Looking for 3d rendering machine(s) - AMD or Intel?

scarredtaka

Weaksauce
Joined
May 21, 2007
Messages
112
Hi,

I've got a laptop I use for 3d rendering, and I am unhappy with the heat it puts out, and am afraid it will fail like my last one did. So, I am thinking about selling it, and using the funds from that to build a new desktop.

I will be using this machine pretty much only for 3d rendering work. The gaming I do is inconsequential.

My budget is $900-1200 with shipping. Shipping will be to Fairbanks, AK, so expensive.

I will be looking for the whole setup. Case, MB, CPU, Video Card, RAM, PSU, HD, CD/DVD.

I already have speakers, Keyboard, Mouse, 27" monitor @ 1920x1200, 2x 19" monitors @ 1440x900 .

I plan on building the computer at the end of June or sometime in July.

I just need the basics in a motherboard, except for the ability to run 2 video cards, but NOT in SLI.

I will need an OS. Windows Vista basic is fine.

First off, my experience so far.

My laptop has a 2.4 ghz Intel Core 2 Duo chip, 4 gb of ram and an 8600GTM w 256mb ram.. Doing 3d rendering heats it up to 80-85C. This laptop is actually a warranty replacement for my last laptop, which died due to heat issues. This is with the laptop propped up on an active cooling pad.
This is the machine that would be sold.

My desktop is a AMD Phenom x4 9550 @ 2.2 Ghz stock, with 4gb ram, and a 9500GT 512mb card, which easily takes care of all my needs, both for rendering, and gaming.

Despite my laptop only having 2 cores, and my rendering software able to take advantage of up to 4 cores, my laptop is consistently 20-30 seconds per frame faster when rendering.

This may not seem like much, but when applied to 2-3000 frames, the difference adds up.

So, what I am looking for is a speed increase. I thought simply moving to a quad core machine would give me that, but it appears not.

The other thing I would like with this machine is the ability to run 3 or 4 monitors. They would be running at 1920x1200, but only one would be displaying any significant motion at a time. I.E. palletes/workspace/palletes. My program is 90% cpu when rendering, and only uses the video card when displaying previews (openGL). So running with 2 9500GT's would be plenty in the video card department.

Finally, if I spend a little bit of money on my graphics program, (don't include in budget) I can gain the ability to run a "render farm" and distribute work over multiple computers.

So, with this in mind, where should my money go?

Should I upgrade my phenom system to a phenom II (Probably a 940 @ 3Ghz), and then build a second Phenom System like it?

Or should I build/buy an i7 system?

The i7 is faster, yes, but what if used in conjunction with the phenom?

Would an i7 / phenom setup be faster than a Phenom II/PhenomII setup?
i.e. would the bottleneck of the original phenom vs the i7 even out against the phenom II's being faster than the phenom I?

My program would "hand out" frames to the other computer to render (over the network), which would render the frame, send it back to the "master", then ask for a new frame.

I guess what I'm really asking is whether to go with AMD or Intel. Whatever I build I will be stuck with for a while. (3-4 years, probably.)

Doesn't have to be scientific. If you have a general idea, please share.

Thanks very much

-ST
 
The Phenom and the i7 are close performers in gaming. But, the i7 is 50% faster clock for clock in editing and rendering. The Hyperthreading really comes in use.
 
%50? love to see some numbers to back that up, sure it is application dependant

i7 is dam fast, i would go i7 for sure either way.
 
Thanks for the replies.

Kinda what I thought would end up happening. From what I've seen the i7 is a bit faster than the PII, which is itself at least a bit faster than the PI.

So, i7 it is, then.

Do you think it is worth saving up and dropping in a PII chip into my current system in place of the PI?

Thanks,

-ST
 
With your budget, I'd say Core i7 build. Even if you O/C an X4 920 Phenom II Deneb, it still can't touch a Core i7 920 running at stock 2.66ghz speeds. No doubt man the Core i7 are superior in rendering and video encoding.
 
Thanks for the replies.

Kinda what I thought would end up happening. From what I've seen the i7 is a bit faster than the PII, which is itself at least a bit faster than the PI.

So, i7 it is, then.

Do you think it is worth saving up and dropping in a PII chip into my current system in place of the PI?

Thanks,

-ST

yeah your looking at a 15-20% increase clock-for-clock from PI to PII, and seeing as how PII chips are pretty cheap, and come in SKU's well in excess of 2.2GHz, it might be worth it. Your call of course.

From what I've read and heard, the i7's are pretty superb renderers.
 
While everyone is saying i7, you may want to look at AMD's multiprocessor offerings or see if the programs you use can network render. If so, it may be cheaper to build a small render farm instead of having a single buff computer.
 
While everyone is saying i7, you may want to look at AMD's multiprocessor offerings or see if the programs you use can network render. If so, it may be cheaper to build a small render farm instead of having a single buff computer.

Yes, I can upgrade to a network render for a bit of money.

-crazjayz
That's just what I was looking at. For about $1400 I could build 2 Phenom II x4 systems (AM2+) @ 3Ghz w 4GB of Ram. I think those, plus my Phenom 9550 @2.2 Ghz would outrun an i7 plus my phenom.

12 cores, 12GB Ram, ~32Ghz of processing power. That is tempting!
 
The Dell outlet has some T5400s on sale for that price range (albeit the lower end configs.) For the most part they're all quad core systems, you can add on a second cpu for around $200 and up. I do believe shipping is free from them as well (might have to check on that though.)
 
what motherboard do you currently have your phenom 9550 running on? because honestly.. if it was me.. id save the money and slap in a phenom II 955.. overclock it to 3.8-4.0ghz.. should be 150%+ in performance over your laptop and about 100%+ in performance over the 9550.. the higher L3 cache makes a huge difference with the phenom II's over the old phenom's.. not to mention the old phenom processors are a piece of junk compared to the phenom IIs
 
what motherboard do you currently have your phenom 9550 running on? because honestly.. if it was me.. id save the money and slap in a phenom II 955.. overclock it to 3.8-4.0ghz.. should be 150%+ in performance over your laptop and about 100%+ in performance over the 9550.. the higher L3 cache makes a huge difference with the phenom II's over the old phenom's.. not to mention the old phenom processors are a piece of junk compared to the phenom IIs

-sirmonkey

I am currently running a MSI AM2+ microatx board

http://www.newegg.com/Product/Product.aspx?Item=N82E16813130187

It looks like this chip (AMD Phenom II 940 @3.0Ghz) would work in the board I currently have.

http://www.newegg.com/Product/Product.aspx?Item=N82E16819103471

However, dropping in a new chip would also leave me with an extra chip doing nothing, which I don't like the thought of.

Also, this board only has one PCI X16 slot, and therefore can't drive all 3 of my monitors, without me buying this card, which doesn't seem to be getting the best reviews.

http://www.newegg.com/Product/Product.aspx?Item=N82E16814102813

That would also probably require a new power supply as well, as I'm currently only running a 400Watt.

I think it may be more logical to build a second computer around a Phenom II, and be able to use it in conjunction with my current computer, since by updating my software, I can gain the ability to run a "render farm", which I would think would be faster than just the one machine alone?

No matter how fast any one single computer is, I'm beginning to think that 2 or 3 computers, slightly slower, lets say even %50 slower, would end up being faster in the end, since they would be able to work 2 or 3 frames at once.

Does that make any sense?

Thanks,

-ST
 
The Dell outlet has some T5400s on sale for that price range (albeit the lower end configs.) For the most part they're all quad core systems, you can add on a second cpu for around $200 and up. I do believe shipping is free from them as well (might have to check on that though.)

-Keiichi

Took a look at that, and am not sure how well that would work. My program will only take advantage of up to 4 cores, so building a machine with more than 4 doesn't seem like a good use of resources. There is virtualization, but I'm not sure how well that works when dealing with intensive processes.

What I would have to do to get maximum performance out of a dual quad core machine is to run one network renderer on the host OS, and have a second renderer running on a "guest OS". I have no idea what kind of performance hit I would see while using a VM in this kind of situation.

Ah!
The other thing I see now, is to run 2 instances of the renderer on the same computer, and just set processor affinity?

Can you do that? Can I tell 1 instance to run off the first chip, and the second one to use the second chip?

Also, I'm not sure if I can run two instances of the network renderer on the same computer, as the render is limited to 3-10 computers, (depending on the version) and has it's own license key.

That's research I would need to do.

If anyone has experience with using VM's with intensive processes, or using a dual CPU machine to run more than one task at once, and can chime in thoughts or ideas, that would be great.

Thanks,

-ST
 
Back
Top