Testing Procedure
Methods
To test the performance of video cards at eTeknix we run a variety of tests at four different resolutions, where supported and appropriate, across a suite of games and benchmarks. We run each of these benchmarks three times to take an average and use the latest WHQL-certified video card drivers from AMD and Nvidia to test with.
To attain Furmark load noise levels we hold a decibel meter approximately 3 inches away from the card after running Furmark for five minutes. To attain 3DMark Load noise we do the same in the last stage of a 3DMark 11 benchmark run. To achieve idle noise we allow the system to stay idle at the desktop for 5 minutes before taking a noise reading.
To achieve Furmark load temperatures we take the maximum temperature of the GPU core during Furmark load as recorded by CPUID HW Monitor. To achieve idle temperatures we take the minimum recorded GPU core temperature in CPUID HW Monitor after desktop-idling for 5 minutes. We also do the same in 3DMark 11 to simulate a more realistic load. We then convert these temperatures into Delta temperatures – that is = actual recorded temperature minus the ambient temperature of the room. Note we do not alter the default fan profile in any way, all graphics cards are left to run at their stock settings.
To measure power consumption we take the maximum power consumption at idle on the desktop over a 5 minute period, and for load we take the maximum stable power consumption during Furmark load. We also do the same during a 3DMark 11 run, taking the maximum stable power consumption from the last combined test only. To address recent developments in thermal throttling (and Furmark throttling more specifically) we take our measurements within the first 30 seconds to a minute on cards that dramatically clock down when hot (e.g. the R9 290X) as this causes there power consumption to appear much lower than it actually can be.
To formulate our performance metric we use a weighted average method and each benchmark or test gets a weight of 1 out of a total calculation with 9 weights (meaning all resolutions and variations of the same test are only ever one weight). Those 9 weights are 1 for each test we do excluding Battlefield 4 (so 6 games and 3 benchmarks) which is omitted as we do not have results for all the graphics cards in that game. We then calculate the relative score for each card in each test (by dividing the cumulative frame rate/score of each card by the average cumulative frame rate/score from all the graphics cards in group). We then take those 9 relative scores and average those for each card, before dividing those averages by the average of the card being reviewed in the test so that the value of the card being reviewed is 1 and every other card’s performance metric is in relation to it. The end result are numbers that vary around 1 with 0.5 being half the performance of the tested card and 2 being double the performance of the tested card. We then convert those numbers into percentages to make them more readable so 1 becomes 100% and 0.5 becomes 50%.
Test System:
- Motherboard – Asus Rampage IV Extreme X79 LGA 2011 (chipset fan disabled)
- Processor – Intel Core i7 3960X at stock clock speeds of 3.3GHz with Turbo Mode disabled.
- RAM – 16GB (2 X 8GB) Corsair Vengeance Pro Series 1866MHz at 9-10-9-27
- CPU Cooler – Corsair H100i with Quiet Fan Profile
- Power Supply – Corsair HX1050W
- Main Storage Drive – Kingston HyperX 240GB SSD over SATA III interface
- Chassis – Lian Li T60 Test Bench
- Displays – Dell U2711 Ultra Sharp for 2560 by 1440 and we use the LG IPS 234 & LG IPS 224 with the Dell U2711 Ultra Sharp to run 5760 by 1080.
- Operating System – Windows 7 Ultimate 64 Bit
We would like to thank Asus, Corsair, Kingston, Lian Li and all our other partners who supplied us with test equipment and hardware. Their generosity makes our testing possible and without them we wouldn’t be able to produce the reviews we do, so thank you!
Games Used
- Aliens Vs Predator
- Dirt Showdown
- Battlefield 4
- Metro Last Light
- Sleeping Dogs
- Bioshock Infinite
- Tomb Raider
Resolutions Used (where possible)
- 1680 x 1050
- 1920 x 1080
- 2560 x 1440
- 5760 x 1080
Hardware Used
- Plug “killawatt” style electricity usage meter
- Benetech GM1351 decibel meter
Software Used
- 3DMark 11
- 3DMark 2013
- CPUID HWMonitor
- Unigine Heaven Benchmark 4.0
- Windows 7 Professional 64-bit
- Furmark 1.10.5
nice card but going from a 650 ti to a 750 ti dont really seem like much of an upgrade for me to pay out $150 i would get a 650 ti boost over this card 😛
if only when the 860 comes out it could be like only $200 but yes i know it wont be 😛
Yeh it’s not meant to be much of an upgrade for GTX 650 Ti Boost users, it is more aimed at people with GTX 5XX series cards or those building new small form factor/low power PCs
It’s about 15-20% quicker than the 650Ti while consuming less power at typically common screen resolutions. If you’re in the market for either card the smart money would be on the 750Ti.
but i have the 650 ti already was just saying it isnt much of an upgrade for me 😛
This is weird! If the 750Ti is faster than the 650Ti, Then how did Nvidia do it? Because both cards memory is the same speed at 86.4 gb/s , Next the 650Ti has 64 TMU’s giving a speed of 59392 Mtexels/sec
vs the 750Ti’s 40 TMUs Giving a speed of 40800 Mtexels/sec, Both cards have 16 ROP’s, Next the 650Ti has 768 cuda cores vs the 750Ti’s 640 cuda cores. So if it is faster I am not seeing how. I saw the benchmarks. Did they use older bench scores vs the recent 750Ti bench scores? If I am not mistaken the 650Ti improved in performance with later driver updates. Something about a V-Sync issue with the older drivers. I think other 600 series cards also had the same issue until the driver update.
If the 750Ti is faster I will buy it, Because I am going to eventually need to get a card with at’least 2GB in the future. And I want a card close in performance and price to what I got, I just hope they are properly comparing them.
Nvidia GPU Cores are not comparable across generations, Nvidia’s GTX 750 Ti is Maxwell while the GTX 650 Ti is Kepler. To use an AMD example You wouldn’t say the HD 6670 that has 480 (VLIW5) cores so why is not faster than the HD 7730/R7 240 which ONLY has 384 (GCN) cores. So why do the same thing with the Nvidia cards here. The GTX 750 Ti is SIGNIFICANTLY faster than the GTX 650 Ti, there is absolutely no disputing that in any way. Any drivers will give you the same results – the GTX 750 TI is at least 10% faster. The GTX 650 Ti BOOST on the other hand is faster than the GTX 750 Ti. So please ensure you do not confuse the two.
Oh and if you need verification from another site please see below:
http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_750_Ti/25.html
I know the difference between the 650 Ti boost and Non Boost 650 Ti, (I do own a non boost after all) However I did finally come to a conclusion, The 750 Ti is slightly faster but not enough to make any difference at all from a regular NON BOOST (so you can read) 650 Ti, Also both cards only have a memory bandwidth of 86.4 GB/sec meaning neither can really utilize 2GB , They can however utilize a little over 1GB but the difference is not enough to matter much. Looks like if anyone wants a cheap upgrade this year from the 650 Ti 1GB they will have to go with the R7 265, It has 2GB @ 179.2 GB/sec that’s enough bandwidth to utilize all 2GB. It stinks having to go to a AMD card because the crap drivers, But for $149 it is the same price as the 750Ti and a much more logical choice.
The R7 265 is a good option but currently it isn’t really available to buy at many places, and in the UK at least it is selling from £125+. The GTX 750 Ti starts at £106. I’d rather pay £106 for a 750 ti than £125 for an R7 265 when the difference in performance is so small.