Fees and rewards
Intelligent Computing Platform service fee policy (2025)
This is an English rendering of the published Chinese policy of the Tianjin Institute of Industrial Biotechnology, Chinese Academy of Sciences. Amounts are in RMB. Where the source is ambiguous or inconsistent, the issue is identified below for confirmation with the platform.
1. General provisions
Article 1. This policy regulates the use of institute computing and storage resources, promotes efficient support for research, education, and discipline development, and safeguards the computing needs of major national research projects. It is established under the Technical Support Center's equipment-use rules as the trial service-fee policy for the Intelligent Computing Platform.
Article 2. The policy applies to all platform users, including institute staff, students, and external users.
2. Operation and management
Article 3. The Biodesign Center manages and maintains the platform under the principles of resource sharing, open collaboration, innovation support, safety, and efficiency. It maintains stable operation, develops technical staff, and provides training, technical exchange, and international collaboration.
Article 4. Platform infrastructure is centrally planned and built by the institute. Research groups should not independently purchase comparable equipment without institute permission.
Article 5. Major projects may require temporary exclusive node reservations or suspension of paid queues. Users must be notified promptly and given sufficient time to respond.
3. Financial management
Article 6. Services are billed after use, with resources allocated through unified scheduling.
Article 7. Internal users should normally pay platform service fees from testing and processing expenditure accounts.
Article 8. Revenue is primarily used for equipment repairs, consumables and parts, training, and exchange activities.
4. Rates
Article 9. Some resources are free and others are paid. Internal users have priority; some capacity may be made available to external users.
Article 10. Resources are offered in queued and dedicated modes. The policy describes dedicated-mode charges as 20% higher than queued-mode charges. Internal users pay the standard rate; external users pay 1.5× the standard rate. The published base-rate table is reproduced below; confirm dedicated quotations with the administrator because hourly and monthly billing units differ.
Computing
| System / node type | Main configuration | Queued rate (RMB/CPU core-hour or GPU-hour) | Dedicated monthly rate (RMB/node or GPU) |
|---|---|---|---|
| Free trial: virtual CPU | 4 × 2.4 GHz vCPU, 16 GB RAM | 0 | 0 |
| Free trial: virtual GPU | 4 × 2.4 GHz vCPU, 32 GB RAM, 1 Tesla P4 | 0 | 0 |
| Taihang I: dual 4116 | 2 × Intel Xeon 4116, 2.1 GHz, 12 cores each; 64 GB RAM | 0.05 | 1,000/node |
| Dual 8358 | 2 × Intel Xeon 8358, 2.6 GHz, 32 cores each; 256 GB RAM | 0.10 | 5,500/node |
| Quad 8380H | 4 × Intel Xeon 8380H, 2.9 GHz, 28 cores each; 1.5 TB RAM | 0.15 | 12,000/node |
| 8 × A800 | 2 × AMD EPYC 7663, 2.0 GHz, 56 cores each; 1 TB RAM; 80 GB per GPU | 4.8 | 28,800/8 GPUs; 3,600/GPU |
| 8 × L20 | 2 × Intel Xeon Gold 6530, 2.1 GHz; 512 GB RAM; 48 GB per GPU | 3.2 | 28,000/8 GPUs; 3,500/GPU |
| 8 × A40 | 2 × AMD EPYC 7763, 2.45 GHz, 64 cores each; 512 GB RAM; 48 GB per GPU | 1.8 | 12,000/8 GPUs; 1,500/GPU |
| 8 × 3090 | 2 × AMD EPYC 7763, 2.45 GHz, 64 cores each; 512 GB RAM; 24 GB per GPU | 1.0 | 6,400/8 GPUs; 800/GPU |
| 8 × 4090 | 2 × AMD EPYC 9654, 2.4 GHz, 96 cores each; 512 GB RAM; 24 GB per GPU | 1.5 | 9,600/8 GPUs; 1,200/GPU |
| 8 × 5090 | 2 × AMD EPYC 9654, 2.4 GHz, 96 cores each; 512 GB RAM; 32 GB per GPU | 2.4 | Not specified |
| Office virtual host | 4 CPU cores, 8 GB RAM, 60 GB system disk, 200 GB data disk; Windows 7–11 options; remote access | — | 150/host |
| Service-support virtual host | 8 CPU cores, 16 GB RAM, 60 GB system disk, 500 GB data disk; Windows or Linux | — | 300/host |
| Web-service virtual host | 16 CPU cores, 32 GB RAM, 60 GB system disk, 500 GB data disk; Windows or Linux | — | 600/host |
Billing units and specifications
CPU usage is billed per core-hour; GPU usage per GPU-hour. Virtual configurations and base rates can be negotiated. Each GPU includes 8 CPU cores; additional CPU cores cost RMB 0.1 per core-hour. The published fee table's memory/CPU specifications differ from the inventory for some nodes; check the assigned hardware before running.
Storage
- Free accounts: 50 GB home directory per user.
- Paid accounts: 1 TB home directory included; additional storage costs RMB 40/TB/month, rounded up to whole TB.
- Paid accounts can request backup storage at
~/backup: RMB 30/TB/month, rounded up to whole TB. - Independent backup services: online backup RMB 30/TB/month; offline tape-library backup RMB 450/TB/year.
Storage use in fhome, fproject, and fbackup is sampled at a fixed time each day. The billing period is normally a calendar month. The maximum measured usage during that period is used, rounded up to whole TB.
The published monthly formula is:
fhome billed TB × 0 + (fproject billed TB − 1) × 40 + fbackup billed TB × 30
Clarify storage allocation and billing
The source describes the included allowance as a home-directory quota, but its formula subtracts 1 TB from fproject. It does not specify a lower bound for that subtraction. Confirm the account-specific allowance and charge with the administrator; do not interpret the formula as promising negative charges or credits. The separate resource-limits page also gives a general 1 TB default, so free-account users should check their actual quota.
Equipment hosting
| Equipment | Annual rate |
|---|---|
| Full rack | RMB 240,000 (26 kW, 42U) |
| Standard rack server | RMB 10,000 per 2U |
| High-density server | The higher of RMB 240,000 × occupied U/42 and RMB 240,000 × actual kW/26, or a separately assessed rate |
| Supporting equipment (switches, management/login/storage nodes, etc.) | RMB 5,000/U |
For each research group, capacity above two hosted racks is charged at 1.5× the base rate. Equipment hosted for more than five years is charged at 2× the previous rate for the period beyond five years.
5. Discounts and rewards
Article 11. Monthly volume discounts for large-scale computing and storage:
| Resource | Monthly volume | Percentage of base rate payable |
|---|---|---|
| CPU | At least 500,000 core-hours | 75% |
| CPU | 300,000 to less than 500,000 core-hours | 80% |
| CPU | 100,000 to less than 300,000 core-hours | 85% |
| GPU | At least 30,000 GPU-hours | 75% |
| GPU | 10,000 to less than 30,000 GPU-hours | 80% |
| GPU | Source prints [5000, 1) 万卡时; interval requires clarification | 85% |
| Storage | At least 100 TB | 80% |
| Storage | 50 to less than 100 TB | 85% |
| Storage | 20 to less than 50 TB | 90% |
The GPU 85% row contains inconsistent units/bounds in the original. Confirm its threshold before estimating a bill. Percentages above represent the amount payable, not the percentage deducted.
Article 12. Urgent projects with significant societal benefit, especially major national projects in frontier or fundamental research and major economic applications, may receive preferential treatment. Users must apply; the platform expert committee decides the applicable terms.
Article 13. Rewards are available for reported publications whose acknowledgment explicitly credits the platform. The supplied English wording is:
This work received computational resources support from the Intelligent computing platform at the Tianjin Institute of Industrial Biotechnology, Chinese Academy of Sciences.
| Publication category | Points per paper |
|---|---|
| Cell, Nature, or Science main journals | 100 |
| Journals in those families, or institute-recognized high-impact papers | 50 |
| SCI Zone 1 under the institute's classification | 20 |
At year-end settlement, 1 point equals 1,000 CPU core-hours or 30 GPU-hours. Rewards can offset future queued-computing charges or proportionally offset dedicated hosts, cloud hosts, or hosting charges. They are not cash-redeemable or transferable. See Acknowledgments and applications.
6. Rights and obligations
Article 14. Users may request resource changes according to their needs.
Article 15. Users must:
- Follow the service agreement. Do not perform unrelated computations (including mining), deliberately waste compute/network resources, or disrupt the environment.
- Pay fees on time. Deliberate nonpayment may result in suspension; resulting losses are the user's responsibility.
- Protect account credentials and bear responsibility for losses caused by their disclosure.
- Promptly request updates when user information changes.
- Identify research outputs produced using the platform, as required by this policy.
Article 16. Users must regularly back up their own data unless a special backup arrangement has been agreed with the platform. Users are responsible for their stored data; the platform does not assume responsibility for data lost on the cluster.
7. Supplementary provision
Article 17. The policy takes effect upon publication.