# Question about EPYC access for benchmarking workloads

**URL:** <https://devcommunity.amd.com/t/question-about-epyc-access-for-benchmarking-workloads/412>\
**Category:** Workloads\
**Created:** [March 13, 2026, 7:17pm UTC](https://devcommunity.amd.com/t/question-about-epyc-access-for-benchmarking-workloads/412 "2026-03-13T19:17:16Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![Osei\_Harper](https://avatars.discourse-cdn.com/v4/letter/o/a8b319/32.png) [@Osei\_Harper](https://devcommunity.amd.com/u/Osei_Harper)\
**Post date:** [March 13, 2026, 7:17pm UTC](https://devcommunity.amd.com/t/question-about-epyc-access-for-benchmarking-workloads/412/1 "2026-03-13T19:17:16Z")

</div>

I have been developing a high-throughput data ingestion engine and am seeing some unexpectedly high throughput numbers on a small workstation.

Early tests suggest the architecture scales very well with concurrency, but my development machine is a poor benchmark environment due to background workloads and consumer hardware constraints.

I would like to characterize how the system behaves on EPYC under controlled conditions, particularly around concurrency scaling and NUMA behavior.

Does AMD have any developer or research programs that allow independent developers to access EPYC systems for benchmarking purposes? Even remote test access would be sufficient.

The goal is to measure real throughput limits rather than rely on theoretical scaling assumptions.

---

<div class="post-metadata">

**Author:** ![Hakob\_Arzumanyan](https://dub1.discourse-cdn.com/amd/user_avatar/devcommunity.amd.com/hakob_arzumanyan/32/214_2.png) [@Hakob\_Arzumanyan](https://devcommunity.amd.com/u/Hakob_Arzumanyan)\
**Post date:** [March 18, 2026, 4:51pm UTC](https://devcommunity.amd.com/t/question-about-epyc-access-for-benchmarking-workloads/412/2 "2026-03-18T16:51:17Z")

</div>

Hello Osei,

AMD AI developer Program will be perfect option for you, here are some characteristics

- **Infrastructure:** While much of the cloud is GPU-centric (MI300X), these instances are hosted on flagship EPYC 9005 or 9004 systems.

- **Evaluation:** Unlike a standard VPS, these environments are often pre-tuned for high-performance computing (HPC) and include tools like **AMD uProf** and the **Zen Software Studio**.

- **The Program:** You can apply for the **AMD AI Developer Program** to receive **$100 in credits** or up to **25-100 complimentary hours** on the AMD Developer Cloud.

Here is the link to join the program → [AMD AI Developer Program](https://www.amd.com/en/developer/ai-dev-program.html)

Thank you and do not hesitate to ask any questions that come in mind.

Hakob
