Habana Labs- AI Chip 3x faster than Nvidia

I am a investor in Nvidia, so have been closely following this type of stuff. How it will affect Nvidia remains unknown but as soon as I see Nvidia losing momentum in AI, I might start selling Nvidia down.

I think it might take a few years before Nvidia sales to Hyperscalers, car companies, etc starts getting affected by companies like Habana Labs. There is another company that Chamath Palihapitiya https://en.wikipedia.org/wiki/Chamath_Palihapitiya invested in called Groq https://news.crunchbase.com/news/groq-a-stealthy-startup-fou… that’s also in stealth mode that will also probably come up with something similar or better within the next several years.

By Tal Shahaf, en.globes.co.il:

Israeli company Habana Labs claims to have developed a processor three times faster than that of its rival Nvidia.

Israeli startup Habana Labs has come out of stealth and presented what it claims is the world’s fastest artificial intelligence (AI) chip at a conference in the US. Similar technologies to Habana’s Goya chip have been announced by other companies, including Israeli startups, but most of them are still in the development stage.

cont’d

https://en.globes.co.il/en/article-israeli-co-habana-labs-un…

The above article mentions that the founder of Habana Labs, Avigdor Willenz, also sold chipmaker Annapurna Labs to Amazon in 2015, so this means that Willenz probably still has contacts with Amazon.

The company has developed one chip for inference called Goya and plans another chip for AI training in 2019 which gets mentioned in this article by Michael Feldman, top500.org:

Habana is currently sampling Goya for select customers. It plans to do the same for Gaudi, its first AI training processor, in the second quarter of 2019. For this chip, the company is promising that training performance will “scale linearly to thousands of processors.” Stay tuned.

cont’d

https://www.top500.org/news/ai-startup-sets-high-water-mark-…

Starrob

14 Likes

I think it might take a few years before Nvidia sales to Hyperscalers, car companies, etc starts getting affected by companies like Habana Labs.

“Many are called but few are chosen.” That’s how “one takes most” businesses work. Nvidia is not forever but for long enough.

Denny Schlesinger

3 Likes

There are lots of people gunning for both the inference and the training space, trying to topple the GPU. An ex-colleague of mine from ARM went to work for a start-up doing this, Tenstorrent. A few of the old SPARC people from Oracle started SambaNova Systems after that division was scrapped. And ARM (my current employer) is working on an AI/ML acclerator for inference in IoT devices. This is a crowded space because the market is going to be huge and the rewards will be high. This is why I stay invested in NVDA, and why I think, once the SaaS gas is gone the next big growth area is going to be IoT. And that IoT is going to be AI dependent.

It will take years to possibly topple the GPU. And its entirely possible NVDA is smart enough to not put all the future eggs in the GPU basket anyway. Architecturally there are great reasons why the GPU is king over the CPU in this space. But there are also some issues with the GPU that are making people look for a new architectural alternative to both.

Bill Jurasz
CPU development for eons (TI, Motorola/PowerPC, AMD, Oracle/SPARC, ARM) and a good dose of GPU (Apple mobile GPU development)

22 Likes

Chinese firms Alibaba or Huawei may ultimately post bigger threat, as trade war going, Chinese starts to have urgency in this area.

1 Like

Chinese firms Alibaba or Huawei may ultimately post bigger threat, as trade war going, Chinese starts to have urgency in this area.

I’m not sure but the chips being made in Taiwan are not from the US so trade war does not apply. I read something to that effect.

Denny Schlesinger

1 Like

Warning!

I have followed Macro Economics for 15 years.

Macro Economics is an excellent way to make a small fortune. . . out of a large one.

While there are some immediate threats, real estate in target zones, i.e. Iowa, the longer term threats are simply not calculable.

Ben Hunt has an excellent article titled “Three Body Problem” where he discusses a math certainty. It is worth the read.

I will not speculate (here) on how this ends, but will say I am pretty sure that anyone that says they know is either lying, misinformed or an idiot.

If you want to make a large fortune out of a small one, analyze the companies, only dance with the pretty girls, and “do not have a big dilemma about your big leg Emma.”*

Cheers
Qazulight

*Frank Zappa

2 Likes

Ben Hunt has an excellent article titled “Three Body Problem” where he discusses a math certainty. It is worth the read.

If Ben Hunt would cut it by half I might get to the end. I’ve tried to read several Epsilon Theory articles to no avail. Can’t reach the end.

Denny Schlesinger

1 Like

Ok, The Three Body Problem, that was a fascinating read. His first suggestion is simply diversification. That kinda runs counter to what we do here. :wink: His second suggestion, to minimize maximum regret, I agree with. Its why I only invest 40-45% of my money in the growth stocks, and why I don’t have more than 9% in any single name – it minimizes my maximum regret. His third suggestion, to cut back on computer directed investing, I think everyone hear would agree with. On the other hand, I have 15% of my money in AIEQ… His fourth, that chaotic systems are not knowable, but we still need to try, dead-on.

Thanks for pointing out that article.

2 Likes

Interview with Eitan Medina, Habana Labs’ Chief Business Officer:

I recently sat down with Eitan Medina, Habana Labs’ Chief Business Officer, to discuss the development of this new class of AI processors and what it means for the cloud business.

cont’d

https://www.convergedigest.com/2018/09/interview-habana-labs…

Starrob

1 Like

What do you suppose Nvidia has in the lab that they’ve not made public? My guess is they already in prototype form chips that are 3x, 5x or more than what they have in production.

6 Likes

What do you suppose Nvidia has in the lab that they’ve not made public? My guess is they already in prototype form chips that are 3x, 5x or more than what they have in production.

brittlerock

The issue would not be that Nvidia can’t also build chips that perform as well as Habana Labs but the issue seems to be that the AI inference chip market looks to be developing in a way in which the market might fracture and the chips become commoditized.

GPUs seems not be a area that has gotten totally commoditized. Nvidia can charge premium pricing in many areas in which GPUs get used because Nvidia seems to have a narrow moat around GPUs because of CUDA and other technical advantages. Not only does Nvidia get considered to make the best GPUs but they often lock customers in to using Nvidia chips. It’s not impossible for customers to switch to AMD but it can be a hassle…enough of a hassle for Nvidia to have a narrow moat.

Many people have been and still would be assuming that Nvidia will become as dominant in AI chips as they would be in GPUs because people assume that GPUs will become the dominant AI chip well into the future. The problem would be that the GPU chip has deficiencies that make GPUs to likely only be a short term answer for AI applications, not a long term answer.

GPUs have become dominant for use in AI Training but slowly but surely it’s becoming obvious that the GPU will likely not dominate inference and that’s not just me saying that…industry experts have been saying that and the actions by various players that will use AI inference chips from Hyperscalers like Google, Facebook, Microsoft to companies intending on playing on the edge have been indicating that.

If people truly look at the Facebook Glow project https://facebook.ai/developers/tools/glow , at it’s essence GLOW does for AI chip inference makers what Android does for hardware manufacturers not named Apple. Android relieves manufacturers of having to build a operating system for their hardware, while GLOW would in a similar fashion relieves the need of hardware manufacturers of building their own compiler software and instead rely on GLOW which would be community-driven open source compiler software.

The end result for customers of AI chips would that it prevents or makes it harder for AI chip companies to lock in customers because theoretically, it makes AI chips easier to be replaced.

Some news about Facebook Glow:

At Facebook’s 2018 @Scale conference in San Jose, California today, the company announced broad industry backing for Glow, its machine learning compiler designed to accelerate the performance of deep learning frameworks. Cadence, Esperanto, Intel, Marvell, and Qualcomm committed to supporting Glow in future silicon products.

https://venturebeat.com/2018/09/13/intel-marvell-qualcomm-an…

Why would Facebook create GLOW? I would guess that Facebook has no desire to be locked into buying chips from only one vendor. I mean currently Nvidia’s data center inference chip cost almost $9,000 bucks. What company in their right mind wants to be locked in to paying Nvidia $9,000 buck/chip?? Totally not scalable.

Facebook and other hyperscalers desire a environment in which competition drives chip prices more down into the hundreds of dollars level instead of the thousands of dollars level. Facebook and other hyperscalers want to easily switch to the latest & newest chips as they become available and not become dependent on only one company to technologically advance the field. The Hyperscalers do not want to be locked in.

With big companies like Cadence, Esperanto, Intel, Marvell, Nvidia and Qualcomm, along with around 50 start-ups, along with some “customers” (like Google, Apple, etc.) deciding to build their own inference chips, the AI inference market will likely fracture.

The problem with the GPU as a inference chip would be that a pure GPU chip can not compete with the TPU as a inference chip. Nvidia says that in so many words because Nvidia’s inference chips have a Tensor core (the idea “borrowed” from Google as TPU stands for tensor processing unit):

NVIDIA® Tesla® GPUs are powered by Tensor Cores, a revolutionary technology that delivers groundbreaking AI performance. Tensor Cores can accelerate large matrix operations, which are at the heart of AI, and perform mixed-precision matrix multiply and accumulate calculations in a single operation. With hundreds of Tensor Cores operating in parallel in one NVIDIA GPU, this enables massive increases in throughput and efficiency

https://www.nvidia.com/en-us/data-center/tensorcore/

I have seen some call that the “TPU-ification of the GPU”. Now, while Nvidia might have a moat around their GPU technology, if & when they come out on the field a compete with other players using similar technology to the Google’s and Habana Labs, they won’t have that same advantage. Nvidia will likely then be competing on a level playing field

I think Nvidia locking lots of customers into their AI inference platform while becoming the “Intel of AI Chips” has a low likelihood of happening. The dreams of having Nvidia selling AI chips for thousands of dollars at huge margins for long periods of time…ummmm…well, I would not model that in to any projections about Nvidia because I view such a thing as very speculative to the point where that scenario has a low likelihood of happening.

Will Nvidia be a player in AI chips? Almost certainly but Nvidia might not gain as much out of AI as the hype suggests. The hype suggests little competition in AI far into the future and the reality might be in three to five years there might be a vastly different scenarios occurring as far as competition might be concerned…

Starrob

10 Likes

Starrob,
Thanks for that post. Very informative. At present, NVDA is still primarily a graphics GPU game (pun intended). Their gaming market being more than twice the size of any other segment. And as you said, The hype suggests little competition in AI far into the future and the reality might be in three to five years there might be a vastly different scenarios which pretty much puts my ind at ease for the immediate future. Seriously, no one can predict where these tech companies will be in 3 - 5 years. I kind of look at things quarter to quarter.

3 Likes