this build is mainly for blender.. but I want to do some occasional gaming too..

THIS IS ACTUALLY AN OLD THREAD OF MINE… i REVIVED IT NOW… BECAUSE i’M BUILDING A NEW PC FOR BLENDER AND SKYRIM.

Items that I have:
Speaker
750 watt Power supply
Monitor
keyboard
mouse
DVDr
ASUS direct CU II 6950 2GB (reccomended by Arexma :slight_smile: thanks man :wink: )
i7 2600k

Items that I need for this new build:

  1. OS: I’ll probably will stick with WIN 7…

2.Motherboard: I want a Motherboard that can make use of every feature on my GPU, one that benefits Blender (if there’s any) and one that won’t bottleneck my ASUS 6950 2gb graphic card…

3.RAM: suggest me a good one??

  1. Hard Drive: suggest me a good one??

5.case: I’m going to do a lot of rendering… so… a case with a good air flow would be nice :slight_smile:

P.S,… is it true that ASUS B3 revison motherboards still consist of a few bugs?? I heard some rumours about it effects the CPU?? thus making a slower rendering??

recieved my paycheck… blow it on Computer Hardware again!! :smiley:

You should be ready to go. That’s very good actually. Just check if you can use those memory sticks in dual channel.

thank you for your opinion.

I’d go 2600k - more cache, faster overall. But, more $$ too.
Otherwise, very solid. I’ll be going that route soon. I didn’t do it for you, but check http://www.harddrivebenchmark.net/ for your hard drive to make sure it’s performance is up. A Slow hard drive will kill your system’s speed - it’s already the bottleneck - might as well loosen it up as much as possible.

I’d get another harddrive. If you do a lot of uncompressed video editing, it’s always better to render from one physical HDD to the other. This is far faster than reading and writing to the same harddrive at the same time.

bump bump bump :slight_smile:

I’ve edited my first post above :slight_smile:

NEW BUILD. need Advice from experts here.

  1. Get W7x64 Professional SB. The systembuilder version costs around 100 Euro, you can activate it as often as you want, hence the name systembuilder. The only 2 downsides: If you sell the license, YOU are responsible to give support (as if it ever happens) and you got neither free tickets for MS telephone or email support… useless anyways if you´re not copletely lost.

  2. There´s no way a mainboard can bottleneck your graphicscard. You might want to spend a tad more on the mainbard though and be sure to have SATA6, USB3.0 and maybe already PCI-E 3.0, in case you want to keep the mainboard and CPU longer than 2 years and just upgrade the graphics card.

  3. Seeing you get a Ci7-2600k, it´s almost certain you´ll OC it. You can easily OC the CPU to 4GHz+, so you might want to choose high clocked DDR3 too. The point is, you OC the chip with the FSB CPU multiplier but if you want to increase the FSB as well, the memory clocks go up too. Also the benifit rendering complicated scenes, or using physics where you have a lot of memory operations, you benefit from 1600MHz+ memory. And it´s not really more expensive. I´d get memory modules though with heatspreaders, or heatpipes, it can get hot while rendering, and it´s prone to err if hot. It´s not EEC memory, so you´ll just bluescreen if it gets too hot.
    I can recommend OCZ and Corsair memory currently.

  4. I´d not get one, I´d get 3. A SSD for system and programs, and 2 datatanks in raid1 (mirror) for secure storage, most mainboards support raid1 already.
    If a SSD is too expensive, I´d get a cheap raidcontroller and run raid 5. Raid 5 is striping with a parity disk. You have to have 3 identical HDDs, or at least identical partitions and end up with the storage amount of 2 of the disks.
    It will be fast because it writes on 2 disks “simultaniously”. It will be secure, because if any of the 3 disks fails with the parity disk, you can restore all data on the broken disk by just replacing it. And lastly you´ll have the space of 2 disks “striped” together to one.

Coolermaster HAF-X ~160Euro
or
Nox Extreme Hummer ~120 Euro

Both have superior airflow to many other cases, are practical and ready to install watercooling.

PS) There are usually a lot of bugs left in chipsets, cpus and mainboards due to the complexity they´ve reached. They are not really listed though as bugs, because they are not really repruduceable and the chances that they might occur are incredibly small.
Like the SATA6 bug in pre REV3 sandybridge. It was there but highly unlikely to happen.
I know of no bug currently but there might be one.
Look at AMDs Phenom2 series… ALL cpus have the TLB bug. It´s very common to build “faulty” hardware, simply because you can´t test all possible states the machine can be in to find them.

hth.

  1. by system builder… you meant OEM right? :confused:

  2. hmm… I know just the motherboard to buy now! :slight_smile:

  3. EEC = slower. right? I’ll take my chances having a blue screen :mad::slight_smile:
    I’ve bought a corsair Vengeance and dominator before…though I don’t really know what’s their differences… :spin: never tried OCZ though…:no:
    btw… how fast is fast enough?

  4. not sure if I’ll jump in with the SSD bandwagon yet :stuck_out_tongue: does SSD has any real benefit on blender or PC gaming? (SSD’s are kinda expensive :frowning: I’ll stick with HDD and raid em… lol… I actually have little to none knowledge regarding all these HDD and Raid stuffs… I had to google on this topic to reply on this particular answer :slight_smile: )

  5. sweet Case :eyebrowlift: I’m checking them out right away :slight_smile:

P.S, do you also OC your system,(Gpu,Cpu) for use in blender? or is it for PC games only?
thanks again man… your answers are the kind of answers I was looking for! :yes: instead of just pointing on what hardware to buy… you educate me with what’s what and tell me what I’m supposed to be looking for when buying computer hardware for use with blender. :slight_smile: very informative :ba:

hey, I’ll try to support this Blender community too :slight_smile:

  1. yes.

  2. No. EEC is Error Correction Code. Those are memory modules for servers mostly, or systems that need a high relyability. Its a redundant parity bit error correction method. The memory modul more less is selfaware if an error occured storing information and knows what the correct information is.
    Usually Socket1156 boards (for Core-i5/i7) have a default memory “idea” of 1066MHz. AFAIK this results from the FSB being a default of 106 MHz and having a multiplier of 10. I would have to check on the Ci5 system but I am too lazy atm :wink:
    So you can buy any memory from 1066 upwards. The usual buy is 1333 MHz memory, personally I´d go for 1800 or 2000. If you get the memory to run with that speed it will benefit you for memory operations, not linear though, the benefit should be between 15-30%
    Costwise it´s not much difference but odd for the 1333, random pick, the cheapest modules listed on geizhals.at for 24GB modules:
    AData 1333, 2
    4: 101 Euro !?!?
    AData 1800, 24: 65 Euro
    AData 2000, 2
    4: 70 Euro

Better timings get expensive though, especially T1 memory and fast (high clock) CL5-DDR3 modules.
T1 means that the memory is able to read or write each clockcycle of the CPU. Standard is T2, which means every second cycle. I am not sure if there are any or many T1 DDR3 modules.
And CL is the CAS Latency. It´s the time in clockcycles it takes for the memory to respond, or more precise. It´s the time from sending a column address to the memory and getting a correct data response.

  1. SSD more less is luxury. Having the OS on the SSD is not a huge boost, your system will boot faster, an OS like Windows barely needs the HDD after booting, most is ran as services and loaded in the memory after booting. What benefits is if you have your pagefile/swappartition on a SSD, it boosts performance greatly, and if you have your daily tools on the SSD. Many tools take ages to load (e.g. 3dsmax) or often need to load stuff from the HDD. That´s where you benefit.

Raid is no magic. It´s straight forward. I can generalize it for Linux Software Raid, which you can´t use for windows though, you´d need a raidcontroller for it.
You boot your linux installer and partition your disk. Software raid can only boot from Raid0 or Raid1 but you want Raid5.

You create on your three disks:
1 partition on each disk with 256 mb, choose use as raid, initialize Raid1 (mirror with 2 paritions) with one spare partition
1 partition on each disk with 2048 mb, choose use as raid, initialize Raid5 (striping and parity) with 3 partitions
1 partition on each disk with the rest of the disk, choose use as raid, initialize Raid5 (striping and parity) with 3 partitions
Apply settings.

After that the raid will show as 3 partition:
One with 256mb, and if one of the disks dies, the data will be re-initialized on the spare partition until you change the broken disk. So raid1 will be operable even if one disk dies and you can boot from it.
One will have 4096mb for swap. Two 2048mb partition are striped together to 4069mb, you´ll get a speedboost too if you use NCQ and AHCI as you can write to both disk “at the same time”, while the 3rd partition is the parity patition. Simplyfied said, it stores the logical operation of the data of the first 2 disks. Usually it is an exclusive OR, or XOR. The HDDs store BITs.
DiskA stores 0, DiskB stores 0, so the parity disk stores 0 as well.
0 XOR 0 = 0
DiskA stores 1, DiskB stores 0, so the parity disk stores 1.
0 XOR 0 = 1
DiskA stores 0, DiskB stores 1, so the parity disk stores 1.
0 XOR 0 = 1
DiskA stores 1, DiskB stores 1, so the parity disk stores 0 as well.
0 XOR 0 = 0

So you can see, if DiskB for instance dies, the RAID knows that DiskA has stored either 0 or 1 and the parity disk has stored either 0 or 1 and with reversing the XOR, the RAID can reinitialize all the data on the broken disk after you change it to a working one.
There are even higher redundancy raids if more than one disk at once dies, but that´s not practical for a home machine.

And the last partition shows up the same, just with XXXX space for your root filesystem.
Now you set
Raidpartition A to be used as ext4 and /boot
Raidpartition B to be used as /swap
Raidpartition C to be used as ext 3 and /

Done. The rest of the raid is automated and you don´t have to care again. If a disk dies, you replace it and when it asks to reinitialize you say yes. Speedbost noticable, dataloss impossible as long as only 1 disk dies at once.

Some mainboards today offer decent raid controllers, but only a few are capable of raid5.
A hardware raidcontroller costs from 20 Euro upward to a few thousand. The price sets the quality and speed usually, then again, for homeusers a cheap one is good enough.

PS) It´s better to educate, else people keep bugging you :smiley:
Jesus anyone with fish and fishing? :smiley:
Give people hardware advice, they’ll buy it and ask again, educate them to choose your own hardware and you’ll have peace :smiley:

Yes, OC benefits you overall. You just have to be aware where it benefits you, and that the beneift is at most linear, usually less.
Overclocking the CPU directly effects for instance rendering and does so linear.
Take a Ci7 with 4*3.40 GHz takes 30 minutes to render a frame, pure raytracing, not render preperation methods that don´t use all cores fully.
If you OC it to 4.5GHz, which should be rather easy, it runs with 132% performance. The rendering will now take 22.7 minutes.
Games barely use more than 2 cores, so there it´s beneficial too, because the 2 cores are faster now. There was a time when games performed better on a Core2Duo heavily OCd than on a Core2Quad, simply because the 2 cores where able to be OC´d higher than 2 cores in a Quadcore.

For the graphics card, it soley benefits you in GPGPU in CUDA/OpenCL. Like for the CPU it´s the one thing that has full load on a GPU. So it matters if you render with Cycles for instance and if the image takes 10 minutes or 7 minutes.
It doesn´t really matter if your game runs tih 70fps or with 80fps. Where it starts to matter is if it´s 25fps or 31fps or if you play at high resolutions with huge textures and high AF/AA levels.

Personally I OC my systems right from the start to have all available power right away and not as an act of desperation when the system gets too slow after 2 or 3 years just to squeeze out a year longer.

You should be aware though, that the energy consumption usually grows exponentionally while the performance grows linear.
The additional thermal stress is no issue given proper cooling. It doesn´t matter (to me) if a chip lasts 20 or just 18 years due to thermal stress.
Besides that (under load):
My C2Q and stock clocks with reference cooling had ~58°C at reference voltage of 1.225V
My C2Q now OC´d by 33% and aftermarket cooler runs with ~42°C at lowered 1.208V
My GTX470 at stock clocks with reference cooling had ~90° at 0.975V.
My GTX470 OC´d by 20% and aftermarket cooler runs with ~58°C at the same voltage.

So I dare to say although OC´d my components will hold longer now but the system consumes more power :smiley:

Now I am going to lean back and watch Bulldozer, SandybridgeE, GTX600 and HD7000 series and see if my new system will be AMD or Intel and if I keep the GTX470 or get something new.
The sad truth is, my sytem is more less 3 years old now and still sufficient for most tasks and there´s no justifyable reason for me to get a new one. Maybe something breaks soon :smiley:
Personally I think with the system you chose you should be good the next 2 years at least unless there´s a breakthrough in semiconductors and after that it might be enough to upgrade the graphics to a nextget PCIe 3.0 card. That´s why I said you might want to check a board that already has it, after all it actually doubles the bandwith of PCIe 2.0.

  1. lol… I know what EEC stands for :wink: a year ago…I stumbled across an article regarding EEC memory module stating that it’s slower and what not…(I think :P) lol… I’m not really sure now… my brain is a bit foggy at the moment :spin:

2.I guess I could burn a little bit more money one those RAM :ba: No food for me for a whole week!! :ba: this Hobby of mine… it’s killing me… :smiley:

  1. OC it is then :yes: too bad I can’t use CUDA…:frowning: and looking at Small LUX gpu development had been stopped and all… sigh cross fingers OpenCL cycles is there hope? I’ll continue dreaming.

4.SSD is a luxury I can’t afford :mad: :wink: I’m poor… :frowning: :slight_smile:

  1. haha… my dad often used to give me the Fish and the fishing analogy when I was a kid. :slight_smile:

  2. I’m still a bit blurry on the whole raid topic :spin: don’t worry though… I’ll make sure to research more on the topic on my own till I get it :slight_smile: I’m a very persistent learner :yes:
    but… should I really worried that much about Hard drive failure?

  3. Open GL issues will probably be fixed on their 6xx cards… or so I heard… I’ve read somewhere on a forum that Nividia has some interest with Blender? if they do… then 6xx will definately have major stability with blender & open GL :slight_smile: here’s hoping :slight_smile:

P.S, lol… thanks for the WALL of text answer haha… :yes: I was joking :slight_smile: I thank you for your reply which is filled with answers worth of days of searching on the web :yes:
indeed you have been very helpful :wink: appreciate it mate :slight_smile:

First, you shouldn´t always fullquote, it makes the thread very uncomfy to read really.
Then,

  1. The faster clocked RAM is cheaper really but it´s a thing with Sandybridge. The CPU itself is meant to support DDR3-1333 and run with it. It´s rather easy to get 1600MHz memory to run with it, but the gain is minimal. For higher memory clocks you need some skills OCing. But it´s no loss, your next platform might be AMD, or you sell the memory at one point, it´s better to have modules supporting higher clocks. As for cutting edge timings and T1 memory, I don´t think you should invest in it. It´s really the performance sector, where you spend double the money to gain 10% more performance for stuff where every second counts.

  2. OC is easy as pie today. Especially if you have a board with UEFI instead of BIOS. Move slider from normal to performance/OC. Done.
    Smallluxgpu is Lux’ sandbox. Smalluxgpu wasn´t meant to be developed, it was a feasability test of OpenCL raytracing. All stable and working features from Smallluxgpu got implemented into Luxrender, or are about to be implemented.

sorry bout that.

good to hear some news on SLG!

I’ll heed your advice sir! :slight_smile:
P.S, have you ever attend Suzaane animation festival?