks reported being able to STABLE overclock more.
 Like I said, cards VARY.

Do you force your cards to have P0 state when using compute applications? You will never get stable high memory OC without it unless you you have samsung memory.

[wacko]

Do you force your cards to have P0 state when using compute applications? You will never get stable high memory OC without it unless you you have samsung memory.

Well, to be honest the P2 issue is universal, and cards with Samsung memory suffer from it the same way as the cards with Micron. The problem, as you probably know it, is not that the card mines in P2 mode, but that when you close the miner window the memory clock jumps up for just a moment, and in that moment the driver/rig crashes. You set the memory to +500 in the MSI AB, but when you close the miner, the clock jumps to +700 for a bit. That can be avoided when you force cuda apps to run in P0 via nvidia profile inspector. It doesn't just change the mode for cuda apps to P0, but gets rid of that weird P2=>P0 jump so that the memory clock doesn't ever go higher than you set it with the +xxx offset.

[Kompik]

Do you force your cards to have P0 state when using compute applications? You will never get stable high memory OC without it unless you you have samsung memory.

Well, to be honest the P2 issue is universal, and cards with Samsung memory suffer from it the same way as the cards with Micron. The problem, as you probably know it, is not that the card mines in P2 mode, but that when you close the miner window the memory clock jumps up for just a moment, and in that moment the driver/rig crashes. You set the memory to +500 in the MSI AB, but when you close the miner, the clock jumps to +700 for a bit. That can be avoided when you force cuda apps to run in P0 via nvidia profile inspector. It doesn't just change the mode for cuda apps to P0, but gets rid of that weird P2=>P0 jump so that the memory clock doesn't ever go higher than you set it with the +xxx offset.

Yes we are on the same page

It was just my bad wording, i was reffering to the +550 clock offset regarding the samsung memory. You will get +550 with samsung even on the P2 state, since most samsung cards will not fail when they jump to +750 for a second.

[bigjee]

Set up my EVGA 1070 Ti this afternoon (Micron memory  )and adjusted settings to 80% TDP +200 Core +600 memory.  Ran fine with about 525-530 sol/s for about 45 mins then artifacted on me on and crashed system as my monitor was hooked up to this.  This something I should RMA if it happens again?

If you do so while overclocking then you are abusing the RMA system.
I dont think its the cards fault. Just need to tweak it better. Try 100 and 600 and run for an hour. Also try 85% TDP.
If you get artifacts on stock setting then I would RMA as that is a card issue. 
they also might send the card right back to you since they wont find any issues with it. 

[bigjee]

Micron garbage memory and the temp is fine, never gets above 55C.  I tried lowering OC settings and it still errors out of EWBF.  Its basically a worse 1070.  My 3x 1070s are pushing +200/+600 no issues @ 70 TDP (and they have Micron mem as well).  This 1070 Ti cant hold +50/+200 at 80 TDP for more than a hour maybe two.  Hashrate is equal to or a little less than the 1070s. Disappointing.

Even tried putting all my 1070s on the risers and the 1070 Ti directly on the MB but still same issue.

Did you try overclocking them separately? It's always +xx/+xx @ xx TDP in your posts, so it's hard to be sure what's what. Set the powerlimit to 100, don't touch the core clock and simply increase the memclock to check the actual memory OC limit. Cause +200 max doesn't sound plausible, it should be able to OC more.

All 3 of my MSI 1070 run stable at +200 core +600 Mem at 70 TDP

The EVGA 1070 Ti is my only issue. I will try tonight to leave TDP at 100 and then slowly up the memory to see when it fails.  My concern is the same, it should be able to handle +200 memory with no issue.

 Depends in part on where it sets the STOCK clock of the memory - part of the reason my EVGA 1070 ti SC cards overclock memory so well is that the BASE MEMORY CLOCK is underclocked when they are running a mining program.

 Card models VARY, and even individual CARDS of the SAME model can vary - memory brand seems to make a big difference, and EVGA (like ALL card makers) tends to use whatever memory it can get ahold of to make a batch of cards, even if it's not the SAME brand as what went into a previous batch of the same card model.


 MSI Armor Gaming 1070 card is a VERY GOOD example of the "variation in TDP between models" issue - 240 watt TDP by factory default, where MOST 1070 models are 151 or 180 watts.


 For the record, I've NEVER had a 1070 model that could do "+600" in afterburner settings AT ALL, most were unstable at +550 and at least a couple unstable at +500 - even on models other folks reported being able to STABLE overclock more.
 Like I said, cards VARY.

+1 There is variation within the same card, same manufacturer and same model.
I think Micron memory is very capable and overclocks like a champ.
I have and do run some (not all) 1070s at +675 memory at 85TDP and 100 TDP (24-48 hours stable) so it really depends on the cards "personality" some are athletes others are divas.  

[lunobird]

My Msi 1070 ti Titanium color edition runs great.

My current semi aggressive setting.  I haven't even dialed it in, just used your guys recommended start point.

tdp 75 percent
+200 core clock
+700 mem clock
= 500 sol/s equihash.

Stable for days,  I have not tried to plug in the monitor to it.  too scared it might start to artifact if I'm taxing the gpu more to render a screen as well as hashing.
My advice is to slightly go more conservative on the gpu that needs to render to monitor.


If I were to build a large farm I would go with 1070 ti all the way.  It's highly efficient.  It's the new standard. 1070 is phased out (older card, won't have a good resale value since everyone will want the 1070 ti, 1070ti overclocks better and runs cooler)

As much as I like my 1080 ti i feel they struggle dollar for dollar to keep up with the 1070ti.

For example, I go by the $1 dollar = 1 sol or higher
My 1080 ti at 75 tdp +125 core clock does 680 sols about. But the cards cost $700-750 (depending on deals)
My 1070 ti  at 75 tdp +200 core/700mem does 500 sols about.  Those cost only $460-480

The 1070 ti performs higher than 1 sol per dollar whereas the 1080 ti struggles to get 1 sol per dolllar spent.

[sikke1]

Asus Strix A 1070 TI
I have been able to push it to 3.9 Sol/W but now looking most stable setup. It's 505-520 Sol. Best bang for the $ anyway. Because they run Wayyy!! cooler than Vega or 1080 TI no competition. You can dismiss cooling needs with these compared to Vega or 1080 Ti so clearly killer GPU for next 12 months. They outperform 1080 gddr5x easy.

bit OC tweaking and it's great

   GPU1 65C  Sol/s: 501.7  Sol/W: 3.63  Avg: 506.4  I/s: 270.9  Sh/s: 0.033
   GPU0 64C  Sol/s: 506.1  Sol/W: 3.62  Avg: 504.2  I/s: 271.4  Sh/s: 0.039
   GPU2 61C  Sol/s: 505.2  Sol/W: 3.63  Avg: 505.7  I/s: 269.2  Sh/s: 0.035  +
   GPU3 57C  Sol/s: 517.3  Sol/W: 3.63  Avg: 505.9  I/s: 270.9  Sh/s: 0.042
   ========= Sol/s: 2030.3 Sol/W: 3.63  Avg: 2022.3 I/s: 1082.4 Sh/s: 0.149

And for the LONG run these can do 4.9 Sol/W underclocked if you think of running them 4 years. So their ROI + profit is clearly beyond xxx%
Build a rig and run it forever, win. Nothing beats their efficiency now for future 12 months. They run most efficient sol /w there is.

[wacko]

I have been able to push it to 3.9 Sol/W but now looking most stable setup. It's 505-520 Sol. Best bang for the $ anyway. Because they run Wayyy!! cooler than Vega or 1080 TI no competition. You can dismiss cooling needs with these compared to Vega or 1080 Ti so clearly killer GPU for next 12 months. They outperform 1080 gddr5x easy.

I wouldn't say they outperform 1080. They are good of course, but nothing special for equihash. I took this screenshot when I was testing one of my new 1080s, it was in a test rig and it's not even the best it could do (pushed it a bit further later on, although not very much):



So 1070 Ti is just what it is, a good card between the regular 1070 and the 1080. For ETH and Neoscrypt it's great (although there were initially problems with the latter, not sure whether they were fixed). For everything else it's just fine.

[QuintLeo]

A well dialed in 1080 can pretty close match a 1070 ti on efficiency - but it's also going to do about the same hashrate when it does so (WELL under 5% difference and commonly 1-2%) - which is NOT worth the ballpark 10% higher price on comparable models for no better efficiency and such a tiny gain on hashrate.

[lunobird]

A well dialed in 1080 can pretty close match a 1070 ti on efficiency - but it's also going to do about the same hashrate when it does so (WELL under 5% difference and commonly 1-2%) - which is NOT worth the ballpark 10% higher price on comparable models for no better efficiency and such a tiny gain on hashrate.

Yep I agree,  I got a 1080 msi gaming x for fun to test it against the 1070 ti msi gaming.  At around 75 tdp for both the 1080 only pushes 20 more  sol hashes more than the 1070 ti.  However, the price tag difference for 1080 msi gaming vs 1070 ti gaming was quite a big gap still for not much more hash power at efficient settings.

$460 after rebate
https://www.amazon.com/MSI-GTX-1070-TI-GAMING/dp/B076Q6GYKY/ref=sr_1_2?ie=UTF8&qid=1511562480&sr=8-2&keywords=1070+ti++msi+gaming


$515 after rebate
https://www.amazon.com/MSI-GTX-1080-GAMING-8G/dp/B01GLYD7MG/ref=sr_1_3?ie=UTF8&qid=1511562446&sr=8-3&keywords=1080+++msi+gaming