There is a lot of truth to this. Find "design partners" early in the process and build the MVP for them. After several iterations of making requests and providing feedback that you action on, they're a lot more willing to buy, especially if you give them a sweetheart discount.
Unless it's some sort of bauble or obviously useful tool to them, I think it'll be difficult to sell MVP without any of their input.
No idea but I suspect the "unified" on chip memory is very very quick.
Some friends and I were BSing about the "pro" level parts, it you can graft 2 or 4 M1s together, use off chip RAM and then treat that onboard 16GB like cache? We're talking about some game changing stuff.
The Xeon Phi had up to 16GB of fast memory in the same package as the main die. IIRC, it could be used as memory or as cache for external memory (which was much slower).
If Apple integrates two more memory chips, it'll be able to power a pretty solid desktop or laptop.
On the performance, Rosetta is most likely doing JIT so that most of the time it's running native ARM code. It did this with PPC binaries and DEC had it for Alpha.
We could also entertain the idea that if they would find some instruction particularly hard to emulate - they could have added new instructions on their own chip to cover it.
It is not on-chip memory, the dies are separate, they're just in the same package. They seem to use standard LPDDR4 connectivity, so I don't think its actually faster. The "unified" bit seems to matter more: having a single address space for both CPU & GPU, but this is pure speculation. I don't know if AMD or Intel APUs do this too.
The fact they are on the same package means that the electrical signals have a lot less far to travel from memory to cpu, and therefore you don’t have the signal losses or interference from the board having to route memory lines externally.
As a result you would be able to drive a higher bandwidth because you don’t need to be as limiting with the transfer time of signals.
Or you could use less power for the same speed. Hard to tell what Apple did, without some detailed benchmarks. I suppose one could bench memcpy and derive the clock rate from that.
It's unlikely you're going to transfer data any faster - they're using commodity drams like anyone else - they will however be able to save a clocks's worth of latency here and there which is useful
Didn’t Intel have a similar idea with Skylake? Those had very fast albeit smaller eDRAM die glued to the processor. It was dropped on subsequent generations.
It’s actually still surprisingly relevant in terms of performance [1], and I see it as a precursor to the gigantic caches we are seeing in the latest chips.
It might make sense to use very fast SSD as the main memory and on-chip RAM as cache. Huge amounts of RAM make only sense if your disks are slow or your workload actually needs the whole RAM which is rare.
It's more than morale, it's just being a good human.
If a promotion is coming down to how someone got thanked by a peer, if that's the differentiator like some of these poster are suggesting, then it's probably not a great place.
Be a good human, thank people and express gratitude whenever they help you. Our culture has become so jaded in some many ways that it actually sort of disarms some folks. they will help you more in the future. When someone asks for help, help them, stretch to do it, it can be hard to ask for help so make it worth their while to take that risk. Give praise, it costs you nothing. It will pay dividends in the long run, big ones that you probably can't imagine yet. Being someone that people like to work with is huge...
I'll get myself some down voting here... I kind of like asking questions about data structures and algorithms. I don't see them as simple black and white questions though. I see several values: 1) If you're a "computer science" degreed person, I expect basic understanding of basic algorithms and data structures. As for your basic job stuff, things like Java have multiple implementations of maps or hash tables you should have some idea when to pick one vs the other. I don't think that's off limits. 2) I like to ask a couple challenges or questions that sort of make the interviewee a little uncomfortable, not to get the answer but to work with them on it and see how they work through that situation. I intentionally try not to be cruel but I want them to think, I want to communicate with them as they sort through it, I want to see how that stuff goes. I've had 60+ minute conversations about engineering a professional grade hash table and the different things you have to consider, that's good stuff. The 3rd thing, and this is maybe more subtle, there is a gigantic difference between implementing quick sort and qsort in glibc, the same can be said about most algorithms and data structures and I think at least a cursory knowledge of that is an indication of some wisdom. Should you be asked to implement data structure x or classical algorithm y, understanding the mechanics beyond simply copying it out of a text book indicates some wisdom.
After a grueling interview process at Goldman Sachs, with 7+ technical interviews that required me to solve very specific questions on college-level Maths, Stats and Computer Science (admittedly, I was applying for a quant job position), I was eventually asked to interview candidates myself. While I did not feel entitled to change the current interviewing culture at the company by asking questions of a completely different nature from what I got asked in the first place, I conducted my interviews in a very similar way to what you described. By no means did I expect applicants to reach a definitive answer, but I instead worked on the problems with them to see how far their intelligence, creativity, curiosity and, most importantly, their ability to well communicate their thought process would take them.
Such interviews used to take a while hour of my time (which is a lot to afford when you work in a bank), but by the end of each I believed to confidently ascertain the candidate’s ability to thrive on our team. In retrospect, it has served me really well.
All in all, the problems posed (and the solutions given to them) might not carry as much weight in the final decision as the discussion held with the applicants. As long as the questions asked give them some material to work on, and DS and algorithms usually serve this purpose very well for us engineers and developers in general, one should be able to effectively select candidates given some time investment.
I think this program is generally well intentioned. I directly know organizations that benefited from it and it absolutely helped cushion the blow of initial lock-down. I also think that any time the US federal government is offering out free or low cost money it is imperative that we look closely and carefully, especially when the program was thrown together in a week or two.
"Need" is probably not the way to look at this, anytime there is "free money" there will be incredibly wealthy businesses and organizations with "need." There is a school of thought that you should "always need" free money. I think that's just business. I think the actual distribution is where you look.
It would be completely foolish, ATMO, to believe that every organization, bank, and business were treated equally or fairly in this process. We already know that banks like Cross River handled what appears to be a disproportional amount of the approved "loans," presumably because they are a slick fintech business. When the dust fully settles, I would be shocked if the distribution of the funds or the forgiveness was in any way egalitarian, the country just isn't. I fully expect that it will be another mechanism in which certain classes benefit from the US system and others will have been overlooked. And maybe there will be another excuse, minority owned small businesses may not have been as "well banked" as others or something and your choice of bank will be another differentiator. I've heard a couple stories of orgs that didn't even get applications from their banks until the fund was depleted. There will be some great success stories too, I'll reiterate that I think the program is based in good intentions. I just can't see it being executed fairly though, not when they're actually debating on whether the Bubba Wallace noose was a "hoax" or not and the president chimes in...
That is exactly what will happen if QI goes away, and taxpayers will ultimately pay the premiums. Maybe that would be better, or maybe not. But in general, few policemen could withstand the risk of a single ruinous suit, and without insurance of some kind, would be forced to flee the occupation.
The problem is that the legal system has a quite noticeable element of randomness to it. Despite what you may have read, justice is not always done. (source: have been on the receiving end of this)
As a practical matter, if there is possible liability, everyone will have to have liability insurance, just as in the medical professions, for example.
Unless it's some sort of bauble or obviously useful tool to them, I think it'll be difficult to sell MVP without any of their input.