What will be the maximum achievable flop utilization on the next generation of Nvidia server chips?
Plus
35
Ṁ14312025
1%
<30%
1.5%
30-40%
6%
40-50%
9%
50-60%
34%
60-70%
23%
70-80%
11%
80-90%
14%
90-100%
Concretely, what is the best FLOPS (Floating Point Operations per Second) the next generation of Nvidia server cards will be able to achieve on fp16 matrix multiplications on matrices generated by the normal distribution, divided by the maximum theoretical FLOPS that Nvidia reports?
For example, for A100s, it's possible to achieve about 280+ TeraFLOPS out of a maximum of 312 TeraFLOPS, for a maximum flop utilization of ~90%.
On H100s, it seems to be closer around 700 TeraFLOPS, out of a maximum of 1000.
Will resolve when values seem clear after HNext cards are released, or a maximum of one year after Nvidia announces it.
To clarify, this will be for the B100, not the B200.
This question is managed and resolved by Manifold.
Get
1,000
and3.00
Sort by:
Related questions
Related questions
Will a machine learning training run exceed 10^25 FLOP in China before 2025?
77% chance
When will a US government AI run overtake private AI compute by FLOP?
Will Nvidia still retain over 50% market share of PC gamer GPU usage by 2030
59% chance
Will the Groq chip inspire Nvidia/AMD to produce radically new AI chips before 2025?
15% chance
When will Nvidia's GH200 "Grace Hopper" superchip be released.
Will the Groq chip inspire Nvidia/AMD to produce radically new AI chips before 2026?
45% chance
Will Nvidia report gross margins lower than 60% in any year by 2028?
68% chance
If China invades Taiwan in 2023-2030, what will FLOP/s per dollar of top-ML GPUs be 5 years later?
If China invades Taiwan in 2023-2030, what will FLOP/s per dollar of top-ML GPUs be 10 years later?
Will AI accelerators improve in FLOPs/watt by 100x of an NVidia H100 by 2033?
90% chance