Attached Link: http://compiler.free.fr/CPU_test/CPU_test_english.htm

I have been very interested in Jim Burton's survey about rendering time for a Poser 5 scene, using various machines. Although layback gave a very interesting interpretation, I thought I could put my nose in it too. The original thread : http://www.renderosity.com/messages.ez?Form.ShowMessage=1403770 The results http://www.renderosity.com/messages.ez?Form.ShowMessage=1410910 The analysis http://www.renderosity.com/messages.ez?Form.ShowMessage=1414423 My 2 cents : http://compiler.free.fr/CPU_test/CPU_test_english.htm For those not wanting to go through it, the conclusions are : 0- The number of answers to the survey isn't sufficient to give more than very general ideas. 1- The survey cannot give results for Mac users (Poser 5 being just issued on this platfom, so there were not enough answers). 2- Pentium 3 and Windows 98 equip older machines, with also less RAM and a lower CPU speed which all lead to a longer rendering time. 3- RAM size influence over rendering time follows a thresold effect : below a minimum value, it is an important issue and can lead to greatly augmented rendering times, above this value, the increase in RAM size is not a very significant factor in rendering time. This threshold value was 512 GB RAM in the survey, but it may be image dependent, rising with the complexity of the scenes to be rendered. 4- Above this threshold level, and for modern systems (Pentium4 or AMD, running on Win2K or Win XP) (MACs could not be tested because of too few answers) lower rendering times are strongly related to higher CPU speeds, especially for the Pentiums 4. 5- A CPU speed of 2,4 seems the most cost-efficient for Pentium 4 CPUs. 6- For Pentium 4 users, it seems that Win2K is a quicker system, but it can be caused by various interactions, none of which could be studied here. Basically, to answer the usual question "what PC should I have to use Poser 5", the answer could be : Pentium 4, 2.4 GHz (if cost effectiveness is of importance) Win 2K 512 GB RAM are a minimum for moderately complex scenes (just above the "simple portrait" level). Of course, these are just my 2 eurocents : please feel free to comment, amend, criticize... I hope this helps somehow. Thanks Jim for this idea. Compiler

Your CPU speed/renderine time chart is incorrect. An Athlon 2800+ is not a 2.8Ghz CPU, it is 2.08Ghz. Same goes for all the other Athlon xxxx CPUs in the chart. The fastest Athlon on the market right now is the 2.2Ghz AthlonXP 3200. So when you bump the dot back to the speed range it is supposed to be in you will see it basically takes a 2.8-3Ghz P4 to equal or slightly better a 2.08Ghz AthlonXP. Another correction, in regards to layingbacks analysis, the number rating on the Athlons are not how they relate to the P4. They are how they relate to the old model of Athlons. Basically, an older Athlon would have to run at 3.2Ghz to equal the performance of the 2.2Ghz Athlon XP 3200+. Its confusing, but it gets worse. The new Athlon64 coming out this month will be running at 1.8Ghz, but will be called the Athlon64 3200. It is 400Mhz slower than the current Athlon 3200+, but the better performer of the two. This isnt the first time they did that however. The AthlonXP 2800+ is a 2.083GHz processor, but the Athlon 2700 is a 2.17Ghz processor! As far as value goes, the AthlonXP 2800+ costs about the same as the P4 2.4Ghz, and is the better performer (in many other applications as well).

Gee, I don't know, for one thing, I don't see any magic in the fact that an AMD chip is faster than an Intel chip of the same clock speed, because the Intel numbers go higher, after all, it is the end result that counts. Like you said, we didn't really get enough high-end AMD results to prove much, but I'd go out on a limb and say the current P4 2.8 GHZ will beat any current AMD in a fair Poser 5 test, and prices aren't that far apart, $190 for a 2800+ AMD, without a fan and heatsink, $270 for the P4, with a fan and heatsink. It isn't like when AMD prices were 1/3 of the Pentium prices for slower-performing chips. Things will change down the road, of course, in addittion to the AMD the P5 is supposed to be out by the end of the year. Oh- Compiler, thanks! I agree with what you say, too. And I'd guess Win2K speed advantage isn't over 1 or 2% at best over WinXP- I saw 10% difference in the old test between Win2K and Win98, but Win98 is awfull, WinXP just has too much running in the background for pure speed.

cool and interesting thread :)

The best thing to increase performance under WinXP is to go into Control Panel->System->Advanced->Performance Settings ->Visual Effects and set it to "Adjust for best performance". This won't get rid of all of the other background processes, but it eliminates the eewi-gui and makes it more like Win2K, improving overall performance greatly. First thing I do when installing/reinstalling WinXP. :) Still, there seemed to be a speed advantage with Win2K, which may be related all of these new background processes and services, some imparted from SP1/2 or the myriad of security updates. ;)

C makes it easy to shoot yourself in the foot. C++ makes it harder, but when you do, you blow your whole leg off.

 -- Bjarne Stroustrup

Contact Me | Kuroyume's DevelopmentZone

In all the posted results and comments I've ssen reguarding this test nobody has mentioned anything about bucket size when rendering. I have litterally increased my rendering times by 50% in some cases with playing with the bucket size. does anybody think this would help with times for the test file?? Am I completely spaced?

"Gee, I don't know, for one thing, I don't see any magic in the fact that an AMD chip is faster than an Intel chip of the same clock speed, because the Intel numbers go higher, after all, it is the end result that counts." Plain english: AthlonXP 2.08Ghz is as fast or faster than a P4 2.8Ghz in Poser and just about everything else (except Quake3). If that AthlonXP was running at 2.8Ghz it would be miles ahead. If you wanna compare "clock to clock", look at how slow a P4 2Ghz is compared to the same Athlon. You are right, it is the end result that counts.. If you want to spend $100 more to get the same performance, or $200 more for a 3Ghz to get a little better performance, thats your choice. My comment wasnt intended to be turned into a "who's CPU is better" one. The speed/performance chart is simply incorrect. It shows AMD processors as 2.5, 2.8Ghz etc.. They arent running at that speed. Thats all.

Also, I was not impressed with the render times of the 1.5g ram machines. Not only was there "minimal" gain, I'm wondering if there is actually a hindrance. Also, since we didn't have very many high end AMD's, I'd not be willing to say that the high end Intel's are faster. I think AMD still has the speed and the price. Also, there was not a test of what MB was in use. This can be a huge difference. Low end Intel boards IMHO are way better than low end AMD boards. High end AMD boards can make the AMD chip run like a raped ape (which is why AMD has an approved board list). Of course, there are other factors like the number of TSR's running and whatnot. Plus, I wonder if we timed it right. I posted my times with a caveat and after compilation, my 2.1 amd was way faster than it should have been, so that leaves me suspicious of the accuracy because I KNOW that my AMD chip is not running at peak performance. Also, some of the ram is pc2100 DDR (like mine) and other higher end ram on the higher end Intels run at faster bus speeds. But still, it was a good data set in that we can see some general trends. I thought Win2k had the faster times in a particular chip class than any of the other operating systems. But I know XP Pro is also a very fast OS.

MachineClaw- a couple / four guys did run times with bigger buckets, saw big improvements. I'd guess this is memory related, but if you have 1GB or more you should probably 2x or 4X your bucket size. The test is useful for seeing the difference, but only time the first run, the bigger buckets seem to take awhile to clear out the old render on the second or later try. The bigger buckets are invalid for comparing CPUs, of course, but useful for getting Poser 5 to run faster.

1Freon1, I guess it's all in how you want to say it. The AMD's could have been classed based on true clock speed, or the "marketed" clock speed (in "" 'cos Intel started this game in this case not AMD). I might have chosen the former if I'd done the graphing, but the only point I was trying to make in the earlier thread was that the 1 AMD "fast" outlier was being classed differently than the other AMD XP's. (BTW, will the PC crowd allow me to use "classed" on even an inanimate object? ;-) But what really did surprise me was the closeness of the tracking of performance of varying CPU's - even if you break out AMD's as 1Freon1 suggests the result would be similar, just 2 distinct performance bands across the graph vs. a single one. As EricofSD said there is wide variation in AMD boards due to chipset differences, e.g. Via's are generally good but suffer from PCI bottlenecks. Yet none of this seems to have a significant bearing on the outcome. (Possibly because Poser5 doesn't stress much of anything beyond the CPU and memory - which is does to extreme, you can see it on a CPU temperature even when Poser is idle!!! Let alone rendering.) There is variation across the band, but no more than you could likely get by just fine tuning your Windoze for perfomance, vs. messing it up e.g. Yes, more testing and analsys could certainly be done, but to me the current data shows way too much conformity across the entire set, to give any hint that there might be a hidden overlooked factor present. Beyond Loser5 being badly written of course ;-) And I've slammed Billyware as much as anyone, but I do have to give them some credit for backward compatibility - it's not far short of amazing that an application which appears to still be compiled and linked as a Win 3.1 app can run at all on the latest Win XP.

@1Freon1 : thanks for the input on AMD's, I really don't know a thing about them. I'd just like to stress the low number of answers, so that the confidence interval of our survey is quite wide indeed : one cannot draw definitive conclusion out of it : it must be confronted to other results, and to each one's own experience. Thanks for answering.

Well since I did the graph: 1) This was just a quick visualisation aid I knocked together in about 10 minutes. 2) I didn't have a list of AMD's real clock speeds for various model numbers to hand 3) Some people quoted x.xGHz and others quoted model numbers - for simplicity I assumed they were the same. 4) No attempt made to check original posts to make sure figures were based on "pure" test and not tweaked. 5) Looking at the graph, even on this basis the AMD stands up well against the P4 at each clock speed (look carefully at 2800, the best result is an AMD) 6) As compiler says the sample size is too small, all we can do is identify some tentative trends. -- Mark

Mark