Epyc 7402P Build

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.

Shafarevich

New Member
Aug 15, 2019
18
1
3
This is the first time I built a PC and I have learned a lot from this forum. It is now working as intended. A big thank to everyone!

Build’s Name: Epyc Simulation Workstation
Operating System/ Storage Platform: Ubuntu 18.04 LTS
CPU: Epyc 7402P
Motherboard: Supermicro H11SSL-i (Rev2.0)
Chassis: Cooler Master HAF XB Evo
Drives: Samsung 860 EVO 1TB (SATA3) + Samsung 970 EVO 1TB (M.2)
RAM: 8 x Crucial 32GB DDR4-2933 RDIMM (CT32G4RFD4293)
Add-in Cards: Audio -- Creative Sound Blaster Audigy Fx; WIFI -- Gigabyte GC-WB1733D-I
Power Supply: Corsair AX-1600i
GPU: EVGA GeForce RTX 2080 Ti XC
CPU Cooler: Noctua NH-U12S, using 2 x Noctua NF-P12 redux-1300 fans

Usage Profile: Numerical simulation, using both CPU and GPU

Other information:
  • I could not get the Noctua PWM fans to work with the Supermicro motherboard. Had to use 3-pin Noctua fans in the end but they were quiet enough.
  • Due to the CPU socket orientation on the Supermicro board, the CPU cooler blows air towards the side of the case instead of the rear. Not an ideal scenario, but for now it seems to work fine. Stress testing all cores of the CPU reaches a temperature of 57 deg celsius.
 

fp64

Member
Jun 29, 2019
71
21
8
What r u simulating? What kind of cpu operations are most used (integer,floating point,logical)? is the code a commercial buy or own-written? why is the 7402P a suitable choice? any comparison benchmarks?
--
 

Shafarevich

New Member
Aug 15, 2019
18
1
3
What r u simulating? What kind of cpu operations are most used (integer,floating point,logical)? is the code a commercial buy or own-written? why is the 7402P a suitable choice? any comparison benchmarks?
--
It is for physics simulation in computer graphics applications, formulated as an optimization problem. I write a parallel numerical solver for the problem, using OpenMP and CUDA for parallelization. The solver is memory bound, hence the need for Epyc Rome that supports 8-channel memory. 7402P is a good choice because of its high core count and large cache size (both good for the solver) as well as relatively low price.