NVIDIA P106-090 vs NVIDIA RTX PRO 6000 Blackwell Max-Q Comparison

The NVIDIA P106-090 and the NVIDIA RTX PRO 6000 Blackwell Max-Q are two GPUs that could hardly be further apart in intent, and the recorded data makes the gap unmistakable. One is an end-of-life mining card from the Pascal era with no display outputs and a PCIe 1.0 x1 link; the other is an active workstation flagship built on Blackwell 2.0 silicon with 96 GB of GDDR7 and a full PCIe 5.0 x16 interface. The database contains exactly one shared benchmark between them, and it is a landslide: the RTX PRO 6000 Blackwell Max-Q wins it, 11088 to 509. That margin, a 95.4 percent gap in favor of the Blackwell card, is among the widest the database records, and it frames everything that follows.

Head-to-Head Benchmarks

The only test in which both cards appear is 3DMark Steel Nomad (DirectX 12). The RTX PRO 6000 Blackwell Max-Q posts a score of 11088; the P106-090 manages 509. The Blackwell card is therefore more than twenty times faster on this workload, and the database records the P106-090 as trailing by 95.4 percent. There is no nuance to extract here and no scenario in the data where the mining card recovers ground. The P106-090 simply has no answer to a GPU with 24064 shading units against its 768, and 109.7 TFLOPS of FP32 throughput against its 2.352 TFLOPS.

The P106-090 does carry two additional benchmark entries that the Blackwell card lacks: a Geekbench OpenCL score of 21304 and a Geekbench Vulkan score of 18596. These cannot be compared directly to the RTX PRO 6000, since the two cards share no results in those tests, but they do place the P106-090 in context on its own merits. Its average benchmark score across all recorded tests is 13470, and its nearest rivals sit almost exactly on top of it: the GeForce GTX 570 averages 13515 (the P106-090 trails by 0.3 percent), the Radeon Pro 555 averages 13407 (the P106-090 leads by 0.5 percent), the Radeon HD 7770M averages 13536 (a 0.5 percent deficit), and the Radeon RX 9070 XT entry in its rival list averages 13543 (again a 0.5 percent deficit). In short, the P106-090 performs at the level of much older mainstream hardware, dead center of a cluster it neither escapes nor falls behind.

The RTX PRO 6000 Blackwell Max-Q has an average benchmark score of 11088 from its single Steel Nomad result. Its nearest rivals include the RTX PRO 6000D Blackwell Max-Q at an identical 11088 (a dead-even tie), the Radeon RX 550 at 11075 (the Blackwell card ahead by 0.1 percent), the GeForce GTX 1650 SUPER at 11047 (ahead by 0.4 percent), and the FirePro W4300 at 11225 (trailing by 1.2 percent). Every head-to-head in the database goes to the RTX PRO 6000: one win to the Blackwell card, zero to the P106-090.

Architecture Differences

The architectural gulf begins at the foundry. Both chips are TSMC products, but the P106-090's GP106 die is built on a 16 nm process while the RTX PRO 6000's GB202 die uses a 5 nm process. The consequences are stark. The GP106 packs 4,400 million transistors into a 200 mm² die, for a density of 22.0 million transistors per square millimeter. The GB202 holds 92,200 million transistors across 750 mm², a density of 122.9 million per square millimeter. That is more than twenty times the transistor count on a die less than four times larger, and roughly five and a half times the density.

The compute resources scale accordingly. The P106-090 has 768 shading units, 48 TMUs, and 48 ROPs, with no RT cores and no tensor cores. The RTX PRO 6000 Blackwell Max-Q has 24064 shading units, 752 TMUs, 192 ROPs, 188 RT cores, and 752 tensor cores. The presence of dedicated ray tracing and AI hardware on the Blackwell part, and its complete absence on the Pascal-era mining card, is a categorical difference that no clock speed can bridge.

Clocks actually run lower on the newer card at base: 1035 MHz for the RTX PRO 6000 versus 1354 MHz for the P106-090. But the Blackwell boost clock reaches 2280 MHz against the Pascal card's 1531 MHz. Half-precision throughput is another decisive contrast. The P106-090 manages FP16 at just 36.74 GFLOPS, running at a 1:64 ratio to FP32, while the RTX PRO 6000 delivers full-rate FP16 at 109.7 TFLOPS with a 1:1 ratio. Output rates follow the same pattern: 73.49 GPixel/s and 73.49 GTexel/s for the P106-090 against 437.8 GPixel/s and 1,714.6 GTexel/s for the RTX PRO 6000.

Memory is perhaps the widest gap of all. The P106-090 ships 3 GB of GDDR5 on a 192-bit bus with 192.2 GB/s of bandwidth, clocked at 2002 MHz for an effective 8 Gbps. The RTX PRO 6000 Blackwell Max-Q carries 96 GB of GDDR7 on a 512-bit bus with 1.79 TB/s of bandwidth, running at 1750 MHz for an effective 28 Gbps. That is thirty-two times the capacity and more than nine times the bandwidth.

Platform and feature support diverge as well. The P106-090 connects over PCIe 1.0 x1, a single-lane link that would cripple any modern workload, and it has no display outputs at all, consistent with its mining origin. The RTX PRO 6000 uses PCIe 5.0 x16 and provides four DisplayPort 2.1b outputs. On the API side, both support OpenGL 4.6 and Vulkan 1.4, but the Blackwell card supports DirectX 12 Ultimate (12_2) while the P106-090 stops at DirectX 12 (12_1). The RTX PRO 6000 succeeds the Workstation Ada line and remains in active production; the P106-090 is end-of-life, released on 2017-07-30, while the Blackwell card arrived on 2025-03-17. Its launch MSRP was 8,565 USD.

FAQ

Q: Is there any benchmark where the P106-090 beats the RTX PRO 6000 Blackwell Max-Q?

A: No. Of the tests where both cards have recorded results, the Blackwell card wins the only shared benchmark, Steel Nomad, by 11088 to 509. The overall head-to-head record is one win for the RTX PRO 6000 and zero for the P106-090.

Q: How much faster is the RTX PRO 6000 in raw FP32 compute?

A: The RTX PRO 6000 delivers 109.7 TFLOPS of FP32 against 2.352 TFLOPS for the P106-090. The Blackwell card also runs FP16 at full rate (1:1, 109.7 TFLOPS), while the P106-090's FP16 runs at a 1:64 ratio and manages just 36.74 GFLOPS.

Q: Which card has ray tracing and AI acceleration?

A: Only the RTX PRO 6000 Blackwell Max-Q. It includes 188 RT cores and 752 tensor cores. The P106-090 has neither, which excludes it entirely from ray tracing and modern AI workloads in the data.

Q: Can the P106-090 drive a monitor?

A: No. The database lists no display outputs for the P106-090, reflecting its design as a mining GPU. The RTX PRO 6000 offers four DisplayPort 2.1b connections.

Q: How do the two cards compare in power and physical requirements?

A: The P106-090 has a 75 W TDP, one 6-pin connector, a dual-slot design, and a 250 W suggested PSU. The RTX PRO 6000 Blackwell Max-Q has a 300 W TDP, one 16-pin connector, a dual-slot design, and a 700 W suggested PSU. Both occupy two slots, though the Blackwell card is longer at 267 mm versus 250 mm.

Q: Which DirectX level does each card support?

A: The P106-090 supports DirectX 12 with feature level 12_1. The RTX PRO 6000 Blackwell Max-Q supports DirectX 12 Ultimate with feature level 12_2.

Specification Differences

The two cards differ on nearly every recorded field. Process node: 16 nm versus 5 nm. Transistors: 4,400 million versus 92,200 million. Die size: 200 mm² versus 750 mm². Density: 22.0M versus 122.9M transistors per mm². Shading units: 768 versus 24064. TMUs: 48 versus 752. ROPs: 48 versus 192. Memory: 3 GB GDDR5 on a 192-bit bus at 192.2 GB/s versus 96 GB GDDR7 on a 512-bit bus at 1.79 TB/s. Base clock: 1354 MHz versus 1035 MHz. Boost clock: 1531 MHz versus 2280 MHz. Bus interface: PCIe 1.0 x1 versus PCIe 5.0 x16. TDP: 75 W versus 300 W. Power connector: 1x 6-pin versus 1x 16-pin. Suggested PSU: 250 W versus 700 W. Length: 250 mm versus 267 mm, with the Blackwell card also recording a height of 111 mm and a width of 40 mm. Production status: end-of-life versus active. The P106-090 offers no display outputs; the RTX PRO 6000 provides four DisplayPort 2.1b ports.

The Verdict

For any workload represented in this database, the RTX PRO 6000 Blackwell Max-Q is the only rational choice. It wins the sole shared benchmark by a 95.4 percent margin, holds a percentile ranking of 50 against all GPUs versus 54 for the P106-090, and delivers orders of magnitude more compute, memory, and bandwidth. Workstation users needing large-memory professional workloads have exactly one option here: the 96 GB Blackwell card with its 188 RT cores and 752 tensor cores. The P106-090, with its single PCIe lane, absent display outputs, end-of-life status, and performance clustered alongside the GeForce GTX 570 and Radeon Pro 555, is not a contender in any recorded comparison. Its remaining relevance is historical, as a snapshot of Pascal-era mining hardware.

Where Each One Wins

The RTX PRO 6000 Blackwell Max-Q wins everywhere the data reaches: DirectX 12 Ultimate rendering at 11088 in Steel Nomad, ray tracing via its 188 RT cores, AI and half-precision compute via 752 tensor cores running FP16 at a 1:1 ratio, memory-heavy workloads with 96 GB of GDDR7 at 1.79 TB/s, and multi-monitor setups through four DisplayPort 2.1b outputs. The P106-090 wins nothing in the benchmark data. Its only quantitative edges are situational: a lower 75 W TDP against 300 W, a single 6-pin connector against a 16-pin, a 250 W suggested PSU against 700 W, and a slightly shorter 250 mm board against 267 mm. For constrained power envelopes, that is the full extent of its advantage, and it comes at the cost of 95.4 percent less performance in the one test where both cards have been measured.