Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Advice to Tenstorrent

github.com

Post reply on HN
Print view View original

31–40 of 67 posts

Re: Advice to Tenstorrent

#31

Post by throwawaythekey »

throwaway314155 wrote: post #12

Guy would really benefit from learning some manners. Just comes across as painfully toxic no matter how correct he is. edit: For what it's worth, if you can't see that this language is rude or think it is somehow acceptable for people of a certain caliber to talk this way - you're also probably toxic.

Not trying to flame or anything but what is your age?

As someone in my early 30's who grew up on message boards/gaming the language he is using is fairly mild. I think we just have very different social norms.

Re: Advice to Tenstorrent

#32

Post by 1024bees »

samsartor wrote: post #13

I'm doing my PhD in ML shit. Before that I was a systems programming guy, lots of C++, bit of CUDA, big fan of Rust. On the side I'm obsessed with RISC-V. Own a couple of boards. I made a stupid little cuda-like-compiler on top of the RISC-V vector extensions, just for fun. What I'm saying is, tensorrent couldn't find a more excitable third-party developer if they grew one in a lab. And you know what? I can't make he…

There is a natural tension between developing an API that is nice to use and having a full fledged graph compiler. Most graph compilers, and the hardware that requires them will be complex and difficult to approach. The "original sin" was pytorch vs tensorflow -- tensorflow capturing the entire graph and then compiling it with XLA (or whatever it was before, I'm probably mixing up tf1 and tf2 here) was such an intractable mess to actually hack on (also the runtime had unapproachable complexity, from what I recall). This has probably changed, but pytorch won out because it was both nice to use and develop.

There are clear reasons why a hardware company would use a graph compiler -- they think such an approach is higher performance, and makes tenstorrent look better on price per dollar when compared to competitors (read: nvda).

There is some legitimate criticism of TT here, their hardware is composed or simple blocks that compose into a complex system (5 separate CPUs being programmed per tensix tile, many tiles per chip), and that complexity has to be wrangled in the software stack -- paying that complexity in hardware so there is less of a VLIW model in software might remove a few abstractions.

Re: Advice to Tenstorrent

#33

Post by catgary »

moralestapia wrote: post #22
Aurornis wrote: post #20

Earlier quoted context omitted.

> Huh? Yeah, that's what I wrote. You cut out the part of my post where I made my point. Earning “cred” for past accomplishments doesn’t give someone a free pass forever to be a loud mouth.

Then I don't get the point you're trying to make. >Earning “cred” for past accomplishments doesn’t give someone a free pass forever to be a loud mouth. What the ... ? Sorry, but you don't have to earn that. I suggest you familiarize yourself with human rights, particularly if you won the lottery and were born in a country like the US; it is very clearly written, in unambiguous terms, the rights their citizens enjoy.…

I don’t know what you’re talking about here. You can run your mouth as much as you like…you just become a laughingstock at a certain point. Nobody is calling for geohot to be silenced.

Re: Advice to Tenstorrent

#34

Post by catgary »

Havoc wrote: post #17
bigyabai wrote: post #2

With all due respect Mr. Geohot, you've got an awful lotta nerve to throw stones from your glass house. This guy needs to drop the middle school tough-guy tone and post numbers. The central thesis of this article is sound, lead with it: > You aren't going to get better deals on tapeouts/IP than NVIDIA/AMD. You need some advantage. > If you want a dataflow graph compiler, build a dataflow graph compiler. Now explain w…

>This guy needs to drop the middle school tough-guy tone and post numbers. Pretty sure comma is profitable? Not particularly, but for a hardware startup selling multiple iterations and not getting wrecked is a sound start

Wasn’t he pushed out?

Re: Advice to Tenstorrent

#35

Post by vasco »

throwaway314155 wrote: post #12

Guy would really benefit from learning some manners. Just comes across as painfully toxic no matter how correct he is. edit: For what it's worth, if you can't see that this language is rude or think it is somehow acceptable for people of a certain caliber to talk this way - you're also probably toxic.

> you can't see that this language is rude or think it is somehow acceptable for people of a certain caliber to talk this way

Who's the people of a certain caliber? Are there people who can talk like this and others that are better people which shouldn't? What weird thing to say.

Re: Advice to Tenstorrent

#36

Post by georgehotz »

latchkey wrote: post #19
imtringued wrote: post #10

Earlier quoted context omitted.

I think he is right about AMD but completely misses the mark when it comes to tenstorrent. He is ranting about exponential linear unit (elu), which hardly seems to be something that could possibly hold an AI company back. If the hardware is running and training models reliably, then it's just a matter of pricing to stay competitive. Poor optimization cuts your margins, but the incentives are aligned. With AMD the exp…

People refuse to believe it, but the AMD experience is getting better every day by leaps and bounds. Over the last few months, there is a brand new focus on improving the software. There is still a long way to go, but the company is absolutely trying to save itself.

AMD has legitimately been making great progress. They still have a long way to go, and I appreciate SemiAnalysis taking up the mantle of calling them out, but I ran:

pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/rocm6.3

today on stream and it just worked. No external ROCm install, and just the amdgpu driver that's in Arch.

We also have our own complete driver/runtime now in tinygrad; it's so much nicer to build on a foundation when you can blame yourself for the bugs.

Re: Advice to Tenstorrent

#37

Post by latchkey »

georgehotz wrote: post #36
latchkey wrote: post #19

Earlier quoted context omitted.

People refuse to believe it, but the AMD experience is getting better every day by leaps and bounds. Over the last few months, there is a brand new focus on improving the software. There is still a long way to go, but the company is absolutely trying to save itself.

AMD has legitimately been making great progress. They still have a long way to go, and I appreciate SemiAnalysis taking up the mantle of calling them out, but I ran: pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/rocm6.3 today on stream and it just worked. No external ROCm install, and just the amdgpu driver that's in Arch. We also have our own complete driver/runtime now in ti…

Regarding SA, I’m all for holding AMD accountable, but let’s at least get the facts right, and maybe don’t come at it with a history of cheerleading for Nvidia.

Maybe SA might set their sights on you next?

Re: Advice to Tenstorrent

#38

Post by empiricus »

htrp wrote: post #7

Has geohot done anything since the original iphone jailbreak? The ventures he has started (I can think of tinygrad and comma ai) all seem like half finished tech demos.

Well, at the minimum he writes more code and is more competent than me, so he has my respect and I take his opinions into consideration.

Re: Advice to Tenstorrent

#39

Post by Aurornis »

henning wrote: post #6
add-sub-mul-div wrote: post #4

Earlier quoted context omitted.

Don't forget he was also going to save Twitter but noped out after a few weeks.

Maintaining and improving existing software sure is boring and often thankless compared to starting flashy new projects where you get to make and understand all the major decisions up front.

> Maintaining and improving existing software sure is boring and often thankless

There was a big public episode where he appealed to Elon Musk for a job at Twitter, was given an “internship”, tried soliciting public submissions for the code he was tasked with, left a lot of people shocked that he was struggled with basic FE work, and then resigned 4 weeks in: https://news.ycombinator.com/item?id=34074344

Re: Advice to Tenstorrent

#40

Post by Aurornis »

moralestapia wrote: post #22
Aurornis wrote: post #20

Earlier quoted context omitted.

> Huh? Yeah, that's what I wrote. You cut out the part of my post where I made my point. Earning “cred” for past accomplishments doesn’t give someone a free pass forever to be a loud mouth.

Then I don't get the point you're trying to make. >Earning “cred” for past accomplishments doesn’t give someone a free pass forever to be a loud mouth. What the ... ? Sorry, but you don't have to earn that. I suggest you familiarize yourself with human rights, particularly if you won the lottery and were born in a country like the US; it is very clearly written, in unambiguous terms, the rights their citizens enjoy.…

> What the ... ? Sorry, but you don't have to earn that. I suggest you familiarize yourself with human rights, particularly if you won the lottery and were born in a country like the US; it is very clearly written, in unambiguous terms, the rights their citizens enjoy.

This is obtuse. I obviously wasn’t talking about human rights or the 1st amendment.

I was saying it’s illogical to suggest that because someone did one impressive thing in another domain a long time ago, we should therefore continue to value their input on every topic forever.

Post reply on HN