Post
Topic
Board Mining (Altcoins)
Re: [ANN] cudaMiner - a new litecoin mining application [Windows/Linux]
by
madman1980
on 01/03/2014, 07:38:31 UTC
Code:
[quote author=cvax link=topic=167229.msg5441080#msg5441080 date=1393657934]
[quote author=madman1980 link=topic=167229.msg5440626#msg5440626 date=1393654990]
[quote author=cvax link=topic=167229.msg5439853#msg5439853 date=1393650682]
[quote author=madman1980 link=topic=167229.msg5439718#msg5439718 date=1393649926]
Hey guys,
  Am new here, and new to mining with Nvidia cards.  I've been following this thread for the past week, finally got my hands on a few of the 750Ti cards last night.  Picked up 3 different brands, but have only had a chance to tinker with the MSI twin frozr (no pcie plug) so far.  I am definitely no expert, and am learning cudaminer as I go, but wanted to share my experience so far:

  I've only run one card at a time so far.  Biostar H81S2 mobo, 4gb ram 750w psu.   and mostly using auto-tune, as I dont understand the blocks and warps yet..
System idle = 42.2W (per killawatt meter)
gpu = +0, mem= +0, load @99% = 114W.  hashrate 268 kh/s T5x19
gpu = +0, mem +300, hashrate 252kh/s (T230x1.  this may have been a bad setting by auto-tune, i'm not sure.  i initally was trying gpu and mem clocks others were reporting, and was getting alot of instant artifacting, or 'rapid scrolling' error. so i left it at that, and moved on to doing the TDP raising bios mod.

  With TDP bios mod:

gpu = +0, mem= +0, hashrate 272khs, T10x24, gpu utilization 75% (?), and system load increased to 124W (from 114W)
gpu = +0, mem= +300, hashrate 291khs, T10x24, system load = 130Watts.

  This last setting was stable for 6+ hrs while I was at work today.  I have tried, briefly, with higher gpu and mem clocks, and highest i've hit was ~322khs. but i need to keep testing for stability, as i was getting errors along the way.  Interestingly, the bios mod has definitely increased power draw, and has allowed near 300khs with a modest +300 mem overclock.  I have seen others on this thread with much higher clocks at this speed.  Christian, are you using a modified bios on your 750 Ti?  I feel there is more there.  Also, after the bios mod, the power % on Precision X drops to 75%.  I dont understand this, or if it may be reporting it incorrectly.
   I'll continue seeing how far i can push this card.  I am concerned about the scalability issues others are having, and will assess this as well, when i have time to add other cards (hopefully tonight).  Will keep you guys posted on my findings. 
[/quote]Hi madman,

I believe your findings are in line with mine. I would say try the non-BIOS mod +300 mem overclock again with the same T10x24. My results yielded no additional hashrate with the power mod, but increased power consumption.
[/quote]

Re-ran it on stock bios, +300 mem, T10x24, and at 268-269khs, unchanged from stock clock settings.  Another note, this was on the pci-e 16x slot, no riser.  I just tried the card on 2 different powered riser cables (ribbon and usb), and seeing reduced hash rate ~255khs, and reduced system wattage on the killawatt meter to around 114W, even with the above mem +300 overclock.  ???
[/quote]Believe risers may be expected to reduce some hashing. Did you use the riser in PCI-E 1x slots or 16x to 16x? I am about to try my risers as well. What 69charger said for T5x24 being optimum I can confirm that as I just finished profiling my 8 Zotac GTX 750 Ti's.

[/quote]

  So far, thats the only explanation i can come up with.  i tried both the ribbon and usb 1x to 16x powered risers, both ~30cm in length.  the board im using (biostar h81S2) has one 16x slot and 5x 1x slots.  i used the 16x slot when testing the single card.  Has anyone else noticed reduced performance with risers tested on a SINGLE card??  Currently adding one card at a time, and seeing how they scale.  just finished adding the 4th, and so far, they are scaling well.  255-258khs on bone stock settings.  not sure if i'll get through the 6 card tonight.  gotta work early in the am..