Predictika · Lenovo servers and racks

An AI sales agent's server and rack quote, checked before it goes out

An AI sales agent prepares a quote for Lenovo ThinkSystem servers and racks: platform, processors, memory, GPUs, adapters, drives, power supplies, racks, power and cooling, support. Before it goes out, the agent sends it to Predictika, which checks it against a declarative model of the configuration rules and returns a verdict, every reason, the figures it computed and, for a quote that can proceed, the line-item quote itself.

Scope: Lenovo ThinkSystem SR630 V4, SR675 V3, SR680a V4 and SR780a V3 (Neptune) servers, with racks, power, cooling, support and deployment. See everything this validator covers

This validator models the Lenovo ThinkSystem SR630 V4 rack server and the SR675 V3, SR680a V4 and Neptune liquid-cooled SR780a V3 GPU servers, together with the racks, power distribution, top-of-rack switches, cooling, support and deployment quoted around them — not just the scenarios shown below.

  • Processors and memory — socket and per-processor power limit of each platform, processor counts, memory slots and channel balance.
  • Power supplies — the draw of the whole server against the capacity of its power supplies with the redundant ones held back.
  • GPUs and fabric — GPU form factor (HGX, AMD OAM or PCIe) against the platform, per-GPU power limit, full-baseboard population and network adapters for the GPU fabric.
  • Racks and facility — racks needed from space and power, PDU capacity, facility water for liquid cooling and power per rack.
  • Slots and drives — network adapter slots and drive-bay type.
  • Quote totals — volume discount band, support plan, deployment, and the total price and power the agent states, compared with what the validator computes.

Any quote for these four Lenovo ThinkSystem servers is in scope — any combination of platform, processor, memory, GPU, adapter, drive, power supply and rack that touches the checks above, not only the scenarios pictured here. It does not cover other Lenovo product lines such as ThinkAgile or ThinkSystem DE storage, or networking quoted on its own.

The AI sales agentBuilds the server and rack quote from the conversation and says it is ready.
Predictika validatorChecks every rule at once and returns the verdict, each finding, and the computed figures.

8 scenarios in 4 groups · each group runs from the agent’s first attempt to the outcome · click one to switch

What the agent proposes

The AI sales agent says

Illustrative agent statement. The validator receives the structured request below.

Proposing agent: AI sales agent (scripted)

Before sending, the agent must call Recorded run

Predictika validator


      

        
        
        
      

      
Waiting for the call.

      
    

  

  
  

    
Where rules-based CPQ gets it wrong
Click to expand
    
Our assessment of what a typical rules-based CPQ (product rules, price rules and validation rules, as in Salesforce CPQ) does with the same cases. It is a judgment about effort and reliability, not a claim that nothing can be built.

    
What the validator caughtTypical rules-based CPQPredictika
Processor socket and power limit that differ by platformOnly with care
A lookup rule per platform for socket and per-processor limit. Works, but every new platform or processor adds rules that someone must keep in step.		One catalog table; the check follows from the platform chosen, in any order.
A server that draws more than its power supplies can carry once redundant ones are held backOnly with workarounds
Needs the draw of the processors, memory, GPUs, adapters and drives added up per server, then compared with supplies minus the redundant ones. Doable with chained formulas and summary variables, and fragile when any input changes.		Computed from the model and checked on every line of every call.
A GPU baseboard ordered with fewer GPUs than it ships withEasy
A straightforward product rule once the baseboard size is a field on the platform.		Same, from the catalog.
Liquid-cooled servers quoted for a site with no facility waterOnly with care
A validation rule against a site field. Fine if the site field is always filled in; silent if it is left blank, or if the cooling type is not recorded on each platform.		The cooling type comes from the catalog and is checked against the site on every call.
Racks needed from space and power, with PDU capacity halved for A and B feedsOnly with workarounds
Needs rack units, power per server, PDU capacity, switch space and the site limit combined, then rounded up. Rules-based CPQ tools rarely compute this; reps often guess.		Computed from the model; the racks chosen are checked against it.
A quote that states a total or power different from what the build givesOnly with care
A validation rule can compare a stated figure with a calculated field, but only if the stated figure is a field on the quote, which it rarely is.		The stated figures travel with the quote and are compared with the computed ones on every call.
Every problem returned in one answerOnly with workarounds
Rules tend to fire as the quote is saved or a step is completed, so a rep often fixes one error and meets the next.		All findings come back together, whatever order the choices arrived in.

  

  

    
What this validator catches: findings from its live runs
Click to expand
    
Every rule is checked on every call, in any order the choices arrive; one problem never hides another. Blocking findings stop the quote; warnings and notes are computed alongside. These are the findings returned in the live runs behind this page: the eight scenarios shown.

    
1Processor does not fit this server. Lenovo ThinkSystem SR630 V4: AMD EPYC 9655 uses socket SP5 but this platform takes LGA4710.
2Processor runs too hot for this server. Lenovo ThinkSystem SR630 V4: AMD EPYC 9655 is rated 400 W; this chassis cools up to 350 W per processor.
3Quoted power does not match. The quote states 10.5 kW total power; the validator computes 11.02 kW.
4Quoted total does not match. The quote states a total of $412,300.00; the validator computes $423,466.60.
5Power supplies too small for the server. Lenovo ThinkSystem SR675 V3 (dense GPU module): the server draws up to 6290 W but its power supplies give 5200 W with 2 kept as redundant spares.
6GPU baseboard not fully populated. Lenovo ThinkSystem SR680a V4: 6 GPUs chosen; the GPU baseboard comes fully populated with 8.
7Air-cooled racks above the comfort zone. Air-cooled racks average 37.04 kW; above 30 kW air cooling usually needs rear-door heat exchangers or hot-aisle containment.
8Liquid-cooled servers need facility water. 8 liquid-cooled servers are on the quote but the site has no facility water for liquid cooling.

  

  

    
Scenarios and results
Click to expand
    
The recorded result of the validator for each scenario on this page: the request the AI agent sent, the verdict, and every finding, in the order they appear above.

    
1. Processor limits: Ten 1U servers, one processor choice

    
The agent's request. Please quote ten Lenovo ThinkSystem SR630 V4 servers, each with two AMD EPYC 9655 processors, sixteen 32 GB DDR5 memory modules, one Lenovo 25 Gb OCP network adapter and two 3.84 TB U.2 drives, with Premier Support 3-year. The quote shows a total price of $412,300.00 and a total power of 10.5 kW. It is ready to send.

    
Verdict: Blocked.

    
    
2. Processor limits: The agent resubmits after fixing the errors

    
The agent's request. Please quote ten Lenovo ThinkSystem SR630 V4 servers, each with two Intel Xeon 6 6787P processors, sixteen 32 GB DDR5 memory modules, one Lenovo 25 Gb OCP network adapter and two 3.84 TB U.2 drives, with Premier Support 3-year. The quote shows a total price of $394,379.60 and a total power of 10.02 kW. It is ready to send.

    
Verdict: Allowed.

    
    
3. GPU servers and power supplies: Six SR675 V3 GPU servers, eight GPUs each

    
The agent's request. Please quote six Lenovo ThinkSystem SR675 V3 GPU servers, each with two AMD EPYC 9355 processors, twenty-four 64 GB DDR5 memory modules, eight NVIDIA H200 NVL GPUs, two NVIDIA ConnectX-7 400 Gb adapters and two 7.68 TB E1.S drives, with Premier Support 5-year. The quote shows a total price of $1,950,000.00 and a total power of 36 kW. It is ready to send.

    
Verdict: Blocked.

    
    
4. GPU servers and power supplies: The agent resubmits after fixing the errors

    
The agent's request. Please quote six Lenovo ThinkSystem SR675 V3 GPU servers, each with two AMD EPYC 9355 processors, twenty-four 64 GB DDR5 memory modules, eight NVIDIA H200 NVL GPUs, two NVIDIA ConnectX-7 400 Gb adapters, two 7.68 TB E1.S drives and four Lenovo 3200 W power supplies, with Premier Support 5-year. The quote shows a total price of $2,125,096.80 and a total power of 37.74 kW. It is ready to send.

    
Verdict: Allowed.

    
    
5. HGX GPU baseboards: Four SR680a V4 HGX servers, GPU count

    
The agent's request. Please quote four Lenovo ThinkSystem SR680a V4 servers, each with two Intel Xeon 6 6787P processors, sixteen 64 GB DDR5 memory modules, six NVIDIA HGX B300 GPUs, eight NVIDIA ConnectX-8 800 Gb adapters and four 3.84 TB U.2 drives. Place them in a Lenovo 48U rack with 57.5 kVA PDUs and an NVIDIA Spectrum-X SN5600 switch, at a site with 60 kW per rack, with factory rack integration and deployment and Premier Support 5-year. The quote shows a total price of $1,850,000.00 and a total power of 36 kW. It is ready to send.

    
Verdict: Blocked.

    
    
6. HGX GPU baseboards: The agent resubmits after fixing the errors

    
The agent's request. Please quote four Lenovo ThinkSystem SR680a V4 servers, each with two Intel Xeon 6 6787P processors, sixteen 64 GB DDR5 memory modules, eight NVIDIA HGX B300 GPUs, eight NVIDIA ConnectX-8 800 Gb adapters and four 3.84 TB U.2 drives. Place them in a Lenovo 48U rack with 57.5 kVA PDUs and an NVIDIA Spectrum-X SN5600 switch, at a site with 60 kW per rack, with factory rack integration and deployment and Premier Support 5-year. The quote shows a total price of $2,348,597.76 and a total power of 45.84 kW. It is ready to send.

    
Verdict: Allowed.

    
    
7. Liquid cooling and facility water: Eight SR780a V3 Neptune servers, site cooling

    
The agent's request. Please quote eight Lenovo ThinkSystem SR780a V3 Neptune liquid-cooled GPU servers, each with two Intel Xeon Platinum 8580 processors, sixteen 64 GB DDR5 memory modules, eight NVIDIA HGX B200 GPUs, eight NVIDIA ConnectX-7 400 Gb adapters and four 7.68 TB E1.S drives. Place them in Lenovo 48U racks with four 57.5 kVA PDUs per rack and NVIDIA Spectrum-4 SN5400 switches, at a site with 80 kW per rack, with factory rack integration and deployment and Premier Support Plus with 4-hour response 3-year. The quote shows a total price of $3,800,000.00 and a total power of 80 kW. It is ready to send.

    
Verdict: Blocked.

    
    
8. Liquid cooling and facility water: The agent resubmits after fixing the errors

    
The agent's request. Please quote eight Lenovo ThinkSystem SR780a V3 Neptune liquid-cooled GPU servers, each with two Intel Xeon Platinum 8580 processors, sixteen 64 GB DDR5 memory modules, eight NVIDIA HGX B200 GPUs, eight NVIDIA ConnectX-7 400 Gb adapters and four 7.68 TB E1.S drives. Place them in Lenovo 48U racks with four 57.5 kVA PDUs per rack and NVIDIA Spectrum-4 SN5400 switches, at a site with facility water and 80 kW per rack, with factory rack integration and deployment and Premier Support Plus with 4-hour response 3-year. The quote shows a total price of $3,923,576.32 and a total power of 87.52 kW. It is ready to send.

    
Verdict: Allowed.

    
  

  

    
Narration transcript
Click to expand
    
The text of the narrated walkthrough, step by step.

    
    
      
  1. Welcome to a demo of Predictika's Validator. An AI sales agent prepares a quote for Lenovo ThinkSystem servers and racks: platform, processors, memory, and more. Before it goes out, the agent sends it to Predictika, which checks it against a declarative model of the configuration rules and returns a verdict, every reason, the figures it computed and, for a quote that can proceed, the line-item quote itself.
    2. 
      
  2. I will start with the first request by the agent.
    3. 
      
  3. The agent says. Please quote ten Lenovo ThinkSystem SR630 V4 servers, each with two AMD EPYC 9655 processors, sixteen 32 GB DDR5 memory modules. The quote shows a total price of $412,300.00 and a total power of 10.5 kW. The rest of the request is on screen.
    4. 
      
  4. Before it goes out, the agent calls the validator.
    5. 
      
  5. Blocked, with four findings, all in one answer. Lenovo ThinkSystem SR630 V4: AMD EPYC 9655 uses socket SP5 but this platform takes LGA4710. Lenovo ThinkSystem SR630 V4: AMD EPYC 9655 is rated 400 W; this chassis cools up to 350 W per processor. The quote states 10.5 kW total power; the validator computes 11.02 kW. The other one is on screen.
    6. 
      
  6. The validator returned four blocking findings in one answer. A check that reads one field at a time would pass most of this; it shows only when the choices in the quote are checked against each other and against the stated total.
    7. 
      
  7. What this shows: for processor socket and power limit that differ by platform, a rules-based CPQ is error-prone for this.
    8. 
      
  8. Now the agent resubmits.
    9. 
      
  9. The corrected request. It changes the processor to Intel Xeon 6 6787P.
    10. 
      
  10. It calls the validator again.
    11. 
      
  11. This time the validator clears it.
    12. 
      
  12. The validator also returns the line-item quote. It starts with Lenovo ThinkSystem SR630 V4 and Intel Xeon 6 6787P; the other seven lines are on screen.
    13. 
      
  13. All four earlier findings are cleared, and the validator's total, $394,379.60, matches the quote.
    14. 
      
  14. Four problems caught in one call, fixed in one pass, and a clean quote. That is the value of Predictika's validator: catching mistakes before they reach the customer. Now you can try the other scenarios yourself.
    15.