Support Website Contact Support Blog

Optimal hardware configurations

I have been building some benchmarks lately to determine where and if performance scale is relative to pure clockspeed vs cores.

 

My experience so far is showing that dedicated workstation components are yiedling minimal gains over consumer grade hardware. I.E. XEON vs i7 or QUADRO vs GTX.

This is my current build and I am seeing some decent render times through each step.

CPU: i7 4790k

Ram: 16gb DDR3 PC1600

HDD: 850 EVO 512 GB (OS) and 3TB 10k RPM HDD’s (Storage)

Motherboard: ASUS Sabre tooth x99

GPU: EVGA GTX 980TI superclocked

I am toying around with a couple of other component options to see if and where we can optimize and possibly halve our times by enabling SLI components or dual CPU solutions. So far, my conclusions draw me to the answer being no and that dedicated single component systems far exceed the money invested.

Mind you, the overclocking per component is roughly 30% over stock speeds. However, only gives me roughly 10% extra time increase.

Has anyone come across the same datapoints or similar results?

I’m interested in your research.

My config is Dell Precision i7-4940mx, 32g ram, k5100m.

If I could, think I would swap the k5100m for the most gtx card possible.

Price difference would go into more ram and fastest SSDs possible.

Sofar this Dell is performing well, albeit under Windows 10, grrrrr!!!

Real prefer Linux environments for this type of processing work.  

 

dual CPU or SLI never increases the performance by 2 fold on intel pentium/i series (except 2011/2011v3 sockets) because of the pci-express bandwidth bottleneck. if you want to see true performance increase using SLI I recommend you to use Intel Xeon E5 V3 series, they can scale up to 2 - 4 cpu’s and can support multiple GPU’s without sacrificing the bandwidth (but of course only the cpu’s are expensive then your whole setup), but does it worth it? uh… no. I have a similar setup to yours, and I’m always against the overclocking because it reduces the life time of the system and doesn’t increase the performance as advertised. Switching to Titan series doesn’t increase the performance on your case too, because they have almost the same CUDA core count as GTX980Ti.

Would you mind joining us in the thread New Xeon Computer as we would value your input regarding Xeon issues in P4D:

https://community.pix4d.com/t/5306-New-Xeon-Computer#gsc.tab=0