GPU Comparison

The NVIDIA Tesla M60 and Intel Arc A370M are separated by seven years of GPU architecture, yet their average benchmark scores land within 4.5% of each other. The Tesla M60 posts an average score of 30,490, while the Arc A370M averages 29,175. This places both in the mid-70th percentile of all GPUs, with the Tesla at 75% and the Arc at 74%. The head-to-head data shows a split decision: the Arc wins one test, the Tesla wins the other, and the margin of victory tells a more complex story than the overall average suggests.

Head-to-Head Benchmarks

The two GPUs trade wins in the Geekbench suite, but the magnitude of each victory differs sharply. In the OpenCL workload, the Intel Arc A370M edges out the Tesla M60 with a score of 29,676 against 29,506, a margin of just 0.6%. This is a near-photo-finish result, indicating that raw compute throughput in this API is effectively identical between the two. The Arc’s newer architecture and higher clock speeds offset the Tesla’s larger memory bus and older Maxwell design.

The Vulkan test tells a different story. The NVIDIA Tesla M60 wins decisively with a score of 31,473, beating the Arc A370M’s 28,673 by 9.8%. That is a substantial gap, roughly ten times larger than the OpenCL delta. The result suggests the Tesla’s Maxwell architecture handles Vulkan’s driver overhead and draw-call patterns more efficiently, or that the Arc’s Xe-HPG implementation struggles with this particular workload. For Vulkan-based applications, the Tesla M60 is the clear choice.

Looking at the nearest rivals for context, the Tesla M60’s average score of 30,490 puts it essentially tied with the NVIDIA CMP 70HX (30,476, 0% delta) and just 0.2% ahead of the AMD Radeon RX 6700 (30,433). The Arc A370M’s 29,175 average is similarly clustered, sitting 0.1% below the AMD Radeon RX Vega M GH (29,197) and 0.6% above the AMD Radeon RX 470 (28,996). Both GPUs occupy a performance tier where rival products are separated by fractions of a percent, meaning real-world differences will depend heavily on the specific application and API.

Where Each One Wins

The NVIDIA Tesla M60 is the pick for Vulkan-heavy workloads. Its 9.8% lead in Geekbench Vulkan is the largest performance gap between the two cards in any test. This advantage likely stems from the Tesla’s 256-bit memory bus and 160.4 GB/s of bandwidth, which provides more headroom for the memory-intensive operations common in Vulkan rendering. The Tesla also holds a marginal edge in pixel rate (75.39 GPixel/s vs. 65.60 GPixel/s) and texture rate (150.8 GTexel/s vs. 131.2 GTexel/s), which can help in fill-rate-bound scenarios.

The Intel Arc A370M wins in OpenCL, albeit narrowly, and offers capabilities the Tesla simply lacks. The Arc includes 8 ray tracing cores, enabling hardware-accelerated ray tracing that the Maxwell-based Tesla cannot perform. It also supports DirectX 12 Ultimate (12_2), while the Tesla is limited to DirectX 12 (12_1). For applications that leverage ray tracing or require the latest DirectX feature set, the Arc A370M is the only option between the two. The Arc also consumes dramatically less power, 35 W versus 300 W, making it feasible for portable devices, while the Tesla requires a dual-slot cooler and an 8-pin power connector.

In floating-point compute, the two are close: the Tesla delivers 4.825 TFLOPS FP32, while the Arc delivers 4.198 TFLOPS. However, the Arc adds FP16 support at 8.397 TFLOPS (2:1 rate), which the Tesla does not provide at all. For mixed-precision workloads, the Arc has a distinct advantage. The Tesla counters with a higher memory bandwidth (160.4 GB/s vs. 112.0 GB/s) and double the memory capacity (8 GB vs. 4 GB), which matters for large datasets that exceed the Arc’s smaller frame buffer.

Architecture Differences

The architectural gap between these two GPUs spans a full generation shift in process technology and design philosophy. The Tesla M60 uses NVIDIA’s GM204 chip built on Maxwell 2.0 architecture, fabricated on a 28 nm process at TSMC. The die measures 398 mm² and contains 5,200 million transistors, yielding a transistor density of 13.1 million per mm². The Arc A370M uses Intel’s DG2-128 chip on Xe-HPG architecture, fabricated on a 6 nm process, also at TSMC. This die is far smaller at 157 mm² but packs 7,200 million transistors, achieving a density of 45.9 million per mm², more than three times the transistor density of the Tesla.

The core configurations reflect these differences. The Tesla M60 has 2,048 shading units, 128 TMUs, and 64 ROPs, with no ray tracing or tensor cores. The Arc A370M halves the shading units to 1,024 but adds 8 ray tracing cores, while keeping 64 TMUs and 32 ROPs. Clock speeds favor the Arc: its base clock of 1550 MHz and boost of 2050 MHz dwarf the Tesla’s 557 MHz base and 1178 MHz boost. Memory configurations also diverge sharply. The Tesla uses 8 GB of GDDR5 on a 256-bit bus at 1253 MHz (5 Gbps effective), while the Arc uses 4 GB of GDDR6 on a 64-bit bus at 1750 MHz (14 Gbps effective). The Tesla’s wider bus gives it higher bandwidth, but the Arc’s faster memory clock partially compensates.

The Tesla M60 is a dual-slot, 300 W card with a PCIe 3.0 x16 interface and no display outputs, it is a compute accelerator. The Arc A370M is an integrated GPU (IGP) with a 35 W TDP, PCIe 4.0 x8 interface, and display outputs that are portable-device dependent. The Tesla supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. The Arc supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The Tesla was released in August 2015, succeeding Tesla Kepler and preceding Tesla Pascal. The Arc arrived in March 2022 as part of the Alchemist generation, with no direct predecessor or successor in the data.

FAQ

Q: Which GPU has a higher average benchmark score?

A: The NVIDIA Tesla M60 averages 30,490, which is 4.5% higher than the Intel Arc A370M’s 29,175.

Q: How do the two GPUs compare in Vulkan performance?

A: The Tesla M60 wins Geekbench Vulkan with a score of 31,473, beating the Arc A370M’s 28,673 by 9.8%.

Q: Does the Intel Arc A370M support ray tracing?

A: Yes, the Arc A370M includes 8 ray tracing cores, while the Tesla M60 has none.

Q: What is the memory capacity difference?

A: The Tesla M60 has 8 GB of GDDR5 on a 256-bit bus, while the Arc A370M has 4 GB of GDDR6 on a 64-bit bus.

Q: Which GPU has a higher transistor density?

A: The Intel Arc A370M has a density of 45.9 million transistors per mm², compared to the Tesla M60’s 13.1 million per mm².

Q: Can the Tesla M60 be used for display output?

A: No, the Tesla M60 has no display outputs; the Arc A370M’s outputs are portable-device dependent.

Specification Differences

The two GPUs differ in nearly every measurable specification. The Tesla M60 uses the GM204 chip on Maxwell 2.0 architecture, while the Arc A370M uses DG2-128 on Xe-HPG. The process node shifts from 28 nm (Tesla) to 6 nm (Arc). Transistor counts are 5,200 million (Tesla) versus 7,200 million (Arc), with die sizes of 398 mm² and 157 mm², respectively. Transistor density is 13.1M/mm² for the Tesla and 45.9M/mm² for the Arc.

Clock speeds favor the Arc: base clock is 557 MHz (Tesla) versus 1550 MHz (Arc), and boost clock is 1178 MHz (Tesla) versus 2050 MHz (Arc). Memory clocks are 1253 MHz (5 Gbps effective) for the Tesla and 1750 MHz (14 Gbps effective) for the Arc. Memory capacity is 8 GB GDDR5 for the Tesla and 4 GB GDDR6 for the Arc. Bus width is 256-bit versus 64-bit, and bandwidth is 160.4 GB/s versus 112.0 GB/s.

Shading units are 2048 (Tesla) versus 1024 (Arc), TMUs are 128 versus 64, and ROPs are 64 versus 32. The Arc has 8 ray tracing cores; the Tesla has none. Pixel rate is 75.39 GPixel/s (Tesla) versus 65.60 GPixel/s (Arc), and texture rate is 150.8 GTexel/s versus 131.2 GTexel/s. FP32 performance is 4.825 TFLOPS versus 4.198 TFLOPS, with the Arc adding 8.397 TFLOPS FP16. TDP is 300 W versus 35 W. The Tesla is dual-slot with a 1x 8-pin connector and 700 W suggested PSU; the Arc is an IGP with no power connector or PSU requirement. Bus interface is PCIe 3.0 x16 versus PCIe 4.0 x8. The Tesla has no display outputs; the Arc’s are portable-device dependent. DirectX support is 12 (12_1) versus 12 Ultimate (12_2). Release dates are August 2015 versus March 2022.