GTX 750 Ti mining - Page 14 - Overclock.net - An Overclocking Community

Forum Jump: 

GTX 750 Ti mining

Reply
 
Thread Tools
post #131 of 1291 (permalink) Old 03-02-2014, 02:57 PM
New to Overclock.net
 
HardwareDecoder's Avatar
 
Join Date: Apr 2012
Posts: 4,514
Rep: 142 (Unique: 106)
Quote:
Originally Posted by xartic1 View Post

"-H 1" Won't that have some affect on the CPU power consumption?

Yep apparently he has a wrong setting:


--hash-parallel [-H] scrypt also has a small SHA256 or Keccak component:
0 hashes this single threaded on the CPU.
1 to enable multithreaded hashing on the CPU.
2 offloads everything to the GPU (default)

HardwareDecoder is offline  
Sponsored Links
Advertisement
 
post #132 of 1291 (permalink) Old 03-02-2014, 02:57 PM
New to Overclock.net
 
dph314's Avatar
 
Join Date: Jun 2011
Location: Gaming...or waiting to
Posts: 3,593
Rep: 114 (Unique: 92)
Quote:
Originally Posted by AlphaC View Post

Quote:
Originally Posted by dph314 View Post

Why 'don't get EVGA FTW'? I bought 2 of them, they seem pretty nice so far. Do Asus/MSI get better hashrates?
snip (Click to show)
The power components used are less efficient / get hotter (thermal wall when overclocking) and the whole reason you get a Maxwell card is efficiency (whether it is GFLOPs/W , hashrate/W , FPS/W , etc.)

D-paks and other older designs are less efficient than newer DirectFET (from International Rectifier) used on EVGA Classified / MSI Lightning / R9 290 series and R9 280 series reference cards, Fairchild Semiconductor FDMF6823A used on GK110 which is an integrated driver+high side+low side (similar to International Rectifier PowIRStage), Power QFN / PQFN or " Power Quad Flat No Leads" (Fairchild Semiconductor / International Rectifier) , etc.

Simpler explanation: A mosfet ("metal–oxide–semiconductor field-effect transistor") is an electrical switch. You know those lightbulb dimmer switch experiments in elementary school where you have a switch you change the resistance of? Like that but without someone actually moving the switch. Low RDS(on) means low resistance "drain to source". Basically the lower the resistance when it is on, the less power loss you (power is lost to heat). The second issue is thermal resistance : lower thermal resistance = better conductor of heat , so it doesn't get as hot since it will conduct heat to the circuit board or ideally the heatsinks.

I give them the benefit of the doubt (EVGA warranty + stepup / etc.) but reference design uses a Magnachip MD1502 & MDD1501 which the reference datasheet says is a D-PAK. http://www.magnachip.com/application/down.php?filename=MDD1502.pdf , http://www.magnachip.com/application/down.php?filename=MDD1501.pdf

It's clearly used for memory and PLL.

D-Paks were known for blowing on AMD 800 series motherboards and AMD A75 Asrock boards. Of course, the power draw of memory/PLL is much less , maybe 10-15W or so.

Thank you, nice explanation. I'll probably only have these a few months though, and you're right, EVGA's warranty service is top-notch. I think I'll be good for the short time I have these, but thanks again for the info.


Quote:
Originally Posted by Freekers View Post

snip (Click to show)
Alright guys, I've played around with one of my Gigabyte 750TIs (GV-N75TOC-2GI) for a few days and here are my findings:
- I'm getting a stable 325 kh/s in cudaminer with an OC of +135 on the core and +635 on the memory. Pushing the memory clock any further would make cudeminer crash.

- I'm using these cudaminer arguments:
-l T5x24 -C 1 -i 0 -H 1
Any other arguments than this would yield a lower hashrate. Even a slight/subltle change such as H1 to H2 would cost me about 20kh/s.

- The chip does not get hotter than 45 degrees celcius with the fan speed at it's lowest setting (56%). It really is whisper quiet!!

- I flashed the BIOS using the instructions and zip file provided by cryptomining-blog.com. Please do not that you need to save/dump and flash the BIOS in DOS, NOT in Windows. It will _not_ work, believe me. Additionally, only use nvflash to dump the bios. GPU-Z did not dump mine properly.

- Using the 32bit version of cudaminer yields 5 kh/s more than the 64 build... Yes you read that right.

- Power usage is as follows:
Complete System Power Usage
Idle without 750TI installed: 32W
Idle with 750TI installed: 37W
Load with 750TI Installed: 115W
Idle with 750TI OC +135/+635: 37W
Load with 750TI OC +135/+635: 125W
These readings were taken from my Kill-A-Watt meter straight out of the wall.

This tells us that the 750TI alone uses:
Idle: 5W
Load: 83W
OC Idle: 5W
OC Load: 93W

I.e. a system with 6 of these cards overclocked will use 590W while mining and yields 1950 kh/s.

I'm still awaiting my Powered Risers so these results are based upon just 1 GV-N75TOC-2GI plugged straight into the 16x slot of the mainboard.
Full System Specs:
Intel Celeron G1820
ASRock H81 Pro BTC
4GB DDR3
250GB HDD
be quiet! System Power 7 700W
Windows 7 Ultimate 64-Bit
Driver Version: 334.89

I hope this helps.

Kind regards,
Freekers.

Great info. I'll try playing around with mine some more, see how it goes. I'm getting just over 290kh/s on each of mine using the default config.

Quote:
Originally Posted by maarten12100 
Quote:
Originally Posted by dph314 
Is it weird that I'm slightly aroused?
No, it isn't. Epeen is rising!


dph314 is offline  
post #133 of 1291 (permalink) Old 03-02-2014, 03:05 PM
New to Overclock.net
 
dotaduck's Avatar
 
Join Date: Oct 2011
Posts: 74
Rep: 6 (Unique: 6)
Quote:
Originally Posted by ivanlabrie View Post

I've seen Zotac cards oced do better than that...they are pretty darn good.
Quote:
Originally Posted by cam51037 View Post

From what I've seen from Zotac cards you should be happy and expect around 295KH/s per card.

Thank you both! I will wait patiently for these cards to arrive :3

dotaduck is offline  
Sponsored Links
Advertisement
 
post #134 of 1291 (permalink) Old 03-02-2014, 03:14 PM
New to Overclock.net
 
dph314's Avatar
 
Join Date: Jun 2011
Location: Gaming...or waiting to
Posts: 3,593
Rep: 114 (Unique: 92)
I broke 600kh/s at only 1343/2950 with this config. Brought me up from 580kh/s at slightly higher clocks. Will keep testing. This is the x86 version of cudaminer also-
Quote:
cudaminer.exe -o pool:port -u user.user -p password -d 1,2 -l T5x24,T5x24 -C 1,1 -H 1,1
pause

Was one of the configs posted a few pages back, just thought I'd mention it again since it worked quite well. Actually I think it was from another thread maybe. All these mining threads are blending together redface.gif

(Edit: Don't forget to change the -d flag if your setup is different, and the other flags if you don't have two cards.)

Quote:
Originally Posted by maarten12100 
Quote:
Originally Posted by dph314 
Is it weird that I'm slightly aroused?
No, it isn't. Epeen is rising!


dph314 is offline  
post #135 of 1291 (permalink) Old 03-02-2014, 03:52 PM
New to Overclock.net
 
gfgrimm's Avatar
 
Join Date: Mar 2014
Posts: 37
Rep: 0
Quote:
Originally Posted by Freekers View Post

Alright guys, I've played around with one of my Gigabyte 750TIs (GV-N75TOC-2GI) for a few days and here are my findings:
- I'm getting a stable 325 kh/s in cudaminer with an OC of +135 on the core and +635 on the memory. Pushing the memory clock any further would make cudeminer crash.

- I'm using these cudaminer arguments:
-l T5x24 -C 1 -i 0 -H 1
Any other arguments than this would yield a lower hashrate. Even a slight/subltle change such as H1 to H2 would cost me about 20kh/s.

- The chip does not get hotter than 45 degrees celcius with the fan speed at it's lowest setting (56%). It really is whisper quiet!!

- I flashed the BIOS using the instructions and zip file provided by cryptomining-blog.com. Please do not that you need to save/dump and flash the BIOS in DOS, NOT in Windows. It will _not_ work, believe me. Additionally, only use nvflash to dump the bios. GPU-Z did not dump mine properly.

- Using the 32bit version of cudaminer yields 5 kh/s more than the 64 build... Yes you read that right.

- Power usage is as follows:
Complete System Power Usage
Idle without 750TI installed: 32W
Idle with 750TI installed: 37W
Load with 750TI Installed: 115W
Idle with 750TI OC +135/+635: 37W
Load with 750TI OC +135/+635: 125W
These readings were taken from my Kill-A-Watt meter straight out of the wall.

This tells us that the 750TI alone uses:
Idle: 5W
Load: 83W
OC Idle: 5W
OC Load: 93W

I.e. a system with 6 of these cards overclocked will use 590W while mining and yields 1950 kh/s.

I'm still awaiting my Powered Risers so these results are based upon just 1 GV-N75TOC-2GI plugged straight into the 16x slot of the mainboard.
Full System Specs:
Intel Celeron G1820
ASRock H81 Pro BTC
4GB DDR3
250GB HDD
be quiet! System Power 7 700W
Windows 7 Ultimate 64-Bit
Driver Version: 334.89

I hope this helps.

Kind regards,
Freekers.

I built a six-card system via PCIe x1-x16 powered USB risers. I get 525W at the wall running the cards OC'd at full load.

Let's look at this a little more closely--assume my system, w/o GPUs at load dissipates 60W. Also take into account 10% losses estimated from PSU conversion.

That puts me at 70W per OC'd card, which aligns more with other numbers people are posting.

So you can fret about whether or not people are measuring current being delivered by powered riser molex cables + 6-pin PCIe cables, but at the end of the day, you're paying for what is being drawn at the wall.

I guess if you're concerned with heat dissipation, then the 70W figure matters more, but I doubt any of you are thinking about re-designing a 750 Ti heat sink assembly.

Finally, my best has rate with a single card is 325kH/s, but my six-card scaled rate is 1630 kH/s. It took some work to get it this good, and I had to use Ubuntu x64 over WIndows 8.1 x64, as I was seeing about 230Kh/s (or ~1400kH/s) for six cards. Cudaminer does NOT scale well with PCIe x1 risers. Hell, it starts taking a hit when you change the mode of a single x16 slot to x8, and the performance hit becomes larger when you reduce to x1 mode. I'm hoping the author can find a way to optimize his PCIe bus BW utilization.
gfgrimm is offline  
post #136 of 1291 (permalink) Old 03-02-2014, 03:55 PM
New to Overclock.net
 
ccRicers's Avatar
 
Join Date: Oct 2013
Location: Chicago, IL
Posts: 2,791
Rep: 72 (Unique: 66)
Quote:
Originally Posted by gfgrimm View Post

I built a six-card system via PCIe x1ox16 powered USB risers. I get 525W at the wall running the cards OC'd at full load.

Let's look at this a little more closely--assume my system, w/o GPUs at load dissipates 60W. Also take into account 10% losses estimated from PSU conversion.

That puts me at 70W per OC'd card, which aligns more with other numbers people are posting.

So you can fret about whether or not people are measuring current being delivered by powered riser molex cables + 6-pin PCIe cables, but at the end of the day, you're paying for what is being drawn at the wall.

I gues if you're concerned with heat dissipation, then the 70W figure matters more, but I doubt any of you are thinking about re-designing a 750 Ti heat sink assembly.

Well, that helps answer some of my questions. I have a EVGA 500B that I plan to use with 4 cards. I wonder how much power is added to the 70w per card for every powered molex riser you use.

Also, how is the noise level in these cards?
ccRicers is offline  
post #137 of 1291 (permalink) Old 03-02-2014, 03:58 PM
New to Overclock.net
 
gfgrimm's Avatar
 
Join Date: Mar 2014
Posts: 37
Rep: 0
Quote:
Originally Posted by ccRicers View Post

Well, that helps answer some of my questions. I have a EVGA 500B that I plan to use with 4 cards. I wonder how much power is added to the 70w per card for every powered molex riser you use.

You may be OK, as the 525W is measured at the wall after switching losses, plus my understanding is that many quality PSU's are over-designed. I would expect to not have the best PSU efficiency though. My experience with DC-DC conversion is that typically the efficiency curves peak around 40-50% of rated load. I'm less versed with AC-DC switching though. smile.gif
gfgrimm is offline  
post #138 of 1291 (permalink) Old 03-02-2014, 04:39 PM
New to Overclock.net
 
HardwareDecoder's Avatar
 
Join Date: Apr 2012
Posts: 4,514
Rep: 142 (Unique: 106)
Quote:
Originally Posted by gfgrimm View Post

I built a six-card system via PCIe x1-x16 powered USB risers. I get 525W at the wall running the cards OC'd at full load.

Let's look at this a little more closely--assume my system, w/o GPUs at load dissipates 60W. Also take into account 10% losses estimated from PSU conversion.

That puts me at 70W per OC'd card, which aligns more with other numbers people are posting.

So you can fret about whether or not people are measuring current being delivered by powered riser molex cables + 6-pin PCIe cables, but at the end of the day, you're paying for what is being drawn at the wall.

I guess if you're concerned with heat dissipation, then the 70W figure matters more, but I doubt any of you are thinking about re-designing a 750 Ti heat sink assembly.

Finally, my best has rate with a single card is 325kH/s, but my six-card scaled rate is 1630 kH/s. It took some work to get it this good, and I had to use Ubuntu x64 over WIndows 8.1 x64, as I was seeing about 230Kh/s (or ~1400kH/s) for six cards. Cudaminer does NOT scale well with PCIe x1 risers. Hell, it starts taking a hit when you change the mode of a single x16 slot to x8, and the performance hit becomes larger when you reduce to x1 mode. I'm hoping the author can find a way to optimize his PCIe bus BW utilization.

how did you get more on linux, I thought you couldn't overclock on linux.

HardwareDecoder is offline  
post #139 of 1291 (permalink) Old 03-02-2014, 04:43 PM
New to Overclock.net
 
gfgrimm's Avatar
 
Join Date: Mar 2014
Posts: 37
Rep: 0
Quote:
Originally Posted by HardwareDecoder View Post

how did you get more on linux, I thought you couldn't overclock on linux.

At stock settings, cards were getting more on Linux, so there is an issue with Cudaminer+WIndows 8.1 x64+running cards at PCIe x1 mode. Whatever issue is, its effect is less on linux for me.

I OC'd the MEM CLK by changing the value in the VBIOS for the cards after figuring out in Windows with MSI AB, what was stable.

If ANYONE could inform me of the best way to adjust the GPU CLK in the VBIOS, that would be wonderful. So far, any attempts to adjust base clock or boost clock results in much fail.
gfgrimm is offline  
post #140 of 1291 (permalink) Old 03-02-2014, 05:23 PM
New to Overclock.net
 
UNOE's Avatar
 
Join Date: Nov 2010
Location: So Cal
Posts: 3,589
Rep: 128 (Unique: 116)
Quote:
Originally Posted by gfgrimm View Post

At stock settings, cards were getting more on Linux, so there is an issue with Cudaminer+WIndows 8.1 x64+running cards at PCIe x1 mode. Whatever issue is, its effect is less on linux for me.

I OC'd the MEM CLK by changing the value in the VBIOS for the cards after figuring out in Windows with MSI AB, what was stable.

If ANYONE could inform me of the best way to adjust the GPU CLK in the VBIOS, that would be wonderful. So far, any attempts to adjust base clock or boost clock results in much fail.

In Windows you have to keep Chrome open for some reason while mining to get cards at 99% load.


UNOE is offline  
Reply

Quick Reply
Message:
Options

Register Now

In order to be able to post messages on the Overclock.net - An Overclocking Community forums, you must first register.
Please enter your desired user name, your email address and other required details in the form below.
User Name:
If you do not want to register, fill this field only and the name will be used as user name for your post.
Password
Please enter a password for your user account. Note that passwords are case-sensitive.
Password:
Confirm Password:
Email Address
Please enter a valid email address for yourself.
Email Address:

Log-in



Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)
 
Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may post new threads
You may post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off