NVIDIA H100 vs Intel Gaudi 3
Intel Gaudi 3 comes out ahead of NVIDIA H100 on 4 of the 7 things that decide this, going by what each maker publishes. The one that matters for you may be one the leader loses.
The verdict
Intel Gaudi 3 takes it, 4–3 on rounds.
On the published figures, Intel Gaudi 3 takes 4 of 7 rounds. Treat that as a starting point and check the current figures on both pages, because prices and plans move.
Choose NVIDIA H100 if…
- • You want NVIDIA H100's edge on how fast it runs fp8 ai math and how much power it draws
Choose Intel Gaudi 3 if…
- • You want Intel Gaudi 3's edge on how much memory it has and how fast its memory is
Round by round
- Round 1How much memory it hasIntel Gaudi 3 wins
Intel Gaudi 3 wins this round on what the company itself publishes.
- Round 2How fast its memory isIntel Gaudi 3 wins
Intel Gaudi 3 wins this round on what the company itself publishes.
- Round 3How fast it runs FP8 AI mathNVIDIA H100 winsIntel Gaudi 35/10
1,678 TFLOPS
fp8 mme (matrix) tflops
- 1,678 TFLOPS, against 1,979 TFLOPS
Source: cdrdv2-public.intel.com ↗
NVIDIA H100 wins this round on what the company itself publishes.
- Round 4How fast it runs FP16 AI mathIntel Gaudi 3 winsNVIDIA H1005/10
989 TFLOPS
fp16 tensor core
- 989 TFLOPS, against 1,678 TFLOPS
Source: dam-cdn.nvd.orangelogic.com ↗
Intel Gaudi 38/10459 TFLOPS
FP16 (signed) matrix (MME) peak, Gaudi 3 technical paper Table 5; 1,678 is the BF16/FP8 figure
- bf16 mme (matrix) tflops
Source: cdrdv2-public.intel.com ↗
Intel Gaudi 3 wins this round on what the company itself publishes.
- Round 5How much power it drawsNVIDIA H100 wins
NVIDIA H100 wins this round on what the company itself publishes.
- Round 6Which memory generation it usesNVIDIA H100 wins
NVIDIA H100 wins this round on what the company itself publishes.
- Round 7How new it isIntel Gaudi 3 wins
Intel Gaudi 3 wins this round on what the company itself publishes.
Side by side
- bf16 mme (matrix) tflops
- 80 GB HBM3, against 288 GB HBM3E
- 3.35 TB/s, against 8 TB/s
- 1,979 TFLOPS, against 5 PFLOPs
- 989 TFLOPS, against 2.5 PFLOPs
- 128 GB HBM2E, against 192 GB HBM
- 3.7 TB/s, against 7.37 TB/s
- 1,678 TFLOPS, against 4,614 TFLOPs
- 1,678 TFLOPS, against 2,307 TFLOPs