Budget Range: $1000 (plus/minus $50, preferably under $1000 but as close as possible)
System Usage from Most to Least Important: scientific computing on linux (double-precision floating-point important), matlab on linux, watching hd videos, gaming
Parts Not Required: keyboard, mouse, monitor, speakers, DVD, OS
Preferred Website(s) for Parts: newegg.com, amazon.com
Country of Origin: US
Parts Preferences: None
SLI or Crossfire: No
Monitor Resolution: dual display (one 1920x1080, one 1920x1200)
I'm shooting for 16GB RAM not for performance but for heavy data-sets and potential ramdisk usage.
Added SSD just cz I read if you're getting z68 better get an SSD
Why I'm getting z68? no idea ... future proofing? or would P67 be good enough?
Wait for Bulldozer? Not sure ... I'm fine waiting until Oct-12 (new release date) but I don't feel safe buying newly released product that doesn't have user reviews, feedback, and might have unforeseen issues, product recalls?, etc (stabilization would take a few months?). i5-2500k looks like a stable, tested, recommended option. Am I right?
Xeon instead of i5? don't know what would be the benefit (plus I still wanna do some gaming, would Xeon+GPU be okay for gaming?).
Quadro or Tesla? Not in budget range (might find some old quadro under $500 but they don't support dual-precision floating-point I guess).
GeForce? DPFP performance capped at 1/8th that of SPFP. On Tesla same chip has DPFP 1/2 of SPFP (read that on nvidia forums). I'm assuming Radeon 6950 would also be around 1/2 instead of 1/8.
I don't have immediate need for GPGPU but I wanna have a GPU card just to have that option open.
I have a budget of $1000, so I wanna make full use of it instead of bringing total price down. If a part in my list is excessive, please suggest some upgrade in another part so that total stays around $1000.
A newegg combo has most of what I'm lookin for but Seagate Barracuda instead of Samsung Spinpoint. Reviews tell me that's not good.
You mentioned potential GPGPU usage though. If you are debating getting into that in the future you will want to get an NVIDIA card for CUDA. OpenCL on AMD products is not nearly as refined or highly used.
So I received all components and did an external build with MOBO, RAM, CPU, and GPU ... it has worked
however ... the Hyper 212+ cooler is too large for 1 of 4 corsair vengeance DIMMs to fit under it on Gigabyte z68-d3h-b3 ... what that means is that vengeance RAMs are 1 cm (about half inch) too high (or heat sink is 1 cm too low)
orienting the cooler 90 degrees is out of the question cz then two RAM slots are overtaken
what do you guys suggest? i have the following options:
1 - put fan on the opposite side (so fan sucks from heat sink instead of blowing into it)
2 - put fan on proper side but 1 cm higher ... now the fan won't be covering bottom 1 cm of the heat sink plates (i don't plan to overclock, do you guys think this would be a problem?)
3 - do both 1 and 2 (would have to buy another cooler master 120mm fan)
4 - cut about 1 cm of the heat sink of 1 of the 4 vengeance DIMMs
5 - return the Hyper 212+ and stick to the cooler that came with i5-2500k
6 - return the motherboard and get one that has more clearance (doubt I'll be able to find one)
That is the trouble with RAM that has started coming with funky fins. If you won't be overclocking then you may be able to get away with either 2 or 5. I wouldn't cut up the ram (What if it fails and you need to RMA?). Another option would be to get RAM without silly heat fins.
The case you chose has pretty good cooling depending on case fans so as long as you monitor your temperatures you should be good.