The New Nvidia GeForce GTX 750 Ti Scrypt Mining Performance

18 Feb


Nvidia has just announced a new GPU architecture called Maxwell and their first GPU to use it – GeForce GTX 750 Ti. We’ve had a card for testing and decided to give it a try and see how it will perform for Scrypt mining as normally Nvidia GPUs are not providing as much hashrate as AMD graphics cards for crypto mining. The next Maxwell architecture from Nvidia however is interesting, because it is focused on optimization for better performance per watt – something that crypto coin miners are also interested a lot in. The new GeForce GTX 750 Ti GPUs have a power consumption of just about 60W and the price range for these cards is $150 USD. When you add the fact that a GeForce GTX 750 Ti based on the Maxwell architecture without overclock makes about 265 KH/s using CUDAminer these become an interesting solution for crypto miners.

CUDAminer startup parameters for 265 KH/s:

cudaminer.exe -o stratum+tcp:// -u yourworker.cuda -p password -i 0 -l T5x24 -C 1


Furthermore after overclocking the GeForce GTX 750 Ti you can get close up to 300 KH/s hashrate out of the card without problems. There is a chance if the card allows higher overclocking frequencies and higher voltage operation as well as if it has external power and you can increase the power limiter to over 100% to get even higher performance out of it. Furthermore have in mind that CUDAminer is hot yet optimized to support the new Maxwell architecture and the hashrate we are getting out of it is with the kernel for the older Kepler architecture that is used for GTX 780 cards. So there is a chance for even higher performance with a specially optimized CUDAminer for the new Maxwell architecture…

Other Similar Publications:

28 Responses to The New Nvidia GeForce GTX 750 Ti Scrypt Mining Performance


February 18th, 2014 at 19:15

What driver version did you use? There’s a new one (334.89) out today for Maxwell. Also thanks for the results, good info!


February 18th, 2014 at 21:17

The driver version we used was 334.69, though we just also tried with the new 334.89 whql and the results are pretty much the same in terms of performance.


February 18th, 2014 at 21:40

Did you try the card with power pci-e 16x risers? Ones with molex / sata connectors?


February 18th, 2014 at 23:11

No, we do not have an extender with additional power connector. The x16 PCI-E extender was without additional power connector and the card had issues working with it properly. Using x1 extenders without external power is not recommended at all as they limited to 25W of power as opposed to x16 that supports up to 75W.


February 26th, 2014 at 01:48

Can I use MSI Nvidia GeForce GTX 750 Ti Classic Video Card – 2GBDR5 OC, 5400MHz? and what about multiple cards?


February 26th, 2014 at 21:11

We haven’t tested the MSI model, but we see no reason why it will not work and perform similar to the reference model.


February 28th, 2014 at 02:28

what version of nvidia drivers did you use?


February 28th, 2014 at 18:10

The we have used both 334.69 and the newer 334.89 WHQL driver and no notable difference between them in terms of performance.


March 2nd, 2014 at 15:57

I have the msi gtx 750 ti and I am having a trouble keeping it stable around 300ish… even at lower clocks I can’t get it stable and I am wondering if there is something silly I am missing. After like 30 min it gets unstable but temps fine and eveything in afterburner doesn’t look strange not too sure what is up. Using the same switches as you do in the bat file, they do seem to work the best.
Did you have trouble installing cuda toolkit 5.5? Was wondering if it had something to do with that? I even raised pcie voltage in bios to max. Tried the 2 most recent versions of cudaminer.

In afterburner did you select or not select the following:

unlock volatage control
unlock voltage monitoring
fouce constant voltage

Spent a ton of time messing with this and learned alot but getting kinda old fast hehe

Great article btw thank you.


March 2nd, 2014 at 16:19

soulrain, have you tried lowering a bit the video memory frequency? Our experience show that if the video card is unstable after some time of mining overclocked, going 5-10 MHz down on the video memory could make it stable for 24/7 mining.

Also are you using a modified BIOS to increase the maximum power target limit to 65W, this can also help ad maybe even unlock some extra performance if your card is hitting the 100% power target all the time.

A guide for modifying the BIOS is available here:

We’ve seen a boost with up to about 320 KHS after flashing a modified video BIOS wit increased power limit on a card with the stock BIOS hitting around 300 KHS, so you might want to try that as well.


March 3rd, 2014 at 06:56

Ya using modified bios even tested to see if the dual bios switch of the gfx card works which it does pretty neat (just takes a restart to take effect).

Basically with any overclocking after an hour I get cuda scrolling errors I have no idea why. Tried varius nvidia drivers, went from win_7 to win_8 (couldn’t get xp installed).

Here is my rig:

3 x MSI N750TI-2GD5/OC GeForce GTX 750 Ti 2GB 128-Bit
1 x Cooler Master HAF XB EVO – High Air Flow Test Benc
1 x CORSAIR HX Series HX1050 1050W ATX12V / EPS12V SLI
1 x ASRock 970 EXTREME4 AM3+ AMD 970 + SB950 SATA 6Gb/
1 x SAMSUNG 840 EVO MZ-7TE120BW 2.5″ 120GB SATA III TL
1 x CORSAIR XMS 4GB 240-Pin DDR3 SDRAM DDR3 1333 Deskt
1 x AMD Sempron 145 Sargas 2.8GHz Socket AM3 45W Singl

I have no idea over the course of three days have basically exhausted every conceivable combination of afterburner settings possible to fix this.

Ya the link you posted is exactly what I used to alter and flash my bios. I confirmed it stuck with nvflash -b blah.rom then read with kepler to make sure was what I wanted.

I did have trouble installing Cuda toolkit as the Hardware ID was not in the display.inf of Cuda toolkit 5.5 so I had to add it.

I am kinda new to mining but not to comps so not sure if this is crucial.

Thanks for the reply and once again great article and if you can think of anything else let me know!


March 3rd, 2014 at 07:56

the more that I think about this… it sounds temperature related as what really changes over time. Added more case fans and see what that will do. Have some copper after market gpu coolers coming so will see how that works as well.

Not sure if you have taken off the fan connectors on these cards but boy is it hard not to rip off the female connector!


March 3rd, 2014 at 08:58

Threw some fans in now gfx temp is lower. I guess it was hitting 80C in GPU Shark but 80C under load I didn’t think was that bad…

Gonna have it run a bit and see :)


March 3rd, 2014 at 12:08

80 degrees C is quite high for these cards, are they placer too close to each other – you are not using extenders?

By default Nvidia sets temperature target of 80 degrees Celsius, so if the cards get to that temperature they might get throttled (operating frequencies lowered) and that can be the cause of thee instability you are getting.

You can increase the temperature target, however the better solution is to get the cards cooler… this temperature is really high for the 750 Ti.


March 3rd, 2014 at 13:16

Right now I just have 1 card in I am experimenting with it solo thus no risers or other cards at the moment.

Ya 80C was while running MSI Kombuster under full load not sure if that is still unacceptable or not. I ordered aftermarket cooling anyway so see what that does.

The weird thing is this: Why would I run stable under Kombuster for like 1hr than switch over to cudaminer and crash after a few minutes?

This is a pretty weird issue but I will focus on getting temps lower but now while mining with a moderate O/C (just +20 Core Clock) temps are hovering around 60C…

This doesn’t seem right and I have a spacious case and lots of fans.

Also GPU Shark constantly says:

Limiting policies (NVIDIA):
– GPU overvoltage limit reached

Anyway back to investigating.

Thanks for your insight and guidance.


March 3rd, 2014 at 18:18


So I swapped the gtx 750 ti into another one of my boxes and it is working flawlessly.

On the one hand that’s great but on the other not sure what is wrong with the rig I pulled it out of…

I am not sure if this is a conflict or something but the card just seemed unstable.

Can you think of anything I should check or perhaps the mobo/psu/ or God knows what just didn’t want to play well together.

Once again thanks for all the keen insight.


March 3rd, 2014 at 22:23

Another update so 3 eva gtx 750 ti’s came today and the extra power connector in addition to the pcie seems to do wonders.

The msi ones without additional connector I could not even get stable in another box no idea why. Most likely will just run them stock but guess time will tell.


March 4th, 2014 at 09:22

I’d invest money to create a rig like this but not sure about results in mining?!
I wonder if performance will be at least the same with 4 x AMD 280x rig I have now…
Have you made some tests and comparison?
What board is the best for this GPUs?


March 8th, 2014 at 03:45

i have buy pci x1 to x16 riser for 750ti cards wich have and 6pins power connector , i must buy a riser with extra power conector or not ?
i true that riser drop the hashrate ? i can do something about this ?

Robert Adger-Barton

March 8th, 2014 at 04:40

Great article sir! I just bought a Geforce GTX 750 TI, and popped it into an old Dell computer I had lying around. It worked flawlessly out of the box on install and I was getting around 250k hash. This was on a fresh install of Win 7 Pro x86. I installed the Afterburner program becasues I wanted to monitor my CPU temperature, and I didn’t change anything, but now I’m lucky to get 30k hash. I’ve tried the coniguration script you have: -i 0 -l T5x24 -C 1, but it didn’t change anything.

Any thoughts on why this is happening?


March 8th, 2014 at 22:22

Teodor, for what performance you can expect from a 6-card 750 TI mining rig check this:

alex, if the cards have the extra 6-pin PCI-E power connector there should be no need for powered risers, if the cards have no PCI-E power connector on them you will need to use powered risers.

Robert Adger-Barton, if the cudaminer gives you an error and crashes you might need to restart the system in order to get proper hashrate again. You can also try to run cudaminer with autotune instead of passing it the T5x24 kernel to see what results you will get.

David K

March 20th, 2014 at 05:45

Which version of CUDA SDK are you using?


March 20th, 2014 at 16:59

David K, the version we are using is 5.5.


March 25th, 2014 at 09:55

Just a tip

Use H81 intel motherboards chipset and an Intel I3 just like this one

MSI gtx 750 Ti 1gb/2gb or any GTX 750 will be stable at 320 khps. There is also an driver update for Maxwell chips with unlock oc potential

Robert Adger-Barton

March 25th, 2014 at 18:05


How do you run cudaminer with autotune?


March 25th, 2014 at 22:01

Robert, if you don’t specify a kernel in the command line the autotune feature is automatically activated.


January 10th, 2015 at 16:00

Hi I have just bought two geforce gtx 750ti cards to mine with and I am only seeing a hash rate of around 180 kh/s mining litecoins. I have them on an older motherboard and I was wondering if this would slow them down as I was expecting around 250 kh/s?


January 12th, 2015 at 20:18

Oracle1, it is a bit low, but there is no longer any point in mining litecoins with that video card, you should go for something like X11 or X13 mining for example with the latest ccMiner.

Leave a Reply

Your email address will not be published. Required fields are marked *