Ian Mapleson
Forum Replies Created
-
There are a number of different acceleration plugins for C4D (iRay, Octane, etc.), and which one is right for you depends on what you’re doing, your budget, and so on. Each product has its strengths and weaknesses. There isn’t a single best answer for everything. See:
https://www.maxon.net/en/products/workflow-integration/render-engines/
https://greyscalegorilla.com/2017/08/what-renderer-should-i-use-in-cinema-4d/I suggest searching forum sites for discussions on the topic, shove “best renderer for C4D” into a search engine, that’s how I found the above links and there were many more.
Ian.
——–
SGI Guru -
Tenchi writes:
> Here in germany you will pay much more than US customers will pay.One of many reasons why some do not like the subscription model.
> The first year of the subscription is much cheaper then going into the second year.
Can’t help thinking of it being analgous to getting addicted to a drug. ???? Get the user hooked, then they’re stuck with the cost bump.
> Even the “cheap” teacher & student version cost first €232,05 than in the second year over €356,-
> Thats 30 €/per month thats a lot for students.They can certainly do better if they wanted to. Twenty years ago I tried to organise an advanced SGI lab in the north of England which would have a mix of O2s, Octanes and an Onyx2 rack loaded with Alias/Wavefront apps. This was at the university where I was a sysadmin (I ran a student lab of SGI Indys); A/W was willing to give an astonishing 99.5% discount on their software, which back then was very expensive in the commercial space. A staggeringly enormous discount; their management could see the benefit in students being able to learn these tools early (I think it led later to the Alias PLE release). Alas, the plan never went ahead, too many lazy lecturers who just didn’t care (over 1 million of grant funding was available). Pity, what a difference it might have made to students in the region to have free access to every single A/W there was (Alias, PA, Maya, Dynamation, Kinemation, and dozens more).
> I know the us citizen are more the people which rent things. Germans are more own things people.
Wouldn’t surprise me. ????
> And atm i did not see that CC2017 has improved technolgy in CPU & GPU using.
> It runs like a slug.A common criticism of Adobe, they’re slow to update things, and then stuff just gets ditched entirely.
Mind you, Autodesk does the same thing. I’m not sure about now, but in 2008 there were bugs in Flame that had been there for years.
> Adobe should take care of this for getting every year much money.
Ah but that would mean fewer luxury parties for their management. :}
> 7980XE CPU only rendering (Multicore set to 16 CPU cores):
Try using a thermal monitor program to record what the CPU is doing during the render. Keep an eye on whether any throttling is occuring. Use HwINFO64, etc.
Do you have it overclocked? If not, I would have thought an EPYC build would be a better buy (would have said XEON but they’ve put the prices way up recently), but anyway.
Hmm, just to compare, what score do you get for running the Cinebench R15 test?
> Yes! I replaced the optix.1.dll file of cs6 with version 3.9.2.0 and the GTX 1080TI’s works perfectly with AE CS6 !
I guess Teddy’s point is that even though this works, it nevertheless is very likely the case that such newer GPUs are not being exploited properly.
RT3D does have some very weird behaviour with multiple GPUs, and sometimes certain elements in the scene can ruin the load spreading. I don’t understand why Adobe never added round-robin rendering as an option, as that would often be far more efficient, and faster, especially with partial scene data reuse, something ILM used to do with their Origin-based CPU rendering.
> 4960X@4Ghz / Asus REIV Black Edition/ 64GB RAM
Ironic that three 780 Tis would give the same result but have cost a lot less back then.
> I tried CC 2018, now i found out that i missed to change to set the Raytracer to Cinema 4d (cpu only).
To enable CUDA you need to buy the separate plugin, which is rather expensive. Another reason why some I talked to did not like the changes.
Ian.
——–
SGI Guru -
(apologies Tedday, I missed the earlier-post context of what you said before!)
Tenchi Muyo wrote:
> I got a new System 7980XE with MSI GTX 1080TI SLI and the gpu rendering in CC2017 does not work or working horrible.(SLI is not necessary or used for GPU rendering)
Pardon me for asking, but if you’re using CC2017, doesn’t that mean you’d now be wanting to use the C4D system instead? If so, I think that only gets CUDA acceleration if you buy the separate plugin.
> CPU only Rendering only is horrible too.
Shoulda bought a 1950X. ???? The XE overheats too much for this sort of task, downclocks to manage the thermals.
> I thinking about to replace the TIs with my 3 Titan Black Cards and hope this will works much better.
Bad idea, I would expect three TBs to be slower than two 1080 Tis.
Ian.
——–
SGI Guru -
Teddy writes:
> Jesus I can’t believe people are still trying to use this benchmarkI think there are still quite a few users who have not “upgraded” from their earlier standalone versions. I put that in quotes because I have talked to solo pro users who do not like the subscription cost model at all. I say solo pro because the bigger companies I know don’t use AE (seems to be mostly Flame, Smoke, Nuke and a few others).
> Guys, the AE raytracer is far beyond end of life, it’s no longer supported, stop using this benchmark or trying to hack
> your AE to get results you can’t even compare accurately to others, and that do not utilize any new graphics tech.For those who are still using CS6, it continues to be relevant. What you say is true only if one is now using the CC setup. Some still don’t. I’ve talked to people who really like how CS6 works. I can’t convey the specifics that well, but they like the way its 3D mechanism operates. However, I reckon the number of people in this situation is very low, hence the rareity of native-CS6 posts here these days.
You’re right though that it’s kinda nuts to bother with this benchmark if one is using proper CC 2017. ????
> Embrace the future, move on from this and use modern technology to the fullest.
That would need the stragglers to be convinced to accept the subscription payment model. Some don’t like it.
> Explore actual 3D programs like C4D. …
How much does the CUDA plugin cost?
> And may I also suggest moving on from Creative Cow. This place hasn’t been relevent for 5 years.
Blimey. ???? What is relevant these days? I guess things change so fast. Must admit I only still post because I keep receiving the email notifications for this thread (I don’t check the site in general).
> … May I suggest Redshift or Cycles4D? They would blow your minds if you’re still doing 3D in AE.
I suspect CS6 would die overnight if Adobe had a straight purchase option for CC aswell as subscription.
Ian.
——–
SGI Guru -
I think peope are using modified libs to enable the relevant later CUDA support, hence my earlier question. Really though it contradicts the subject line of the thread. I just wish newer submissions would include details of how any mod was done so others can benefit from the info.
Ian.
——–
SGI Guru -
Marc, did you change the setup somehow so RT3D could function?
Ian.
——–
SGI Guru -
Walter Soyka writes:
> I think it’s a tradeoff. Presumably the Z840 slot configuration is as complicated as it is because
> the HP’s Gen3 PCIe slots are CPU-direct, not switched like the X99E-WS. …The switching is more for broad multi-GPU support, etc., but I don’t think it adds any significant latency. I suspect bandwidth, and the ability to use broader configurations at all (such as three
GPUs and a RAID card), are probably more useful (has anyone tested the latency factor?). But then,
OEM systems have always had a less flexible PCIe setup compared to top-end consumer boards. That’s
why I rather like the Z10PE-D8 WS, kinda the best of both worlds, though even there it’s not quite
as good as the X99E-WS.> That said, I’d doubt there’d be any measurable performance difference for our applications.
Only insofar as a board with more PCIe slots can allow one to install extra CUDA cards for faster
rendering, or use spare slots for further PCIe SSDs, etc.Btw, I’ve talked to someone who has tested an HP Z Turbo Drive Quad Pro for raw 8K work, he was
able to obtain approx. 4.6GB/sec write and 8.2GB/sec read sustained. Not cheap though. 😀 See:https://www8.hp.com/h20195/v2/GetPDF.aspx/4AA6-1667ENW.pdf
Ian.
——–
SGI Guru -
Thomas Leong writes:
> … (which I suspect the ASUS also has though not mentioned by ASUS).Yes, it has them, I posted a link to a picture showing them earlier.
> For storage, I would think the ASRock offers superior features in that there are 2 x 32Gb/s M.2 slots
> (though they share PCIe lanes with a SATA3 connector when occupied) versus 1 similar speed M.2 slot
> for the ASUS. Additionally, the ASRock has 18 x SATA3 (8 x SAS3 12.0 Gb/s + 10 x SATA3 6.0 Gb/s versus
> the ASUS 8 x SATA3 6.0 Gb/s).I do like the presence of SAS on the Asrock board, and it’s nice that it has two M.2 ports (caveats
with lane sharing notwithstanding), but I’m not sure moving someone onto a path of using SAS devices
with a consumer level board is such a good idea, given the cost of SAS storage (it’s a feature set
which seems out of place on a single-socket board). Also, some of the effectiveness of SSD performance
is lost when connected via the 3008, partly because speed is limited by the SAS controller’s max IOPS,
but mostly because the 3008 has no cache RAM AFAIK (ruins RAID0 when I tried it). The Asrock is also
a fair bit more expensive.I’ve used both Asrock and ASUS boards, and for a long while was a fanatical Asrock user (I have sooo
many of their P55 Extreme/Deluxe boards, X58 Extreme6, etc.), but I jumped ship when the X79 Extreme11
came out, couldn’t believe they’d put SAS on a mbd without any cache RAM, and it cost a fortune (well
over 600 UKP).I wouldn’t put a XEON on either board, an oc’d 5960X would be faster (same overall threaded speed,
much better IPC once oc’d). XEONs only really shine once 2 or more are put on a multi-socket board.Trouble is of course, if Asrock uses a SAS controller which did have some cache RAM, that would add
a lot to the cost.Don’t get me wrong, I love Asrock boards, I have loads, but the feature set of the Extreme11 isn’t
a natural match for Adobe apps IMO, and the ASUS will handle an oc’d 5960X better for sure.Ian.
——–
SGI Guru -
Walter Soyka writes:
> Here’s the PCIe configuration on a Z840:Yikes, what a mess! Yeah, exactly the sort of thing I was thinking of, all sorts of horrible
dependencies & gotchas. Makes deciding what to use rather complicated re SSDs, PCIe, etc.I should have been more specific btw, I meant OEM boards often have nowhere near as many full
x16 slots as a board like the X99E-WS.Hence, the ASUS Z10PE-D8 WS is a far better board for this sort of task than whatever HP is using.
I’d rather do a home build than by a Z840.> Nearline storage on NAS, archive on LTO-5. I am eagerly awaiting LTO-7.
Sweet! 8)
Ian.
——–
SGI Guru -
David Lawrence writes:
> My understanding is that this is the fastest possible internal storage for a PC. In my monster
> system, I’m imagining three PCIe SSDs – 1) System and applications, 2) Adobe cache (AE and
> Premiere), 3) Rendered outputHmm, I’m not sure you’d notice any useful or measurable gain from having a PCIe SSD for the C-drive
as opposed to a normal top-end SATA SSD like an 850 Pro. Maybe you would, but so far I’ve seen no
data, ie. to a large extent people are just assuming this would be the case but without any results
to back that up. I would do tests, but I don’t have the relevant parts atm, can’t afford them either.What I mean btw is someone comparing a system for AE where it’s as you described (PCIe for the cache,
PCIe for the output, etc.), but one has an 850 Pro for the C-drive while the other config uses PCIe.
Change just the variable in question, see what happens. All too often comments come from people with
entirely different builds (CPU, mbd, gfx, etc.), so conclusions are iffy at best.Remember too it’s a good idea to have a separate SSD for the Windows paging file, just a medium
model like a 120GB. For a system with 64GB RAM, the paging file should be 96GB, leave the rest unused
for spare area. Frees up lots of space from the C-drive, which makes quite a difference if the C-drive
SSD is only 256GB. For a system with 128GB RAM, then a 250GB EVO would suffice (paging file should be
1.5X RAM capacity).Lastly, consider this alternative: PCIe for the cache and output, but a PCIe for the source input and
normal SATA SSD for the C-drive. Would having a PCIe SSD instead of a SATA SSD for the source files
be faster? I don’t know for sure, but annecdotal evidence suggests it may be.> A fourth, larger fast SATA drive would be for storing final output. …
Do consider using Enterprise SATA, not consumer SATA. I’ve had good success obtaining new/unused
drives of this kind, mostly 2TB Hitachi/Seagate.> Backup would be some external >solution TBD.
Recommend LTO, or if cost is an issue then RAID1 at a minimum, RAID10 for an extra speed edge.
> Does this seem like a reasonable plan?
If you can use PCIe for all of them then sure, but it’ll gobble up PCIe slots pretty quick (most mbds
will only have one M.2 port you can use). The more normal PCIe slots you use up for SSDs, the fewer
will be available for extra GPU expansion.Hmm, is it possible to buy a x16 PCIe to M.2 multiport adapter card yet? I would have thought there’d
be a market for something like that. There are single port cards, but I couldn’t find one with multiple
M.2 ports (not even on a site like span.com).Ian.
> Re: PCIe – when I go through PC configuration options (for example on HP’s Z840 customize page)
> I find I’m only allowed one or two PCIe storage choices. I assume this is a limitation of the
> motherboard? …Not necessarily, but probably. OEMs often use custom boards which have nowhere near as many PCIe
slots as a board like the ASUS X99E-WS.> Are there motherboards out there I should be looking for that would allow for additional fast
> PCIe storage (three or more SSDs)?See above, that’s what I’d use for a prosumer build. In the dual-socket XEON world, it’d be the
ASUS Z10PE-D8 WS instead.> Thanks again for sharing your expertise!
Most welcome! 🙂
Ian.
——–
SGI Guru