Would this be a good option for 6 GPU system?
http://www.miningrigs.net/?product=gray ... tem-batch2
I would be replacing the Motherboard and other components with:
Asus Rampage IV black edition.
32 GB ram
6 core Hanswell
6x 980s or 780ti (if I can get them)
What really concerns me in the use of USB PCIe risers. That does not seem like it will handle bandwidth well? Anyone have experience with a similar scenario or have insight?
Thanks Octane community!
Need Wisdom, 6 GPU server?
Forum rules
Please add your OS and Hardware Configuration in your signature, it makes it easier for us to help you analyze problems. Example: Win 7 64 | Geforce GTX680 | i7 3770 | 16GB
Please add your OS and Hardware Configuration in your signature, it makes it easier for us to help you analyze problems. Example: Win 7 64 | Geforce GTX680 | i7 3770 | 16GB
Octane for 3ds Max v2.21.1 | i7-5930K | 32GB | 1 x GTX Titan Z + 2 x GTX 980 Ti
Why don't you just go ahead with 8x GPUs?
TYAN FT77A should fit your need
http://pcfoo.com/2013/06/cg-workstation-baby-titan/2/
TYAN FT77A should fit your need
http://pcfoo.com/2013/06/cg-workstation-baby-titan/2/
3x GTX Titan - Win7 Pro x64 - Intel i7-3930K - ASUS Rampage IV Formula-X79
- mib2berlin

- Posts: 1194
- Joined: Wed Jan 27, 2010 7:18 pm
- Location: Germany
Hi, I have no personal experience with riser but the PCIe speed does not affect render performance.
It affects loading time to the card, x1 push 500 MB/s, x4 2000 MB/s.
Search the forum for riser, example:
http://render.otoy.com/forum/viewtopic. ... sb#p199213
CHeers, mib
It affects loading time to the card, x1 push 500 MB/s, x4 2000 MB/s.
Search the forum for riser, example:
http://render.otoy.com/forum/viewtopic. ... sb#p199213
CHeers, mib
Opensuse Leap 42.3/64 i5-3570K 16 GB
GTX 760 4 GB Driver: 430.31
Octane 3.08 Blender Octane
GTX 760 4 GB Driver: 430.31
Octane 3.08 Blender Octane
hi
on http://www.octane.it we have tried to figure out with a simple test the speed difference between PCIe x8 vs PCIe v1.
it's absolutely far to be scientific since a lot of elements are involved during the compilling phase, but it could give a rough idea of the difference.
Here is the link for downloading the ocs for the test.
Procedure:
Open the standalone and load one of your ocs and start to render for a couple of seconds. Then open the test ocs file, click on the geometry node and start the stopwatch. Stop the time as soon as the scene is loaded and it's visible in the render viewport.
the results so far:
i7 3770k 3.5 Mhz Asus P8Z77WS HD 7200 2TB PCIe 3 x8
1':55",30
1':55",96
i7 4770k 3.5 Mhz Asus Z87WS SSD 250GB PCIe 3 x8
2':09",11
2':09",79
i5 3570k 3.4 Mhz MSI Z77A-GD65 HD 7200 1TB PCIe 3 x8
2':36",96
2':35",21
The bitcoin setup at x1:
I7 3,4 Mhz ASROCK H61 PRO PCIe 2 x1
2':44.89
2':42.24
on http://www.octane.it we have tried to figure out with a simple test the speed difference between PCIe x8 vs PCIe v1.
it's absolutely far to be scientific since a lot of elements are involved during the compilling phase, but it could give a rough idea of the difference.
Here is the link for downloading the ocs for the test.
Procedure:
Open the standalone and load one of your ocs and start to render for a couple of seconds. Then open the test ocs file, click on the geometry node and start the stopwatch. Stop the time as soon as the scene is loaded and it's visible in the render viewport.
the results so far:
i7 3770k 3.5 Mhz Asus P8Z77WS HD 7200 2TB PCIe 3 x8
1':55",30
1':55",96
i7 4770k 3.5 Mhz Asus Z87WS SSD 250GB PCIe 3 x8
2':09",11
2':09",79
i5 3570k 3.4 Mhz MSI Z77A-GD65 HD 7200 1TB PCIe 3 x8
2':36",96
2':35",21
The bitcoin setup at x1:
I7 3,4 Mhz ASROCK H61 PRO PCIe 2 x1
2':44.89
2':42.24
Last edited by bepeg4d on Tue Sep 30, 2014 8:03 am, edited 1 time in total.
Hi, Bulwerk.
to buy such parts & get 1x raiser is not the wise move..- as You bottleneck the system.. - try to search for ribbon cables X16-to-X16 instead. Some might cause some problems, but they should work, just do some testing. Using those x1 "connected via usb" is just..plain wrong, unless You intentionally go for low-end mobo..In Your case buying this black beauty would be just a waste of Your hard earned money..
it depends what for are You looking to use Your rig.. if CPU performance matters (let's say You do some Vray, or Maxwell loads from time to time..) then looking at something X99 + Octa Core CPU & pairing with capable motherboard would be a way to go. Here's a nice one..X99EWS https://twitter.com/tomglimps/status/515038953125453824
if You don't need insane CPU power & don't mind buying used stuff, head to some sites as "Guys with too deep pockets" are changing their gaming rigs based on x79 & six cores..- some of these boards will easily power 4cards at 8x & others (equipped with PLX) might be capable to handle 7 GPUs..
in any way, houseing more than 4cards (double slott) will be no go on any case.. so maybe ordering something like dopamine or making custom stand inspired by folders/minerswould be a good idea
last but not least..these gm204 (aka 980) are efficient only on games, for CUDA they draw the same amount of power as gk110 (aka 780/titan) if taxed at max laid - yeah, they cost less, but..You get less..2GB less of vRAM (read more if You wish)
Nvidia phased out 780(ti) & left more costly, but older TitanBacks for a reason..=) they are not trying to make a hole shooting at their foot =) they make good business especially with naming.. as 980 is not a successor of 780, but actually of 680..- 780 is a different beast inside..
ok, that's enough =) hope You get some ideas.
to buy such parts & get 1x raiser is not the wise move..- as You bottleneck the system.. - try to search for ribbon cables X16-to-X16 instead. Some might cause some problems, but they should work, just do some testing. Using those x1 "connected via usb" is just..plain wrong, unless You intentionally go for low-end mobo..In Your case buying this black beauty would be just a waste of Your hard earned money..
it depends what for are You looking to use Your rig.. if CPU performance matters (let's say You do some Vray, or Maxwell loads from time to time..) then looking at something X99 + Octa Core CPU & pairing with capable motherboard would be a way to go. Here's a nice one..X99EWS https://twitter.com/tomglimps/status/515038953125453824
if You don't need insane CPU power & don't mind buying used stuff, head to some sites as "Guys with too deep pockets" are changing their gaming rigs based on x79 & six cores..- some of these boards will easily power 4cards at 8x & others (equipped with PLX) might be capable to handle 7 GPUs..
in any way, houseing more than 4cards (double slott) will be no go on any case.. so maybe ordering something like dopamine or making custom stand inspired by folders/minerswould be a good idea
last but not least..these gm204 (aka 980) are efficient only on games, for CUDA they draw the same amount of power as gk110 (aka 780/titan) if taxed at max laid - yeah, they cost less, but..You get less..2GB less of vRAM (read more if You wish)
Nvidia phased out 780(ti) & left more costly, but older TitanBacks for a reason..=) they are not trying to make a hole shooting at their foot =) they make good business especially with naming.. as 980 is not a successor of 780, but actually of 680..- 780 is a different beast inside..
ok, that's enough =) hope You get some ideas.
That might be true (good info/link, thanks) but then not even 7xx series use maximum TDP under rendering in OCTANE, heck some on this forum reported 750Ti usage around 60W when rendering with Octane (TDP is 75W so that's 21-22% less than MAX) so it might be same as for games for 980 cards in Octane (so they could go below 200W in Octane rendering)'coz it obviously not using full TDP when rendering.glimpse wrote: last but not least..these gm204 (aka 980) are efficient only on games, for CUDA they draw the same amount of power as gk110 (aka 780/titan) if taxed at max laid - yeah, they cost less, but..You get less..2GB less of vRAM (read more if You wish)
Until we see exact 980 series in octane it's hard to say is it better to go with 780(Ti) or 980
--
Lewis
http://www.ram-studio.hr
Skype - lewis3d
ICQ - 7128177
WS AMD TRPro 3955WX, 256GB RAM, Win10, 2 * RTX 4090, 1 * RTX 3090
RS1 i7 9800X, 64GB RAM, Win10, 3 * RTX 3090
RS2 i7 6850K, 64GB RAM, Win10, 2 * RTX 4090
Lewis
http://www.ram-studio.hr
Skype - lewis3d
ICQ - 7128177
WS AMD TRPro 3955WX, 256GB RAM, Win10, 2 * RTX 4090, 1 * RTX 3090
RS1 i7 9800X, 64GB RAM, Win10, 3 * RTX 3090
RS2 i7 6850K, 64GB RAM, Win10, 2 * RTX 4090
performance wise yeah, it's too early..but considering other things like vRam - those 2GB extra might be a big of a deal for some..Lewis wrote: Until we see exact 980 series in octane it's hard to say is it better to go with 780(Ti) or 980.
as for TDP, my cards (kepler 104 & 110) run between 80-90% if taxed 100% in OctaneRender & then performance is capped by various reasons (thermals, power draw..& others I don't understand - as GPUz gives some weird short codes that I haven't tried to nail down..yet)
we have not to forget nVidia targets GTX to gamers..& it's just a matter of time till we have to pay extra (obviously Titan, TitanZ isn't branded just as a gamer cards..but entry level compute devices)
Think in the future instead of Gamer & Pro, we are going to have three lines of nVidia's products..
now all the mettrics online will be based more or less on game workloads - think nVidia even asked to take of some numbers from TomsHardware website as they don't reflect how these new cards behave in "normal" conditions (as they took overclocked card & put it under laod in CUDA)..
