Predictika · Rackmount server orders

An AI sales agent's 2U server order, checked before it goes out

An AI sales agent prepares an order for 2U rackmount servers with the model, processors and cooling, memory modules, storage and PCIe configuration, accelerators and power supplies. Before it goes out, the agent sends it to Predictika, which checks it against a declarative model of the specifications QCT publishes on its product pages and returns a verdict, every reason and the figures it computed.

Scope: QCT QuantaGrid 2U servers (D54Q-2U, D53XQ-2U, D43KQ-2U and D52Y-2U): processors and cooling, memory, storage SKU, PCIe configuration, accelerators and power supplies. See everything this validator covers

This validator models orders for four QCT QuantaGrid 2U rackmount servers (D54Q-2U, D53XQ-2U, D43KQ-2U and D52Y-2U), not just the scenarios shown below.

  • Processors and cooling — two processors per server, the highest processor TDP each model is listed for (and the higher figure the D54Q-2U lists with liquid cooling), and the memory generation of the platform.
  • Memory — memory slots, module sizes listed for each model, and the memory per server compared with the figure the order states.
  • Storage — the storage SKUs listed for each model, with their front bays, the front bays that take NVMe drives, and the rear drives.
  • Expansion and accelerators — the PCIe configurations listed for each model, dual-width and single-width accelerator counts and the GPU configuration each needs, and the power supply wattages listed for each model.

Any order for these four models that touches these rules is in scope — not only the scenarios pictured here. Not covered: prices (the QCT product pages list no prices, so this validator produces no price build-up or line-item quote), other QuantaGrid models, specific processor, GPU, NIC and drive part numbers (QCT lists those in compatible components lists and datasheets that sit behind a registration form and were not read), and support contracts. Assumptions: a GPU configuration is read as supporting the accelerator type in its name, and accelerators on the D52Y-2U, or single-width ones on the D43KQ-2U, go to a person because the product page does not detail them.

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

6 scenarios in 3 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
An accelerator configuration that only one PCIe layout supportsOnly with care
A product rule per model that ties the accelerator type to the PCIe configuration, and a second one for the storage SKU a GPU layout is listed with.		The model, the PCIe configuration, the storage SKU and the accelerators are in one request and checked together.
A processor TDP that depends on the coolingOnly with care
A lookup per model, plus an exception for the one model listed with liquid cooling.		The limit comes from the model and the cooling together on every call.
A stated memory figure that disagrees with modules times sizeOnly with workarounds
The stated figure is rarely a field on the order, and the check needs a calculated field that has to be kept in step with module counts and sizes.		The stated figure travels with the order and is compared with the computed one.
Every problem returned in one answerOnly with workarounds
Rules tend to fire as the order 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 test cases
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; approval findings need a person to sign off; everything else is computed.

    
1The quote has processors rated 351 W, but the selected model with this cooling supports at most 350 W.
2The quote has processors rated 386 W, but the selected model with this cooling supports at most 385 W.
3The quote has processors rated 271 W, but the selected model with this cooling supports at most 270 W.
4The quote asks for liquid cooling, but only the QuantaGrid D54Q-2U is listed with a liquid-cooled TDP; the selected model is listed for air cooling only.
5The quote has processors rated 281 W, but the selected model with this cooling supports at most 280 W.
6The quote has processors rated 206 W, but the selected model with this cooling supports at most 205 W.
7The quote has 3 processors per server; these QuantaGrid 2U models take 2 processors.
8The memory type on the quote does not match the selected model; the D54Q-2U takes DDR5 and the D53XQ-2U, D43KQ-2U and D52Y-2U take DDR4.
9The quote has 33 memory modules per server, but the selected model has 32 memory slots.
10The quote has 256 GB memory modules, a size not listed for the selected model.
11The quote has 17 memory modules per server, but the selected model has 16 memory slots.
12The quote states 2,048 GB of memory per server, but the quoted modules give 1,024 GB.
13The quote has 1,200 W power supplies, a wattage not listed for the selected model.
14The quote has 2,200 W power supplies, a wattage not listed for the selected model.
15The storage SKU on the quote is not one of the storage SKUs listed for the selected model.
16The PCIe configuration on the quote is not one of the expansion configurations listed for the selected model.
17The quote has 25 front drives per server, but the selected storage SKU has 24 front bays.
18The quote has 9 NVMe front drives per server, but the selected storage SKU takes at most 8.
19The quote has 2 NVMe front drives per server, but the selected storage SKU takes no NVMe drives.
20The quote has 2 rear drives per server, but the selected storage SKU lists none.

  

  

    
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. Accelerators on a D53XQ-2U: Three dual-width accelerators on a D53XQ-2U order

    
The agent's request. Please quote 24 QuantaGrid D53XQ-2U servers, air-cooled, each with 2 processors rated 270 W, 32 DDR4 memory modules of 128 GB, the SFF All Flash storage SKU with 24 front drives (24 of them NVMe), the General PCIe configuration, 3 dual-width accelerators, and 1,600 W power supplies. The quote states 4,096 GB of memory per server. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • More than two dual-width accelerators. The quote has 3 dual-width accelerators per server; the 2U models are listed with at most 2.
    • 
      
  • Accelerators without the matching GPU configuration. The quote has accelerators, but the PCIe configuration is not the GPU configuration listed for that accelerator type on the selected model.
    • 
    

    
2. Accelerators on a D53XQ-2U: The agent resubmits after fixing the errors

    
The agent's request. Please quote 24 QuantaGrid D53XQ-2U servers, air-cooled, each with 2 processors rated 270 W, 32 DDR4 memory modules of 128 GB, the SFF All Flash storage SKU with 24 front drives (24 of them NVMe), the Dual Width GPU PCIe configuration, 2 dual-width accelerators, and 1,600 W power supplies. The quote states 4,096 GB of memory per server. It is ready to send.

    
Verdict: Allowed.

    
    
      
  • Server summary. 24 servers, each with 4,096 GB of memory (32 modules of 128 GB), 98,304 GB across the order, 24 front bays on the storage SKU, 270 W processors, 1,600 W power supplies.
    • 
    

    
3. GPU server on a D43KQ-2U: A D43KQ-2U GPU order with tiered storage and 256 GB modules

    
The agent's request. Please quote 12 QuantaGrid D43KQ-2U servers, air-cooled, each with 2 processors rated 300 W, 32 DDR4 memory modules of 256 GB, the LFF Tiered storage SKU with 12 front drives (8 of them NVMe), the GPU PCIe configuration, 2 dual-width accelerators, and 1,600 W power supplies. The quote states 8,192 GB of memory per server. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Processor TDP above the model limit. The quote has processors rated 300 W, but the selected model with this cooling supports at most 280 W.
    • 
      
  • Memory module size not listed for the model. The quote has 256 GB memory modules, a size not listed for the selected model.
    • 
      
  • D43KQ GPU configuration needs a GPU-capable storage SKU. The D43KQ-2U GPU PCIe configuration is listed only with the SFF Passthrough and SFF Expander Pure SATA/SAS storage SKUs.
    • 
    

    
4. GPU server on a D43KQ-2U: The agent resubmits after fixing the errors

    
The agent's request. Please quote 12 QuantaGrid D43KQ-2U servers, air-cooled, each with 2 processors rated 280 W, 32 DDR4 memory modules of 128 GB, the SFF Passthrough storage SKU with 24 front drives (8 of them NVMe), the GPU PCIe configuration, 2 dual-width accelerators, and 1,600 W power supplies. The quote states 4,096 GB of memory per server. It is ready to send.

    
Verdict: Allowed.

    
    
      
  • Server summary. 12 servers, each with 4,096 GB of memory (32 modules of 128 GB), 49,152 GB across the order, 24 front bays on the storage SKU, 280 W processors, 1,600 W power supplies.
    • 
    

    
5. Liquid-cooled D54Q-2U order: A D54Q-2U order with 385 W processors

    
The agent's request. Please quote 40 QuantaGrid D54Q-2U servers, air-cooled, each with 2 processors rated 385 W, 32 DDR4 memory modules of 128 GB, the SFF Tiered storage SKU with 24 front drives (8 of them NVMe), the General PCIe configuration, no accelerators, and 1,200 W power supplies. The quote states 4,096 GB of memory per server. It is ready to send.

    
Verdict: Blocked.

    
    
      
  • Processor TDP above the model limit. The quote has processors rated 385 W, but the selected model with this cooling supports at most 350 W.
    • 
      
  • Memory generation does not match the model. The memory type on the quote does not match the selected model; the D54Q-2U takes DDR5 and the D53XQ-2U, D43KQ-2U and D52Y-2U take DDR4.
    • 
      
  • Power supply wattage not listed for the model. The quote has 1,200 W power supplies, a wattage not listed for the selected model.
    • 
    

    
6. Liquid-cooled D54Q-2U order: The agent resubmits after fixing the errors

    
The agent's request. Please quote 40 QuantaGrid D54Q-2U servers, liquid-cooled, each with 2 processors rated 385 W, 32 DDR5 memory modules of 128 GB, the SFF Tiered storage SKU with 24 front drives (8 of them NVMe), the General PCIe configuration, no accelerators, and 2,700 W power supplies. The quote states 4,096 GB of memory per server. It is ready to send.

    
Verdict: Allowed.

    
    
      
  • Server summary. 40 servers, each with 4,096 GB of memory (32 modules of 128 GB), 163,840 GB across the order, 24 front bays on the storage SKU, 385 W processors, 2,700 W power supplies.
    • 
    

  

  

    
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 an order for 2U rackmount servers with the model, processors and cooling, memory modules, storage and PCIe configuration, accelerators and power supplies. Before it goes out, the agent sends it to Predictika, which checks it against a declarative model of the specifications QCT publishes on its product pages and returns a verdict, every reason and the figures it computed.
    2. 
      
  2. I will start with the first request by the agent.
    3. 
      
  3. The agent says. Please quote 12 QuantaGrid D43KQ-2U servers, air-cooled, each with 2 processors rated 300 W, 32 DDR4 memory modules of 256 GB. The quote states 8,192 GB of memory per server. The rest of the request is on screen.
    4. 
      
  4. Before it goes out, the agent calls the validator.
    5. 
      
  5. Blocked, with three findings, all in one answer. The quote has processors rated 300 W, but the selected model with this cooling supports at most 280 W. The quote has 256 GB memory modules, a size not listed for the selected model. The D43KQ-2U GPU PCIe configuration is listed only with the SFF Passthrough and SFF Expander Pure SATA/SAS storage SKUs.
    6. 
      
  6. The validator returned three 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.
    7. 
      
  7. What this shows: for a stated memory figure that disagrees with modules times size, a rules-based CPQ can do this only with workarounds.
    8. 
      
  8. Now the agent resubmits.
    9. 
      
  9. The corrected request. It changes the Processor TDP to 280 W, the memory module size to 128 GB, the Storage SKU to SFF Passthrough, the front drives per server to 24 and the stated memory per server as to 4,096 GB.
    10. 
      
  10. It calls the validator again.
    11. 
      
  11. This time the validator clears it.
    12. 
      
  12. All three earlier findings are cleared, and the validator's figure matches the quote.
    13. 
      
  13. Three 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.
    14.