GPU vs CPU benchmark results - same time = equivalent processing power?

[CityWrecker]

Wondering, if I get the same time result for gpu vs cpu benchmark does that essentially mean they have similar processing power, with regards to Vray?

i.e. no point buying a gpu that gets the same result as my cpu, if I'm hoping for faster renders? Of course, assuming on the same scene.

[oglu]

i see it from a diffrent perspective.
For 1k you get more GPU power than CPU power.
And its easier to fit four GPUs in your system.

[CityWrecker]

Very curious about this....hopefully someone from Chaos can chime in.

[zero-13]

Wondering, if I get the same time result for gpu vs cpu benchmark does that essentially mean they have similar processing power, with regards to Vray?

i.e. no point buying a gpu that gets the same result as my cpu, if I'm hoping for faster renders? Of course, assuming on the same scene.

It really depends what kind of CPU setup you currently have and how many GPUs you plan on upgrading to?

I had a similar situation... I have crappy dual Xeon E5-2620 v3 giving me 24 slow threads, I had two GTX Titan Blacks, and in certain situations I could get V-Ray CPU to outperform the old Titan's, so they were somewhat comparable in rendering or processing power CPU vs GPU.

I just recently upgraded to a 2080 Ti and plan on adding another, but for a rough comparison, this single 2080 Ti is more than 2x faster than my old Titans or my Dual Xeons.

So eventually adding a second 2080 Ti will end up providing me with a huge speed boost compared to what I had going.

If you already have a crazy fast, and densely threaded CPU setup, with plenty of RAM for each thread, you may very well have much faster rendering capabilities than say, even three 2080 TI.

[Alex_M]

It is not always about speed but also feature set. Some things just still aren't possible or good enough with the GPU engine. Besides, these days heavily threaded CPUs like Threadripper can be considered very affordable, especially compared to just a year or two ago when heavily threaded CPUs were really, and I mean REALLY expensive. For example, for the same price as a 2080 Ti you can now buy a 24-core 48-Threaded Threadripper 2970WX which is a beast of a CPU and supports all features of Vray completely. I've ran a few tests on my 16-core Threadripper and my 1080 Ti and most of the time I found that my CPU renders things faster for the same quality and with less fiddling. Especially interior scenes. I might change my mind if someone shows me compelling evidence that GPUs are a better choice.

[danio]

I just went in on two 2080 Ti’s to add to my 1950x and 1080 Ti primary system. I’ll chime in on how it performs when it arrives next week. I’ve been doing a mix of Corona and VRay, VRay mostly in CPU mode, so I’ll be curious how things go.

[CityWrecker]

Guys, appreciate the feedback,but it's not a thread about GPU vs CPU rendering....tradeoffs, benefits etc.


It's one question, about the benchmark and hardware equivalency... if I get the same time result for GPU vs CPU on benchmark does that essentially mean the GPU has similar processing power to the CPU, with regards to Vray?

[M.Max]

Because the benchmark is using 2 totally different scenes for cpu vs gpu I would say no, the same timing doesn't mean you have the same power in both. I would like to hear from Chaosgroup team if this is being considered for future release even tho I think making such thing isn't easy without spreading flase consumptions (over / underestimating) but let's see what they will say about it )

[CityWrecker]

Hoping someone from Chaos can chime in here and comment....

If I get the same time result for GPU vs CPU on benchmark does that essentially mean the GPU has similar processing power to the CPU, with regards to Vray?

[joconnell]

Tomasz from Dabarti thinks that a gpu is worth about 4 cpu's but only if you're building your scenes for gpu to start with - it's a different set of rules about textures, geometry and features but if you build a scene that runs on gpu, it'll be way faster than the same scene on cpu.

[savage309]

Hoping someone from Chaos can chime in here and comment....

If I get the same time result for GPU vs CPU on benchmark does that essentially mean the GPU has similar processing power to the CPU, with regards to Vray?

No. V-Ray Bench measures CPU and GPU using completely different scenes.

[Alex_M]

Tomasz from Dabarti thinks that a gpu is worth about 4 cpu's but only if you're building your scenes for gpu to start with - it's a different set of rules about textures, geometry and features but if you build a scene that runs on gpu, it'll be way faster than the same scene on cpu.

That's a strange and vague statement. What model GPU and what model CPU are we talking about? It really depends, you can just say this and stop there. For example, a GTX 1080 Ti is certainly not worth 4 x 16-core Threadripper 1950X's. I'm interested to learn what he meant by that.