I'll get them more soon. For now I need to wait for my custom designed watercooling parts to be produced. 2-3 weeks and I'll show you what I gotTutor wrote:Yes - Your drawing is correct. Your risers are same size as mine and many of my 90+ risers of that type came from the very same source that you referenced last week.smicha wrote:You are the man, Tutor. I am sorry if I repeat myself
Is my drawing correct so?
Gold! Gold!milanm wrote:Compilation and upload times should not be underestimated. Sometimes It's better to have more machines and licenses. Running two instances of Octane with different GPUs assigned to each on the same machine can result in 2X faster compilation (depending on the CPU ofc.). That also works with C4D for me.
Regards
Milan
Since you posit that speed and cost of production would remain the same, that makes no. 2 especially more desirable if the time to breakthrough wasn't excessive, given that I'm driven by the need to always learn new things and to apply that knowledge. Moreover, I develop sentimental/familiarity attachments to my gear and I enjoy tweaking my gear to keep it working and to make it faster and more featureful than it was upon my aquisition of it - no matter how old it is - like my self-upgraded, self-repaired and still used Video Toaster 040s, Atari 040s and classic Macs.Notiusweb wrote:Tutor, a philosophical question for you....
Which would you rather have, if 'speed' of production would be equal and cost would remain same:
(1) a revolutionary technology appears created by someone else that makes it completely unnecessary to have rigs (GPUs, cooling, hardware, etc), and you would easily have access to the new technology to such a degree that you could still accomplish all your work, or
(2) you achieve a learned personal breakthrough where you solve your own rig's questions and discover a new potential / ability to run everything (GPU number, power, speed) that and as you ever wanted, using your same gear
Hi,teknofreek wrote:Thanks for the reply Tutor. I've just received the amfeltec switch, so I'll see how that works..
Unfortunately I need to run a Wacom tablet and 3D Connexion spacenav, so maybe they're also taking some IO ( don't really know though).
I might try to remove the MSI 680 lightnings from the main MB. They are not branded ti, but maybe are hogging resources as most of this problem came about when I replaced the 660ti's with those.
Anyway thanks again.
Hey itou31, have to ask, is this with tonemapping turned off on the splitter cards? Is there a mobo-attached card that can run the tonemapping, or can you spare a USB riser and have one card attached to the Asus P9X79 on a riser (probably won't crash upon using tonemapping...)itou31 wrote: 2) 4x splitter (4 GPU) (4X PCIe)
The second, I have big problem with my Asus P9X79 deluxe and octane V3 : freeze
Re: PCI-E Speed?
Postby juanjgon » Sun May 01, 2016 4:17 am
Octane 3 has change the samples integration and frame buffers architecture, that is now done in the CPU RAM. This makes possible render large resolution images and a lot of passes without fill the GPU VRAM, but the problem is that there are a lot more traffic in the PCIe buses. The PCIe 1x buses are not fast enough, so a 10% to 30% performance hint has been reported while working with GPUs using low speed PCIe buses.
-Juanjo
Thanks for information.itou31 wrote: Hi,
Just a little note on riser (USB3 1X) or other splitter and clusters. I have the 3 types :
1) 1X PCIe USB3
2) 4x splitter (4 GPU) (4X PCIe)
3) Dual Cluster amfeltec (2 x 4 GPU) (4X PCIe)
The first is compatible on my 3 rig.
The second, I have big problem with my Asus P9X79 deluxe and octane V3 : freeze
The third, no pb yet.
About speed loss since octane V3 (more comminucation with host CPU and RAM), I have here some test from my scenes :
Test with an 980Ti:
- inside 16X PCIe : 7min16s
- with cluster (only the 980Ti) : 9min35
- with splitter (only the 980Ti) : 10min49
- with USB3 connected to 1X PCIe : 8min45
The scene is 4K size in PT kernel (parallel samples : 16, max tile : 32, adaptative) and V3.0.6 test4.
Render with network (slave) is better, but not test yet with the same scene and the 980Ti.
I don't remember this episode, but haven't test again the splitter with my P9X79 with lastest V3.0.6 as it is well integrated with the cluster, and the splitter is on my MSI 1155 socket.Notiusweb wrote:Hey itou31, have to ask, is this with tonemapping turned off on the splitter cards? Is there a mobo-attached card that can run the tonemapping, or can you spare a USB riser and have one card attached to the Asus P9X79 on a riser (probably won't crash upon using tonemapping...)itou31 wrote: 2) 4x splitter (4 GPU) (4X PCIe)
The second, I have big problem with my Asus P9X79 deluxe and octane V3 : freeze
Hi itou31, with your 4x splitter (4 GPU) (4X PCIe), how were you able to isolate the speed of the cards if they cause a freeze. Is it that the freeze happen sometimes and not others?itou31 wrote:I don't remember this episode, but haven't test again the splitter with my P9X79 with lastest V3.0.6 as it is well integrated with the cluster, and the splitter is on my MSI 1155 socket.Notiusweb wrote:Hey itou31, have to ask, is this with tonemapping turned off on the splitter cards? Is there a mobo-attached card that can run the tonemapping, or can you spare a USB riser and have one card attached to the Asus P9X79 on a riser (probably won't crash upon using tonemapping...)itou31 wrote: 2) 4x splitter (4 GPU) (4X PCIe)
The second, I have big problem with my Asus P9X79 deluxe and octane V3 : freeze
And with V3 and above (i think), it's not worth to go with 1X PCIe as you lost 10% to 30% as I experience myself.
And with some scene, I noticed that adding the 7th GPU (Titan), it adds only a little to rendertime, and adding the 8th GPU (still Titan), the rendertime doesn't decrease.
So networking seems better, but need one more license.
