Hacker Newsnew | past | comments | ask | show | jobs | submit | david-gpu's commentslogin

Just because it is obviously true to you, it doesn't mean it is obviously true to others.

And just because it is obviously true to you, it doesn't mean that it is true, either.


And just because harvard researchers say something doesn't make it true either. Especially when it's like 6 months late and reviews are probably not the bottleneck anymore - testing, ci, parallelization are (imo).

Paraphrasing one of my favorite movie quotes: "Well ain't this a temporal anomaly. 6 months from any criticism!"

No, they don't.

(I am testing the hypothesis)


All contrarians are deliberately wrong.

Not all, not deliberately, and not wrong.

"All of them, deliberately wrong," said the contrarian as he raised his shovel

Nada. Knee jerk online contrarianism is the lowest effort path to an “I will be heard!” dopamine hit the Earth has ever seen. The act is the reward.

(Sorry, couldn’t stop myself.)


I worked on AMD GPUs post-ATI acquisition, and many of my coworkers were former ATI employees. Later in my career I joined NVidia.

With all respect to my former ATI/AMD coworkers: the folks at NVidia are on a different league. It wasn't even close. My years at NVidia were a very humbling experience.


    > the folks at NVidia are on a different league
What made them great? Can it be learned?

There was no big secret sauce. They were simply smarter in average, as a cohort. At ImgTech, AMD and Qualcomm we had more variation: some people were just as smart as folks at NVidia, many were not.

Everything else, I think, was downstream of that. It was a culture of excellence. Like I said, it was very humbling. I'm glad I had an opportunity to sneak in.


It has some redeeming qualities.

I would guess its running time is as consistent as the memory hierarchy allows. Having tight bounds on execution can be valuable in real time systems.

Simple to verify that the implementation is correct. Simple to implement in assembly.

Could be worse! I still prefer bogosort.


It has some cache locality that may make it beat smarter algorithms for certain arrays. For example, it can perform a lot of swaps while an O(n log n) algorithm is waiting for data to be fetched from memory.

Of course, an O(n log n) algorithm can also be optimized for cache locality.


> No more affairs, and he can come back.

Please walk me through it. What makes the first three known affairs tolerable, but then a hypothetical fourth affair unacceptable?

In my mind, when a person betrays their own partner, I have no reason to place any trust in them. Not in their judgement, not in their character.

As a former employee that had a rather positive view of the man, I was massively disappointed in the revelations. If I was still at NVidia, I would not want him back; NVidia has lots of excellent managers that can replace him.


These traits might be important if your goal was to marry a person. But I'm pretty sure the track record with NVidia from the very start speaks for itself plenty.

>I'm pretty sure the track record with NVidia from the very start speaks for itself plenty

Would that be the nepotism or the market manipulation?


I only had to assemble instructions by hand three times in my career. It is not difficult if you are already familiar with the ISA of that particular processor, but it is tedious.

The first time was in my first job writing OpenGL ES 2.0 blending operations for the Imagination Technologies SGX 535, which was a remarkably advanced little GPU for its time. Alpha blending happened through special instructions in the fragment/pixel shader. We did have a compiler, but it didn't handle blending, for whatever reason.

Through testing I noticed that NVidia GPUs performed blending incorrectly, according to the specification, which was a big surprise. Only years later I figured out that the specification had some errors; I now would bet that NVidia was doing the correct thing.

The second time I had to assemble instructions by hand happened while working on a stripped down version of the Radeon 7xxx series that shipped on some mobile phones. I don't recall what the problem was, whether we were adding a feature or optimizing the microcode generated for a particular OpenGL operation, but it involved a rarely-used feature called user-defined clipping planes. We didn't have an assembler for that GPU; all microcode was assembled by hand.

The third time happened at NVidia, and I will be vague, as not enough time has passed for me to be comfortable going into any detail. I had an idea for how to provide hardware acceleration for persistent CUDA kernels, which eventually was marketed as Cluster Launch Control (CLC). In order to quantify the benefit of adding some hardware, when it was possible to approximate that functionality in software, I needed to test the most efficient software implementation that was possible, and the compiler wasn't quite giving the assembly I wanted, so I had to manually patch some instructions post-compilation. Why did I not use the assembler that was available? I couldn't, because the assembly language did not provide a way to express the exact feature I had to patch.


I was such a nerd, happily learning 6802 assembly on the II+...those were the days...I miss it.

Control codes had you down? ;)

Shush!!

I find Ramani to be rather contemptuous of or bitter about people she doesn't like. In other words, to me she comes across as somewhat hypocritical.

If you enjoy Ramani, you may also want to read the works of Dr. Lindsay Gibson. She does a great job of guiding people who suffer from having emotionally immature parents, without dehumanizing them. She doesn't even use clinical terms in her books, even though she is a clinical psychologist. The principles are the same even if the people you are struggling with are not your parents.


Therapists can get burned out, especially those who work with NPD people, and Ramani seems over it. Unfortunately that comprises her usefulness as a therapist and as a source of expert advice. Dr. Mark Ettensohn of the YouTube channel Heal NPD is a better source, imo.

Ramani specializes in NPD. People with NPD are not able to provide for themselves emotionally so they rely on others, their “narcissist supply”, to which they become dependent. It’s very much like addiction, except instead of abusing a substance, they end up abusing people in order to meet their needs; people become objects and resources instead of being people. That objectification, if not treated, is how narcissism spreads. That’s the great benefit of grey rocking; they aren’t attracted to people who don’t give them energy and will leave them alone. Yellow rocking is a more positive way to interact with the world, but you have to be really good at setting boundaries, otherwise someone with NPD will misjudge polite engagement as a willingness to give them energy. With NPD people, imho it’s more kind to grey rock, as yellow rocking can make them feel rejected, which adds to their trauma. Like addiction, the thing that helps them recover is to be cut off from their supply, so they can learn to provide for their own emotional needs instead of relying on external sources. Very often empathetic people unintentionally end up acting as enablers which causes more harm. Objectification harms the objectifier as well as the person being objectified; if they see others as objects, they internalize that themselves as well, and become the source of their own trauma.


> Dr. Mark Ettensohn of the YouTube channel Heal NPD is a better source, imo.

Thanks for the tip, I will check him out.

I will admit that these days I do not consume any material that deals with NPD, because while I found it useful early on to contextualize the behavior of a couple of people who have abused me, it is not my responsibility to either diagnose or treat them. Instead, my attention has shifted to folks like Tim Fletcher and Lindsay Gibson, who elaborate on the path to recovery.

> Like addiction, the thing that helps them recover is to be cut off from their supply

There appears to be quite some debate around whether or not folks with NPD can be treated at all, with only a few therapists seemingly confident that they can genuinely help them, such as Diana Diamond et al.


I think Ramani is useful for people in the midst of dealing with people with narcissistic traits, especially if they cannot immediately escape the relationship. There are probably others who are better at understanding from a greater distance.

Brute force is not an option when you have 640Kb of RAM and your processor's clock runs at single-digit MHz.

Yes! Even 640K was a crazy luxury. Microsoft BASIC ran with a tiny fraction of that.

I mean brainpower instead of good design to manage the complexity. Not brute force algorithms.

When I switched from the Pro to the Plus subscription, they did not state anything to that effect. I expected a smaller quota of their most recent flash models, and instead I got models that are so old they are useless to me, so I canceled the subscription.

It is the lack of transparency that gets me. With PAYG open-weight models, I know exactly what I get for my money.


> It is the lack of transparency that gets me.

The Google One site lists all the tiers and their access levels. You didn't bother to check, and call the lack of a nag-screen a "lack of transparency?"


I did check. Did you?

It literally says "Access to our Flash models" with no indication that they would see no updates for... How many months has it been?

I think you can understand why I was expecting the version of the Flash models to keep up with updates.

Edit: And back when I paid for my yearly subscription they also promised access to the Pro models. They simply took that feature away and offered no compensation. So, understandably, I am disappointed and disinterested in their offerings.


I would be interested in where you are looking, because the first page I found [1] says Plus includes "Access to Gemini 3.6 Flash and Deep Research in Gemini." and I don't see any "Access to our Flash models" on that page. Can't find anything similar on the wayback machine either.

Removing access to the Pro models is pretty messed up though, I agree.

[1] https://one.google.com/intl/en_ca/about/google-ai-plans/


I think they are changing the language, because my quote about Flash models was verbatim from that same page when I posted my comment. It now reads a vague "Access to Gemini 4 Hydrogen and Deep Research in Gemini.".

Edit: And opening the Gemini app, I see indeed Hydrogen 4 rather than Flash 3.6, which is what I had yesterday when I posted that. It would not surprise me if this is a reaction to the constant complaining in HN about what the had been doing for months now.


People see

"Access to 3.1 pro" and they read it as "Access to Gemini Pro models".

Turns out when Google names a model in that subscription, they mean that model, not necessarily that class of model :/


Very appropriate use of British understated sarcasm. Philomena Cunk would approve.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: