Replace 1 Titan X Pascal + 2 GTX 1080's with 4 GTX 1070's ?

[art3mis]

Seriously thinking about selling my 1 yr old Titan XP and 2 1080’s and replacing with 4 FE 1070’s  (for GPU rendering).

Not 100% happy with the power consumption/rendering performance value.

Although its 1.5 year old the 1070 still represents THE sweet spot in terms of price/performance of any GPU out there.

And from the bench marks over on redshift3d  2 1070’s are roughly equal to 1 Quadro GP100 but at 1/10th the cost!

Two hesitations are
1-  OpenCL  apparently can only take advantage of 1 GPU, which you would normally assign to whatever your fastest is.
2- The extra 4GB of VRAM on the Titan

Am I crazy to even consider this?

Anyone know would be the real world (not TDP ratings) total RS rendering power draw of 4 1070’s vs 4 1080’s? (not 1080TI’s)

Edited November 26, 2017 by art3mis

[schwungsau]

with gpu you get diuble renderpower. 3nd card 80% more speed and 4th card 60% more speed. with 4 cards your bottleneck will be the motherboard and cpu. the cpu more core to feed the gpu cards with data.

Quote

If you’re building a computer for Redshift today and anticipate adding more GPUs to it in the future, we recommend choosing a motherboard that has 4 PCIe3.0 x16 slots or more. Please note that some motherboards will claim to have 4 PCIe3.0 x16 slots but their specifications will say something like (x16, x16), (x8, x8, x8, x8). This means "if you have two GPUs, they’ll both run at x16 speed but if you have 4 GPUs, each will run at x8 speed". In other words, even though the motherboard has 4 slots, they can't all be running at full x16 speed at once.

 

 

 

Edited November 25, 2017 by schwungsau

[schwungsau]

Not all CPUs can drive 4 GPUs at full PCIe x16 speed. CPUs have a feature called "PCIe lanes" which describes how fast data can be communicated between the CPU and the GPU. Some CPUs have fewer PCIe lanes than others. For example, the Core i7-5820K 3.3GHz has 28 PCIe lanes while the i7-5930K 3.5GHz has 40 PCIe lanes. This means the 5930K can drive more GPUs and at higher speeds

opencl runs all cards and cpu's. not houdini opencl but if have renderer like porrender it renders on everything.  you even can mix nividai and amd.

why not get the AMD dual pro, 2x 16MB 2X GPU on one card for houdini openGL and hodini-OpenCL for sims and nividia20170 for cuda renderers. 2 cards for 3 jobs.

 

Edited November 25, 2017 by schwungsau

[Guest tar]

Yeah - I would consider the 1070ti as it's 25% faster and 22% more in cost. You also save on PCIe lanes for more rendering power.

 

For the CPU, the ThreadRipper 1900X also has 64PCE lanes which is much better than the 1700, at 24, and is 55% of the price of the 1950X

1. 16 for PCI-E, 4 for M.2, and 4 for the chipset.

 

[art3mis]

Interesting product. AMD GPU's were not even on my radar. So what's the consensus with this card, performance benchmarks, etc as compared with say the similarily priced Nvidia  1080ti?

There is also the newer Vega 56 and 64.

Haven't been able to find many benchmarks. And see that AMD is marketing the duo for 'VR' use.

Watching this review

my takeaway is that AMD's 56 and 64 only really shine in OpenGl performance over the comparable 1070 and 1080 with the 56 taking a rough 5-7% lead over the 1070 in most benchmarks but consuming 25% more power. In general kind of underwhelming and agree with the comment about these cards having more in common performance wise with older Nvidia Maxwell architecture.

 

In any event cannot seriously consider AMD GPU's since my GPU renderer ONLY supports Nvidia  

 

Edited November 26, 2017 by art3mis

[schwungsau]

amd dup pro  (2x  9100 gou on on card)  its like 2x 1080s

 

Edited November 26, 2017 by schwungsau