There is never a single computer build that will suffice for everybody so I’ll explain each components options and in another tutorial ill build a pretty generic build for 3d work and suggest where tweaks would be needed for certain applications.

There isn’t an order of importance of where to throw your money at however there are 3 areas that you want to focus most of your research on.

-The CPU
-RAM
-Video Card

This knowledge base got REALLY long. Incase you didnt know, most browsers have the ability to search text by hitting the key combo "ctrl+f" and it will bring up a search box. On OS X it's "open apple+f"

The CPU is the workhorse of the computer, it’s the main brain and it does all the calculations that you tell it to do via the programs on the computer.  In 3d, rendering is extremely demanding on a CPU.  As of right now there aren’t any “good” renderers that utilize your graphics/video card’s processor thus, everything, as far as final image processing, is left up to the CPU.    Currently the best processor/CPU for the dollar is the Intel Core 2 Quad Q6600 Kensington.  Its, essentially, two Intel Core 2 Duo Conroe E6600’s jammed onto one circuit, but that’s beside the point.  The Q6600 is currently priced at 189.99 (as of September 30, 2008) and it will probably continue to slowly drop every month as it has been, and all technology does. 

Are there better CPUs out there? Of course, there are always bigger, badder, faster CPUs but most 3d people are forced to either upgrade or replace their computer every 2 years or so because of the stresses that are employed by rendering, so getting the best performance while not spending over a grand for a cpu is important.  And besides, Intel just announced the 6 core processor and according to Moore’s Law there should be a 16 core processor coming around January of 2009.

There is a ton of information about CPUs and how they work, and in order to ensure you are getting a good processor you should make sure to compare and confirm those settings/features will have a positive effect on your work (they actually mean enough to warrant the price increase.)

A few things to watch out for when choosing a CPU.
Socket Type – This HAS TO MATCH the socket receptacle on the motherboard.  The Q66600 is a socket type: LGA 775.  LGA 775 stands for “Land Grid Array” and all that means for you is that the CPU pins are on the receptacle NOT on the CPU.  Back in the day the CPU itself had all the pins on the little chip.  This was risky when building your own computer as it was easy to bend these pins, and in turn, ruin the CPU.  If you manage to break a CPU pin on the motherboard it’s almost always cheaper to replace the motherboard than the CPU. And CPU returns/RMAs are never easy anyway.

L2 Cache - This is a CPU cache that is used to store repeatedly used data.  This is different from ram.  There is indeed a performance boost when you have a larger L2 cache, HOWEVER, not when rendering 3d.  It sounds weird I know, The CPU has to continuously take in data and output data, having a larger cache should benefit it.  It should, but it doesn’t.  Pretty much everything a CPU has to read and write while rendering is too large or not accessed enough to warrant storing it in the CPU’s cache so it is stored in the ram.  That’s not completely true, some stuff is stored but again, not enough to warrant a performance boost while rendering, the benefits are too insignificant.  Now I’m not saying disregard this statistic because for other processes it does benefit.  Compression, reformatting, CPU aided graphics, etc will be enhanced with this larger cache, just don’t throw a ton of money at a chip with a lot of L2 cache.

FSB- Front Side Bus- Front side bus is the bus (no not the big yellow thing, it’s a piece of hardware that can logically carry data across the same set of wires) that carries data between the CPU and the north bridge.  This is normally measured in MHZ and this number is the “clock frequency.”  It is used to calculate how many data transfers occur per clock cycle.  Basically, every time the CPU makes a calculation; this number represents the amount of data that is transferred as a result.  All in all a higher number is preferred but going overboard is not worth it with 3d.

Northbridge- This is the memory controller hub, it controls the flow of the data between many of the critical processing chips/parts, ie… CPU, RAM, AGP or PCI-E, and the southbridge.

64 Bit support- This feature is CRUCIAL, assuming you’re actually doing any 3d that is even remotely challenging.  I’ll explain more later and will touch on it in the RAM section.

Hyper Threading- This is an Intel exclusive feature.  It was something for Intel to brag about and has now been, for the most part, phased out.  In its simplest, form it allowed data to be sent and received at the same time and that allowed parts of the processor to be used that would otherwise be sitting idle.  It is now only available on the xeon processors used in servers but was in the Pentium 4s and Atom processors.  Many people claim it is highly inefficient, requiring much more power than actually warrants its benefit.  To sum it up- it’s not worth it right now and if you follow my guidelines you’ll get better performance anyway.

Voltage- This is pretty straight forward, it’s how much power the CPU uses during its normal functions, usually the low number is idle and the high number is under full load.  This is a handy number when planning how much power you will need for your computer.

Operating Frequency – Back in the day (the arbitrary day) this number was actually usable when comparing the speed of a processor.  Not so much anymore, if at all.  CPU manufacturers have begun separating these numbers by CPU class, and with a legitimate reason.  As CPUs began taking on multiple cores you couldn’t just double the operating frequency, this number is an actually quantifiable number, it’s the number of clock cycles per second, or the number of calculations performed per second.  Take, for example, the Q6600 vs the E6600.  Both processors have an operating frequency of 2.4 Ghz.  As I stated earlier the Q6600 is essentially two of the E6600’s jammed on one die (socket) and as one would expect, doing the same process on both processors results in the Q6600 performing at TWICE the speed.  And I have both of these processors and have confirmed this, or you can look at benchmarks elsewhere. Either way, take this number however you will, I no longer hold it with much merit as it is only useful for comparison within the same processor class, and those classes are hard to distinguish, unless you’re Intel. 

Ram is, more or less, the servant or even servants to the CPU. The acronym R.A.M. stands for ‘Random Access Memory.’  The RAM’s job is to hold temporary data for the CPU, either to process or that it has just finished processing.   I made a pretty simple diagram below that shows the order of data transfer when you request the CPU perform a task.  No matter what you request the CPU to do it has to go through the RAM.

The arrows just indicate the flow of data, that line cant be interrupted, there isn’t any way to just skip the RAM and write directly to disk.  As processors speed up, RAM is often the first component that PC repair guys/gals assume is lacking, and for good reason as it is often overlooked when buying/building a computer.  There are definitely limitations to how much RAM a computer can handle but ill explain that in the things to know about RAM section.  When a request is sent via input device (mouse and keyboard, or touch screen)  the first thing the CPU does is request data that is being stored on the hard drive.  Because the hard drive is not fast enough to read and write to the CPU it hands that data off to the RAM, which in turn, hands a portion of that data or all of it off to the CPU.  Because the CPU can hold a very limited amount of data (anywhere from 1-24 megabytes) it has to keep handing data back and forth to the RAM.  This may seem inefficient but it’s really not all that inefficient with the speeds of the CPUs now.  If you think of the CPU as a small child with hand size respective to their age and the RAM as 3 or 4 servants standing waiting and the hard drive as a filing cabinet.  The child requests something and the servants instantly grab it from the cabinet and hand parts of it to the child, the child can only hold so much and has to begin handing stuff back in order to grab other stuff.  Now multiply that by a few million times a second!  As the CPU finishes calculations it hands it back to the RAM who then tells the hard drive to write that data to storage. RAM can quickly become a bottleneck and it wont even be obvious, you will just think your computer is slowing down because of too many files or a virus or something.  When you run out of available RAM the RAM has to write to a temporary location on the hard drive in order to take more data, when that data it wrote to the hard drive is recalled it has to dump enough of the data it is currently holding to the hard drive so it can pick up the data that is being requested. A general rule of thumb, don’t have less than 1 gig of RAM per core of CPU power.  That being said read the next section about ram as it is VERY important.

Things To Know About RAM.

Windows XP Pro, XP home, Vista Mac OSX 10 Tiger, and before, all ONLY see 3 gigs of ram.  Yea, you’re telling me.  Ok fine, don’t blame the software, blame the developers that figured we would never need more ram than a 32 bit operating system can support.  Any operating system that is 32 bit can only handle 4096 megabytes of RAM.  Yes, that is 4 gigs, so why will it only see 3?  That’s because the normal operation of the hardware in the computer utilizes RAM as well, usually this is somewhere under a gig, leaving, you guessed it, 3 gigs for everything else. 

There is currently one way around this, and one way only.  You must use a 64 bit operating system.  Windows XP Pro x64, Vista x64, OSx Leopard are designed to run on 64 bit architecture (hardware.) ALSO, the programs you use on 32 bit, while they will run (most of them) on 64 bit, they are almost all programmed to not even look for more than 3 gigs of RAM.  If you have a need for more than 3 gigs of ram then they will usually make a 64 bit version of that software, ie… 3D Max 64 bit. Now this doesn’t mean it’s not worth it to put in more than 3.  Say you want to run 3 or 4 processor intensive programs at a time as I often do, ie… Premiere, Photoshop, Max, each of these programs can see 3 gigs, if I have 8 gigs of RAM each program has 2 and 2/3s gigs of its own RAM.  This means much less writing and reading to disk, aka saving time!  Now, onto the technical words to look out for when buying RAM.

RAM is also called Memory, for short.

Type

This is usually a Pin count followed by the common ‘type’ of memory setup. Ie… 240 pin DDR2 SDRAM.  The 240 pin matches the number of receiver pins on the motherboard’s memory recepticles, always look to see what the motherboard manufacturer suggests you use for RAM.  The DDR2 SDRAM is a long term (Double Data Rate two Synchronous Dynamic Random Access Memory) that basically means it is setup to have two (synchronous) data blocks working at the same time, in turn speeding up the processes.  There is SDRAM (DDR2) and the older DRAM (DDR) and there is now DDR3. Most of the motherboards sold now take SDRAM (DDR2.)

Capacity

Usually presented in a “#GB (# x #GB)” sort of fashion.  This shows the total Gigabytes and then how it is divided between the stick/s.  So if I bought a pair of 2 gigabyte sticks it would be shown as “2GB (2 x 1GB)”  or maybe they would present it as “2GB (2 x 1024MB)” 

Speed

This is usually a number that the motherboard requires and dictates.  For example, DDR2 800, the number on this represents the number of data transfers per second, NOT the MHZ.  This is a common misconception, in fact the MHz is ½ of this number, so DDR2 800 would have a clock speed of 400 Mhz.

Cas Latency

Cas is an acronym as well, standing for (at least what I was told) Column Address Strobe, apparently there are other acronyms used.  Either way the acronym isn’t important, the number is.  The number tells you the amount of time (in clock cycles, NOT seconds) it takes from when the memory controller tells the RAM which Column and Row the data it needs is to the time that data hits the pins on the motherboard.  The higher the number the longer the RAM takes to retrieve the requested data.  Do some googling on these numbers as depending on the CPU and RAM type a one number variation can be either huge or tiny in terms of time spent waiting.

Timing

These numbers, displayed in a 4 number sequence, each represent a certain timed process that the ram performs.  One of these numbers is the CAS latency as just covered.  As a side note you need to watch out that the ram isn’t listed as a say, CAS Latency 4 but all the other timings are 5 and the last is 15, They are just charging you more for the faster CAS latency but the ram isn’t any significant amount faster.  Anyway, the numbers, if you really wanna know what they stand for, they are; tCAS – CAS Latency, tRCD -RAS to CAS Delay, tRP - RAS Precharge, tRAS.  If you have some crazy hankerin to know exactly what each one stands for, google it!  Each one is so specific it would take its own page to explain.  The jist is – the lower the numbers the faster the ram.

Voltage

This is the amount of power each stick requires in order to perform at the advertised speeds.  This is helpful in calculating for a power supply and also must be known when setting up your computer.  Some motherboards require a certain voltage or cant handle above a certain voltage so it is important to look at the motherboard manufacturers numbers first.

Heat Spreader

This just indicates if it has a heat spreader of any sort to radiate heat.  As is with all electronics, the cooler you can get them to run the better and faster they can perform.  (If you ever get bored and wanna see what some people do to get their CPUs cooled so they can overclock them search for “CPU Liquid Nitrogen” its pretty cool!)

Warranty

The GPU (Graphics Processing Unit) is quite important when doing 3d creation or image manipulation or gaming, among a few other things. 

Don’t think having a 2000 dollar SLI’d GPU setup will make your rendering AWESOMELY fast, it wont.  (SLI I’ll explain later if I confused you.)

When working in 3D the GPU is used for real-time presentation of the workspace.  So when you’re toolin around your model that’s 4 million polys and it starts to lag, that’s your GPU hitting its limit.  Think about everything it’s doing, it’s calculating and sending data to be displayed at the fastest rate it can.  It has to figure out anti aliasing on all the 3d objects, calculate the wires in the wireframe mode, output the software’s GUI, calculate some sort of crude lighting that most 3d packages have setup.  (You can get better performance out of your current gpu by turning down viewport anti aliasing and having it display textures at a lower resolution, all should be controllable within your CG package’s  display settings.)   

If it wasn’t obvious yet, the GPU does a lot to make everything look purty for you while you’re working.  The good news, as of late (the past year or so), the price of desktop GPUs has come way down and the performance has exploded upwards.  As a comparison, my current video card cost me about 160 bucks a few months ago and handles 6 million polys in my viewport like a champ.  My school had 1500 dollar cards that were 2, or so, years old……. Opening a scene with a little over a million polys could take quite a while, and good luck navigating it! 

Before you go a drop a grand or more on a video card because you think you need it check out some 3d specific benchmarks, theyre all over the place.  (Sorry if that sounded mean, I know too many people who have spent more on their video cards than they do on their processors and they will never utilize that GPU’s power!  If you are one of them, shame on you.)

Things to look for in a Video Card

As a beginning note I need to mention, Video cards can be manufactured by different companies and still utilize the same chip, for example, the 8800 GT (my card) is made by BFG, PNY, Asus, etc.  The chip itself is manufactured by NVIDIA, to their specs, and is then placed on a board with circuitry manufactured by the card maker company.  Think of the chip like the CPU of the computer and the manufacturers board like the motherboard of the CPU, now miniaturize it and dedicate it to graphics display.

Interface

There are 2 different interfaces, PCI Express (aka PCI-E) and VGA (aka VGA.)  PCI Express is the newer of the 2 and has, for the most part, much better performance for a significant price decrease.  You CANNOT interchange the two and NO, you CANNOT use a PCI Express video card in a PCI slot, it has to be a PCI Express Slot.  This interface is specific to the motherboard; check which receptacle the motherboard has, though I strongly suggest you get a mobo (motherboard) with a PCI-E slot.

Chipset Manufacturer

This tells you who made the Video Processor chip.  I have used both NVIDIA and ATI, as a blatant slap to ATI’s face, I will NOT use their product anymore.  Their Driver support is terrible and I have spent too much time rolling back drivers than would ever make the decreased price worth it.  But that’s my opinion; take it for what you will.  Should I hear they are supporting their drivers better maybe ill try them someday again, but that day is far away.

GPU

This is the graphics processing unit, its often used as a synonym for video card, as you have seen me do.  This is, in reality, the chip that does almost all the processing of the graphics data.

Core Clock

This is the speed of the GPU.  Measured the same way the CPU core clock is. (Higher = faster)

Stream Processors

This number represents the number of stream processors.  Stream Processing is relatively new (in execution, though its long been theorized) and allows for the GPU to be used for general computing.  It used to be that the GPU was solely used for graphics, with stream processing the GPU can be used for specific applications upon which a GPU can be utilizes, at times performing MUCH faster than any available CPU could.  Folding@Home, a volunteer program that makes your computer available for calculations for protein folding whenever it goes idle, utilizes the GPU.  Maybe you’ve seen the option on your PS3, maybe not.  It’s a pretty cool idea and program.

Memory Clock

This is the clock speed of the memory on the video card.

Memory Size

As it is named, this is the size of the memory installed on the video card.

Memory Type
This is the type of the memory, GDDR2 GDDR3 etc.  These acronyms correspond to the memory classifications just like with the RAM.  GDDR3 is the newest and fastest

Directx

This just tells you if it works with microsofts Directx and to which version.  The current version, as of this writing, is 10.

OpenGL
Indicates if it is OpenGL compatible.  Any card should be, almost all 3d programs use OpenGL so you need to make sure this supports OpenGL.

Ports
This lists all the ports/connections that are on the external interface of the card.  If you want to run dual monitors you have to have 2 ports, preferably DVI.

Max Resolution
Max resolution is the largest (or smallest rather) resolution that this video card can output.  If you want to run a 30 inch monitor those usually require certain video cards.  Most current, decent, video cards can output to 2560x1600, and that’s per monitor/output.  1080p HD TV is 1920x1080.

SLI Supported
SLI is a “Scalable Link Interface’ and that means you can hook up two identical cards that can be utilized as one.  This boosts performance for applications and games that can utilize more power.  If you want to do dual monitors and SLI it isn’t possible at this time.  SLI requires a capable motherboard and a rather large power supply, as video cards demand the most power after the CPU (unless you have a crazy amount of fans or hard drives.)

SLI isn’t something I “endorse” right now.  It used to be the only way to get the performance needed in order to play certain games at full resolution and to maximize their capabilities, however, lately the video cards have become extremely powerful and I feel that its too hard to justify the cost of upgrading everything just to have an SLI setup.  And besides, if you’re doing 3D work you should have 2 monitors; it’s the greatest thing ever.

Here is my response to the “what I do,” (in regards to 3D production) question; for all those that understand the lingo.

I graduated college with a heavy modeling, texturing, and rendering focus.  While I still love those three areas, I have branched out and feel it has undoubtedly evolved my work for the better.

The term “scene assembler” has been used more frequently, as of late, and I find myself fitting that mold more and more, most likely due to the demands from freelancing. 

Scene assembler, as I interpret it, usually consists of:

        I enjoy learning and I do a lot of it every day. People often ask "what are you reading," and since it seems like a common question I figured there are other people out there that may wonder it but don't ever ask, or don't see me reading so don't think to ask.

        Also, I'll review each book as I finish it incase you want to buy it or just want to know if it's good!

C++ From the Ground Up - ISBN-0078824052
       Personal Reading. Scripting helps tremendously in 3D production, I learned some C++ back in high school but need to freshen up my skills.

Cocoa Programming for Mac OS X - ISBN- 9780201726831
       Although I use windows most of the time I own a number of Macs and enjoy the interface and design enough that I would like to know how to code for it.

Learning ActionScript 3.0 - ISBN-9780596527877
       Picked this up to use as a reference for a project, It's a well written book and I am enjoying it so I keep reading it!

        I feel I have to make this public knowledge. I spent the better part of 2 hours trying to figure out why, anytime I was in Editable Poly mode and I hit the alt key it would execute the traditional windows alt function and not the max defined alt function. Maybe it's clearer to say, when I hit alt+r while in editable poly mode, I would get the render dialogue box instead of selecting an edge ring. As you may or may not understand, selecting edge rings is pivotal to modeling.

        The solution: You have to delete the 3ds max user profile so max will create a new one. In order to get to this file you have to turn on the "show hidden files and folders" in windows. This file is stored in "C:\Documents and Settings\Administrator\Local Settings\Application Data\Autodesk\3dsmax\9 - 64bit\enu" fun, no?! Go ahead and add the extension ".bak" onto the 3dsmax.ini and restart max, when max opens it will automatically create a new one and now you can reapply all your UI settings! (Adding the .bak just allows you to revert back to the old .ini should it not fix your problem for whatever reason by just deleting the newly created .ini and then removing ".bak" from the original)

TADA. Everything should be fixed, though, the entire UI is reset, a small price to pay to get work done.

It looks like if you're running Max 2009 the 3dsmax.ini is stored in the root directory. AKA - "C:\Program Files\Autodesk\3ds Max 2009"

Hope this saves someone a few hours of searching for a resolution to a problem that doesn't have an obvious solution at first glance.

        I thought I would start off my thoughts page with proof that it’s not a blog, at least in my mind.  I’ve got a little tip/helper that I consider to be too small to be a full tutorial but is still rich with experience and help. 

        We (3D people) often need reference images for whatever the reason is, and for many of us, our first instinct is to go to google images and search.  While this can, and often does, turn up decent results, they usually aren’t as elaborate as we had hoped for. Alas, a solution.

Use popular, community embraced, photo sharing web sites.  Google doesn’t search flickr when you type in “Office Building,” it just searches web sites and image names.  So if someone puts an image up on their site and doesn’t name it with a description of what the image is (who would do that? Cough cough… shifty eyes) google will not return with that image.  Tags, which are what flickr and many other popular photosharing sites use, are human entered keywords that are used to return results when you use the search function.

The consumer cameras are now at a point where they shoot 10, 12, 14 mega pixel images which is usually plenty for personal reference. The community doesn't worry about how many images of something are already online, they will take a photo anyway. This awesome attribute of humans allows us to pick and choose from the best photos. Light too washed out on that one? Not enough detail on that one? No worries, click the next page and 10 more versions of that photo will show up!

        Now here’s the real tip- and I am not condoning illegal use of these community posted images, but many of the images on said photosharing sites are blocked by the dreaded spaceball.gif.  Within firefox there are ways around this, yes there are programs and other tools out there that will download a “hidden” image but I find this to be simple and quick and you can do it on a computer that you don't have authorized access to install software. Go to Tools>Page Info and then click the “Media” tab and then highlight an image name in the list and just scroll through the list while watching the preview window.  Once you get to the image you want just click “Save as” in the corner and you’re home free.

I suggest doing this, not as a way to break any copyright rules or anything, but more of a way to free up bandwidth and save on many resources.  Not having to be on the internet all the time saves a lot of wasted energy on the use of servers and switching hubs and all that jazz.  Also, you may work without an internet connection on so viewing those images would be impossible while offline.  Again, these are for personal use, produce your own stuff to sell if that’s what you’re after.

To continue to the assembly tutorial click this.

This was a lot of typing, if theres a typo or something please forgive me and email me to make me aware of it. Danmanther@gmail.com