Gigabyte GeForce GTS 250 512MB (GV-N250-512I)
Exhibit Type: Consumer Hardware Specification & Architectural Dossier
•
Microarchitecture: Tesla 2.0 / G92b (55 nm TSMC)
•
VRAM & Bus: 512 MB GDDR3 (256-bit Bus, 64.0 GB/s)
•
Museum Inventory: 1 Unit (Single Cooler, Installed in Dell Precision T7610)
•
Source Authority: TechPowerUp, Gigabyte & PassMark
•
Microarchitecture: Tesla 2.0 / G92b (55 nm TSMC)
•
VRAM & Bus: 512 MB GDDR3 (256-bit Bus, 64.0 GB/s)
•
Museum Inventory: 1 Unit (Single Cooler, Installed in Dell Precision T7610)
•
Source Authority: TechPowerUp, Gigabyte & PassMark
1. Technical Overview & Key Specifications
| Category | Property | Specification |
|---|---|---|
| GPU Identity | Model Name | Gigabyte GeForce GTS 250 512MB (GV-N250-512I) |
| Board Partner / Vendor | Giga-Byte Technology Co., Ltd. / NVIDIA Corporation | |
| GPU Code Name / Variant | G92B (G92-428-B1 / G92-426-B1) | |
| Architecture Generation | Tesla (Tesla 2.0 / G92b 55 nm Die Shrink) • GeForce 200 Series | |
| Market Segment & Launch | Desktop / Performance Mainstream • March 3, 2009 (CeBIT 2009) | |
| Product Part Number & Revision | GV-N250-512I (Single Cooler Edition, Ultra Durable 2 PCB) |
|
| Physical Silicon & Foundry | Foundry & Process Size | TSMC (Taiwan Semiconductor Manufacturing Co.) • 55 nm CMOS |
| Transistor Count | 754 Million transistors | |
| Die Surface Area | 260 mm² (Down from 324 mm² on original 65 nm G92) | |
| Transistor Density | 2.9 Million transistors / mm² | |
| Package Type | FC-BGA (Flip-Chip Ball Grid Array, 1144 contacts) | |
| Compute Pipeline & Clocks | CUDA Cores / Shaders | 128 Stream Processors (Unified 1D Scalar Shaders) |
| Compute Units / TPCs | 8 Texture/Processor Clusters (TPCs) • 16 Streaming Multiprocessors (SMs) | |
| Texture Units (TMUs) | 64 TMUs (Address & Filtering) | |
| Render Output Units (ROPs) | 16 ROPs (4 independent 4-wide ROP partitions) | |
| Core Engine Clock | 675 MHz (Gigabyte GV-N250-512I profile • 738 MHz Reference) | |
| Shader (ALU) Clock | 1,620 MHz (Decoupled 2.4× shader domain • 1,836 MHz Reference) | |
| Single-Precision Compute (FP32) | 414.72 GFLOPS (0.415 TFLOPS) | |
| Fillrates (Pixel / Texture) | 10.80 GPixels/sec • 43.20 GTexels/sec | |
| Memory Subsystem | VRAM Capacity & Type | 512 MB GDDR3 SDRAM |
| Memory Bus Width | 256-bit (4 × 64-bit crossbar memory channels) | |
| Memory Clock Frequency | 1,000 MHz (2,000 MHz Effective GDDR3 Data Rate) | |
| Peak Memory Bandwidth | 64.0 GB/s sustained bandwidth | |
| Memory Chips Configuration | 8 × 64 MB (16M × 32) high-speed BGA GDDR3 memory packages | |
| Thermal & Electrical | Thermal Design Power (TDP) | 150 W Maximum Board Power |
| Suggested System PSU | 450 W Minimum (Installed in Dell Precision T7610 with 1300W Gold PSU) | |
| Auxiliary Power Connector | 1 × 6-pin PCIe auxiliary power connector (75W PCIe slot + 75W 6-pin cable) | |
| Cooling Solution & Form Factor | Dual-slot width • Single axial fan cooler with circular aluminum fin array and copper core | |
| Display & Host I/O | Host Bus Interface | PCI Express 2.0 x16 (5.0 GT/s per lane • 8.0 GB/s bidirectional throughput) |
| Physical Display Connectors | 1 × HDMI (Gold-Plated) • 1 × Dual-Link DVI-I • 1 × D-Sub (VGA) | |
| Maximum Resolutions | Digital: 2560 × 1600 @ 60 Hz • Analog (VGA): 2048 × 1536 @ 85 Hz | |
| Multi-Display & Multi-GPU | Dual independent active displays • 2-Way and 3-Way NVIDIA SLI (dual MIO bridge fingers) | |
| Software & API Support | DirectX Support | Microsoft DirectX 10.0 • Feature Level 10_0 • Shader Model 4.0 |
| OpenGL / OpenCL | OpenGL 3.3 (Driver 340.xx legacy) • OpenCL 1.1 | |
| NVIDIA Compute Ecosystem | CUDA Compute Capability 1.1 • NVIDIA PhysX • NVIDIA PureVideo HD (VP2) | |
| Vulkan / Modern APIs | Not supported (Vulkan requires Kepler architecture or newer) | |
| PassMark Benchmarks | PassMark G3D Mark | 563 (PassMark PerformanceTest composite 3D graphics score) |
| PassMark G2D Mark | 47 (2D Vector and UI rendering score) | |
| Category Rank & Population | Rank #1,801 across all tested desktop video cards (3,277 submitted samples) |
2. Silicon Topology & Tesla G92b Microarchitecture
Silicon Scale
754M
Transistors • 260 mm² die (55 nm TSMC)
Compute Pipeline
128 Cores
16 SMs across 8 TPCs @ 1,620 MHz
Raster & Texture
64 TMU / 16 ROP
43.20 GTexel/s • 10.80 GPixel/s
Memory Throughput
64.0 GB/s
256-bit GDDR3 crossbar @ 2,000 MHz
Internal Silicon Die Topology & Compute Pipeline Map
+---------------------------------------------------------------------------------------------------------+
| NVIDIA G92B SILICON DIE TOPOLOGY (55nm TSMC • 754M TRANSISTORS) |
| |
| +---------------------------------------------------------------------------------------------------+ |
| | HOST INTERFACE: PCI EXPRESS 2.0 x16 (5.0 GT/s) | |
| +---------------------------------------------------------------------------------------------------+ |
| | |
| +---------------------------------------------------------------------------------------------------+ |
| | GigaThread ENGINE (HARDWARE THREAD SCHEDULER & DISPATCH) | |
| +---------------------------------------------------------------------------------------------------+ |
| | |
| +---------------------------------------------------------------------------------------------------+ |
| | TEXTURE / PROCESSOR CLUSTERS (TPC 0 - 7) | |
| | | |
| | +---------------------------+ +---------------------------+ +----------------------------+ | |
| | | TPC 0 | | TPC 1 | ... | TPC 7 | | |
| | | +---------------------+ | | +---------------------+ | | +----------------------+ | | |
| | | | SM 0: 8 SPs (ALUs) | | | | SM 2: 8 SPs (ALUs) | | | | SM 14: 8 SPs (ALUs) | | | |
| | | | 16KB Shared Memory | | | | 16KB Shared Memory | | | | 16KB Shared Memory | | | |
| | | +---------------------+ | | +---------------------+ | | +----------------------+ | | |
| | | | SM 1: 8 SPs (ALUs) | | | | SM 3: 8 SPs (ALUs) | | | | SM 15: 8 SPs (ALUs) | | | |
| | | | 16KB Shared Memory | | | | 16KB Shared Memory | | | | 16KB Shared Memory | | | |
| | | +---------------------+ | | +---------------------+ | | +----------------------+ | | |
| | | | 8 TMUs (Addr & Filt) | | | | 8 TMUs (Addr & Filt) | | | | 8 TMUs (Addr & Filt) | | | |
| | +---------------------------+ +---------------------------+ +----------------------------+ | |
| | [ Total: 8 TPCs • 16 Streaming Multiprocessors • 128 CUDA Cores • 64 Texture Mapping Units ] | |
| +---------------------------------------------------------------------------------------------------+ |
| | |
| +---------------------------------------------------------------------------------------------------+ |
| | HIGH-SPEED INTERCONNECTION CROSSBAR & 512KB L2 CACHE | |
| +---------------------------------------------------------------------------------------------------+ |
| | | | | |
| +---------------+ +---------------+ +---------------+ +---------------+
| | ROP PART 0 | | ROP PART 1 | | ROP PART 2 | | ROP PART 3 |
| | 4 ROP Units | | 4 ROP Units | | 4 ROP Units | | 4 ROP Units |
| | 64-bit MC 0 | | 64-bit MC 1 | | 64-bit MC 2 | | 64-bit MC 3 |
| +---------------+ +---------------+ +---------------+ +---------------+
| | | | | |
| +---------------------------------------------------------------------------------------------------+ |
| | 256-BIT MEMORY BUS (4 x 64-bit CHANNELS • 512 MB GDDR3 @ 64.0 GB/s) | |
| +---------------------------------------------------------------------------------------------------+ |
| | |
| +---------------------------------------------------------------------------------------------------+ |
| | PureVideo HD (VP2 ENGINE) | DISPLAY ENGINE (DUAL RAMDACs, HDMI, DVI, VGA) | |
| | - Full H.264 / VC-1 / MPEG-2 Bitstream | - 1x Gold-Plated HDMI w/ Internal S/PDIF Audio Header| |
| | - Hardware Motion Comp & Inverse DCT | - 1x Dual-Link DVI-I (2560x1600) • 1x D-Sub VGA | |
| +---------------------------------------------------------------------------------------------------+ |
+---------------------------------------------------------------------------------------------------------+
3. Memory Subsystem & PCI Express Interconnect
256-Bit GDDR3 Crossbar Architecture
- Memory Topology: 4 discrete 64-bit memory controllers aggregated via a non-blocking crossbar network into a unified 256-bit bus.
- High-Speed GDDR3: Operates at 1,000 MHz base frequency (2,000 MHz effective double-data-rate transfer rate).
- Substantial Bandwidth: Yields 64.0 GB/s of sustained memory throughput, preventing rasterization stalls during anti-aliased 1080p and 1200p rendering.
- Ultra Durable 2 Design: Gigabyte incorporates Lower RDS(on) MOSFETs, Ferrite Core chokes, and All-Solid Japanese capacitors for stable VRAM signaling.
PCIe 2.0 Host Bus & SLI Interconnect
- PCIe 2.0 x16 Interface: Delivers 5.0 GT/s signaling rate, doubling bandwidth over original PCIe 1.0 to 8.0 GB/s in each direction (16.0 GB/s aggregate).
- Full Backward & Forward Compatibility: Negotiates flawlessly in modern PCIe 3.0, 4.0, and 5.0 motherboard slots, including workstation PEG slots.
- Multi-GPU Scalability: Dual MIO gold fingers support 2-Way and 3-Way NVIDIA SLI configurations for synchronized alternate-frame rendering (AFR).
- Platform Auxiliary Feed: Draws up to 75 W directly from the PCIe slot pins, supplemented by an auxiliary 6-pin PCIe connection for complete 150 W headroom.
4. Media Engines, Video Codecs, AI & API Ecosystem
| Engine / Technology | Supported Standards | Functional Capability & Limitations |
|---|---|---|
| Hardware Video Decoding | NVIDIA PureVideo HD (VP2) | Full bitstream hardware acceleration for H.264 / MPEG-4 AVC and VC-1; hardware motion compensation, IDCT, and dual-stream picture-in-picture Blu-ray decoding. |
| Legacy Video Codecs | Hardware MPEG-2 and WMV9 decode acceleration. No VP9, HEVC/H.265, or AV1 decode (software decoding required on host CPU). | |
| Hardware Video Encoding | NVENC Engine | Not available. NVENC hardware encoding was first introduced in Kepler (GK104) GPUs. Video encoding on GTS 250 relies on CUDA-based software encoding or host CPU. |
| CUDA Video Encoding | Supports legacy Badaboom and MediaCoder CUDA-accelerated H.264 transcoding pipelines via CUDA 1.1 compute kernels. | |
| Compute & AI Acceleration | NVIDIA CUDA 1.1 | 128 scalar ALUs operate as CUDA compute workers. Suitable for legacy scientific algorithms, Folding@home classic, and vintage CUDA 1.x – 3.x applications. |
| Tensor & RT Cores | None. Predates Volta Tensor cores and Turing RT cores. Modern LLM inference (Ollama/vLLM) and hardware ray tracing are not supported. | |
| Audio & Physical Connectivity | HDMI Audio Passthrough | Onboard 2-pin S/PDIF audio header connects to motherboard/sound card audio output to route multi-channel digital audio over the native gold-plated HDMI port. |
| Legacy Analog & Digital Display | Integrated dual 400 MHz RAMDACs provide pristine analog CRT VGA output alongside digital Dual-Link DVI-I with HDCP compliance. |
5. PassMark VideoCard Benchmark Dossier & Synthetic Performance
PassMark G3D Mark (Composite)
563
Derived from 3,277 submitted test samples (Low margin of error)
PassMark G2D Mark
47
2D UI acceleration, Windows GDI, and font vector rendering
Database Category Rank
#1,801
Ranked across all desktop video cards in PassMark database
PassMark PerformanceTest 3D Graphics Test Suite Breakdown
| Test Suite Routine | Hardware Result Metric | Operational Context & Architectural Note |
|---|---|---|
| DirectX 9 Benchmark | ~48 FPS (Sustained) | Strong legacy DX9 pipeline performance; excellent for vintage era games (Source Engine, Unreal Engine 3) |
| DirectX 10 Benchmark | ~11 FPS (Sustained) | Native Shader Model 4.0 complex geometry and parallax occlusion mapping execution |
| DirectX 11 Benchmark | N/A (Unsupported) | Hardware lacks DirectX 11 tessellation and compute shader support (requires Fermi or newer) |
| DirectX 12 Benchmark | N/A (Unsupported) | Low-level command queue API unavailable on Tesla architecture hardware |
| DirectCompute Benchmark | N/A / Legacy CUDA | DirectCompute 5.0 unsupported; GPGPU compute executed exclusively through CUDA 1.1 runtime |
PassMark Relative Hierarchy: Generational Stack & Museum Peer Comparison
| Graphics Card Model | Silicon Architecture / Bus | PassMark G3D Mark | Relative Variance |
|---|---|---|---|
| GIGABYTE Radeon HD 7770 (Post #2115 #4) | Cape Verde XT (28 nm, PCIe 3.0 x16) | 2,175 | +286.3% |
| GeForce GTS 450 (Immediate Successor) | Fermi GF106 (40 nm, PCIe 2.0 x16) | 1,358 | +141.2% |
| GeForce GTX 260 (GT200 Bigger Sibling) | Tesla 2.0 GT200 (65/55 nm, 448-bit Bus) | 1,206 | +114.2% |
| GeForce 9800 GTX+ (Direct Silicon Predecessor) | Tesla G92b (55 nm, 128 SPs, 256-bit) | 630 | +11.9% |
| Gigabyte GeForce GTS 250 (Exhibit #2975) | Tesla G92b (55 nm, 512MB GDDR3, 256-bit) | 563 | Baseline (0.0%) |
| NVIDIA GeForce GT 120 (Post #2115 #5) | Tesla G96b (55 nm, 32 SPs, 128-bit) | 325 | -42.3% |
| MAC NVIDIA GeForce 7300 GT (Post #2115 #6) | Curie G73 (90 nm, 8 Pixel Pipelines) | 128 | -77.3% |
6. Platform Synergy: Pairing with Dell Precision T7610 Workstation (Exhibit #2973)
Primary Video Adapter Integration
This Gigabyte GeForce GTS 250 512MB graphics card serves as the primary graphics and display adapter inside the Dell Precision T7610 Dual Xeon Workstation cataloged in Exhibit #2973. The pairing offers seamless hardware compatibility across all subsystems:
| Hardware Interface | Gigabyte GeForce GTS 250 (Exhibit #2975) | Dell Precision T7610 Workstation (Exhibit #2973) |
|---|---|---|
| PCIe Bus Slot | PCI Express 2.0 x16 operating interface | Slot 2: PCIe 3.0 x16 mechanical slot wired directly to CPU1 root complex |
| Power Delivery | 150 W TDP (Requires 1 × 6-pin auxiliary PCIe feed) | Dell 1300W 80 PLUS Gold modular PSU with dedicated 6-pin/8-pin graphics power rails |
| Thermal & Acoustics | Gigabyte custom single-fan circular cooler | Precision chassis dual-channel guided front-to-back acoustic airflow baffle |
| Display Connectivity | Native HDMI, Dual-Link DVI-I, and D-Sub (VGA) | Provides immediate multi-standard video output for server console and local workspace monitoring without dongles |
7. Curator’s Assessment & Homelab Applicability
- Museum Hardware Provenance: Mateo’s Museum archive preserves 1 verified physical unit of the Gigabyte GeForce GTS 250 (model
GV-N250-512I). This card is actively deployed as the dedicated graphics adapter in the museum’s enterprise flagship Dell Precision T7610 Workstation. It features Gigabyte’s custom single-fan active cooling solution with a circular radial fin stack and solid copper core. - The Pinnacle of G92 Silicon Refinement: The GTS 250 represents the ultimate 55 nm fabrication shrink of NVIDIA’s legendary G80/G92 architecture. Originally introduced with the 65 nm 8800 GT and 9800 GTX, the G92B revision lowered thermals and die size to 260 mm² while maintaining the complete unharvested 128 stream processor core configuration.
- 256-Bit Memory Bus Advantage: Unlike later budget cards that were choked by narrow 64-bit or 128-bit memory interfaces, the GTS 250 maintains a full 256-bit bus delivering 64.0 GB/s bandwidth. This allows smooth rasterization and responsive high-resolution framebuffering in legacy graphical environments.
- Versatile Homelab Display & Console Role: In a homelab workstation environment, the GV-N250-512I shines as a trouble-free display head. With native HDMI, DVI, and analog VGA ports on the rear bracket, it connects directly to vintage CRT monitors, modern HDMI flat panels, or DVI KVM switches without requiring active DisplayPort adapters.
- Retro Gaming & Legacy CUDA 1.1 Foundation: With full support for DirectX 10, OpenGL 3.3, and CUDA 1.1, the card provides an authentic hardware execution environment for classic PC titles from the late 2000s (Crysis, Half-Life 2: Episode Two, S.T.A.L.K.E.R., Bioshock) and early GPGPU computing experiments.
- Definitive Collection Catalog Placement: Scoring 563 in PassMark G3D Mark, the Gigabyte GeForce GTS 250 enters the museum’s GPU Collection (Exhibit #2115) as official Rank #5, decisively outperforming the GeForce GT 120 (325) while sitting beneath the 28nm GCN Radeon HD 7770 (2,175).