New Posts  All Forums:Forum Nav:

GTX 750 Ti mining - Page 62

post #611 of 1291
Hope this helps anyone thinking of going with 750 Ti's

Alright changed PSU from an EVGA 650w Gold to a Seasonic 760w Platinum, and have got rig up and running flawlessly.

Yes the Seasonic is a bit overkill @ £130 but the only other option was the Corsair RM750 / RM650 Gold, but it had terrible reviews and forums are littered with complaints. I know a few folks on here are using them without issue though.

XFX PSU's should be fine too, basically you want / need single rail PSU's. All Seasonics, most Corsair, XFX and a few others are single rail. Why single rail is needed, because you will be drawing (6 card rig) 6x 5Amp on the 12V rail = 30Amp+ @ around 360/400watts. On multi rail PSU's your Molex/SATA rails simply cant manage that amperage for example like my EVGA 650w Gold can only supply a Maximum of 20Amp per rail (4 rails) max of 53Amp combined.

Further more I went for a 760w Unit because from reading reviews the maximum efficiency on these units / most units is around the 50%-60% load mark, so out of curiosity wanted to see how much energy I would save compared to running at 70% capacity on the EVGA.

Here are the specs + actual costs

Asrock H81 Pro BTC @ £55
Intel G3220 @ £44
8GB GSkill DDR3 @ £60
6 x USB Risers @ £78 (£13 each)
120GB Crucial M5 SSD @ £55
Seasonic 760w Platinum @ £133
6 x MSI 750Ti Gamings @ £702 (£117) (got some ingame voucher which I should be able to sell)
Frame - £75 (can omit as isn't crucial)

Total = £1202

So some pics of the rig, had a frame spare lying around so thought might as well use, if I didn't wouldn't have bothered as just an extra expense.

Warning: Spoiler! (Click to show)
007.jpg

003.jpg

002.jpg

004.jpg

005.jpg

Performance Scrypt

Only had 24 hours to test, so need to do some more. This particular setting crashed after about 4 hours so needs some little tweaking but you get a rough idea. Stock bios

Settings

+175 Core
+450 Mem
Temps - 45c-55c (across all cards 55c being max)
Fans - 35%
Power Draw at wall - 440w
Hash - 1760kh/s (293kh/s each)

Warning: Spoiler! (Click to show)
Untitled-4.jpg

001-1.jpg

As I mention above in regards to choosing this Platinum PSU, it is pulling only 440w at the wall maximum @ 58% load. The EVGA 650w Gold was pulling around 455-460w at the wall @ 70% load. So a minimum 15w saving, lol isn't much but curiosity answered.

Performance Scrypt Jane

Now this setting is rock solid 24hr stable, run over night. This some wiggle room to improve as you can see the OC is lower than the above scrypt settings, so somewhere in between will be right. Stock bios

+150 Core
+400 Mem
Temps - 45c-55c (across all cards 55c being max)
Fans - 35%
Power Draw at wall - 430w
Hash - 375kh/s (62kh/s each)

Warning: Spoiler! (Click to show)
This nets around 53 UTC per day and will take around 8 months to break even. Although UTC is undervalued at the moment, I reakon could get that down to around 6months mining more profitable coins and on Betarigs once mux sorts it out to work with cudaminer.

Prof.jpg

Anyone fancy doing something similar for AMD cards, 270x etc to compare? Whislt 6-8months is not great (280x is around 4-5months) one thing worth mentioning is this rig is completely silent and outputs very very little heat.
post #612 of 1291
Ultracoin diff is on the rise, now up to 7.4. So those prices have to be updated already tongue.gif I'm still on the Ghash's Litecoin pool for the bonuses, and when that's over who knows what'll be the best coin to mine.
post #613 of 1291
Very good info craterloads just posted... +1.

The single rail is really something important to care.
My desktop PSU has 500W but divided in 2 rails.

Quote:
Originally Posted by ccRicers View Post

Ultracoin diff is on the rise, now up to 7.4. So those prices have to be updated already tongue.gif I'm still on the Ghash's Litecoin pool for the bonuses, and when that's over who knows what'll be the best coin to mine.

I've stayed with Ghash while bonus was x2.
Then I spent sometime with UTC.
Then VTC (Better than UTC by now IMO).
... and now I'll stick with AUR 'til its airdrop (looks promissing).



After that I shall return to VTC.
post #614 of 1291
Thread Starter 
ASUS cards have memory cooling issues because the bottom of the heatsink has no airflow and the memory doesn't get contact with it:
http://www.overclock.net/t/1469814/nvidia-gtx-750ti-750-maxwell-owners-club/220#post_21961757
Quote:
After the second TIM application .... +400 mem is now stable in cudaminer. Have not tested higher.
There is a consequence. My gpu temperature increased by 3C, partly due to higher performance in cudaminer, but perhaps also due to added memory heat.
Workstation stuff
(407 photos)
SpecViewperf 12.0.1
(158 photos)
 
Reply
Workstation stuff
(407 photos)
SpecViewperf 12.0.1
(158 photos)
 
Reply
post #615 of 1291
Is it better to run:

A. One instance of all the gpus hashing to same worker?

B. Multiple instance of all gpus hashing to same worker?

C. Multiple instance of all gpus hashing to different workers?
Dragonborn
(14 items)
 
  
CPUMotherboardGraphicsRAM
Phenom II 955 Asus M4A79XTD EVO MSI GTX 460 Hawk G.Skill DDR3 1600 2x2GB 
Hard DriveOSOSMonitor
Seagate 1TB 5400RPM Windows 7 Ubuntu 10.04 ASUS VH236H 
KeyboardPowerCaseMouse
Microsoft French Keyboard OCZ Fatal1ty 550W Silverstone PS04-B Logitech G5 
  hide details  
Reply
Dragonborn
(14 items)
 
  
CPUMotherboardGraphicsRAM
Phenom II 955 Asus M4A79XTD EVO MSI GTX 460 Hawk G.Skill DDR3 1600 2x2GB 
Hard DriveOSOSMonitor
Seagate 1TB 5400RPM Windows 7 Ubuntu 10.04 ASUS VH236H 
KeyboardPowerCaseMouse
Microsoft French Keyboard OCZ Fatal1ty 550W Silverstone PS04-B Logitech G5 
  hide details  
Reply
post #616 of 1291
Quote:
Originally Posted by dotaduck View Post

Is it better to run:

A. One instance of all the gpus hashing to same worker?

B. Multiple instance of all gpus hashing to same worker?

C. Multiple instance of all gpus hashing to different workers?

Same thing...mine with all cards on the same instance, pointing to one worker per rig if you want some basic online monitoring. Otherwise mine with a single worker.
post #617 of 1291
Quote:
Originally Posted by dotaduck View Post

Is it better to run:

A. One instance of all the gpus hashing to same worker?

B. Multiple instance of all gpus hashing to same worker?

C. Multiple instance of all gpus hashing to different workers?

I mine with an instant and worker per a GPU because if I have a crash while I am away only one GPU goes down, it allows me to see the performance of each gpu much easier when Im both at home and away and it also allows me to tweak and tune individual gpus while the others continue mining.
post #618 of 1291
Im currently mining with 5 x Gigabyte.

Here's my set-up

MSI Z77 MPower
3570k
700W Silverstone PSU
3 x GTX 750ti on the mobo
2 x GTX 750ti on 16x to 1x rises (USB/powered)

OC setting +180Mhz on core and +700Mhz on memory. 0,0125 overvoltage. (temp 38C to 58C at 100%)
(OC setting are not optimal yet...)

I'm getting 290khash/s with the riser and 305 on the mobo.

Cudaminer settings I'm not sure yet... The only settings that seems to really do a difference is -l T5x24 and -H 1
Edited by Poena - 3/19/14 at 7:48am
post #619 of 1291
Quote:
Originally Posted by Americonfusion View Post

I mine with an instant and worker per a GPU because if I have a crash while I am away only one GPU goes down, it allows me to see the performance of each gpu much easier when Im both at home and away and it also allows me to tweak and tune individual gpus while the others continue mining.

Good point... never thought that.
Have you noticed some difference on the rate if comparing one cudaminer for all with one for each?

Quote:
Originally Posted by Poena View Post

Im currently mining with 5 x Gigabyte.

Here's my set-up

MSI Z77 MPower
3570k
700W Silverstone PSU
3 x GTX 750ti on the mobo
2 x GTX 750ti on 16x to 1x rises (USB/powered)

OC setting +180Mhz on core and +700Mhz on memory. 0,0125 overvoltage. (temp 38C to 58C at 100%)
(OC setting are not optimal yet...)

I'm getting 290khash/s with the riser and 305 on the mobo.

Cudaminer settings I'm not sure yet... The only settings that seems to really do a difference is -l T5x24 and -H 1

- H 1 gives a good boost because uses CPU also.
post #620 of 1291
Quote:
Originally Posted by Americonfusion View Post

I mine with an instant and worker per a GPU because if I have a crash while I am away only one GPU goes down, it allows me to see the performance of each gpu much easier when Im both at home and away and it also allows me to tweak and tune individual gpus while the others continue mining.
Quote:
Originally Posted by RavenXBR View Post

Good point... never thought that.
Have you noticed some difference on the rate if comparing one cudaminer for all with one for each?
- H 1 gives a good boost because uses CPU also.

When I combined all my cards into one worker, I was getting on average 100 Kh/s to the pool than I would be having one separate worker for each card. I do this as well, to check on all the cards individually in case one of the card's drivers crash and Cudaminer starts throwing errors.

But sometimes even TeamViewer becomes unresponsive when one card crashes, no matter which one, and I have to reboot the whole thing. Never had to deal with that in Linux/SSH tongue.gif
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Distributed Computing - General