Consumercarts

Consumercarts is reader-supported. Learn more

NVIDIA L40S vs NVIDIA RTX PRO 5000 72GB Blackwell: Comparison

NVIDIA L40S and NVIDIA RTX PRO 5000 72GB Blackwell separate most on memory bandwidth: NVIDIA L40S at 864 GB/s, NVIDIA RTX PRO 5000 72GB Blackwell at 1344 GB/s, 36% in favour of NVIDIA RTX PRO 5000 72GB Blackwell. VRAM splits them by another 33%, again favouring NVIDIA RTX PRO 5000 72GB Blackwell. They match on 7 other specifications, including gpu brand, tier and ray tracing support.

Specs

Specification
NVIDIA L40S
NVIDIA L40S
NVIDIA RTX PRO 5000 72GB Blackwell
NVIDIA RTX PRO 5000 72GB Blackwell
GPU BrandNVIDIANVIDIA
SeriesRTX Ada GenerationRTX PRO Blackwell
Board PartnerNVIDIA Reference (data center/OEM only, passive cooling)NVIDIA Reference (workstation, no AIB partners)
VRAM48 GB-33%72 GB
TierFlagshipFlagship
Memory TypeGDDR6GDDR7
Boost Clock2520 MHz2410 MHz-4%
Ray Tracing Supporttruetrue
TDP350 W-17%300 W
ArchitectureAda LovelaceBlackwell
Process NodeTSMC 4NTSMC 4N
Card Length267 mm267 mm
Slot Width22
Memory Bus Width384 bit384 bit
Memory Bandwidth864 GB/s-36%1344 GB/s
Recommended PSU600 W-9%550 W
Power Connector1x 16-pin (12VHPWR)1x 12V-2x6
MSRP85006999
Launch Date2023-082025-09
Base / Game Clock735 MHz1850 MHz
PCIe InterfacePCIe 4.0 x16PCIe 5.0 x16
CUDA Cores1817614080
Display Outputs4x DisplayPort 1.4a (rarely used in server deployment)4x DisplayPort 2.1b
Media Engine3x NVENC (8th gen), 3x NVDEC, AV1 encode/decode4x NVENC (9th gen), 4x NVDEC, AV1 encode/decode

Verdict

NVIDIA L40S

8.9 / 10 Expert Rating

One of the most widely deployed data center GPUs of its generation - a genuinely strong middle ground between pure graphics cards and dedicated AI accelerators.

NVIDIA RTX PRO 5000 72GB Blackwell

8.8 / 10 Expert Rating

A smart memory-capacity-focused variant - genuinely useful for teams whose bottleneck is VRAM rather than compute, at meaningfully lower power draw than the flagship.

More comparisons in Graphics Cards