NVIDIA Quadro FX 370 LP

# NVIDIA Quadro FX 370 LP

The NVIDIA Quadro FX 370 LP is an end-of-life professional graphics card built on the Tesla architecture, featuring the G98S chip fabricated at UMC on a 65 nm process. With 210 million transistors on an 86 mm² die, this single-slot card targets legacy workstation environments where basic 2D/3D visualization is sufficient. The data shows a 50th percentile ranking among all GPUs, placing it squarely in the middle of the performance spectrum—though that percentile reflects its era, not modern workloads. This card carries a launch MSRP of 129 USD.

Who Should Consider It

The Quadro FX 370 LP is not designed for modern gaming or compute-intensive tasks. Its 256 MB DDR2 memory and 64-bit bus limit its utility to low-resolution desktop environments, legacy CAD applications, and multi-monitor setups that require minimal graphical horsepower. Benchmark results indicate that users working with 2D schematics, spreadsheet-heavy workflows, or basic office productivity at resolutions at or below 1080p could find this card sufficient—provided the software does not demand advanced shader models. The card supports DirectX 11.1 (10_0) and OpenGL 3.3, which means it can run older professional applications but will struggle with titles or tools requiring DirectX 11 features beyond the 10_0 feature level.

For 3D modeling or rendering at higher resolutions, the data suggests this card is inadequate. The pixel rate of 2.160 GPixel/s and texture rate of 2.160 GTexel/s indicate that filling high-resolution framebuffers will be slow, especially when compared to even entry-level modern GPUs. Users considering this card should limit expectations to 1280x1024 or lower resolutions for any interactive 3D work. It is best suited for secondary display output, remote desktop environments, or as a replacement part in legacy systems where the original GPU has failed.

Power and Cooling

The Quadro FX 370 LP draws a maximum TDP of 25 W, making it an exceptionally low-power solution. The card requires no auxiliary power connectors, drawing all its power from the PCIe 1.0 x16 slot. A suggested PSU rating of 200 W is listed in the specifications, which is remarkably modest—even older office power supplies can typically support this card without concern. The single-slot design, measuring 198 mm (7.8 inches) in length and 69 mm (2.7 inches) in height, ensures compatibility with most low-profile chassis, though the DMS-59 output requires an adapter for standard dual-monitor setups.

Thermal management is straightforward given the low power draw. The 25 W TDP means that even a basic passive or low-speed fan heatsink can maintain acceptable temperatures. However, the card's end-of-life status means replacement cooling solutions may be difficult to source. The absence of power connectors simplifies installation, but users should verify that their system's PCIe slot provides adequate clearance and that the chassis airflow is sufficient for the card's modest heat output.

Benchmark Performance

The FACT PACK lists no benchmark scores for this card, and the average benchmark score is 0, indicating that no standardized performance data exists in the database. The percentileVsAllGpus value of 50 places it at the median of all GPUs ever tracked—but this is a historical artifact, as modern GPUs have dramatically shifted the performance curve. The FP32 compute capability is rated at 17.28 GFLOPS, which is minuscule by modern standards. For context, this means the card can perform roughly 17 billion floating-point operations per second, a figure that contemporary integrated graphics surpass by orders of magnitude.

Without nearest rivals listed in the data, direct percentage comparisons are impossible. However, the specification sheet tells a clear story: 8 shading units, 4 TMUs, and 4 ROPs working at a memory clock of 500 MHz (1000 Mbps effective) produce 8.000 GB/s of bandwidth. These figures place the Quadro FX 370 LP in the same performance class as early 2000s mainstream GPUs. The 50th percentile ranking is likely skewed because the database includes many older, equally weak cards. In practical terms, this card will bottleneck any modern 3D application, and its DirectX 11.1 (10_0) support means it cannot execute shader models beyond the 10_0 profile.

FAQ

Q: Can this card run modern operating systems' desktop environments?

A: Yes, the card supports DirectX 11.1 (10_0) and OpenGL 3.3, which are sufficient for 2D desktop rendering in Windows 7 through 10, though driver support may be limited given its end-of-life status.

Q: What is the maximum memory bandwidth available?

A: The 64-bit DDR2 bus at 500 MHz (1000 Mbps effective) provides 8.000 GB/s of bandwidth, which is adequate for 2D workloads but severely limits texture-heavy 3D applications.

Q: Does this card support hardware ray tracing?

A: No, the Tesla architecture includes no RT cores or tensor cores. The card is limited to traditional rasterization with 8 shading units.

Q: What power supply is recommended?

A: A 200 W PSU is suggested, and the card itself draws only 25 W with no auxiliary power connectors required.

Q: How many displays can this card drive?

A: The single DMS-59 output can be split via an adapter to drive two monitors, though the exact resolution limits depend on the adapter and driver support.

Q: Is this card suitable for modern 3D games?

A: No, the 256 MB VRAM, 17.28 GFLOPS FP32 throughput, and DirectX 11.1 (10_0) support make it incapable of running most modern titles even at minimum settings.

Ray Tracing and Feature Set

The Quadro FX 370 LP has no ray tracing or tensor cores—these features did not exist in the Tesla architecture. Its feature set is defined by the DirectX 11.1 (10_0) API support, which means it can technically report DirectX 11 compatibility but is limited to the 10_0 feature level. OpenGL 3.3 support is present, which allows compatibility with legacy professional applications that rely on this API. Vulkan is not supported, and no RT or tensor core counts are listed in the specifications.

For professional users, the lack of dedicated ray tracing hardware means that any modern rendering workloads requiring RTX-accelerated ray tracing are out of the question. The card relies entirely on its 8 shading units for all compute and graphics tasks, which explains the 17.28 GFLOPS FP32 rating. This is a pure rasterization card from an era before hardware-accelerated ray tracing became a standard feature. The absence of tensor cores also precludes any AI-accelerated workloads, such as DLSS or neural network inference.

Memory Subsystem

The memory configuration consists of 256 MB of DDR2 memory on a 64-bit bus, running at 500 MHz (1000 Mbps effective). This yields a bandwidth of 8.000 GB/s. For comparison, this is roughly one-fiftieth of what a modern mid-range GPU offers. The small frame buffer is a critical limitation: at 1080p, many textures and geometry buffers will exceed 256 MB, causing the card to spill to system memory over the PCIe 1.0 x16 interface, which itself is an older standard with limited bandwidth.

At higher resolutions, the memory subsystem becomes the primary bottleneck. The 64-bit bus width means that memory access patterns are inefficient, and the 8.000 GB/s bandwidth will saturate quickly in any 3D scene with moderate texture detail. For 2D desktop work, 256 MB is sufficient, but for any 3D rendering at resolutions above 1280x1024, the memory constraints will cause noticeable stuttering and texture pop-in. The DDR2 type also has higher latency than later memory generations, compounding the performance limitations.

How It Compares

The FACT PACK lists no nearest rivals for this card, which means direct percentage comparisons against specific competitors are not available from the data. However, the 50th percentile ranking provides a general reference point. In the broader context of the database, this card sits in the middle of all GPUs tracked—but that is a misleading statistic because the database includes many low-end cards from the same era. The predecessor to this card is the Quadro FX Curie series, and the successor is the Quadro Fermi line, both of which are also end-of-life products.

Without rival data, the comparison must rely on architectural lineage. The Tesla architecture, 65 nm process, and 2.4M transistors per mm² density place this card in the mid-2000s era of GPU design. Modern GPUs using newer architectures, larger dies, and more advanced memory technologies will outperform this card by a wide margin, but the data does not provide specific percentages. The card's 25 W TDP and 200 W suggested PSU make it one of the lowest-power discrete GPUs in the database, which is its primary distinguishing feature. For users needing a minimal-power legacy workstation card, this remains an option, but for any performance-critical workload, the data clearly indicates this card is obsolete.