HP NVIDIA Quadro K4000 3 GB (713381-001)

2026-09-22

Desktop GPU Server

HP NVIDIA Quadro K4000 3 GB (713381-001)

Exhibit Type: Professional Workstation Graphics Accelerator & Silicon Specification Dossier
•
Microarchitecture: Kepler • GK106GL (28 nm TSMC)
•
Compute Pipeline: 768 CUDA Cores (4 SMX) • 64 TMUs • 24 ROPs • 810 MHz Core Clock
•
VRAM & Bus: 3 GB GDDR5 • 192-bit Interface • 134.4 GB/s Bandwidth • PCIe 2.0 x16
•
PassMark G3D Mark: 2,718 (Rank #7 in Museum GPU Collection)
•
Museum Inventory: 1 Physical Unit (HP Spare: 713381-001 • HP Assy: 700104-001 • S/N: 0322314010635)
•
Active Host Pairing: Dell Precision T7610 Workstation (PCIe Slot 2 • Bus 03:00.0)

1. Technical Overview & Key Specifications Table

Category	Property	Specification
GPU Identity	Product Name	NVIDIA Quadro K4000 (HP OEM Workstation Edition)
	OEM Board Partner	Hewlett-Packard (HP Inc.)
	HP Part Numbers	HP Spare: `713381-001` • HP Assembly: `700104-001`
	GPU Code Name & Revision	GK106GL (Variant `GK106-ES-A1` / Revision `A1`)
	PCI Device ID / Subsystem	PCI ID: `10de:11fa` • Subsystem ID: `103c:079c`
	vBIOS Version	`80.06.76.00.07` (HP Factory Workstation Firmware)
	Market Segment & Release	Professional CAD / DCC / CAE Workstation • March 1, 2013
Physical Silicon	Foundry & Lithography	TSMC (Taiwan Semiconductor Manufacturing Co.) • 28 nm Bulk CMOS (HP)
	Transistor Count	2,540 Million (2.54 Billion) transistors
	Die Surface Area	221 mm² (Monolithic Die)
	PCB Markings & S/N	`900-52030-0300-000 L` • S/N: `0322314010635` (HP S/N: `CN42404748`)
	Form Factor	Single-Slot Standard Full-Height (9.5-inch / 241 mm length)
Compute Pipeline	Microarchitecture	Kepler (SMX Generation)
	CUDA Core Count	768 CUDA Cores (4 SMX Units × 192 Cores/SMX)
	Texture Mapping Units (TMUs)	64 TMUs (16 TMUs per SMX)
	Render Output Units (ROPs)	24 ROPs (3 ROP Partitions of 8 ROPs each)
	GPU Core Base Clock	810 MHz
	Single-Precision Compute (FP32)	1.244 TFLOPS (1,244.2 GFLOPS)
	Double-Precision Compute (FP64)	51.84 GFLOPS (1/24 FP32 Rate)
Memory Subsystem	Dedicated VRAM Size	3,072 MB (3 GB) GDDR5
	Memory Bus Width	192-bit (3 x 64-bit Memory Controllers)
	Memory Clock Frequency	1,404 MHz (5,616 MT/s Effective Data Rate)
	Peak Memory Bandwidth	134.78 GB/s (134,784 MB/s)
	L2 Cache Size	384 KB High-Speed L2 Cache (128 KB per 64-bit Controller)
	ECC Memory Support	Non-ECC GDDR5 (ECC reserved for Quadro K5000 / K6000 tiers)
Display & Media I/O	Physical Video Outputs	2 × DisplayPort 1.2 + 1 × Dual-Link DVI-I
	Maximum Digital Resolution	3840 × 2160 @ 60 Hz (DisplayPort 1.2) • 2560 × 1600 @ 60 Hz (Dual-Link DVI)
	Maximum Active Displays	Up to 3 Concurrent Independent Displays (or 4 via DP 1.2 Multi-Stream Transport)
	Hardware Video Acceleration	1st Gen NVENC (H.264 up to 4K) • PureVideo HD VP5 (MPEG-2, VC-1, H.264)
Thermal & Electrical	Thermal Design Power (TDP)	80 Watts Maximum Power Consumption
	Auxiliary Power Connector	1 × 6-pin PCIe Auxiliary Power Connector
	Cooling Solution	Active Single-Slot Blower Fan with Copper Core Heatsink
	Live Operating Telemetry	+44.0 deg C Idle • 2100 RPM Fan Speed • 20.08 Watts Idle Power Draw
API & Platform Support	DirectX Support	DirectX 12 (Feature Level 11_0) / Shader Model 5.0
	OpenGL / Vulkan	OpenGL 4.6 • Vulkan 1.2
	Compute APIs	CUDA Compute Capability 3.0 • OpenCL 1.2 • DirectCompute
	Operating System Drivers	Linux `nouveau` DRM (Kernel 6.8 Verified) • NVIDIA Legacy 470.xx Driver

2. GK106 Silicon Topology & Kepler SMX Architecture

The GK106 silicon chip powers NVIDIA’s mid-range Kepler lineup. Designed for high energy efficiency and workstation reliability, the SMX (Next-Generation Streaming Multiprocessor) architecture replaced Fermi’s high-frequency hot-clock design with an ultra-dense array of 192 FP32 ALU CUDA cores per SMX block:

+======================================================================================================================+
|                         NVIDIA GK106GL [QUADRO K4000] SILICON DIE & SMX COMPUTE TOPOLOGY                              |
+======================================================================================================================+

                          HOST INTERFACE: PCI EXPRESS 2.0 / 3.0 x16 BUS (ROOT PORT 00:02.0)
                                                    |
                                                    v
                                  +------------------------------------+
                                  | GIGA-THREAD ENGINE & HOST DISPATCH |
                                  +------------------------------------+
                                                    |
         +------------------------------------------+------------------------------------------+
         |                                                                                     |
         v                                                                                     v
+------------------------------------+                               +------------------------------------+
| GRAPHICS PROCESSING CLUSTER 0      |                               | GRAPHICS PROCESSING CLUSTER 1      |
|                                    |                               |                                    |
|  +-------------------------------+ |                               |  +-------------------------------+ |
|  | SMX BLOCK 0                   | |                               |  | SMX BLOCK 2                   | |
|  | - 192 FP32 CUDA Cores         | |                               |  | - 192 FP32 CUDA Cores         | |
|  | - 32 Special Func Units (SFU) | |                               |  | - 32 Special Func Units (SFU) | |
|  | - 32 Load/Store Units (LD/ST) | |                               |  | - 32 Load/Store Units (LD/ST) | |
|  | - 16 Texture Units (TMU)      | |                               |  | - 16 Texture Units (TMU)      | |
|  | - 64 KB L1 / Shared Memory    | |                               |  | - 64 KB L1 / Shared Memory    | |
|  | - PolyMorph Engine 2.0        | |                               |  | - PolyMorph Engine 2.0        | |
|  +-------------------------------+ |                               +-------------------------------+ |
|                                    |                               |                                    |
|  +-------------------------------+ |                               |  +-------------------------------+ |
|  | SMX BLOCK 1                   | |                               |  | SMX BLOCK 3                   | |
|  | - 192 FP32 CUDA Cores         | |                               |  | - 192 FP32 CUDA Cores         | |
|  | - 32 Special Func Units (SFU) | |                               |  | - 32 Special Func Units (SFU) | |
|  | - 32 Load/Store Units (LD/ST) | |                               |  | - 32 Load/Store Units (LD/ST) | |
|  | - 16 Texture Units (TMU)      | |                               |  | - 16 Texture Units (TMU)      | |
|  | - 64 KB L1 / Shared Memory    | |                               |  | - 64 KB L1 / Shared Memory    | |
|  | - PolyMorph Engine 2.0        | |                               |  | - PolyMorph Engine 2.0        | |
|  +-------------------------------+ |                               +-------------------------------+ |
+------------------------------------+                               +------------------------------------+
         |                                                                                     |
         +------------------------------------------+------------------------------------------+
                                                    |
                                                    v
         +------------------------------------------+------------------------------------------+
         |                                          |                                          |
         v                                          v                                          v
+-----------------------+                  +-----------------------+                  +-----------------------+
| MEMORY CONTROLLER 0   |                  | MEMORY CONTROLLER 1   |                  | MEMORY CONTROLLER 2   |
| - 64-bit GDDR5 Bus    |                  | - 64-bit GDDR5 Bus    |                  | - 64-bit GDDR5 Bus    |
| - 128 KB L2 Cache     |                  | - 128 KB L2 Cache     |                  | - 128 KB L2 Cache     |
| - 8 ROP Units         |                  | - 8 ROP Units         |                  | - 8 ROP Units         |
+-----------------------+                  +-----------------------+                  +-----------------------+
         |                                          |                                          |
         v                                          v                                          v
    1,024 MB GDDR5                             1,024 MB GDDR5                             1,024 MB GDDR5
   (1404 MHz Clock)                           (1404 MHz Clock)                           (1404 MHz Clock)
+-------------------------------------------------------------------------------------------------------------+
| AGGREGATE MEMORY SUBSYSTEM: 3,072 MB (3 GB) GDDR5 • 192-BIT BUS • 134.4 GB/s THEORETICAL BANDWIDTH           |
+=============================================================================================================+

3. PassMark Synthetic Benchmarks & Performance Metrics

In the standardized PassMark PerformanceTest 3D Graphics benchmark suite, the NVIDIA Quadro K4000 achieves a composite G3D Mark of 2,718, positioning it as a highly capable legacy workstation accelerator with robust DirectX 11, OpenGL, and DirectCompute capabilities:

PassMark G3D Mark

2,718

Rank #7 in Museum GPU Collection

PassMark G2D Mark

614

High 2D Desktop GUI Responsiveness

DirectCompute Score

1,385

GPGPU Physics & Compute Acceleration

Performance vs GTS 250

+383%

Nearly 5× Faster Than Former T7610 GPU

Hardware Model	Silicon Architecture	VRAM Specification	TDP (Watts)	PassMark G3D	Relative Variance
Gigabyte GeForce GTS 250	Tesla (G92b, 55 nm, 128 cores)	512 MB GDDR3 (256-bit)	150W	563	-79.3% (Baseline Replaced)
NVIDIA Quadro 4000	Fermi (GF100, 40 nm, 256 cores)	2 GB GDDR5 (256-bit)	142W	1,822	-33.0% (Prior Generation)
HP NVIDIA Quadro K4000	Kepler (GK106GL, 28 nm, 768 cores)	3 GB GDDR5 (192-bit)	80W	2,718	Current Museum Exhibit
MSI GTX 680 Lightning	Kepler (GK104, 28 nm, 1536 cores)	2 GB GDDR5 (256-bit)	195W	5,610	+106.4% (Museum Rank #6)

4. Platform Synergy & Host Workstation Pairing

The HP OEM NVIDIA Quadro K4000 is installed as the primary graphics accelerator in the museum’s high-performance enterprise compute node:

Host Machine: Dell Precision T7610 Workstation (Exhibit #2973)

Physical Slot Topology: Populated in PCIe Slot 2 (Bus 03:00.0), downstream of CPU 1’s Root Port 00:02.0. The card operates on a dedicated 16-lane PCIe link directly tied to Socket 0’s integrated PCIe root complex.
Single-Slot Form Factor Advantage: With an 80W TDP and a compact single-slot profile, the Quadro K4000 leaves all neighboring PCIe expansion slots unblocked, allowing full deployment of the LSI SAS 2308 RAID controller, dual Intel Gigabit NICs, and the Intel Wi-Fi 6 AX200 desktop PCIe adapter.
Thermals & Acoustics: Operating under Linux Kernel 6.8 with the open-source nouveau DRM driver, the GPU maintains a cool +44.0 deg C idle operating temperature with whisper-quiet fan acoustics (2100 RPM) inside the Dell T7610’s partitioned thermal chambers.

5. Curator’s Assessment & Museum Provenance

Museum Hardware Provenance: Mateo’s Museum collection preserves 1 authentic physical unit of the HP OEM NVIDIA Quadro K4000 graphics card. Optical hardware inspection and DMI telemetry confirm full hardware match:
- HP Spare Part Number: 713381-001
- HP Assembly Part Number: 700104-001
- PCB Part Number: 900-52030-0300-000 L • 699-52030-0500-510 E
- Board Serial Number: 0322314010635 (HP S/N: CN42404748)
- Factory vBIOS Firmware: Version 80.06.76.00.07
Replacement of Legacy GTS 250: The Quadro K4000 replaced a legacy Gigabyte GeForce GTS 250 (512 MB GDDR3) that was originally cataloged in the Precision T7610. The upgrade provides 6× more video memory (3 GB vs. 512 MB), 6× more CUDA cores (768 vs. 128), and a 383% boost in synthetic 3D performance while slashing TDP from 150W down to 80W.
Workstation Multi-Monitor Display Hub: Equipped with dual DisplayPort 1.2 and dual-link DVI-I outputs, the K4000 drives modern 4K UHD panels at 60 Hz with 10-bit color accuracy, making it an ideal display hub for virtualization consoles, administrative monitoring, and technical CAD drafting.