Mobile GTX1050 slow, or are this normal rendertimes?

Hi all,

I have been using blender for quite some time, but only for hobby ‘ish’ occasions. I just like to play around in it and do not consider myself as guru, more as a somewhat experienced hobby-guy that has to look into tutorials again every now and then because I forgot how a certain task was done… (sofar the mindset on my skill level)

So, I recently bought a new laptop, and since I still blender every now and then, i wanted one with a GPU so my renderingspeed would improve, compared to my dell i7 intel-graphics laptop

So, clean install windows 10, Lenovo Yoga 730, i7, 16g ram and GTX1050 with 4GB, downloaded blender benchmark from https://opendata.blender.org/

I ran the BMW and classroom test, and compared the CPU vs GPU results.

I rendered the BMW in about 8 minutes on GPU and 11 minutes on CPU (i do have the results saved, but not at my current laptop)

So, I definetly see some improvement there, but I was hoping to get faster render times.

Is the result within expectations with this card, (cant really find a reference) and should I just lower my expectations, or should (could) I achieve higher render speed?

When checking my resources being used, I can see the card is about 20% used in 3d, and the CPU is hardly used, also the ram is not extremely used ( about 3G) and VRAM neither…
(I do realize its not a TOP-spec’d card and its a laptop, so 1 minute would not be realistic :slight_smile: )

Thanks in advance for enlightening me!

Time for the BMW seems reasonable to me. Haven’t rendered for quite some time on my 1050ti but think to recall, that it was along those lines. Can try it out later today.

Edit: Yes, 5:52.99 on my 1050ti so you’re ok on a normal 1050.
Now I know again why I boughty Vega 2:14.05.

On my laptop with a 6GB 1060, I get 04:09 for the bmw in GPU mode and 11:06 in CPU mode so the same as your CPU time and looks like the 1050 is about half the speed of a 1060.

Note that the benchmark is based on an 8 month old version of 2.80. If you get a current 2.80 and enable both CPU and GPU in Preferences->System->CUDA Devices, then in my case the GPU Compute (which will use both CPU and GPU) time drops to 2:44.

And if I then change it from 32 to 10 samples and enable denoising, I get an image that looks just as good in 0:21 (almost 12 times faster than the original GPU only render), so it’s really just a matter of making he best use of what you have!

Thanks for replying! Seems I just had set my expectations too high :slight_smile:
Well, maybe next time just start saving for an eGPU with lots of horsepower !