Dual 3070 Benchmarks, maybe better than a 3090?

[Rob]

I was fortunate to be able to get 2x 3070's which I utilize for rendering using Maya+Arnold and Octane. I wanted to provide a couple of benchmarks I ran for others thinking about utilizing a couple of 3070's instead of a single 3080 or 3090, as the cost for a couple of 3070's is a little over $1k. My old card is the EVGA 980ti Hybrid, an AIO watercooled version that was top of it's game. I ran the dual cards on a 750W power supply without issues.

Notes, the Zotac is sitting upwind from the PNY, and the monitors are plugged into the PNY, so there is a small performance hit on the PNY due to the overhead of displays and the fact that it is not getting the coolest air like the Zotac. They both seem to perform equally, with the Zotac getting a small edge (which may be it's placement and I am not going to switch cards around to test this).

Cards Tested:
* Zotac Twin Edge 3070
* PNY XLR8 Epic-X 3070
* EVGA 980ti Hybrid (AIO watercooled)

Octanebench 2020.1.5 scores (higher is better)
* Dual 3070: 809
* Zotac: 409
* PNY: 402
* 980ti: 143

Arnold Test (Private Project) (Time, less is better)
4K Test
* Dual 3070: 0:40
* PNY: 1:03
* Zotac: 1:02
* 980ti (old card): 3:55

8k Test
* Dual 3070: 2:45
* PNY: 4:22
* Zotac: 3:58

AntonioBosi.com Arnold-Maya Benchmark
* Dual 3070: 3:26
* PNY: 6:48
* Zotac: 6:44

Summary, doubling up 3070s in Arnold performs equal to a 3090, and easily outperforms a 3090 in Octane based on the scores I have seen.

 

[xbigcheezx]

I am going to give one of my CG buds the second card in a couple weeks, let me know if anyone wants to see any other tests with dual 3070s

 

[Alex Glawion]

Awesome! You got Redshift on hands by any chance?

Oh and how's the slot layout on your motherboard? Are those dual slot open air cooled cards packed tightly, or do you have some space between them?

 

[xbigcheezx]

I don't use redshift, so I can't test, but my friend who is going to receive one of my 3070's does and he will test when he puts his build together. I have an asus x99-e ws MB, which sports 4 pci-16 slots. It originally had 4x 980ti hybrids in it. I spaced out the cards in slots 1 and 5 to provide some room. Also, as I went to overclock the cards after the test, I do not think the downstream heat had much of an effect as I monitored the fans and gpu temps.

I did run a blender test on request, but it could only do one of the cards at a time, results attached for the PNY, CUDA test.