New hyper instances
CFD FEA SERVICE announces the release of the brand new hyper instance group. These are brand new instances performed by the brand new INTEL SAPPHIRE RAPIDS processors. This starts the new premium processors which are capable of our best performances and can decrease considerably your simulation time (up to 50% depending on software used and on simulation settings). We propose here two scalability tests performed as comparison with our current processors. This new processor aims at providing the best price-to-simulation performance among those available on the cloudHPC.
OpenFOAM
As part of the AUTOCFD group, a scalability test has been performed on CASE 2 mesh (transient analysis over approximately 110,000,000 cells mesh. The very same analysis has been performed using OpenFOAM v2212 using two configurations: the 112 vCPU / highcore powered by our AMD EPYC ROME/MILAN and the new 96 vCPU / hypercore powered by the new INTEL SAPPHIRE RAPIDS.
 |  |
The table below shows the results comparison between the two cases: the same analysis has been resolved in approximately 50% of the total simulation time with a cost 12.75% cheaper (due to the lower vCPU amount used and to the lower simulation time).
| vCPU | RAM | TYPE | Sim. Time [h] | vCPU/Hrs | Sim. cost [€] |
| 112 | highcore | SPOT | 56.472 | 6324.89 | 316.25 |
| 96 | hypercore | SPOT | 26.128 | 2508.29 | 275.92 |
| -53.7% | -12.75% |
FDS
A second test cases regarded a 4,000,000 cells analysis performed with FDS. The case used is a simple staircase with a very high resolution of the mesh size, a fire scenario located at the bottom and a VENT opening at the top of the staircase.
FDS can be performed in several configuration and for this the new hypercore have been compared against highcore and highcpu type of simulations in several configurations of vCPU allocation.
The new hypercore resulted cheaper or aligned in terms of simulation cost or slightly cheaper (3% cheaper) but with an overall simulation time 50% lower than the 112 vCPU / highcore.
| vCPU | RAM | TYPE | Sim. Time [h] | vCPU/Hrs | Sim. cost [€] |
| 112 | highcore | SPOT | 5.13 | 575.05 | 28.75 |
| 96 | highcpu | SPOT | 8.65 | 830.77 | 33.23 |
| 48 | highcore | SPOT | 11.94 | 573.03 | 28.65 |
| 96 | hypercore | SPOT | 2.64 | 253.83 | 27.92 |
Conclusions
All in all, the new hyper instances proved to be 50% faster in the two considered test-cases with price-per-simulation aligned with those of highcore (the current equivalent alternative). These results can’t be general, as different software, mesh, modelling approches can modify the scalability performance as found in this blog post.
The price per vCPU/Hour proposed for these new instance types is mentioned in the following table.
| Unit | RAM per CORE | REG Unit price | SPOT* Unit price |
| vCPU/Hour | hypercpu | 0.18 | 0.09 |
| CORE/Hour | hypercore | 0.22 | 0.11 |
CloudHPC is a HPC provider to run engineering simulations on the cloud. CloudHPC provides from 1 to 224 vCPUs for each process in several configuration of HPC infrastructure - both multi-thread and multi-core. Current software ranges includes several CAE, CFD, FEA, FEM software among which OpenFOAM, FDS, Blender and several others.
New users benefit of a FREE trial of 300 vCPU/Hours to be used on the platform in order to test the platform, all each features and verify if it is suitable for their needs
0 Comments