Long time user of Poser, Mac. Not sure what to do now that they are dropping Mac support.  Maybe move to Daz3D, maybe move to PC, maybe just move on.  But please help me.  I am totally unfamiliar with PC world and need guidance on issues like CPU v GPU.  If I wished to optimize Poser performance when buying a new PC, what should I concentrate on?  Happy to spend some money ($2000+) to get something well tailored, pleasant to use, efficient.  I'm ignorant.  For example, I've only today learned the word "Optix".  Should I be sure to get an NVIDIA graphics card for best results?  This will also be a gaming computer. What proportion of Poser render quality/time is due to CPU vs graphics card?  And how vital is large quantity of RAM?  What such questions should I be asking?

Since I'm a newbie to this, I would prefer buying a prebuilt.  Would still appreciate some advice on individual components in case I end up taking the leap of building.

This is an unhappy branching point in the road. I could use a word or two.

Thanky kindly,

-- umber

I can't say much, but since you mentioned games, many new ones seem to want NVIDIA or the AMD equivalent. Superfly renders with a graphics card are so much faster than CPU. I have one of the slower RTX cards (2060) and what I can render in less than 20 minutes with a CPU will take 3 or 4 hours.

The reason for dropping Mac is that the majority of the last year of Poser development has been wasted trying to keep up with what ever Apple broke with the previous OS update. The most recent one is where we reached our breaking point. With OS 15.1 Apple broke OpenGL. OpenGL is how Poser previews the scene. Apple broke OpenGL to the point we had to disable the preview of shadows on Mac. They're trying to force us to move to their completely different, and 100% incompatible preview solution. Apple continues to widen divide between PC and Mac. It's now so big that it's not practical to bridge that divide. Especially when we know the next update from Apple will wreck that effort, making the divide even greater.

To the actual question, The best solution is always a custom built but if you're new to PC that a real stretch. Many of the brand names you know from years of advertisements have outlived their usefulness. (Looking at you Dell). First the biggest deal is the GPU

You definitely want Nvidia because their support of Optix is unparalleled. It can be 40 times faster than a CPU render TIMES. not percent.

Your need to focus on the GPU's dedicated memory. When Poser is rendering on GPU the entire scene and all the textures need to fit in the GPU memory all at once. A GPU with 8GB of dedicated memory should be the minimum. 12GB will be able to handle most any thing you throw at it. Something Like and RTX 4070 will definitely do the job without breaking the bank.

For CPU choice clock speed is more important than lots of cores for Poser. For many operations Poser can't take advantage of multiple cores. Then it's just the base speed of the processor that matters. Something like an Intel i7-13700F.

For main memory no less than 32 GB. 64GB would be ideal. It needs to match the requirements for the chosen CPU. For the mentioned CPU above the memory type is DDR5 5600.

Take those specs to a local computer shop. Any respectable nerd with a screwdriver can build a better PC than the big brands. And, probably for less.

My sympathies regarding Apple's moves and direction. Your decision, however regrettable, makes sense.  Thank you for a crystal clear and precise reply to my question!

Do NOT buy a laptop for poser. Poser is a very machine intensive app. Plus, if you get say a 4070 card for your laptop, they throttle it at 50% to prevent over heat. If you get a new machine it should be a desk top.

A 4070 Ti will get you to the point where OptiX renders everything super fast. I agree with custom build versus store bought completely. 

Definitely a desktop PC base unit, not a laptop. The latest Windows OS is a pain in terms of privacy, but there are several one-click freeware solutions that scamper though the Windows settings and turn off most of the tracking and intrusiveness. 

I would buy pre-built, but where you choose your config from a series of drop-down menus. In the UK that means either PCSpecialist or Overclockers for a super-duper gaming PC. If you may also want to run local LLM AIs (chatbots), then 128Gb of main motherboard memory would future-proof the PC somewhat, and as fast as you can get it. Otherwise 64Gb.

A little over $2000 should get you something with a fast/latest Intel CPU and a NVIDIA 12Gb graphics card in it. This will be more than enough. Also enough if you want to feed Poser renders to software that will enhance them as AI generative images (Invoke AI etc). Also good for videogames, as it's a gamer's card. If you choose a 'workstation' from your vendor, the NVIDIA cards are different (special workstation cards tuned for 3D work and video-editing etc, but maybe not so good for some games). 

You don't say what version of Poser you were using. If you intend to also use the now-free E-on Vue, sending Poser scenes to it, then Poser 11 also needs to be installed on the PC. P11 does not support 40-series NVIDIA cards, but you only need it for the SDK that lets Vue import the scene. 11 will co-exist happily with Poser 12 or 13. If you want to use paid-for scripts in P12/13, you can't install Windows 7 as your OS (it doesn't have the encryption modules).

In the UK for your budget, I can get this for $2,196 US from PCSpecialist. Probably less in the USA, since this is the inflated 'UK prices' in action.

CPU: Intel Core Ultra 5 10-Core Processor 225F (Up to 4.9GHz) 20MB Cache

MOTHERBOARD: ASUS PRIME Z890-P WIFI (LGA1851, DDR5, M.2 PCIe 5.0, Wi-Fi 7)

RAM: 64GB PCS PRO DDR5 4800MHz CL40 (2 x 32GB)

GRAPHICS CARD: 12GB NVIDIA GEFORCE RTX 5070

Not sure if Poser 13 supports the brand-new NVIDIA 50xx series cards yet, though. Check that, and get a 4070 Ti if not.

Corrections: It goes to 96Gb motherboard RAM these days, not 128Gb. 

I recently had to upgrade for work (my old PC was about 8+ years old) and went with a prebuild PC from iBuyPower; Y60
- Ryzen 9 7900X CPU
- NVIDIA GeForce RTX 4070 Ti Super 16GB GPU
- 32GB DDR5 5200MHz RAM
- 2TB NMe

It has 2 slots in the back for additional internal harddrives.  I so def made use of those too. 
I'm sure there are better ones (there always are) but this is working out pretty nice.

I recently had to upgrade for work (my old PC was about 8+ years old) and went with a prebuild PC from iBuyPower; Y60
- Ryzen 9 7900X CPU
- NVIDIA GeForce RTX 4070 Ti Super 16GB GPU
- 32GB DDR5 5200MHz RAM
- 2TB NMe

It has 2 slots in the back for additional internal harddrives.  I so def made use of those too. 
I'm sure there are better ones (there always are) but this is working out pretty nice.

My hardware is too old to update;  I'll need to build anew.  Could you guys discuss/compare these GPUs for Poser (Superfly) use?   None of these have Ti or Super suffixes.  What benefit do those add?

Note:  I have a habit of making large and complex scenes.

I make complex scenes too. I recommend the most vram you can afford. I got 12 gig and I have run out at times. Usually, 10 people is where I start running into issues.

Lol... This reminds me of an experiment started 4 years ago. My old 2080 TI could support up to 13 Vic4's, whereas my 4070 OC with 12Gb can support one more Vic4.
I have to say that it was near too much for it  but wow 

19:02:22 - Analytics: Recording event for Size:1736x937
19:02:22 - Analytics: Starting timer for Render:SuperFly
19:02:30 - SuperFly: Start rendering on device NVIDIA GeForce RTX 4070.
19:23:08 - SuperFly: Rendering time: 1237.96 seconds.
19:23:08 - SuperFly: Rendering memory: 10024 MB.
19:23:10 - Analytics: Ending timer for Render:SuperFly

Which kind(s) of cores does Superfly use?

The 50-- cards have Blackwell shader cores as opposed to the 40-- cards' Ada Lovelace architecture.  Is this a problem?

Are the RTX 5000 series cards confirmed working? Often the new generation breaks compatibility.

That would include the 5070, 5080, 5090, and also everything labeled "Ada Generation" because those are workstation GPUs and therefore it's industry standard to name them in a way that is completely inconsistent with consumer products.

I remember there were some issues with the 4000 series cards when they were first released. (They all work great now though.)

The 50xx cards should all be working. That's not to say driver or AIB problems can't pop up. We haven't had any 50xx problems come through support which is good news.

OPTIX

I'm a dummy about hardware, but I just want to put in my 2 cents about Optix rendering. I bought a new machine last May (see specs in my sig below), and when I render on the preset OptixHigh (Nvidia GPU), I blink and the render is done. It's super super super fast!

So, to evaluate a GPU for Poser Superfly use, are the relevant statistics the number of CUDA cores, clock speed of those cores, and the onboard VRAM?

That is, ignore the tensor core count, shader cores, and ray tracing cores?  (Granting that these tend to correlate with the number of CUDA cores anyway).

From what I understand the last 3 mentioned only counts for gaming. I am humming along with the same GPU VedaDalsette mentioned above and there was literally nothing I threw on it that was touted as benchmark that wasn't done in lass than 15 minutes with OptiX and Superfly. It is INSANE. You do not need the latest version of GPU for Poser.

Rhia474 posted at 11:09 AM Sun, 23 March 2025 - #4494469

From what I understand the last 3 mentioned only counts for gaming. I am humming along with the same GPU VedaDalsette mentioned above and there was literally nothing I threw on it that was touted as benchmark that wasn't done in lass than 15 minutes with OptiX and Superfly. It is INSANE. You do not need the latest version of GPU for Poser.

Going up from a 20xx to a 50xx will only get you incremental increases in performance. Getting on any RTX is revolutionary. Money is better invested in a 20xx or 30xx with more memory that a 50xx because it's "new".

nerd posted at 5:42 PM Wed, 26 March 2025 - #4494535

Yes all that really matters is that it's a RTX and how much dedicated memory it has. RTX anything with 8 GB will render circles around any thing less.

Going up from a 20xx to a 50xx will only get you incremental increases in performance. Getting on any RTX is revolutionary. Money is better invested in a 20xx or 30xx with more memory that a 50xx because it's "new".

Does P13 Superfly (Optix) make use of dual GPUs?  If so, does each card need to hold the entire scene in memory (I would guess that they would)?

Rendering in Superfly can use dual gpus, but you don't get gpu1 + gpu2 in performance... It is about 0.70-1.5 at best.

If they are duplicate cards, you will probably get around 1.5... Other than that, it is a crap shoot.

If they are different generations, it could possibly be slower that the lower generation card alone.

Rendering will also be limited to the card with the least amount of ram.

Dual GPU render doesn't add up, in more ways than one...

Multiple GPUs show up in the render options, as long as CUDA is turned on in the Nvidia Control panel for all cards.

Using Multi Device with a 4070ti and a 2060 at the same time shows very little improvement, if at all, due to massive differences between the cards.

I kept the 2060 for rendering in P11, which doesn't support a 4000 series card....

Useful knowledge, shvrdavid, thank you.  

Also the cards do not need to be capable of, or setup as SLI. Poser doesn't even know what SLI is.

OK, suppose a Mac guy really did build a PC to run his old Poser content under new Poser versions.  How about file format compatibility?  Since I've never used a PC, I don't know what the tricks are to transferring Mac content to a PC disk.  Could I say copy a folder of Poser content, .cr2 files for example, onto a USB drive, insert that drive into a PC, copy them to the PC hard disk then open them with a PC version of Poser?  No file format incompatibilities?

After viewing some videos on building a PC it looks not that hard but I am apprehensive about all the software issues a Mac guy would have to learn about.

Thx for the many useful, clarifying comments above.

umber posted at 3:07 PM Sun, 30 March 2025 - #4494607

OK, suppose a Mac guy really did build a PC to run his old Poser content under new Poser versions.  How about file format compatibility?  Since I've never used a PC, I don't know what the tricks are to transferring Mac content to a PC disk.  Could I say copy a folder of Poser content, .cr2 files for example, onto a USB drive, insert that drive into a PC, copy them to the PC hard disk then open them with a PC version of Poser?  No file format incompatibilities?

After viewing some videos on building a PC it looks not that hard but I am apprehensive about all the software issues a Mac guy would have to learn about.

Thx for the many useful, clarifying comments above.

I have 2 Macs and 2 PC hooked up to a KVM all use the same KB, mouse, monitor, graphics tablet, space mouse and all share the same Poser runtimes stored on NAS. All poser files are completely OS agnostic.

Well, close to finishing bringing a new Win11 P13-capable PC online.  Waiting for a couple of cords (DisplayPort 2.1) to connect the monitor (50" Alienware 4K).  Still need to install the sound card, which has an optical output.  That will drive my stereo system downstairs when I'm lifting weights.  An HDMI 2.1 cable will go from video card to an 83" 4K screen downstairs, so that the PC can play anime music videos through the downstairs home theatre during a workout.

     The GPU is an nVidia RTX5090 with 32GB VRAM and 21,760 CUDA cores.   1500W power supply.   System RAM is 64GB.  CPU is Core Ultra 9 285K (24 core, 76MB cache, 3.7GHz-5.7GHz).  I may be asking for help with reducing Win11 bloatware/snoopware.  I already bought a P13 license.  I'll also be able to install the newest version of Silo (modeling program).  I'll probably name this PC Yuki Nagato.

That is a beast of a PC, congratulations. If still doing some Firefly renders with Poser (Sketch, or pure line-art, or maybe ToonID with randomised colours for masking in Photoshop), don't forget to increase your bucket size from the default 32 buckets.

I simultaneously feel both excited and also a bit nervous (hope this doesn't burst into flame when I power it up).Today I am routing an HDMI video cable and an optical audio cable through walls/ceiling to enable the PC to drive my home theatre screen and stereo.  I would have preferred 128GB RAM, but the motherboard only has two DIMM slots.  

Tech question:  When GPU/Optix rendering in Superfly, does each bucket use one CUDA core?  I'll probably render wallpapers at 4K.

Getting the Superfly render settings nailed down in Poser 13 seems very complex (and is different from Poser 12, so that old advice is not to be followed). You'll likely find it's about far more than the buckets (or as Blender Cycles users call them, 'tiles').  But basically, the manual https://www.posersoftware.com/documentation/13/HTML/Poser_Reference_Manual/Rendering/SuperFly/SuperFly_Render_Settings.htm says...

• Bucket Size: Smaller buckets use less memory. Larger buckets render the image faster.

When using GPU rendering, the Bucket Size setting will help to achieve optimal performance. Start with a bucket size between 100 and 300. Some scenes may benefit with larger settings.

Noted.  Thank you, HartyBart.  I scanned through that web page.  Remaining question:   Should I choose bucket size such that it is an integer exponential power of 2?  (i.e., 2,4,8,16,32,64,128,256,512...)

Again, I will mostly render GPU/Optix in superfly at 4K resolution.

When rendering on a GPU use progressive rendering. It will actually go faster than bucket rendering. The overhead of swapping buckets into the GPU memory out weighs any possible performance gain. Modern GPUs have so much dedicated memory it's no longer a performance benefit.

The opposite is true for CPU renders. We use smaller buckets to try and increase hits in the L2 CPU cache. CPU cache is measured in MB. GPU's dedicated memory is in GB.

I have a 2060 with 12 gig vram. I usually use 2048 for optic rendering. It works well with the size images I render. Most are less than 2k. When I run out of vram and have to switch to cpu rendering I use 64 for the bucket size. My test show that to be the best for me though 32 and 128 aren’t much slower. 

@nerd:  The Superfly render settings in the P13 documentation which HartyBart provided has bucket size specified and progressive refinement ticked.  It shows the render device as being an RTX 4070 Ti.  Is the bucket size irrelevant if using progressive refinement?

@RedPhantom:  Do you mean 2048 as bucket size for GPU rendering?  I intend to render at larger pixel size (4K, i.e., 3840x2160 pixels), with scenes of similar complexity/difficulty as yours.  On the other hand, this GPU has more VRAM (32GB).  Does anyone know of a way to monitor VRAM usage in real time? 

With some trouble, yesterday I succeeded in routing the HDMI and optical audio cables from the workstation room (second floor, front of house) to the home theatre system (downstairs, center of house).  This PC chassis has two currently empty 2.5" laptop drive bays.  I think I'll get a couple of 2.5" internal SSD for use as archival closet storage.

Yes, that's the bucket size for the GPU renders. These are my final quality render settings. The 75 pixel samples might be a little overkill at times. I can get away with lowering them if I use the denoising, but sometimes that messes with the texture details so I go with the higher settings. Most renders are done in 20 minutes or less, and there's no rush for me, so it's fine.

I don't know of a GPU memory monitor.

For GPU memory monitoring (and lots of other GPU monitoring) use the GPU-Z app from https://www.techpowerup.com/gpuz/ (Beware there are counterfeits on the web that are Trojans. Only download from the techpowerup web site.)

As for the actual bucket value it's not used for Progressive render mode. I highly recommend trying some combinations for your self. I every scenario I've tried progressive always comes up as 5-10% faster. And if I user buckets the smaller they are the slower it goes. It's fastest if it's one big bucket.

I suspect the only case where using buckets would be better is if you have two GPUs. Then they can both be rendering at the same time.