I have been looking for a test on a GTX 750 or GTX 750 Ti to see how they fair as these two boards are the first two Maxwell cards from nVidia. I agree with your assumptions if cost is the major option for a new build (quad core class CPU).
Please look at one other often overlooked result of the difference the GPU can make. This complex 7-layer timeline does have a 4K clip which of course does require scaling. If you were to run these same tests without GPU acceleration I would guess it might take ~2000 seconds
GT 730 GDDR5:
H.264 Blu-ray: 366 sec
GTX 750 Ti:
H.264 Blu-ray: 135 sec
I have added the results with MPE Off for the H.264 Blu-ray test.
MPE Off: 1000 sec
GT 730 GDDR5: 366 sec
GTX 750 Ti: 135 sec
I forgot to mention that the particular GT 730 I tested is a Kepler GPU (GK208). The same GPU at the same clock speed as the recent GT 630 v2 of a year ago. Only that this GT 730 has 64-bit GDDR5 memory with a throughput of 40.1 GB/sec rather than the 14.4 GB/sec throughput of the GT 630 v2's 64-bit DDR3 memory. DDR3 versions of the GT 730 have either the same GK208 GPU plus 64-bit DDR3 memory with a throughput of a paltry 11.2 to 14.4 GB/sec, or the ancient Fermi-architecture GF108 GPU (the same GPU at the same clock speeds as the lousy GT 430 from 2010) plus 128-bit DDR3 memory with a throughput of 22.4 to 28.8 GB/sec. In other words, the DDR3 versions of the GT 730 are either a rebadged GT 630 v2 or a rebadged GT 430. Both of those GPUs, had I tested them, would have produced MPEG-2 DVD results of 250 to 300 seconds and H.264 Blu-ray results of 500- or 600-ish seconds. Yet they cost about as much money as (if not more than) my tested GT 730 GDDR5 card.
An update on my GPU testing:
It looks like both my main i7-4790K rig and the older i5-2400 rig (I have not yet sold it) are getting GPU upgrades. My cheapo GT 730 will be relegated to troubleshooting duty. The i5-2400 will be getting the GTX 750 Ti that's currently in my main rig while my main rig will be getting either a GTX 960 or a GTX 970. Unfortunately, the GTX 960 with 4GB of VRAM is currently priced too close to the GTX 970 for its own good. Therefore, the GTX 960 will have only 2GB of VRAM.
Once I get the new card in my main PC, I will test and post the results in this thread.
Now that this is said, Bill, which of the two do you recommend given my CPU platform (Haswell with Z97 chipset)? My current workflow is primarily 1080i/p material at up to 59.94 fps, then exported to Blu-ray and/or DVD. However, in the end I felt that the GTX 750 Ti was a downgrade from the GTX 660 that I had to sell when each card was used during the PPBM8 encoding test from the MPEG-2 DVD timeline. In the encoding test from the H.264 Blu-ray timeline, the two GPUs were equal in performance.
Or, given the platform, should I stay put for the GPU in my main PC and just upgrade the one in the i5to something like a GT 740?
Yet another update:
In late 2015 I gave the i5-2400 with the GTX 750 Ti to my brother. He then swapped out the GTX 750 Ti for a GTX 650 that was in his i7-4770 Dell, and then put the 750 Ti into that Dell.
The i7-4790K in my still current main PC has he GTX 970. I have since built an i5-6500 mini PC with (initially) a GTX 960, which was later swapped for its current GTX 1060. Given that, I may move that GPU into my main PC, and then put in a GTX 1050 Ti (I have not yet purchased one) into the mini PC. This is so that the mini PC will have 4GB of VRAM for any future video projects.
By the way, if one needs a cheap GPU with more than 2GB of VRAM, DO NOT BUY a GT 730 with 4GB of VRAM! The commonly available model from eVGA is based on the same craptacular GF108 (aka GT 430) GPU from 2010, and thus has only 96 CUDA cores and even slower performance than even the integrated Skylake or Kaby Lake IGP on Intel's LGA 1151 CPUs. And to top it off, it sells at retail for a whopping $90. A total rip-off unless one has an older mainstream PC that requires a discrete GPU just to even be usable at all whatsoever.