alt Hacker NewsThe thermal design is irrelevant, and people saying they want insane power density are, in my personal view, deluded ridiculous individuals who understand very very little.
Overclocking long ago was an amazing saintly act, milking a lot of extra performance that was just there waiting, without major downsides to take. But these days, chips are usually already well tuned. You can feed double or tripple the power into the chip with adequate cooling, but the gain is so unremarkable. +10% +15% +20% is almost never going to be a make or break difference for your work, and doing so at double or triple the power budget is an egregious waste.
So many of the chips about are already delivered at way higher than optimum efficiency, largely for bragging rights. The exponential decay of efficiency you keep pushing for is an anti-quest, is against good. The absolute performance wins are ridiculous to seek. In almost all cases.
If your problem will not scale and dumping a ton of power into one GPU or one cpu socket is all you got, fine, your problem is bad and you have to deal with that. But for 90% of people, begging for more power proces you don't actually know jack & my personal recommendation is that all such points of view deserve massive down voting by anyone with half a brain.
Go back to 2018 and look at Matthew Dillon on DragobflyBSD underpowering the heck out of their 2990wx ThreadRipper. Efficiency just soars as you tell the chip to take less power. The situation has not improved! Efficiency skyrockets today at least as much as it did then by telling chips not to go all out. Good chips behave & reward. I believe Apple competent enough to thoroughly disabuse this position that this chip would be far better if we could dump 2x 3x more power into it. Just a fools position, beyond a joke, imo. https://apollo.backplane.com/DFlyMisc/threadripper.txt
Replies
It's been funny to see people move from overclocking to underclocking. Especially for the older AMD gpus. On the RX480 a slight underclock would cut the power usage almost in half!
Oh, we're largely on the same page there.
I was actually looking for benchmarks earlier this week along those lines - ideally covering the whole slate of Arrow Lake processors running at various TDPs. Not much available on the web though.
I learned a lot about underclocking, undervolting, and computational power efficiency during my brief time in the ethereum mining[1] shenanigans. The best ROI was with the most-numerous stable computations at the lowest energy expense.
I'd tweak individual GPUs' various clocks and volts to optimize this. I'd even go so far as to tweak fan speed ramps on the cards themselves (those fans don't power themselves! There's whole Watts to save there!).
I worked to optimize the efficiency of even the power from the wall.
But that was a system that ran, balls-out, 24/7/365.
Or at least it ran that way until it got warmer outside, and warmer inside, and I started to think about ways to scale mining eth in the basement vs. cooling the living space of the house to optimize returns. (And I never quite got that sorted before they pulled the rug on mining.)
And that story is about power efficiency, but: Power efficiency isn't always the most-sensible goal. Sometimes, maximum performance is a better goal. We aren't always mining Ethereum.
Jeff's (quite lovely) video and associated article is a story about just one man using a stack of consumer-oriented-ish hardware in amusing -- to him -- ways, with local LLM bots.
That stack of gear is a personal computer. (A mighty-expensive one on any inflation-adjusted timeline, but what was constructed was definitely used as a personal computer.)
Like most of our personal computers (almost certainly including the one you're reading this on), it doesn't need to be optimized for a 24/7 100% workload. It spends a huge portion of its time waiting for the next human input. And unlike mining Eth in the winter in Ohio: Its compute cycles are bursty, not constant, and are ultimately limited by the input of one human.
So sure: I, like Jeff, would also like to see how it would work when running with the balls[2] running further out. For as long as he gets to keep it, the whole rig is going to spend most of its time either idling or off, anyway. So it might as well get some work done when a human is in front of it, even if each token costs more in that configuration than it does OOTB.
It theoretically can even clock up when being actively-used (and suck all the power), and clock back down when idle (and resume being all sleepy and stuff).
That's a well-established concept that [eg] Intel has variously called SpeedStep and/or Turbo Boost -- and those things work for bursty workloads, and have worked in that way for a very long time now.
[1]: Y'all can hate me for being a small part of that problem. It's allowed.
this is all 100% true and yet the 12 year-old boy inside me still smiles smugly at how fucking cool my dual reservoir water-cooled setup is, and how there was a brief moment in time a couple years ago where i had arguably one of the fastest (consumer) setups in the entire world... was any part of that labor or money "worth" it? no, absolutely not. was the $1k power bill i had to pay PG&E one month worth it? even less so. but do i have any regrets? absolutely not! :)
anyone even remotely on the fence about whether or not they should bother with all this stuff, just read OP or read this tl;dr: the answer is no, it is not.
>> people saying they want insane power density are, in my personal view, deluded ridiculous individuals who understand very very little.
Or they are simply not-rich people who cannot afford to purchase extra hardware to run in parallel. Electricity is cheap. GPUs are not. So i want to get every ounce of power out of the precious few GPUs i can afford to own.
(And dont point at clouds. Running AI on someone else's cloud is like telling a shadetree mechanic to rent a car instead of fixing his owm.)
> Overclocking long ago was an amazing saintly act, milking a lot of extra performance that was just there waiting, without major downsides to take.
Back when you bought a 233 Mhz chip with ram at 66 Mhz, ran the bus at 100 Mhz which also increased your ram speed if it could handle it, and everything was faster.
> But these days, chips are usually already well tuned. You can feed double or tripple the power into the chip with adequate cooling, but the gain is so unremarkable. +10% +15% +20% is almost never going to be a make or break difference for your work
20% in synthetic benchmarks maybe, or very particular loads. Because you only overclock the CPU these days so anything hitting the ram won't even go to 20%.