OTOY Forums

on thing to note =) by hardmoding GPUs to single slot You loose all warranties =) as You cutting out one conector..

if that would be not up to Your taste, You can look for K|NGP|N edition 980Ti (as it has all conectors in a single row) & waterblock from EK - this would allow to to get single slot solution much mroe simple & still keep all warranties.

however..GPUs do not fail that often & if done right that single slot mod is not so risky =)
latelly converted 4x GPU & they work without any issues what so ever.

here's the waterblock (attached bellow)
but..those GPUs might be hard to find as EVGA sells those in limmited quantities =)

Yambo wrote:Hi eyelovemotion!

Actually Sebastian is going to build for me soon a new system based on Z10PE-D8, 7x 980ti׳s & Dual CPUs, so i guess it׳s a good call.
I don׳t know enough so i won׳t talk shit but there are some posts here on related stuff. so please look around.

i׳m also sure you will get some answers here soon

Hi Yambo,
So can C4D take full advantage of multiple cores / hyperthreading / parallel computing? With Xeons there's always the tradeoff between more cores at the cost of lower clock speed. Where is the sweet spot for C4D?
Cheers,
Seeker

The price for KINGPIN is approx 300 EUR more than a regular 980ti but this is so tasty gpu! I wish all cards were without this shitty dual slot approach

Look how many power connectors this crazy beast has - 2x8+6. And LN with 2100mhz

What I've read lately on overclock that gigabyte gamers 980ti OC almost to 1600mhz

My goal for Yam's build is to go with EVGA regular 980 ti and 1200-1400mhz on water and single slot and 7 or even 8 of them:)
I am afraid that power draw for kingpin is also huge and in a single slot even watrercooling would not cool it down to nice temps.

Seekerfinder wrote:

Yambo wrote:Hi eyelovemotion!

Actually Sebastian is going to build for me soon a new system based on Z10PE-D8, 7x 980ti׳s & Dual CPUs, so i guess it׳s a good call.
I don׳t know enough so i won׳t talk shit but there are some posts here on related stuff. so please look around.

i׳m also sure you will get some answers here soon

Hi Yambo,
So can C4D take full advantage of multiple cores / hyperthreading / parallel computing? With Xeons there's always the tradeoff between more cores at the cost of lower clock speed. Where is the sweet spot for C4D?
Cheers,
Seeker

The reason for dual xeon approach is rather for easy handling 7 and more gpus and x10drx allows for 13 (after tutor). This gives true 80 pcie lanes, not PLX problematic solutions. With 2x E5 2620 v3 and 12 cores this is a reasonable investment, but as we see on other new thread would be nice to benefit from manu gpus while compiling.

smicha wrote:The price for KINGPIN is approx 300 EUR more than a regular 980ti but..

You get waht You pay.. I mean after conversion warranty is still there..

smicha wrote:Look how many power connectors this crazy beast has - 2x8+6. And LN with 2100mhz

power conectors is insane..but Iread You do not need to use them all if not LNing..

smicha wrote:My goal for Yam's build is to go with EVGA regular 980 ti and 1200-1400mhz on water and single slot and 7 or even 8 of them:)

that should nicelly round 1000 in OB =)

smicha wrote:I am afraid that power draw for kingpin is also huge and in a single slot even watrercooling would not cool it down to nice temps.

Yeah.. this is another issue..probably pair of moras should go to deal with all the wasted heat. Though I'm not sure You come even close to those mhz under water..it's LN teritory =)

smicha wrote:

Seekerfinder wrote:

Yambo wrote:Hi eyelovemotion!

Actually Sebastian is going to build for me soon a new system based on Z10PE-D8, 7x 980ti׳s & Dual CPUs, so i guess it׳s a good call.
I don׳t know enough so i won׳t talk shit but there are some posts here on related stuff. so please look around.

i׳m also sure you will get some answers here soon

Hi Yambo,
So can C4D take full advantage of multiple cores / hyperthreading / parallel computing? With Xeons there's always the tradeoff between more cores at the cost of lower clock speed. Where is the sweet spot for C4D?
Cheers,
Seeker

The reason for dual xeon approach is rather for easy handling 7 and more gpus and x10drx allows for 13 (after tutor). This gives true 80 pcie lanes, not PLX problematic solutions. With 2x E5 2620 v3 and 12 cores this is a reasonable investment, but as we see on other new thread would be nice to benefit from manu gpus while compiling.

Hi Seekerfinder! Well i wish C4D would take full advantage of multiple cores / hyperthreading. But there are some processes/functions in C4D which works only with 1 core (for example 3D modeling or animation is a single-threaded task) This is actually one of the reasons i'm diving more and more into Houdini in the last months.

But as Sebastian said the main reason for the dual xeons are GPU handling. But i guess i will learn how to use to power of the dual xeons for simulations and stuff more and more.

With regard to Xeons, don't limit your thinking to C4D alone... there are plenty of third party apps that can and do take good advantage of multiple cores, especially where simulations are involved. Also, I have read that XParticles is fully multi-threaded, so plug-ins can as well! I am sure that C4D will make advances in multi-threading/processing as time goes along (though not all processes can be multi-threaded, FYI.)

Yambo wrote:

smicha wrote:

Seekerfinder wrote: Hi Yambo,
So can C4D take full advantage of multiple cores / hyperthreading / parallel computing? With Xeons there's always the tradeoff between more cores at the cost of lower clock speed. Where is the sweet spot for C4D?
Cheers,
Seeker

The reason for dual xeon approach is rather for easy handling 7 and more gpus and x10drx allows for 13 (after tutor). This gives true 80 pcie lanes, not PLX problematic solutions. With 2x E5 2620 v3 and 12 cores this is a reasonable investment, but as we see on other new thread would be nice to benefit from manu gpus while compiling.

Hi Seekerfinder! Well i wish C4D would take full advantage of multiple cores / hyperthreading. But there are some processes/functions in C4D which works only with 1 core (for example 3D modeling or animation is a single-threaded task) This is actually one of the reasons i'm diving more and more into Houdini in the last months.

But as Sebastian said the main reason for the dual xeons are GPU handling. But i guess i will learn how to use to power of the dual xeons for simulations and stuff more and more.

Thanks Yambo,
I'm about to adopt C4D as my mograph (& more) 3D app so this is useful info for me. I have seen some guys tout the multi-core capability of C4D but it seems like it's limited to the C4D renderer and some plugins.
Question: would the lower clock speed of the XEON not slow down operations that are not able to utilize parallel computing?
Cheers,
Seeker

smicha wrote:

Seekerfinder wrote:

Yambo wrote: The reason for dual xeon approach is rather for easy handling 7 and more gpus and x10drx allows for 13 (after tutor). This gives true 80 pcie lanes, not PLX problematic solutions. With 2x E5 2620 v3 and 12 cores this is a reasonable investment, but as we see on other new thread would be nice to benefit from manu gpus while compiling.

Hi Smicha,
I agree the native 80 lanes are easier for plug 'n play with 7 GPU's. However, do you really find PLX chips problematic? Sure there are a few hurdles (as you have also identified) but once you cross them a single CPU board works fine in most cases. I find the additional cost for the Xeon setup hard to justify for myself with most apps (especially legacy 3D apps) not yet able to take full advantage of parallel computing.
Seeker

jayroth wrote:With regard to Xeons, don't limit your thinking to C4D alone... there are plenty of third party apps that can and do take good advantage of multiple cores, especially where simulations are involved. Also, I have read that XParticles is fully multi-threaded, so plug-ins can as well! I am sure that C4D will make advances in multi-threading/processing as time goes along (though not all processes can be multi-threaded, FYI.)

A valid point, thanks JR. Interestingly though, an I7 still outperforms Xeons when it comes to particles in Cinebench. But you're right, we're moving forward and not backward. So hopefully we'll see more multi-core architecture adoption over time. At the moment though, Intel seems to control the market tightly with lower clock speed Xeons...
Best,
Seeker