2021 and still haven't started my 100 bagger move (Spectra7 SEV, Lightwave Logic LWLG)
Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
I can't put it down. Thinking about our investment all the time.
I was looking for more ecosystems players like Texas Instruments. They have redrivers in the box. Isn't this something of an interest for Nvidia? Does that play well with Infiniband and our cables?
Just typing some fancy words hoping I make sense.
In switches they use DCPs, I still have to understand, but going for analog with redrivers in and out of the box is also possible for Infiniband type systems?
Damn, I wish I understood it better. I keep digging....
Found it. At 15:30 the Q starts. To be fair, she didn’t name Microsoft but leading the Q like that makes me think it was intentionally.
Volex Mellanox connection pre Nvidia
https://www.volex.com/volex-receives-supplier-recognition-from-mellanox/
Post nvidia
Re: OK, let's see where it all began. In November of last year, we all still didn't really know how big AI was going to be, nor how soon it would be here
I'm working and have no time to search for it, but when was it again that the interviewer mentioned Microsoft? Around the same time CRDO lost a bunch of their cables
We will take Infiniband, Credo can take Ethernet. Deal?
From Credo's CC:
Okay. That's definitely helpful, thanks. And then just to clarify sort of the landscape you're selling into where some - there is Ethernet and there is some - there is certainly some InfiniBand out there. I mean, just the notion that you guys might at some point serve the InfiniBand market or that wouldn't be a future opportunity and you expect Ethernet to take share versus InfiniBand.
Just color here - some basic color would definitely help us. Thanks.
Dan Fleming
Yes, I think that NVIDIA has done just a fantastic job. I'm not sure if you've seen too many back-to-back quarters like what they reported.
And so, there's no question that the near-term InfiniBand is seeing great success for this the back-end networks within the NVIDIA clusters. With that said, we're involved in many, many AI next-generation deployments, all of which are Ethernet.
And I think if you rely on the market forecasters, they see a very clear coexistence between InfiniBand and Ethernet long term.
I think you can measure it by dollars, you can also measure by ports and the expectation I think is that the number of ports in Ethernet is going to exceed that of InfiniBand sometime in the next few years.
So I think that we remain very bullish on Ethernet within AI clusters, there's no question about the amount of activity. And there's no surprise in the sense that the market is not going to be satisfied with the sole source provider.
With that said, from the standpoint of our ability to do InfiniBand, there's no limitations.
So I can say in the optical space, we work with companies that are delivering both Ethernet and InfiniBand solutions. And as far as the AEC standard would go, it wouldn't be a huge leap for us to develop those cables, we feel like there's a market opportunity.
Do we know how much a DSP adds to the cost compare linear drive?
I remember that, Steven. Our at least I think I did. Terry Turpin, friend of George Gilder.
Thanks for sharing. LWLG and the render network from Jules Urbach, what a time to be alive!
Hey Steven,
How did you learn about OTOY and their RNDR network?
I trying to figure out what Nvidia/Microsoft architecture looks like. Datacenters are shrinking in size and power. I just saw a recent presentation of Nvidia CEO Jensen were he speaks about GPU's in datacenters. Low power, low latency, etc but their is more.
Especially the latency has my interest for active copper cables.
800G is starting to roll. Matching speeds between analog/digital and optics is crucial.
CRDO earnings today. Meanwhile Marvell is launches 800G systems.
https://www.marvell.com/company/newsroom/marvell-launches-industrys-first-800G-zr-zr-plus-modules-for-data-center-interconnects.html
Hi Steve,
The ACC solutions on slide 11 is for both 56G and 112G solutions. The longest length is 7 meters for 56G while 112G is up to 4 meters.
We will add footnote to clarify. Thank you again for your feedback!!
Best regards,
Bonnie
I would like to understand latency better and why it’s important.
Datacenters are mostly using CPUs. Through my one blockchain investment, the rendernetwork, I’ve learned that only 50k GPUs exist in datacenters. The network has a very good working relationship with Nvidia. And they will add millions of GPUs to their network. It’s like Airbnb for GPUs
If a creative person wants to make stuff, simply render/rent your 10 times A100 cards. Hollywood will never be the same again. Not chilling my other investment, i really don’t care.
Im saying this because i feel ( don’t know) that the latency will be crucially important for AI models, generating renders on the fly. AECs are to slow.
I would like to know and understand what Nvidia is building from an architectural point of view
I got already a mail out to them. I will report back
Thoughts on slide 13? comparing AEC vs ACC. It says we can't go higher then 4M, but on slide 11, it says up to 7M.
slide 11, says ...up to 7 meters
I saw another difference. 3V instead of 2.4V on slide 12 or 10 times lower power (previous 12x lower power)
this is another hint us working our way up to Nvidia, imo
For InfiniBand cables and transceivers, specific lengths, or reaches, are chosen in coordination with the switch buffer latency requirements, latency delays in transceiver and optical fiber. This is done to minimize latency delays in very high-speed systems – a critical factor for AI accelerated systems. Most AI and HPC systems are in size less than the area of two tennis courts. Distances are kept short between components to minimize latency delays.
https://docs.nvidia.com/networking/plugins/servlet/mobile?contentId=132470097#content/view/132470097
It depends on what your aim is in stocks/life. It perfectly reasonable to wait for more confirmation. I like potential stocks to go 10 - 100 times what i payed for. And it’s easier then most think possible. Be early or well timed and hold for at least 5 years.
Im looking for mispriced equities.
Just one or two stocks were everything is just right, and you are set for life.
I hold one blockchain. The ceo is a young genius compared to Elon Musk. That’s all i need and a proven service. Now we wait.
I love S7 at this prices. Put in 10k and wait. Losing 10k won’t change your life. Hitting a million in a few years, will.
Interesting fact is that Volex has signed a charging cable deal with Tesla.
Thanks Rkf. Let’s hope are efforts will pay off.
I do expect not much news before Q4. Just hold tight.
Well i got a great email from S7 saying that they think (never 100%) to be the only cable/chip supplier for Volex. That confirms Future’s claims. Well done!
Im sure glad I bought a big chunk of shares this week.
If I'm not mistaken, Future got some sort of confirmation of Volex their partner is Spectra7. Of course I don't know the exact wording.
https://investorshub.advfn.com/boards/read_msg.aspx?message_id=172639774
I talk on a regular basis with a former Needham analyst, now he move up the ladder to Goldman Sacks. He talks with CREDO, sits on their CC. That's how I met him. Not a bad contact to have going forward.
GS did the lead on the Credo IPO. Next week will be the call. I'm not expecting he will be asking hard questions.He also mentioned to me that Marvell is looking at AECs, and thus validates this technology. He also said that Linear drive is hyped. (I don't agree)
Many pieces are falling and I'm struggling to keep up. S7 ACCs really looks promising. Keep the DD coming.
This board can do this if we keep working together.
Lol 3 minutes after i posted a offtopic, bam. Instant negativity.
That will teach me
I used chatGPT asking to search through Nvidia documentation for GC1122. He gave me 3 results, even said on which page i could find it. But I couldn’t find anything.
If you guys would run it again, that would be great
I hope that too Future. Im always amazed why people fight online. We have a common goal here with Spectra7. Let’s share as much DD and thoughts on the matter as we can instead of calling each other out.
It’s really that simple
We don’t have to be friends.
Great post
No, that’s my comment
I learned a lot today.
The volex nvidia deal is just a small piece of a larger pie. Multiple cable suppliers must be in play to provide any large project.
Confirm by the company: S7 uses 3.3v because this is an industry standard ( as shown on the nvidia docs too)
Also:
“Tomahawk5 still has connectivity shortcomings that shorten the useable length of passive copper cables and make active cabling a "must-have" for lengths of 2m and longer - which is a high %ge of the market “
Just bought more stock. Added another 45k shares
Thank you so much for saying this. That makes total sense.
I remember Raouf saying they worked with 22 cable suppliers.
Anyone remember this too?
Depends on the order is placed by the tier1. I think the bigger the PP, the bigger the revenues for S7 next year.
Just guessing
Thanks for sharing that. I will look for them on LinkedIn.
You are a new account here on ihub. First time poster
Macom and semtec aren’t digital, they are analog. What future says, if they sell aacs, they could violate our patents
Semtec ACCs 800G matches the Nvidia 3.3V reference. Of course, it depends on the length of the cable. S7 has a voltage of 2.4v for 4 meters. So i guess to get to that 5m, the voltage goes up.
https://www.semtech.com/products/signal-integrity/signal-conditioners/gn8112
A little history on how Ari and Jules met.
https://capedge.com/filing/1766363/0001193125-21-121901/EDR-FWP