Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →OpenAI’s Jalapeño inference chips are reportedly being deployed with AMD EPYC Turin CPUs as hosts, each with 1.5 TB of memory. OpenAI hardware chief Richard Ho told Tom’s Hardware that Turin was the pragmatic choice: its maturity and partners’ experience with the platform helped reduce risk and move the design forward. Ho said NVIDIA Vera, specifically “as a standalone,” was “a little bit behind” on maturity at the time of the interview—not that Vera is universally slower or inferior.
What is hosting OpenAI’s Jalapeño chips?
Tom’s Hardware reported on October 2, 2026, that AMD EPYC Turin CPUs host OpenAI’s Jalapeño accelerator ASICs, with 1.5 TB of memory per host. The report does not identify the exact Turin processor model, and the configuration is not stated in the OpenAI announcement or results page cited here. It should therefore be treated as a reported deployment configuration, not an independently verified system specification. Tom’s Hardware’s report attributes the explanation to an interview with OpenAI VP and Head of Hardware Richard Ho.
| # | Preview | Product | Price | |
|---|---|---|---|---|
| 1 |
|
AMD Epyc 9554 Processor 3.1 Ghz 256 Mb L3, W128281619 (256 Mb L3) | $3,550.00 | Buy on Amazon |
| 2 |
|
AMD Epyc 9354 Processor 3.25 Ghz 256 Mb L3, W128281623 (256 Mb L3) | $2,819.95 | Buy on Amazon |
| 3 |
|
AMD EPYC 9004 [4th Gen] 9124 Hexadeca-core [16 Core] 3 GHz Processor | $1,034.96 | Buy on Amazon |
Why did OpenAI choose Turin?
Ho described the decision as a matter of execution risk and readiness, not a general CPU performance contest. He said Turin was mature enough for the project, and OpenAI’s partners had experience with it. That familiarity helped the team pursue its performance and cost goals without taking unnecessary design risks.
In the interview, Ho said: “The way we approached that design was really in terms of de-risking and being able to do that design fast. Vera, as a standalone, is a little bit behind on that maturity level. The Turing device is strong. It did what we needed to do, and partly our partners had some experience with it.” The report prints “Turing device”; that wording is preserved here rather than silently corrected. In context, the discussion concerns the Turin platform. Tom’s Hardware
#1 Best Overall
- Sockel SP5, 64 x 3.1 GHz (Boost 3.75) GHz
- 384 MB L3 Cache, 64 cores/ 128 threats
- 12-channel memory support up to DDR5-4800 MHz
- Max. Performance consumption 360 watts (structural width 5 Nm)
- Tray (without cooler)
Jalapeño and Vera serve different roles
Jalapeño is an OpenAI-designed accelerator for large-language-model inference, not a general-purpose CPU. OpenAI says it developed the chip with Broadcom, which worked on silicon implementation, networking, and connectivity, and Celestica, which contributed to boards, racks, and systems. OpenAI described Jalapeño as the first accelerator in a multi-generation compute platform and said the initial deployment was planned for the end of 2026. OpenAI’s announcement also says the team went from initial design to manufacturing tape-out in nine months.
NVIDIA positions Vera differently: as a custom CPU designed for agentic AI tasks such as orchestration, tool-calling, reinforcement learning, analytics, sandboxing, and managing long-context state. NVIDIA says Vera can run in standalone CPU systems or host its Vera Rubin NVL72 system. Those are NVIDIA’s product descriptions; they do not confirm or refute Ho’s project-specific assessment of standalone Vera’s maturity at the time he spoke. NVIDIA’s Vera announcement lists 88 custom Olympus cores and 1.2 TB/s of memory bandwidth.
Rank #2
What OpenAI’s Jalapeño benchmarks show—and don’t show
OpenAI has published inference results for Jalapeño using InferenceX configurations. The results are company-reported measurements, not independent comparative tests, and they do not measure the choice of Turin as the host CPU.
| OpenAI-reported comparison | Published conditions and result |
|---|---|
| GPT-OSS 120B: Jalapeño vs. GB200 | For the stated nominal 8k/1k STP setup, OpenAI lists package power of 700 W for Jalapeño and 1,200 W for GB200. It reports 85,448 versus 44,960 mixed tokens per kW, or approximately 1.9× higher peak mixed throughput per kW for Jalapeño. |
| DeepSeek R1 MXFP4: Jalapeño vs. GB300 | OpenAI lists package TDP of 700 W for Jalapeño and 1,400 W for GB300. It reports 19,641 versus 11,781 mixed tokens per kW, or approximately 1.7× higher peak mixed throughput per kW for Jalapeño. |
These figures apply to the models and operating details OpenAI specifies; they should not be generalized into a claim that Jalapeño is more efficient for every workload or that Turin is responsible for the measured differences. OpenAI’s results page also says production qualification, software maturation, preparation for scale, and validation across additional models were still underway as it prepared for deployment. That status is more qualified than the company’s earlier announcement that initial deployment was planned for the end of 2026. OpenAI’s results page
Free tools Windows power users keep installed
One-click scans. No signup required.
Quick Recap
Rank #3
What remains unknown about the host system
- The exact EPYC Turin SKU used in the reported Jalapeño hosts is not identified.
- The 1.5 TB-per-host configuration is reported by Tom’s Hardware; the OpenAI pages cited here do not independently verify it.
- Ho’s maturity comparison is a time-bound comment about Vera “as a standalone,” not a universal ranking of Vera against Turin CPUs.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
