Thu, Nov 28, 9:57 PM CST

Renderosity Forums / Poser - OFFICIAL



Welcome to the Poser - OFFICIAL Forum

Forum Coordinators: RedPhantom

Poser - OFFICIAL F.A.Q (Last Updated: 2024 Nov 28 11:20 am)



Subject: RAM speed and render times.


isaacnewton ( ) posted Sun, 01 December 2002 at 5:35 AM Ā· edited Tue, 19 November 2024 at 3:16 AM

I was discussing the relative merits of different types of RAM with someone and a question arose which someone here might be able to answer. If I use 266MHz RAM instead of 133MHz RAM on a PC, and assuming enough RAM... 1GB or so and a motherboard which can take the faster RAM, will render times be halved? I have been told that a more likely outcome would be a 25% increase in render speed. Does anyone have experience of using fast RAM and know the impact on rendering speed/times? Thanks in advance, Isaac


thgeisel ( ) posted Sun, 01 December 2002 at 6:24 AM

Dont think that there is such a big difference,only the cpu speed makes the rendertime.If you have enough ram so that there is no swapping to the hd im sure you dont see a difference


isaacnewton ( ) posted Sun, 01 December 2002 at 6:47 AM

Hi thgeisel, If the amount of RAM is less than is needed to hold all the data necessary then there would indeed be HD swapping which will certainly slow things right down. For most situations, I guess that 1GB is enough RAM. The CPU speed is the most important factor in determining render speed. However, if the CPU speed is 10 times the RAM speed say between 1 and 2 GHz, (compared to 133MHz for the RAM) then I assume that the speed at which communication takes place between the CPU and the RAM is the limiting factor. If that is true then doubling the RAM speed should significantly increase the rendering speed as the CPU won't be sitting around waiting for the RAM to pass it the next number in the calculation for quite so long. However, that is only theory. I wonder if anyone has any practical experience of a system with fast RAM? Isaac


thgeisel ( ) posted Sun, 01 December 2002 at 7:08 AM

In a german pc magazine they tested computerspeed with different types of ram.they didnt use poser but several programms and bench marks and the speed difference was very little. no reason to replace the old ram with new. In my opinion!!


volfin ( ) posted Sun, 01 December 2002 at 8:27 AM

My last system was a dual Pentium 3 850 mhz machine. I had 100 mhz ram in it initially. when I upgraded to 133 mhz ram, I saw a HUGE difference not only in render speed, but all applications in general. Ram speed does make a big difference. I'm using a P4 2.4 ghz with 266 mhz ram, and I am wishing I had waited for the new 333 mhz ram.


chohole ( ) posted Sun, 01 December 2002 at 8:38 AM

I have to agree with volfin. I am test driving a new system, which has less ram, but faster, and certainly in bryce, which is all I have put on it, I have halved the render time on a complicated scene. Like under 5 hrs instead of over 10 on my own system. OK the system is also 1.7 instead of 800. but I think the results speak for themselves. After all doubling the so called speed does not in fact halve the time in the real world (well the CG real world anyway)

The greatest part of wisdom is learning to developĀ  the ineffable genius of extracting the "neither here nor there" out of any situation...."



ghengis666 ( ) posted Sun, 01 December 2002 at 8:43 AM

If you are using large textures and more than one character in the scene then it's definitely worth having a lot of ram. I have 1GB of 333mhz and poser has managed to use 900MB of that. Without the extra RAM the rendering crawls to a halt but with it the only thing that takes time is anything with dynamic hair.


Blackhearted ( ) posted Sun, 01 December 2002 at 9:50 AM

no, render times wouldnt be halved theoretically it SHOULD be a huge performance increase, but its not. i have 3 comps running poser at the moment. an amd athlon xp 1800 with 1.5gb ram, an amd t-bird 900 with 512 megs of ram, and an amd k6-2/500 with 256 megs of ram. i do a lot of testing of my own products, so i have 2-3 different versions of poser on each one as well. i can tell you that there isnt a huge difference between rendering times. with the slower machines, youll see a HUGE difference in workspace responsiveness - ie. moving the cameras/figures/etc around in the viewport will be sluggish. youll also see a noticeable difference in the amount of time it takes to import a figure into the scene, load textures, etc. these alone are worth upgrading for. however, render times, suprisingly, are almost identical on all three machines. which is VERY unusual, since the fastest of the machines is at least 5 times as fast as the slowest in most other benchmarks. so if you want to increase rendering times, adding RAM or upgrading your processor isnt going to have much of an impact. the only time i see a bit of a difference is when its rendering the shadow maps - the slower machines tend to take a little bit longer. after running just about every build and version of poser on a variety of machines and operating systems over the last few years, i can tell you that just about the most significant 'upgrade' you can make on your systems is to change your operating system to windows 2000 or XP. it runs faster, renders faster, doesnt 'hang' before it starts to render the shadow maps, and doesnt freeze in large renders. i remember when i used windows 9x and large textures/scenes, i was always in the habit of saving before i hit render, because at least a third of the time it would freeze. but sometimes it would freeze simply from turning a morph dial, or whatever. ever since i switched all of my systems to NT, i hardly ever crash anymore - in poser or any other program. in fact, i can probably count on one hand the amount of times my comp has crashed this year - my current network uptime (since my last reboot) is over 70 days :) and, if i can recall, that was because i popped a 5:1 sound card in the machine so i could watch my dvds with better sound. so instead of ram, or a faster processor, if you want the best out of poser upgrade to 2k or xp as a first step. cheers, -gabriel



Blackhearted ( ) posted Sun, 01 December 2002 at 9:56 AM

btw - let me clarify render speeds wont be halved in any program. youll see some improvements in MOST programs - with the 3dstudio max scanline renderer, mental ray, etc... but not with poser. like i said, for some very strange reason when it comes to render times poser isnt very discerning when it comes to system specs. i really USED to think that posers renderer was blazing fast... but now that i think about it im not so sure. was looking at DOA3 on the xbox, and the graphics that are rendered there, in REAL TIME (60 fps), blow away your average poser render. and thats just a 766mhz processor with 64 megs of ram and the equivalent of a geforce3 rendering them. and im talking reflection maps, depth of field, particles, an entire forest with falling leaves that would make the average bryce render pale in comparison. pfft.



Erlik ( ) posted Sun, 01 December 2002 at 10:59 AM

Blackhearted, XBox, Playstation and the others are completely dedicated to the graphics that's on-screen. Practically nothing else happens in the boxes. There's something more, but I'm not certain about that. Don't the boxes use pre-defined routines, as opposed to rendering everything from the scratch?

-- erlik


volfin ( ) posted Sun, 01 December 2002 at 11:21 AM

I agree with what you are saying, gabriel. However, we weren't talking about adding more memory, but rather the speed of the memory. And the graphics rendered on an XBOX or similar game console aren't anywhere near as detailed (or accurate) as what Poser or any true 3D application produce. But I do agree some 3D acceleration could be used to Poser's advantage.


Jim Burton ( ) posted Sun, 01 December 2002 at 2:12 PM

I had a bad RAM chip, because of which I had to run my now-growing-old AMD 1.4 with "pc100" selected in the BIOS, when I finally replaced it, which allowed me to go back to the "pc133" setting I saw times on my CPU test (look in Freestuff) improve about 2%, wow, are we impressed? I also went from 768 to 1024 Mb of RAM, which might have been part of the change. Going from Win 98 to Win 2K improved things about 10%, incidently. I'd guess XP is about as fast as 2K, though, but it is worth changing if your on 98 or ME.


narsil ( ) posted Sun, 01 December 2002 at 2:46 PM

I think we need to be clear about what is being talked about.

Some processors have a large FSB (pentium 4 issueB has a 533Mhz sideband bus.) What does that mean - in a water analogy the pipe is bigger letting more water through for work.

Thats why Pentium 4's currently beat AMD processors, not because of the speed of the processor but rather the amount of cache ram (letting it do more on-cpu work) and the bigger RAM bus (oops pun there for the geeks)

What does this mean in real life- you need to have a processor and motherboard that accept higher rated DDR ram, the RAM speed must be matched to the motherboard specs, otherwise the increased speed of the ram may lead to instability (particularly in older AGP graphics cards).

Base recommendations- if you are buying RAM check with the dealer for compatibility. If you are building from the ground up, Buy some of the newer chipsets and check the maximum amount of ram you can put in the said board- some motherboards on accept certain amounts of the higher speed ram.

I run a couple of very high spec computers (pentium 4 and AMD ) Speed differences between the machines could only be measured in milliseconds at the higher rates)Poser runs faster on these machines naturally but both have been matched very carefully to the maximum ram speeds that the motherboard can handle.

More Ram seems to be the secret for poser though, its a glutton.

And yes I am a sad old geek

PaulC


narsil ( ) posted Sun, 01 December 2002 at 2:51 PM

Oops meant to say I agree with BlackHearted - Render time are almost the same on the new machines as on my old and trusty AMD 1.2Ghz with 512Mb or PC133 ram. I think there is something at work in the rendering algorithm that chugs along at the same rate. P


PJF ( ) posted Sun, 01 December 2002 at 2:57 PM

It isn't reasonable to expect a substantial improvement in Render times with faster memory because that memory speed improvement only affects a relatively minor part of the overall picture.

A useful analogy might be (or might not be ;-)) to consider the influence of rivets on the top speed of an aeroplane. Using flush rivets instead of domed rivets provides a massive improvement in air flow at the point of the rivet, but the overall effect on air resistance is relatively minor because the aeroplane is still a great big lump to have to shove through the air. Improving aerodynamic efficiency helps, but it always takes a more powerful engine to make a plane go a lot faster.

The importance of the rate of transfer of data between the CPU and memory is minor compared to the importance of the rate of speed of calculation on that data the CPU performs. It still takes a more powerful engine to make things go a lot faster.

Most benchmarks show even the fastest RAMBUS memory giving only a few percent improvement to overall speed, and a similar result is obtained with the various new fangled 'pipeline' arrangements. These 'tweaks' can all add up to a helpful boost in performance, and if you can afford to combine them with the fastest processor - then fine. If not - don't worry about it. You're not missing that much.


chohole ( ) posted Sun, 01 December 2002 at 3:25 PM

Ok so tell me how I got such a difference in render speed on a system with 384 mb of pc 133 ( athlon 1.7) as against a system with 512mb of pc 100 ram (athlon800), both running win2k. OK I only tested so far on bryce. lets face it, render speed times on poser4 would be difficult to test. As I said earlier render times for the same graphic were less than half on the machine with the pc133 ram, despite having less. Need to know, because on the basis of the test dives I am doing on the new machine, I will know what to upgrade on my existing system. I don't believe that all the difference was in the speed of the cpu.

The greatest part of wisdom is learning to developĀ  the ineffable genius of extracting the "neither here nor there" out of any situation...."



Spit ( ) posted Sun, 01 December 2002 at 3:48 PM

Chohole...your time difference is due to processor speed not ram speed. There's almost an exact 1:1 relationship between processor speed and render time in Bryce.


Blackhearted ( ) posted Sun, 01 December 2002 at 6:22 PM

thats bryce :) i said youd see significant increases in rendering speed in most programs... however, for some strange reason, poser seems to render just about the same speed on anything. youll see a difference in workspace responsiveness and in shadowmap rendering and figure loading, however,



PJF ( ) posted Sun, 01 December 2002 at 6:45 PM

It isn't surprising that an AMD Athlon running over twice as fast as another AMD Athlon will render a Bryce scene in half the time. The slightly faster RAM will be helping, but not much (and so long as the scene fits in RAM without swapping to disk, the amount of RAM won't make any difference). Your result is pretty much in line with what should be expected. I was curious about the reports of Poser renders not being much improved by faster machines, so I loaded up Jim Burton's CPU test scene on a 500Mhz P3 with 256Mb of PC100 RAM (win98); and a 2.53Mhz P4 with 1Gb of PC2700 RAM (winXP). The scene fits in RAM on both machine (no swapping to disk), and neither machine is especially fancy in its own right. On the P3 the scene rendered in 105 seconds; on the P4 in 24 seconds (both approx). That's about 4 times as fast. I'm surprised that an AMD K6-2 at 500Mhz has render times almost identical to an Athlon 1800xp. It's not just a case of the three times faster Mhz rate; the K6 is not a fast floating point processor - it's good for office but not 3D rendering. I'd be checking into that thoroughly to see what was choking things up on the faster system. With 1.5Gb of RAM, running out of memory doesn't seem likely. Maybe there's something about the type of scene. Jim Burton's test scene didn't use any complex textures; Blackhearted's scenes most probably do. It's a puzzle.


Blackhearted ( ) posted Sun, 01 December 2002 at 7:01 PM

yeah, all of my textures are generally 3000x3000 plus in my scenes. i didnt mean that either of the renders were SLOW... only that there wasnt much of a difference. the slower machine takes about 1.5-2x as long to render shadow maps, however, and in a scene lit by something like my GI presets (proluma) it can really lag behind because it has about 24 shadow maps to render. the main reason that i keep the k62 around is so that i can test my products on different speed systems to make sure they wont crash or perform horribly. and so far, ive rendered even 4000x4000 pixel high res textures, on high-res models, with generated hair and transmapped clothing on the old k62 without it suffering too much or crashing. it chugs along just fine. i know that, theoretically, the newer machines should blow the old ones away... but there seems to be something about the poser renderer that makes it perform pretty much the same as all the others. btw - if youre going to compare the machines, id be interested in seeing the results of rendering time AFTER the shadow maps have been rendered. shadow maps, unfortunately, are heavily dependant on processor power. and i was using the poser 4 renderer on all of my machines. heres some more food for thought: the poser4 renderer, naturally, is much faster than the firefly renderer in poser 5. however, for anyone who has both poser 4 and poser 5 installed, try this. load the EXACT SAME SCENE in both versions, set up the poser 4 renderer to use the same settings, and hit render. theres a huge difference. in P 4.00, unpatched (again i need it this way for testing), it renders this particular scene in about 15 seconds. the exact same scene with the exact same settings in the p5 poser4 renderer took over 3 minutes. so until this gets fixed, i pretty much still use poser4 for most of my rendering. cheers, -gabriel



PJF ( ) posted Sun, 01 December 2002 at 7:20 PM

It's too late tonight (this morning) to repeat the tests, but as I recall the shadow maps (such as they were) seemed to render much faster too. Everything in Jim Burton's test scene rendered much faster on the faster machine. But then, there were no 4000x4000 pixel textures on anything...


volfin ( ) posted Sun, 01 December 2002 at 10:06 PM

The people who say their renders are no faster on a faster machine need to get their PCs checked immediately! If you're not seeing at least a 20% increase in speed, somthing is wrong. I can overclock my 2.4 ghz to 2.8 ghz and get a 5% increase just from that.


Privacy Notice

This site uses cookies to deliver the best experience. Our own cookies make user accounts and other features possible. Third-party cookies are used to display relevant ads and to analyze how Renderosity is used. By using our site, you acknowledge that you have read and understood our Terms of Service, including our Cookie Policy and our Privacy Policy.