Guides & Resources — GPU Servers & On-Prem AI
TopicAll
TypeAll
New to GPU Servers & On-Prem AI?
Start with the guide hub: the essential reads, reference tables and parts in one place →
Front-End vs Back-End Networking for an AI Server: When an Intel X710 Is Enough and When You Need ConnectX
One inference server needs only a 10GbE front end, and an Intel X710 covers it. A back-end RDMA fabric on ConnectX matters only when a model or training job spans servers. Which card and cable fits each, and why 100G is the ceiling in a...
750W, 1100W or 1600W? Sizing Power Supplies for a GPU Server (Dell R740xd, HPE DL380 Gen10)
Dell requires 1100W-or-higher supplies for any GPU in an R740xd, and a 1600W supply delivers only 800W on a 120V outlet. How to size power supplies for one, two or three GPUs in a Dell R740xd or HPE DL380 Gen10, with worked examples and prices.
The Complete Parts List for Turning a Refurbished R740xd or DL380 Gen10 Into an AI Inference Server
A GPU alone doesn't make a refurbished server an AI box. The seven parts an R740xd or DL380 Gen10 inference build needs, from the GPU enablement kit to power, memory, boot, storage and network, with part numbers and prices.
How Much System RAM Does an On-Prem LLM Server Need? (And Why Your DDR4-3200 Runs at 2933)
System RAM for an on-prem LLM server: at least twice your GPU memory, spread across all twelve channels. Configurations and prices for A40 and V100 builds, why bandwidth matters for CPU offload, and why DDR4-3200 runs at 2933 or 2400 in an R740xd.
Why Your 200G ConnectX-6 Tops Out Near 100G in a Dell R740xd or R640 (PCIe Gen3 Explained)
A PCIe 3.0 x16 slot carries about 100-110 Gb/s of real traffic, so a 200G ConnectX-6 in an R740xd or R640 runs at half speed. The math, card-by-card effects, how to check your slot, and what to buy for Gen3, Gen4 and Gen5 servers.
Why Your GPU Won't Work in Your Server: 8 Installation Problems and the Fix for Each (Dell PowerEdge & HPE ProLiant)
A data-center GPU that won't work in a server is almost never a bad card. Eight installation problems in Dell PowerEdge and HPE ProLiant servers, from BIOS memory mapping to the EPS vs PCIe power cable, and the fix for each.
871828, 871829, 871830: Which HPE GPU Cable Kit Do You Actually Need?
HPE publishes five graphics cable kits for ProLiant Gen10 servers and the part numbers differ by a single digit. Which one you need, which servers and GPUs each covers, and when the single 869820-001 cable is the better buy.
Custom Dell R740xd GPU Servers: Build Options, Real Prices, and the Forum Questions Nobody Answers
Resilient Tec builds custom Dell PowerEdge R740xd GPU servers starting at $2,995, with four tested starting points up to the $12,159 A40 build. Plus straight answers to the questions that sit unanswered on Reddit and Dell's forums: which GPU power cable (4VPD3 vs TR5TP), why...
What Can You Actually Run on a 48GB NVIDIA A40? Measured Model Sizes and Tokens Per Second (2026)
One A40 48GB runs 27–34B models at 23–29 tokens/sec, fits a 70B at Q4 at 12–13.5 tokens/sec, and serves 1,705 tokens/sec of aggregate output on an 8B model at 100 concurrent requests. The measured numbers, the KV cache math behind the 72B cliff, and the...
Putting an NVIDIA A40 in a PowerEdge R740xd: Every Part Dell Doesn't Tell You You Need
An NVIDIA A40 48GB works in a Dell PowerEdge R740xd — but the card is not the purchase. The full bill of materials: GPU enablement kit 330-BBMO (US) / 490-BEIX (EMEA), low-profile heat sinks in both CPU sockets, the high-performance fan kit, air shroud and...
NVIDIA's Blackwell GB200 NVL72: What It Means for the Used H100 Market
Blackwell is here — and for buyers of used H100 hardware, that's the best news of the year. NVIDIA's GB200 NVL72 rack system represents the most significant leap in AI compute density the industry has ever shipped. But its arrival is setting off a chain...
The Cloud Sent You a Bill. Time to Come Home.
Call us at: 888.442-3849 · Same Day Shipping Resilient Tec, LLCGPU Interconnects Network Hardware Servers/StorageView AI System Home › Blog › Workload Repatriation AI Infrastructure · Workload Repatriation The Cloud Sent You a Bill.It's Time to Come Home. How AI upended the economics of cloud...