Nvidia has hit out at Intel claiming that a benchmark test it performed to beat Nvida’s Tesla GPUs at inference workloads was “misleading”.
The company claims that the test required up to $100,000 of energy-hungry hardware in order to best a Nvidia Tesla V100 GPU.
“It’s not every day that one of the world’s leading tech companies highlights the benefits of your products,” wrote Paresh Kharya, director of product marketing at Nvidia, in a blog posting hitting back at Intel.
It’s not every day that one of the world’s leading tech companies highlights the benefits of your products
He continued: “To achieve the performance of a single mainstream Nvidia V100 GPU, Intel combined two power-hungry, highest-end CPUs with an estimated price of $50,000-$100,000, according to Anandtech. Intel’s performance comparison also highlighted the clear advantage of Nvidia T4 GPUs, which are built for inference.
“When compared to a single highest-end CPU, they’re not only faster but also seven-times more energy-efficient and an order of magnitude more cost-efficient.”
Inference, also known as ‘prediction’, “is the ‘pattern recognition’ that a neural network does after being trained. It’s where AI models provide intelligent capabilities in applications, like detecting fraud in financial transactions”, explained Kharya.
AI & Machine Learning Live is returning to London on 3rd July 2019. Hear from the Met Office’s Charles Ewen, AutoTrader lead data scientist Dr David Hoyle and the BBC’s Noriko Matsuoka, among many others. Attendance is free to qualifying IT leaders and senior IT pros, but places are limited, so reserve yours now.
Most AI inference conducted today is done on CPUs, but this is a market that is being targeted by Nvidia, hawking the mathematical capabilities of its GPUs. It also partly explains Intel’s interest in driving into graphics card technology.
Kharya went on to breakdown the Intel benchmark test, putting the Tesla V100 and Turing-based T4 GPUs again a dual-socket Xeon 9282. While Intel wins in terms of raw performance in the ResNet-50 test it takes a lot more power – and therefore cost – to achieve its victory.
The efficiency of the GPUs over the CPUs should be expected, given that GPUs are specialised chips more geared up for handling parallel processing, which has also made them ideal for AI applications and inference, or prediction, in particular.
It’s also worth noting that Intel never positioned its Xeon chips as ideal for inference workloads, unlike Nvidia’s Tesla and T4 GPUs. Rather the Xeon CPUs are general purpose processors for data centre use that can also do a good job at inference.
However, Kharya did admit that “Intel’s latest Cascade Lake CPUs include new instructions that improve inference, making them the best CPUs for inference”.
Delta is a new market intelligence service from Computing to help CIOs and other IT decision makers make smarter purchasing decisions – decisions informed by the knowledge and experience of other CIOs and IT decision makers.
Delta is free from vendor sponsorship or influence of any kind, and is guided by a steering committee of well-known CIOs, such as Charles Ewen, Christina Scott, Steve Capper and Laura Meyer.
Ten crucial technology areas are already covered at launch, with more data appearing and more areas being covered every week. Sign-up here for your free trial of the Computing Delta website.