Please help me decide Scientific and gaming build

[ParticlePhysicist]

Mother Board

 

Asus

Z9PE-D8 WS

Samsung 256GB DDR4 RAM (ECC)

 

Xeon E5 2699V4 OR Xeon E7 8890V4

 

im going with 4 Intel 1.2Tb SSDs and 10 8TB seagate HDDs

 

Corsair 900D tower

 

i believe a Corsair 1600W power supply

 

2 of the new Titan X GPUs coming this August

 

obviously liquid cooled

 

any suggestions esp for the CPU would be greatly appreciated would like it to be optimal for both my research and playing games on when its free : )

 

[ttam]

Every thread like this from a new account is just some kid looking for attention because they found a site like PCPartpicker and it makes them feel special.

 

Considering most physicists will have access to a system from the likes of Cray.. yeah.

[ParticlePhysicist]

Actually I'm building my own system for my own research into Hadronization. As i am currently working with a Team here in China i would like to build a computer for home use to cycle through some data and simulation, as well as real time collisions. But thanks for being a supportive member and checking to see if i was a legit Scientist wanting advice for a home build. Obviously my work system is way better than this system i wish to build. Yes i have seen Cray systems comps, but as its for my personal use for research and leisure i would like to keep the cost a bit lower. So to reiterate my own work on Hadronization is separate from the work I'm currently doing on Neutrinos here in China. Thanks.

[Zyndo]

1 hour ago, ParticlePhysicist said:

Snip

I think the AX1500I is the largest PSU corsair has. If you want a 1600 you're going to need another brand... but lets be honest, you don't need more than the AX1500i.  Almost no one does. 900D is massive and awesome; good choice. You doing intel NVME u.2 drives? those things are hella expensive, but damn are they fast. With all those drives, are you planning on doing any RAID setups or just using them all individually? I don't think that new titans will be available come august, where did you hear this information? As well I do not think they are Titan X, they are Titan P (as far as i know). Also, I do not know what sort of work you're doing or what calculations you need.... but wouldn't a Tesla card or Quadro or some workstation specific GPU be better for your needs than Titans? As good as titans are, they're more designed for gaming rather than high end professional applications like the more tailored cards are.

Other than that, damn dude. this is going to be a crazy system

[ParticlePhysicist]

you are right a TESLA or other GPU would be better as they do have double precision. I'm hoping that the new Titans will also have double just as the last Gen models had as well. Keep in mind my computations are not as heavy as the ones i do at work or over at CERN. When i work there i always use their systems OBV. This system is purely for my own work toward my own hypothesis about Hadronization. I understand that my home system wont be able to render or maybe compute as fast as my work super comps do but again..i don't mind waiting a few days to get results on my home comp as again its my own work. haha yes i hope it will be a kick ass machine, i'm actually unsure which raid config i should go with i'm thinking Raid 1. The exasct Intel SSDs are the Intel 750 Series HHHL (CEM2.0) 1.2TB PCIe NVMe 3.0 x4 MLC. (from newegg)

[DrMikeNZ]

That motherboard is not compatible with either of the CPUs that you listed. If you want E5 2699V4 processors you should be looking at newer boards such as Z10PE-D8 WS.

 

The new Titan X (Titan XP for those who follow LTT), is set to be available directly from Nvidia on 2nd August, no word yet on double precision performance. It is possible Nvidia won't want the Titan XP to compete with their Quadro line.

[ParticlePhysicist]

ahh that's true...and yes i do follow LTT i forgot to call it XP ha! damn yes i just looked again im sorry for my mistake i did mean the Z10 not the Z9. : ( Now another question would two Quadros instead of the two Titans be much different in terms of gaming? If i remember correctly from watching LTT that i CAN game on workstation GPUS...i dont clearly remember what draw backs there were..i will find the vid again.

[DrMikeNZ]

Hypothetically, two quadros would work wonders in calculation tasks. However, in most games you would only be able to use one due to lack of multi GPU profile support, and depending on which quadros it would be slower than a midrange GPU (1060 or 1070).

[ParticlePhysicist]

Ok yeah, im looking around doing some research into which GPU i should get, i do see some people having different GPUS in their systems. such as a guy on youtube that has a M4000 and two 980 in sli. Is it possible to have one for my rendering and computations and another for gaming?

 

[ParticlePhysicist]

ahh i guess this is what he says. " just install the quadro driver over the geforce one and both types of cards will be available for use at any given time. for me, i have the display hooked up to the 980 ti, and have all the workloads go through the quadro. you can set this up in either program configuration files and/or the nvidia control panel. although you can make it easier on yourself by having a monitor with multiple inputs (mine only has a single displayport input thanks to it being an early Gsync monitor). that way you can have one input set to the quadro, the other to the geforce. and it's as simple as changing the input source on the monitor when you need to do work or have some free time to game. although i haven't tested this method. but i believe that's how it would work. if you do get a 970, let me know how it goes. i'd like to know for future reference."

[almondhive]

Wow it is a pretty powerful system the one you are describing. Do you have budget limitations? If you do maybe you could make some tradeoffs depending on what you need specifically. 

¿What simulations/programs are you running? ¿Are you going to use GPU in your computations (I suppose you will if you want two titan XPs)? 

Im also a physicist (recently graduated), not the Particle physics type but I run lots of simulations, mainly finite element for fluid flows, Ising model, bayesian analysis (I am working with a biologist). I once worked with CERN's ROOT in a high energy phisics class but thats all my computing experience in that field. 

 

I can tell you from my experience that I once was running some finite element simulations of multiphase flow in porous media on some computers with an i7(I think it was i7 5700) that were on the engineering department of my university and it ran well enough, It took like 8hr for that particular simulation to complete running in parallel with the 4 cores of that i7. But then I discovered that in the Faculty of Design they had lots of macs and some abandoned dells in a corner that students never used because mac. So I discovered they had 12 core Xeons (E5 2697 i think) and I ran my code using the 12 cores and was a huge speedup. I also ran another simulation but with only one core, in that case it ran similar than the i7.

 

I guess my point is: if your simulations can advantage from multiple cores/cpus invest in a lot of cores, if not better invest in the better per core performance processor. If you can take advantage of GPU speedups (for example bayesian analysis runs like a charm in my GTX770 I can only imagine how would it run in two titan xp)  then go for powerful gpus. 

It seems to me that with a system like that you have all fronts covered.

 

Daniel.

[ParticlePhysicist]

Hey sorry for the late reply yes i use most of the same software as you also ROOT and others. In terms of budget limitations i just got some investors to help me fund my project so as of now I'm looking at 15-20k. Though i will need a lot of memory space for collision data etc..i would go with the E7 series BUT i was told that they require LOTS of RAM to be efficient..as RAM anywhere above 256Gb is VERY EXPENSIVE i doubt i could afford 1TB or maybe even 512GB. (i could be wrong about the E7 line up) as of now i bought two E5 2699v4 i'm thinking should be good enough to render or calculate in decent time.