AMD or Intel ??? please help .!

[MadMax]

No, AMD doesn't support Hyperthreading. It doesn't need it.

HT was a hack Intel came up with to compensate for a poor design. Even with HT, AMD pretty much rules the roost in processor speed. Recent benchmarks had intel's best being trounced by AMD's latest release. Even their main holdouts where Intel ALWAYS did better, Cinema 4D and 3D Studio Max the difference was so negligible that it doesn't matter which CPU you buy anymore.

In media encoding AMD hand Intel it's ass.

With 64 bit and dual core coming, yor money is best spent on the company that pioneered the technology. AMD.

[brudney]

in this particular case you're better off buying p4 3.4 ghz rather than amd64 3200+

unlike many other programs, 3dsmax is an aplication that loves raw clock speed.
if you don't need to buy it NOW wait some time for the athlon64 venice based chips to become more available and consider buying this particular chip. you can read more about it HERE (http://www.xbitlabs.com/articles/cpu/display/athlon64-venice.html)

if you can wait few months do as MadMax said and wait for dual core A64. the problem is that it's unlikely to appear before 2006 (although dc opterons should be released soon if you're interested)

[MadMax]

Venice is supposed to start shipping today.

I would be inclined to look for a 3500+ Venice and bump it up a couple of notches.

[michiel_frankfort]

I've bin strugaling with the same question. The answer is more complex then ever, cause each time Inter or AMD brings out a new processor, the other will folow up real close.

So the real question is: For what goal am i going to use this machine? For Video editing and other sort like apps Intel still is faster, cause in this 'simple' mather Gh's does count.
For more complex and advanced cumputing I advise AMD Opteron processors. These things are the best of the best right now, and when you use 2 or more cpu's its so much more reliable then Intel. Most of the times AMD renders up to 5-10% faster (for example a Inter Xeon 3200 vs 2200 Opteron) and when using multi processors my pc (using Dual Opteron 246 and 1 Gb) is about 10-20% faster then a dual Xeon 3200!

Besides that, when using AMD, your PC is much more 'multitaskable' running standaard apss whyle rendering heavy scenes than Intel.

Btw: Dual Core AMD is the future! Intel is already far behind!

[MadMax]

For Video editing and other sort like apps Intel still is faster, cause in this 'simple' mather Gh's does count.


Nope.

Intel isn't faster anymore. At best it's a tie. depending on the encoding apps used, Intel wins some tests, AMD others.

even on the apps Intel does win at, it isn't enough to tell without a benchmark to prove it.

[cha0t1c1]

dual core athlon64 is coming in two months u can be patient and have urself a beautiful 64bit dual core that will last u(without problems) 3-4 yrs...

[lots]

Just to be nit picky :)

No AMD processor runs at 3.2GHz on stock settings :P

Also, now that we know that dual core AMD CPUs also have Hyper Threading, this difference should be somewhat of a moot point now.

[maninflash]

No, AMD doesn't support Hyperthreading. It doesn't need it.

HT was a hack Intel came up with to compensate for a poor design. Even with HT, AMD pretty much rules the roost in processor speed. Recent benchmarks had intel's best being trounced by AMD's latest release. Even their main holdouts where Intel ALWAYS did better, Cinema 4D and 3D Studio Max the difference was so negligible that it doesn't matter which CPU you buy anymore.

In media encoding AMD hand Intel it's ass.

With 64 bit and dual core coming, yor money is best spent on the company that pioneered the technology. AMD.

Hyper Threading does make a difference in rendering with MentalRay, almost 4 sec faster on a single frame render on a dual xeon. With GI and Final Gather.

http://www.zoorender.com/

If you see the benchmark list, you see the top (fastest) result had the HT on and beat all others including Opteron 250. (I did that test by the way)

...but dual core AMD processors are another matter.

[lots]

HT doesnt benifit all renders though. It seems to be most effective on some kinds of rendering, and less effective on others, and even drops performance on still others. Pretty much it depends on what you are doing.

[maninflash]

HT doesnt benifit all renders though. It seems to be most effective on some kinds of rendering, and less effective on others, and even drops performance on still others. Pretty much it depends on what you are doing.

I've done quiet a lot of tests with MentalRay and yes, it's true, sometimes HT makes BIG difference, sometimes not much, but I've never had any result where HT has made the performance drop in render time within MentalRay. Although, inside Maya, specialy in SubD modeling, having HT on would cause a performance drop in some actions.

It seems that scenes that require heavy RayTracing or GI would benefit the most from HT.

[imashination]

Also, now that we know that dual core AMD CPUs also have Hyper Threading.

Just to be nit picky, they don't ;-)

[prixatw]

... that will last u(without problems) 3-4 yrs...

Shirley, you mean 3-4 months? :rolleyes:

[redsky_m]

There are a lot of good pro

You can buy any one of them. More important is graphics card and system ram.

i recommend intel Xeon or amd atholn 64 bit.

:thumbsup:

[tfortier]

I was wondering... for After Effects... what is the best? single p4 3.8 - dualxeon - dualopteron - dualcore 64amd???haaaaa... cant wait after effects go 64bit...

[lots]

Just to be nit picky, they don't ;-)

I really shouldn't post while I'm sleepy :)

I went back and re-examined some stuff. Turns out the dual core Opterons DO have an HT bit for thier CPU identification within an OS, but this is used to identify the other core, not to use HT :) Being wrong seems to be what im good at the past few days :P

Ah well. AMD still doesnt really need it anyway, its pipeline is much more efficient than Intel's wonderful NetBurst..

[comrad]

I would get the Intel, all the way. Or i would wait for the 64 bit HT Dual core P4, that thing will be crazy. But yeah, Intels beat out AMD's in rendering, no contest, gaming is another story. But since Intels have the higher clock speed and longer pipelines and HT thats what makes them better than AMD's.
AMD= 6 clocks per cycle
P4= 9 clocks per cycle
thats why AMD's are fast at low clock speeds, but in multitasking INtel is better.

[MadMax]

I would get the Intel, all the way. Or i would wait for the 64 bit HT Dual core P4, that thing will be crazy. But yeah, Intels beat out AMD's in rendering, no contest, gaming is another story. But since Intels have the higher clock speed and longer pipelines and HT thats what makes them better than AMD's.
AMD= 6 clocks per cycle
P4= 9 clocks per cycle
thats why AMD's are fast at low clock speeds, but in multitasking INtel is better.


Sorry but absolutely nothing in your reply is accurate.

AMD has a Higher IPC than Intel. Google it if you are unaware of the facts. That means Intructions Per Cycle.

Intel doesn't come close to beating out AMD in rendering unless you buy an Intel processor with more than 1ghz. higher clock that an AMD, and even then just barely.

Intel's longer pipeline is a hinderance, not a benefit. that is why Intel needs HT, to feed a largely starving for data pipeline.

You might want to actually make sure you know what you are talking about before giving people bad advice.

[comrad]

Intels long pipelines benefit the P4 extremely when you go to 3.8Ghz and above. Especially when you overclock them.

[MadMax]

Intels long pipelines benefit the P4 extremely when you go to 3.8Ghz and above. Especially when you overclock them.

And yet the 3.8 can barely hold it's own against a 2.6ghz Opteron. Go figure.

Opteron beats it in most media encoding apps now, Opteron wins most rendering comparisons with the exception of a couple and even then it is a toss up at best.

Opteron and A64 still have a lot of head room to go in the ghz ratings, Intel has hit a brick wall. Intel can't keep up now, how will the fare later when 2.8 and 3.0 ghz Opterons and A64's are released?

Intel hit a brick wall for pushing ghz and HAD to go dual core. Even then, in comparison, AMD dual cores are eating Intel alive in the benchmarks. Of course this is partly to do with the fact that AMD cores are integrated and communicate with each other directly at the SRQ level, Intel cores have to go external to an out of date FSB to communicate with each other.

bottom line Intel is WAY out in back.

[comrad]

We are not talking about the opterons though. An Intel p4 660 will out perform and Athlon 63 3500 in encoding, multitasking, rendering (sometimes AMD will win, usually intel)

[novadude]

The Intel 660 also costs more than twice as much as the Athlon 64 3500. Compare the Intel's performance to an AMD of the same cost, and then tell me which one performs better.

[MadMax]

We are not talking about the opterons though. An Intel p4 660 will out perform and Athlon 63 3500 in encoding, multitasking, rendering (sometimes AMD will win, usually intel)

Sorry, that's already been proven false.

[comrad]

ummm, in the other thread, i posted you a link, look at those scores.

[MadMax]

ummm, in the other thread, i posted you a link, look at those scores.

Yeah, and I replied to them, pointing out how ridiculous they were since no actual apps were used.

Synthetics have no bearing on the performance of real world apps.

[Vertizor]

Hi comrad, couple things for you.

- Do you know what "clock" and "cycle" means? They have a 1:1 ratio. Take 3 GHz for example. That's 3000 MHz, they call that clock speed. What's a MHz? A Hertz is 1 clock cycle, so a MHz is a million cycles, etc. So I don't quite know what you mean when you say " AMD= 6 clocks per cycle P4= 9 clocks per cycle."

- Long pipelines indeed help chip designers to scale up in MHz. But it also be harmful when you have a branch misprediction. If you dig into Assembly language and CPU documents for both Intel and AMD, Intel's CPUs have a penalty of up to 14 cycles every time it misses a branch in code. Here's the kicker: CPUs always miss the first time around. So when you're in a very big loop of code, that penalty really adds up and you start to see that 3 GHz figure dwindle down. AMD chips handle branch mispredictions a lot better, the penalty is more graceful.

There is some truth into what Apple was saying about shorter pipelines being better. But even they succumbed to having to increase the pipelines just to keep up with the MHz race. That's not to say that more MHz = better performance, and by the same token longer pipelines aren't always ideal.

To cope with the penalty of having to flush and refill those long pipelines, having a fast FSB and fast memory might be the saving grace. So your praise of long pipelines is misguided. You have other things to thank for the performance you see at high clockspeeds. Quite frankly, you should be impressed that AMD is able to achive the performance they are getting at such lower speeds compared to Pentium 4s. The P4 can only rack up more MHz to gain performance, there are no more tricks up their sleeves. And already Intel has hit the ceiling of MHz speeds. They chose not to produce a 4 GHz Pentium 4.

[lots]

Indeed, even Intel admits they are playing catch up.

To add to Vertizor's comments...

The reason Intel even designed netburst (the P4's core design) was to hit as high a MHz as possible. Intel realized after AMD hit the 1GHz mark, that MHz sell. Thus, the Pentium4 was created. It featured longer pipelines, lower IPC, and bigger MHz numbers. It is a well known fact, that clock for clock a P3 was much quicker than this first version of the Pentium4. But as the MHz ramped up, the P4 started leaving the P3 behind. This is the nature of longer pipelines. You pretty much HAVE to up the MHz in order to gain any performance once you lengthen the pipeline, however, it works out nicely that longer pipes lend to more easily reached higher MHz. You can basically see this in every core revision of the Pentium4. Usually, you would see the first few "low" MHz varients of a new core come out, and not really out perform the predicessor, but once they upped the MHz, the new chips often left thier previous versions in the dust. Most notibly in recent events, is the performance of Prescott. However this trick of Intel's only works as long as they can up the MHz. Unfortunately for Intel, they reached about as far as they could go with this type of design ideal, as evident in the cancelation of the 4GHz P4. Having appeared to have spent all thier efforts on bigger GHz, Intel has to now make a complete 180 and head down the same path AMD has been trotting for a while. This will take time, hence why Intel will be playing catch up for quite some time. Though Intel does have something going for it, the Pentium M.

I could be wrong here, but I'm fairly certain we wont be seeing a MHz speed bump from Intel for the rest of the year. Insted, they will be focusing on bringing "features" to the table. Whatever that means :)

[maninflash]

Opteron wins most rendering comparisons with the exception of a couple and even then it is a toss up at best.

If you care to check some real benchmarks, you'll see that Xeon with Hyper threading beats the hell out of Opteron in MentalRay, which is by the way, one of the most widely used renderers in production and is placed only second in final image-quality after Pixar RenderMan.

Source:
http://www.zoorender.com/html/benchmark_mental.htm
www.zoorender.com (http://www.zoorender.com/)

00:41.14
Dual Xeon 3.2 with HT

00:43.83
Dual Opteron 250

for a single frame render with GI and final Gather in Mentalray. And yet, this is not the Xeon fastest CPU's. In production where you have to render thousands of frames, 2 seconds on each frame would add up to some serious time saving.

Maybe you should back up your words next time with some real facts.

[NoSeRider]

I'm typing this with a Dual.....a DUAL!!...AMD 2400 CPU system I built myself......which beats out any single cpu system.

My Computer has a higher floating point then a Single CPU 3.2 GHZ Intel System.
PROOF!! Gawd Damn it!!

http://user.aol.com/nervegastester/pics/multigod.gif
http://user.aol.com/nervegastester/pics/singlecpu.gif

It's not faster, I can just load more crap on it without crashing or chuggin.

[MadMax]

If you care to check some real benchmarks, you'll see that Xeon with Hyper threading beats the hell out of Opteron in MentalRay, which is by the way, one of the most widely used renderers in production and is placed only second in final image-quality after Pixar RenderMan.


Exactly what part of "wins most" didn't you comprehend? Apprently someone needs to explain to you that there is a difference between the words "most" and "all" since you seem to not understand that.

However I have seen those benchmarks, and there is hardly any "beat the hell out of" there. Intel needed more than an 800mhz. advantage to gain a whopping 2 seconds.

Assuming some drift in reliability on user supplied benchmarks, and looking at each step down on the Opteron scores, we can see that the possibility for error is there when different users post Opteron 244 and 248 both as 2.2ghz machines. Then further down, a 244 is listed as being 1.8ghz.

However if I take out operator error and poor building techniques and use the 2 BoXX systems listed, there is a several second increase in speed for one speed grade higher, Opteron 248 up to the 250. Since Opteron has proven to be fairly linear in it's increases, it would stand to reason that a 252 is substantially faster that the 250. When Tech Report reviewed the 252's, that was certainly the case on every single benchmark posted. Mental Ray would be no different.


for a single frame render with GI and final Gather in Mentalray. And yet, this is not the Xeon fastest CPU's. In production where you have to render thousands of frames, 2 seconds on each frame would add up to some serious time saving.

And? it wasn't Opteron's fastest either, so what? I didn't say Opteron was hands down fastest at every single app on the market, not my fault you can't read.


Maybe you should back up your words next time with some real facts.


I provided facts. Maybe you should grow up and stop making personal attacks when you lack the intellect or maturity to deal with them like an adult.

[maninflash]

I provided facts. Maybe you should grow up and stop making personal attacks when you lack the intellect or maturity to deal with them like an adult.

I wasn't attacking you, just reminding you there's a differnce between YOUR personal opinion and the real facts that come from benchmark results. and still, you didn't provide any benchmark data in which Opteron has been faster in render time. That's what we call facts.

Opteron wins most rendering comparisons with the exception of a couple and even then it is a toss up at best.

You see, the proper way to state your opinion on a public forum is to give us LINKS to benchmarks that show a faster time for Opteron and also give us links to those few exceptions that opteron has lost, so people can judge for their own!

Because what would seem an exception to you may be in fact the most of the Fx industry, in case of MentalRay which I know for a fact that Xeon's are faster, based on benchmarks at www.zoorender.com (http://www.zoorender.com/) and in-house benchmarks we've done at work (included at the bottom), and its mainly due to xeon's HT and the fact that MR is a multi-threaded renderer and uses all the CPUs (physical and virtual) available in the farm in parallel.

As you may know, Maya and 3DMAX are the top two widely used applications in the game studios and both of them have MentalRay as their top-quality renderer in addition to the normal software renderer. Now it seems that xeon's are faster in only mentalray and it's just one package, but it happens to cover the majority of the game industry!

so it's not a mere exception.

And let me tell you, 2 seconds time-cut on a single frame is not a "toss up" when you have to render a 5 minute short which has about 540,000 frames.
---------------------------------------------

using the scene from the zoorender available at:
http://www.zoorender.com/benchmark/Benchmark_Mental.zip (http://www.zoorender.com/benchmark/Benchmark_Mental.zip)

Both operating systems are Windows XP. Using MentalRay throu Maya6

00:41:75
Dual XEON 3.2GHz, SuperMicro X5DA8 mainboard, 2GB RAM, with Hyper-Thread

00:47:68
Dual Opteron 248, Tyan Thunder mainboard, 2GB RAM

00:50:88
Dual Xeon 3.2GHz, SuperMicro X5DA8 mainboard, 2GB RAM, without Hyper-Thread

Hard drive for both systems were Seagate Cheetah 36GB 15k, the primary bootup drive with XP and Maya6 installed. RAM for XEON was two Kingston 1GB DDR 266 ECC and for Opteron two Corsair 1GB DDR 400.
----------------------------------------------

[lots]

At 2.2GHz that is fairly respectable :)

From those benchmarks, you can see why Intel needs HT :P

Its really too bad we can't do an apples to apples comparison of the two huh ;) There are always other factors that come into play with Intel vs AMD that will change the results. Different motherboards, RAM, etc. However, if we look at it as how our performance is at the end of the day, then maybe thats worth something. Aside from just talking about rendering speeds on one scene, maybe you could render a wide variety of scenes. Scenes with heavy volumetrics, scenes with no volumetrics, GI, and whatever else. Also compare viewport performance (assuming same video cards). Afterall its not just rendering times that matter here.

Also these kinds of things carry different weight depending on the usage. In a big studio, they can afford to bicker over which chip renders faster, and then dump it into a renderfarm. For the single user, freelance, or whatever, they dont have this kind of buying power, and require a flexible chip that can speed up UI performance, do most of thier daily tasks, and render like a madman.

I'm just saying maybe a better comparison will tell us more, not just ONE scene. You have to test every kind of scene. Both Intel and AMD chips accel at different kinds of tasks. If you test most tasks that come up, then we can see which chip holds the cards at the end of the day. And also keep in mind the use of the system in question. I know for most of us here, we tend to be making personal systems for our home to work and play on (be it video games or art). That means we need something that can render fast, and help improve our other daily program usages.

[cha0t1c1]

I own a 2*244 and I tell you, anything that comes close has to reach double the price...
My next step: 2*270 :drool: :love:

[Mike.H]

i´ve posted it before and i´ll post it again whenever this question comes up.

we live today not in 2002, today there´s realy not much to think about, unless you use a few and very specifiq situations amd64 is your best bet and it´s not even close.

Yes in some situations intel do perform better but they are few and far between.
Another note is that people realy shoudnt put so much personal feelings into a comparision like this check some tests from proper unbiased sites that knows what they are doing and base your purchase on that + your budget if that is an issue.

Now this might all change in a year or two but for now amd should be the clear choice (Especialy if you´re interested in dual core)

Anyhow dont take my word for it.

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2397

[MadMax]

I wasn't attacking you, just reminding you there's a differnce between YOUR personal opinion and the real facts that come from benchmark results. and still, you didn't provide any benchmark data in which Opteron has been faster in render time. That's what we call facts.



using the scene from the zoorender available at:
http://www.zoorender.com/benchmark/Benchmark_Mental.zip (http://www.zoorender.com/benchmark/Benchmark_Mental.zip)

Both operating systems are Windows XP. Using MentalRay throu Maya6

00:41:75
Dual XEON 3.2GHz, SuperMicro X5DA8 mainboard, 2GB RAM, with Hyper-Thread

00:47:68
Dual Opteron 248, Tyan Thunder mainboard, 2GB RAM

00:50:88
Dual Xeon 3.2GHz, SuperMicro X5DA8 mainboard, 2GB RAM, without Hyper-Thread

Hard drive for both systems were Seagate Cheetah 36GB 15k, the primary bootup drive with XP and Maya6 installed. RAM for XEON was two Kingston 1GB DDR 266 ECC and for Opteron two Corsair 1GB DDR 400.
----------------------------------------------

as LOTS stated, it's not as black and white as your misleading remarks imply.

What Tyan board is that you used? K8W? Not a particularly valid comparison if you are trying to prove fastest available under proper conditions. Try that test on a K8WE, using the nForce4 chipset since there is a 400mhz difference in bus speed between the two.

Also you fail to mention that there is a very large difference in how much performance is gained per each speed grade when comparing AMD vs. Intel. Intels gains are much smaller per speed grade than AMD.

However, and more to the point, Anandtech compares mental Ray scores between a 2.4ghz. Opteron 150 and a 3.6ghz Nocona, Intel's 64 bit CPU. Interestingly enough, AMD is faster. I'm far more inclined to take the results from a known professional site, compared to a user submitted database which doesn't appear to be maintained for accuracy, and suffers from no detailed information about hardware used, system setup etc. There are far too many non professional variables to take all of that as more than just a "guide" at best..


Mental Ray (http://www.anandtech.com/printarticle.aspx?i=2163)


And THAT, is called facts.

And 2.4ghz isn't even AMD's fastest. You'll also notice that POV Ray and LAME encoding are faster on AMD.

And as I pointed out before, you are using a CPU a full 1ghz faster, with a faster bus to make your point look valid. Where is your results with a 250? a 252? How about dual core? See the argument gets even more interesting then because AMD dual core figures are leaving Intel in the dust. We know Intel hit a wall at 3.8hz. Based on current results, it's easy to predict the likely scores for the upcoming 2.8ghz and 3.0ghz single core Opterons in mental Ray, even based on flawed .

As I said, I posted all this crap with links in another thread. go look at it, I'm not posting the same stuff over and over everytime someone gets bent out of shape and wants to whine about it.

[maninflash]

as LOTS stated, it's not as black and white as your misleading remarks imply.

And as I pointed out before, you are using a CPU a full 1ghz faster, with a faster bus to make your point look valid. Where is your results with a 250? a 252? How about dual core? See the argument gets even more interesting then because AMD dual core figures are leaving Intel in the dust. We know Intel hit a wall at 3.8hz. Based on current results, it's easy to predict the likely scores for the upcoming 2.8ghz and 3.0ghz single core Opterons in mental Ray, even based on flawed .


Of course it's not as black and white! I agree, that was just a benchmark I did at work on the systems we had, as we only have dual opteron 248 and dual xeon 3.2, I would have tested this on 250 and 252 or even dual core if I had access to them.

And let me clear this before going furthur, I'm not a Xeon fan, neither an AMD fan, I hardly care what system im working on as long as it doesn't give me any headaches and runs smooth, my point regarding your first comment on the subject was to make it clear that first of all Xeon's HT DOES make a difference in rendering with mentalray, and second, it's not always a clear win for Opteron over xeon as it was the impression one would draw from your post.

Thanks for posting the link to AnandTech benchmarks, regarding Mentalray results, I'm assuming they had very limited working knowledge with MentalRay. I downloaded the scene from their website http://www.anandtech.com/printarticle.aspx?i=2163 and did a quick test here on my station. I took some screenshots as the render job was being done to illustrate my point about the advantage of Hyper threading when using Mental Ray.

This machine is a dual xeon 3.2, 1MB L3 cache, a supermicro x5da8 board with 533 Mhz FSB and 2GB ECC RAM, I tested the scene with Mentalray through Maya 6.0.1 twice, once WITH Hyper Thread and then WITHOUT Hyper thread. I just opened the scene with Maya and clicked the Render Button, I only changed one setting with MentalRay that i'll explain below,


First test, WITHOUT Hyper Thread.
Wallclock for Rendering: 52.54 seconds

With HT off, I set MentalRay to use "2 threads", this is under Render menu in mental ray render option box,
http://www.sabudesign.com/kaveh//mentalray_benchmark/mental_ray_2_thread_option_window.jpg


and then clicked the Render button and took a screen shot during the rendering job, notice that mental ray takes advantage of the full proccessing power available on the machine,
http://www.sabudesign.com/kaveh//mentalray_benchmark/mental_ray_2_thread_task_manager.jpg

and this is the result as shown in the output window,
time: 52.54seconds


http://www.sabudesign.com/kaveh//mentalray_benchmark/mental_ray_2_thread_wallclock.jpg

This time is already faster than the time (113.58 sec) shown for Xeon 3.6 on anandtech site, and that's becasue their system was a uni-procesor machine, which I think doesn't make any sense becasue Xeon's as well as Opterons are designed to run in a dual setting and that's where they will show their fullest power.

anyway, back to our results, here are the results from my second test,

Second Test, WITH Hyper Threading
Wallclock for rendering: 41.79 seconds

This time, I changed the mentalray render settings to use "4 threads",
http://www.sabudesign.com/kaveh//mentalray_benchmark/mental_ray_4_thread_option_window.jpg

and clicked the render button, took a screenshot as it was being rendered, shown below, notice how mentalray is using 100% of all four CPUs appearing in the task manager, that's one of the advantages of mentalray that it can use virtual CPUs as well as physical, as if it was rendering on a quad-proccesor system,

http://www.sabudesign.com/kaveh//mentalray_benchmark/mental_ray_4_thread_task_manager.jpg

And the final result, wallclock for rendering, as shown in the maya output window,

http://www.sabudesign.com/kaveh//mentalray_benchmark/mental_ray_4_thread_wallclock.jpg

So you see, 10 seconds faster on the same machine WITH Hyper Threading on and mentalray set to use 4 threads instead of the default one, many people are unaware of this and I assume people at AnandTech didn't know this either, they should have set mental ray to use 2 threads when they tested the xeon 3.6, as im sure they would have gotten a better result.

so my point is, contrary to what many think, Xeon's Hyper Threading ability is not a useless feature. If you know how and when to take advantage of it, it would give you faster results, im sure you would agree that 10 seconds on a single frame render is a considerable time-cut,
of course there are times that having HT enabled would cause a drop in performance, for instance when working with Maya UI specialy in SubD modeling, becasue Maya uses only ONE cpu during the UI calculations no matter how many you have on your system.

so again, many factors come into play when comparing Xeon vs. Opteron, but for the sake of our discussion, and as far as rendering with mentalray, xeon's HT will give you a boost in the race vs. Opteron when it comes down to rendering your film.

over and out!

[NoSeRider]

How do you find that "mental ray render option"?.....to change number of threads. Where is it in Render Globals?...if it's there.

Doh!....hey, I take Maya Community College courses....what do they know.

[MadMax]

so my point is, contrary to what many think, Xeon's Hyper Threading ability is not a useless feature. If you know how and when to take advantage of it, it would give you faster results, im sure you would agree that 10 seconds on a single frame render is a considerable time-cut,


HT. The real issue with HT is this.... Intel went for longerpipelines to ramp up higher clock speeds. The end result is that the processor suffers from basically an idle pipeline that is being starved for Data. That is why with HT off, a 3.8 Xeon is slower in most instances than a 2.2ghz Opteron. Opteron went for short pipleline, higher IPC, at the expense of being able to ramp up high clock speeds easily.

HT is giving an Intel processor what it should have already. It's really a hack work around to solve a design flaw. It works and that is all people care about, but it's a hack none the less.

I did some reading yesterday and I see that Intel's future plans ae for shorter pipelines to improve efficiency and get higher IPC's. That is going to throw their numbering system right out the window when they are trying to explain why a I-920 3.0 ghz, is faster than a P4 Model 666 3.99ghz., (just making up numbers there) Same conundrum AMD had with numbering systems.

Because I have some comments on benchmarks that I want to address separately, I started a new thread about the subject here:

BENCHMARKING (http://www.cgtalk.com/showthread.php?p=2253451#post2253451)

[Raffi_1]

Greetings all,
I am lost now, please I need some help from the professionals here .. :thumbsup:

we are going to buy a new PC, it is for our architectural office, I will use it for architectural renderings, and the main programs used are:
3d studio max, mental ray, Autocad, Photoshop

and maybe maxwell soon . . . :scream:

So I need something fast in rendering, especially for mental ray ....

We are going with 1 processor for now, so what do you suggest, please help ..
I see that you have some experience in hardware ...

I am lost between these 3 choices:

1)
Intel Pentium 4 with HT 3.60 GHz - 660
http://www.intel.com/products/processor/pentium4/index.htm

Asus motherboard 925XE Express
3 GB of RAM DDR2 533 Kingston
Asus 6800GT 256MB PCI Express
H.D. Maxtor 300 GB 16 MB buffer
PSU 500W or 600 W

2)
Intel Pentium 4 Processor Extreme Edition with HT 3.73 GHz
http://www.intel.com/products/processor/pentium4HTXE/index.htm

Asus motherboard 955X Express
[/url] (http://www.intel.com/products/chipsets/925xe/index.htm)
3)
Intel Pentium Processor Extreme Edition 3.20 GHz Dual-Core - 840
[url]http://www.intel.com/products/processor/pentiumXE/index.htm (http://www.intel.com/products/processor_number/info.htm#pXE)

Asus motherboard 955X Express

or maybe AMD if it is better ? ? ?

So what do you think is the best for fast rendering with mental ray ?! ?!

and what about Windows x64 ? ? ?
is it better for mental ray & 3ds max ? ?
will all the 32 bit applications work smoothly in it wihtout any problems or crashes ? ?


Thank you for being patient with me . . :)

[Raffi_1]

Greetings all,
I am lost now, please I need some help from the professionals here .. :thumbsup:

we are going to buy a new PC, it is for our architectural office, I will use it for architectural renderings, and the main programs used are:
3d studio max, mental ray, Autocad, Photoshop

and maybe maxwell soon . . . :scream:

So I need something fast in rendering, especially for mental ray ....

We are going with 1 processor for now, so what do you suggest, please help ..
I see that you have some experience in hardware ...

I am lost between these 3 choices:

1)
Intel Pentium 4 with HT 3.60 GHz - 660
http://www.intel.com/products/processor/pentium4/index.htm

Asus motherboard 925XE Express
3 GB of RAM DDR2 533 Kingston
Asus 6800GT 256MB PCI Express
H.D. Maxtor 300 GB 16 MB buffer
PSU 500W or 600 W

2)
Intel Pentium 4 Processor Extreme Edition with HT 3.73 GHz
http://www.intel.com/products/processor/pentium4HTXE/index.htm

Asus motherboard 955X Express

3)
Intel Pentium Processor Extreme Edition 3.20 GHz Dual-Core - 840
http://www.intel.com/products/processor/pentiumXE/index.htm

Asus motherboard 955X Express

or maybe AMD if it is better ? ? ?
and what about the other components ?? ?? ??

So what do you think is the best for fast rendering with mental ray ?! ?!

and what about Windows x64 ? ? ?
is it better for mental ray & 3ds max ? ?
will all the 32 bit applications work smoothly in it without any problems or crashes ? ?


Thank you for being patient with me . . :)

[mustique]

OK Raffi_1. The short answer is: Buy the best AMD Athlon64 or Athlon64FX cpu you can get.

Or wait a few weeks and get yourself an Athlon64X2. (2 cpu's on one chip)

Best bet would be to get a quad cpu machine with 2 AMD Opteron 275's.

[Vertizor]

ummm, in the other thread, i posted you a link, look at those scores.
I can't seem to find that other thread we all were involved in. Anyways, I do remember seeing those scores from that site you linked.

Now I have a site with benchmarks for you to see.

http://www.xbitlabs.com/articles/cpu/display/athlon64-x2_6.html

http://www.xbitlabs.com/images/cpu/athlon64-x2/winstone-2.png

http://www.xbitlabs.com/images/cpu/athlon64-x2/winstone-3.png


How's that for you're idea that AMD64 (not even Opteron) is worst at multi tasking than P4's? My point is, take benchmarks in reviews with a grain of salt. You're guarenteed to get different results from a different site.

[MadMax]

WOW.

X2's just kick all kinds of ass.....


TECH REPORT (http://techreport.com/reviews/2005q2/athlon64-x2/index.x?pg=1)

[1000101]

X2 is going in my next system. I'm hoping the price on the 4800+'s drop to around 700 for the CPU when I'm ready to buy.

[skoklyo]

well, maybe it's a bit off-topic, but someone here mentioned 540,000 frames for a 5-minute short.

that would be something like 1800 frames per second, man, are you kidding ?!?

even if a typo occured and the number is 54,000 isn't this still a bit of overkill ?

i thought 30 fps is good enough, so with a 2 sec difference this would make 5 hours more for the 5 min short, a considerable amount of time, but as a whole it's only about 5%. 100 against 105 isn't a great improvement nor "beating the hell out of", especially considering the unfair comparison.

[stuartcookson]

I have read right through this thread with interest. I too am interested in buying a new machine in the relatively near future. Thank you, Vertizor, for the link to benchmark tests of the anticipated AMD X2, comparing it to single core processors - it made interesting reading.

However, I am yet to find any comparisons between the AMD X2 and any dual Xeon or dual Opteron systems.

I was going to head in the direction of a dual Xeon setup, but I would prefer to buy AMD if they have good things to offer, and if the price is right. Does anyone have a comparison between the X2 and any dual processor setups? (I use 3ds Max, Mental Ray, Photoshop, Premier etc).

I presume the X2 will cost silly money at it's launch, and therefore is it worth buying/ building a lower-specced machine with a socket 939 motherboard, then just replacing the processor with an X2 in a few months' time when they've come down in price?


Any advice is more than welcome.

[MadMax]

However, I am yet to find any comparisons between the AMD X2 and any dual Xeon or dual Opteron systems.


Well then check this link as it has comparisons against Opterons, Xeons, P4's DUAL core Intel processors etc.

X2 pretty much blazes through everything.

Tech Report (http://techreport.com/reviews/2005q2/athlon64-x2/index.x?pg=1)


And an interesting thing when people go "oh, but it is expensive!!"

You'd spend anywhere from 300.00-500.00 for a dual CPU board. Minimum of 500.,00 each for CPU's.

With X2 it's actually cheaper. 150.00 for a motherboard, vs. 300-500. Rigtht there you have saved as much as 350.00 that can go towards the cost of the CPU.

Plus, the X2 is likely going to be cheaper than the 2 CPU's you would buy anyway, or very close in price.

Since the motherboard is saving you 300.00 or more, you have the bucks to buy 2gb of PC3200 (2x1024) OCZ or similar performance memory for the system.

X2 is easy to justify if you really look at what you are spending.

[IanE]

Well then check this link as it has comparisons against Opterons, Xeons, P4's DUAL core Intel processors etc.

X2 pretty much blazes through everything.

Tech Report (http://techreport.com/reviews/2005q2/athlon64-x2/index.x?pg=1)


And an interesting thing when people go "oh, but it is expensive!!"

You'd spend anywhere from 300.00-500.00 for a dual CPU board. Minimum of 500.,00 each for CPU's.

With X2 it's actually cheaper. 150.00 for a motherboard, vs. 300-500. Rigtht there you have saved as much as 350.00 that can go towards the cost of the CPU.

Plus, the X2 is likely going to be cheaper than the 2 CPU's you would buy anyway, or very close in price.

Since the motherboard is saving you 300.00 or more, you have the bucks to buy 2gb of PC3200 (2x1024) OCZ or similar performance memory for the system.

X2 is easy to justify if you really look at what you are spending.

I totally agree...

Here's an article from AnandTech.com about the new X2's and how the performed against the Athlon FX and the Intel counterparts... X2 and the FX were in the lead for most of the testing, check it out. It also includes AMD's official prices for the chips.

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2410

Ian

[IanE]

Well then check this link as it has comparisons against Opterons, Xeons, P4's DUAL core Intel processors etc.

X2 pretty much blazes through everything.

Tech Report (http://techreport.com/reviews/2005q2/athlon64-x2/index.x?pg=1)


And an interesting thing when people go "oh, but it is expensive!!"

You'd spend anywhere from 300.00-500.00 for a dual CPU board. Minimum of 500.,00 each for CPU's.

With X2 it's actually cheaper. 150.00 for a motherboard, vs. 300-500. Rigtht there you have saved as much as 350.00 that can go towards the cost of the CPU.

Plus, the X2 is likely going to be cheaper than the 2 CPU's you would buy anyway, or very close in price.

Since the motherboard is saving you 300.00 or more, you have the bucks to buy 2gb of PC3200 (2x1024) OCZ or similar performance memory for the system.

X2 is easy to justify if you really look at what you are spending.

Well then check this link as it has comparisons against Opterons, Xeons, P4's DUAL core Intel processors etc.

X2 pretty much blazes through everything.

Tech Report (http://techreport.com/reviews/2005q2/athlon64-x2/index.x?pg=1)


And an interesting thing when people go "oh, but it is expensive!!"

You'd spend anywhere from 300.00-500.00 for a dual CPU board. Minimum of 500.,00 each for CPU's.

With X2 it's actually cheaper. 150.00 for a motherboard, vs. 300-500. Rigtht there you have saved as much as 350.00 that can go towards the cost of the CPU.

Plus, the X2 is likely going to be cheaper than the 2 CPU's you would buy anyway, or very close in price.

Since the motherboard is saving you 300.00 or more, you have the bucks to buy 2gb of PC3200 (2x1024) OCZ or similar performance memory for the system.

X2 is easy to justify if you really look at what you are spending.

I totally agree.

I found this article at AnandTech.com and it's really great. Covers a lot of aspects on the X2s, comparing it to the Intels, to the Athlon FX and so on. It also lists prices from AMD and other stuff.

Check it out http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2410

Ian

[maninflash]

Well then check this link as it has comparisons against Opterons, Xeons, P4's DUAL core Intel processors etc.

X2 pretty much blazes through everything.

Tech Report (http://techreport.com/reviews/2005q2/athlon64-x2/index.x?pg=1)


And an interesting thing when people go "oh, but it is expensive!!"

You'd spend anywhere from 300.00-500.00 for a dual CPU board. Minimum of 500.,00 each for CPU's.

With X2 it's actually cheaper. 150.00 for a motherboard, vs. 300-500. Rigtht there you have saved as much as 350.00 that can go towards the cost of the CPU.

Plus, the X2 is likely going to be cheaper than the 2 CPU's you would buy anyway, or very close in price.

Since the motherboard is saving you 300.00 or more, you have the bucks to buy 2gb of PC3200 (2x1024) OCZ or similar performance memory for the system.

X2 is easy to justify if you really look at what you are spending.

Not meaning to start another war here, as this thread has had it's own share already. But another thing to consider about a 2 CPU system is that it's much more flexible than a one-cpu system, on a dual machine you can assign different tasks to each CPU as they work completely seperatly from eachother (I usualy set one to render the tests while I work with the other in Maya) or if you're more of a programmer, you can write a little script that would set aside portions of memory for each CPU, so if you're working on a heavy scene that eats memory, have more RAM assigned to CPU one and let the other have the rest, this would become even more handy if you work in Linux.

[MadMax]

Not meaning to start another war here, as this thread has had it's own share already. But another thing to consider about a 2 CPU system is that it's much more flexible than a one-cpu system, on a dual machine you can assign different tasks to each CPU as they work completely seperatly from eachother



Dual core shows up in task manager the same as if it were 2 CPU's. You can still assign tasks and theads on a per Core basis, so I'm not quite sure what you are getting at here?

[Vertizor]

Memory allocation (should be) assigned per process, not per CPU. Why go through the trouble of managing memory manually? Your OS takes care of that. I don't see your approach as beneficial unless you were running a virtual machine.

[stuartcookson]

Well then check this link as it has comparisons against Opterons, Xeons, P4's DUAL core Intel processors etc.

X2 pretty much blazes through everything.

Tech Report (http://techreport.com/reviews/2005q2/athlon64-x2/index.x?pg=1)





Many thanks for that link, MadMax, I think it firmly puts the X2 at the top of my shopping list.


Stuhttp://www.cgtalk.com/images/icons/icon7.gif

[maninflash]

Dual core shows up in task manager the same as if it were 2 CPU's. You can still assign tasks and theads on a per Core basis, so I'm not quite sure what you are getting at here?

sorry, I didn't know that.

[1000101]

Also this differs from hyperthreading in that it is a second phyiscal CPU core and not merely idle portions of a single core p4.

[bumskee]

AMDx2 is going to be really good, I decided to skip venice core upgrade and wait for dual core. It is really impressive, and looks to overclock close to 2.6~2.7ghz for 4400. the price diff between that and 4200 is too small for an extra 512 cache.

some reviews?
http://www.lostcircuits.com/cpu/amd_x2/
http://www.bit-tech.net/hardware/2005/05/09/amd_a64x2_4800/1.html
http://firingsquad.com/hardware/athlon_64_x2_4800/
http://www.hardwaresecrets.com/article/136
http://www.hexus.net/content/reviews/review.php?dXJsX3Jldmlld19JRD0xMTcw
http://www.pcworld.idg.com.au/index.php?id=1961161743&eid=-108
http://www.simhq.com/_technology/technology_042a.html
http://www.sudhian.com/showdocs.cfm?aid=672
http://www.techwarelabs.com/reviews/processors/amd_x2/
http://www.trustedreviews.com/article.aspx?art=1370
http://www.xbitlabs.com/articles/cpu/display/athlon64-x2.html

[NoSeRider]

http://www.sabudesign.com/kaveh//mentalray_benchmark/mental_ray_4_thread_option_window.jpg

All I want to know is how do you open up that Multi Thread Render Display in Maya?

I don't care about which cpu is better, since I'm not gonna rip apart my machine and get the latest stuff.

[imashination]

Whats the ETA on shipping for consumers?

[MadMax]

Whats the ETA on shipping for consumers?

Apparently they are already being shipped in quantity to OEM's. Systems will be available in June.

It has been myexperience that vendors such as Monarch tend to get at least limited quantities of CPU's fairly shortly after release.

in other words it's not going to be in 2006. It's not going to be 4th Q this year.

I'd expect to be able to get some, even if pricey in late June early July.

[imashination]

Might have a to grab a couple and load up the ram slots, especially as they now remove the 333mhz speed limit of using a fourth memory slow :-)

I wonder how long it will take the mobo makers to update the bios on them though. Ive got an msi k8n neo2 if anyone knows. The latest bios readme makes no mention.

[MadMax]

I wonder how long it will take the mobo makers to update the bios on them though. Ive got an msi k8n neo2 if anyone knows. The latest bios readme makes no mention.


Uh, Not sure how to break this to you, but I read something at another forum that the MSI boards wouldn't support dual core. And there was a link to MSI forums regarding the topic which I did take a quick glance at.

Not being a big MSI fan, I just don't know what models it entailed, and I didn't bother to research too deep. it might be a different version than yours............

[maninflash]

All I want to know is how do you open up that Multi Thread Render Display in Maya?

I don't care about which cpu is better, since I'm not gonna rip apart my machine and get the latest stuff.

Choose "Rendering" from the menu pulldowns on your status bar, or press alt+h+LMB and select "rendering", in the main menu, under "Render" go to "Render Using" and select mental ray (if you don't have mental ray as an option there, you should first load it, under windows, setting/prefrences, plugins, check mentalray plugin "loaded".)

once you set mental ray as the renderer, under Render menu, select either single frame render option box or batch render option box to tell mentalray how many threads to use.

[iambic]

Thank god there are at least two major cpu manufactures out there striving to make better cpus for us. The competition between amd and intel is fantastic and we the end users reap the rewards...I'm an intel user but that doesn't stop me from suspecting amd make a better cpu.. but I would just like to point out that in terms of volume intel dwarfs amd and their fabrication technology is also more advanced...they got a bigger wafer and can thus produce their chips more cost effectively. It will be intel driving down the prices of the about to arrive dual cores not amd who will entering at the top because they can't produce the numbers. Amds problem is not their cpu technology everyone agrees it's great it's their manufacturing capacity...All the major review sites seem to thing that the availabilty of dualcore opterons and x2's at reasonable prices will be patchy.

[novadude]

AMD has a new fab coming online soon, and they know that the only thing keeping them from taking over the market is production ability.

[iambic]

Intels strategy to combat amd tech advantages will be to try and totally comoditise the chip market so price and volume is the central issue,not technology. I don't think there are any magical cpu cores inside intels hat. Apart from the improvements that will come from pentium M tech, intels and amds single core cpus are close to maturity, which leaves us with price and volume....intel will try and squeeze amds margins by selling chips as cheap as they can which is good news for us but bad if amd goes under...and unfortunately I think that is what intel hopes...Make no mistake intels strategy has always been to try and wipe amd completely out of the chip business. With the current development of cpus maturing an end game could be closer than people imagine. Amd cannot survive as a niche/boutique player, they have to compete in volumes and which company do you think can survive longest by making very little on their chips? The overheads of chip companys are enormous and can suck them down the gurgler with startling alacrity if they are not making money....

[MadMax]

but I would just like to point out that in terms of volume intel dwarfs amd and their fabrication technology is also more advanced...they got a bigger wafer and can thus produce their chips more cost effectively

A slight correction here, Intel's fabrication technology is absolutely NOTY more advanced. AMD's Dresden facility has been hailed as the most advanced fabrication center on the planet. AMD is building a sister plant which goes online in 2006.

Also, it has been fairly widely reported in the past about Intel and lower yeilds on their wafers.

It will be intel driving down the prices of the about to arrive dual cores not amd who will entering at the top because they can't produce the numbers.


That is accurate, but not for the reasons you think. Intel doesn't have a true dual core solution yet. These so called dua' chips are 2 prescott cores literally glued together.
To be able to communicate with each other, they have to go out to the FSB to talk, they are not an actual integrated solution.

Intel doesn't really have much in the way of manufacturing and R&D costs to recover, so htey can pass them off on to unsuspecting buyers cheaply.


Amds problem is not their cpu technology everyone agrees it's great it's their manufacturing capacity...All the major review sites seem to thing that the availabilty of dualcore opterons and x2's at reasonable prices will be patchy.


True. However AMD has a second major league fab opening next year. this will double their capacity to produce. They have been gaining marketshare with Opteron and A64, pushing 17%, and with a second fab in Dresden, they will be able to double their production and transition to new processes faster than before. It's much harder to go from 130nm to 90nm when you only have the one facility, or from 90nm to 65nm or smaller.

When you have 2 fabs side by side, it's much easier to re tool one, when the other is still producing CPU's.

[novadude]

With recent development over in Japan, Intel might be rethinking their bully strategy. Granted it was just a slap on the wrist, but other countries will follow and the slaps will get progressively harder as every country tries to exploit them on this.

AMD also needs to start a new fab so they are ready to produce more of their quad core chips. They should have started that fab last year if they want to meet their 2007 projected deadline, but they need that as soon as they can.

[iambic]

Intel doesn't have a true dual core solution yet. These so called dua' chips are 2 prescott cores literally glued together.
To be able to communicate with each other, they have to go out to the FSB to talk, they are not an actual integrated solution..

You know this, I know this, as does intel and amd...my point is...is that intel will try to shift the focus away from technology and towards price and volume. I think intel freely concedes amds tech is better...they gave them an award the other day....what they're saying is that we'll give you a chip that's almost as good but significantly cheaper and we can supply you the volumes. Dells decision not to go with amd the other day had more to do with not being able get the volumes not dislike of the tech or worrying about pissing intel off. Intel will play to its strengths which is capacity not necessarily technology.

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.