unconfirmed news that there's been layoffs in the vulcan team... Hope this isn't true and if it is, that's sad; heard they had parts working.. but with Hocknomics nothing's for certain
33 replies (most recent on top)
broadcom cavium vulcan arm server team - is it gone?
Brcm to cavium is like Titanic to Louisiana..
Does anybody know what the terms for Vulcan under the Cavium deal?
If only selected few people are being offered spots at Cavium, how strong would the roadmap be?
Does anyone know how many of Vulcan team were offered job by Cavium?
We interviewed a ex-vulcaner today and likely will make an offer if reference check out ok. According to him, only tens of engineers got offers from Cavm. He is not one of them.
I have a Broadcom document from early 2013 showing promised Vulcan sampling date of Dec 2014. Do they have samples out now? Being 2 years late certainly don,t make your case
Well I'll be... Rick Merritt of EETimes has an article today saying a source from Cavium was considering the purchase of IP from Broadcom's Vulcan or Applied Micro's X-Gene program.
Macom is purchasing Applied Micro but divesting the X-Gene product. So the market is a little full of server-class ARM CPUs for sale at the moment.
From where did you hear 2014 release? FYI 2014 the team was busy with previous MIPS processor.
It looks, you have very little information and you keep on demoting the success of the Vulcan team.
by original target, the chip should have been completed by end of 2014. they did a layoff at 2015 shouldn't matter anyhow if the project was on schedule.
also heard the reason for late design changes were due to poor execution. they had no confidences some major blocks can be completed at all thus had to do some workaround.
the 10x power was for comparison against initially number to Broadcom marketing. not the last revised numbers.
A1 is sampling? Announced at ISC? Please provide a link.
the delay of the tape out is because there is big cut in staffing at early 2015. About 35% people was laid off at that time. a lot of key people were gone all of sudden. And there is no hand over on the work. And then there was a big reorg and change of design. This kind of thing lasted over a year. It is amazing that vulcan chip is such a big success considering what happened.
the power consumption is 1.6 times of the original target. Now 10X. And the solution of addressing the power issue is already there. And it is ready to be implemented in B0
The availability of the vulcan A1for sampling is just announced in supercomputer conference.
In the high voltage, it is able to run over 3Ghz. And with 32 cores and 128 threads, the performance of the vulcan chip is beating the best Intel server chip on the market badly
heard from Broadcom marketing A0/A1 power consumed is almost 10x of initially promised. chip is more than 1.5 years later than original target.
It is 'haswell' not 'hashwell'. it seems you guys know nothing about your illustrated competitor!
They are selling the IP and it is CONFIRMED they are getting all engineers as well.
They have roadmap to make B0 tapeout soon.
a few of the engineers come to my company for interview last week. I heard A1 is working to a degree, but not reliably, power is way too much. Broadcom think it need significant investment to make a product out of it and decided to bail. the product takeout was late by more than 1 year, some potential customers become very pissed. they knew A1 can only for demo purpose even if all fixes worked. all people were cut a few weeks ago but keep 15 engineers on temp base to bring up A1.
are they going to call back layoffed engineers and sell them to Cavium? or just sell IP?
Yes, it does work - with more than 150 benchmarks running on it.
Just to tell you, Cavium is no where when performance numbers are compared with Vulcan-A1 Silicon (These are the words from Cavium)
Do you know any of the ARM processor chip with this much performance. Yes, power numbers could be an issue and they were being addressed in B0. Remember this is the first Silicon.
Just to tell you one thing - Cavium is buying this product. You may see the announcement in a day Or two.
If Cavium believe they have better product why Syed Ali will buy this product ALONG with engineers?
Hashwell Quality !? Are you sure, do you feel shame? Do you remember how many silicons the same team made for XLP-II? Does the cache issue fix?
As a team origin form RMI, you must know Cavium, do you still think it (vulcan) is the only ARM (server)chip? I can list more: APM, Hisilicon, Qualcomm and more, actually with AMD.
Even 'hashwell' quality, how to compete with 'Skylake', (don't tell me you don't know the huge power consumption of A0 vulcan), how to compete with ARM groups, the real in market APM Camvium and Hisilcon products. Oh and Qualcomm already send out its sample to Facebook/google/Lenovo already.
I have to say Mr. Hock's decision is so wise, this project will cost huge RD expend and, more important, with no hope for 'hashwell' quality.
You even think the 'Vulcan A1 chip' work, interesting.
Just for the guys who think silicon is not working - The A1 silicon is working as expected with 2.5 Ghz frequency. Definitely a hashwell quality (if not better).
Mr. Hock is just trying to make some money by write-off millions of dollars to make the helathy balance sheet for investors. He does not look for anything else.
This is the only ARM server chip in the market with working silicon.
Need to keep stock price higher and higher so to acquire company like Xilinx.
No surprise. Hock-t-sizing principle is well known and nobody buys it.
It should have been cut long ago. I actually surprised Hock kept them this long.
It was a good business unit, why could not it be sold?
What a joke! Haswell quality. It was not even working. This project should have been cancelled a year back.
Word is the silicon was top notch. Like Haswell quality for ARM.
About 400 impacted globally in the Vulcan team, they've been given time till this Friday..Apparently the Silicon is working, but the team goes anyway.
The roles were not defined correctly. So many people with circuit-only experience were doing PD or became PD managers. There was no dedicated circuit team to design cells for such a high performance chip. The internal politics caused the key peoples and very few experts they had on the project to move elsewhere. The new management specially PD didn't have guts or leadership to handle a project of this magnitude. No good collaboration between the teams working under different directors.
Sad that they sent few good people a year ago and couldn't manage to do the work without these folks.
Management should have some brains. Earlier the team used to work under one man and with no process it was delivering all the things needed. Now with so many managers & crappy process, they seriously mis-managed.
The old man was better in handling the things.
Yes its true news. All the vulcan team will have to go in another 2 days
it is true
yes, word is this Friday. they feel lucky decision was made just now, it should've been done long time ago.