Hacker News new | past | comments | ask | show | jobs | submit login

On AMD, with Ryzen Master, you can set the TDP-envelope of the processor to what you want. Then the boost/frequency/voltage envelope it chooses to operate in under sustained load is different.

IMO, shopping by performance/watt makes sense. Shopping by TDP doesn't. (Especially since there is no comparing the AMD and Intel TDP numbers as they're defined differently; neither is the maximum the processor can draw, and Intel significantly exceeds the specified TDP on normal workloads).




TDP matters a fair bit in SFF(Small Form Factor) PCs. For instance the 3700x is a fantastic little CPU since it has a 65W TDP but pretty solid performance.

In a sandwich style case you're usually limited to low profile coolers like Noctua L9i/L9a since vertical height is pretty limited.


Performance/watt matters. You can just set TDP to what you want with throttling choices.

If you want a 45W TDP from the 3700X, you can just pop into Ryzen Master and ask for a 45W TDP. Boom, you're running in that envelope.

I think shopping based on TDP is not the best, because it's not comparable between manufacturers and because it's something you can effectively "choose".


How do you do that? Is it a setting in the bios? Or can it be done runtime? If so, how? It sounds interesting if I can run a beefy rig as a power efficient device, for always-on scenarios, and then boost it when I need.


> How do you do that? Is it a setting in the bios? Or can it be done runtime?

On AMD, it's a utility you run. I believe you may require a reboot to apply it. On some Intel platforms, it's been settings in the BIOS.

> It sounds interesting if I can run a beefy rig as a power efficient device, for always-on scenarios, and then boost it when I need.

This is what the processor is doing internally anyways. It throttles voltage and frequency and gates cores based on demanded usage. Changing the TDP doesn't change the performance under a light-to-moderate workload scenario at all.

Ryzen Master lets you change some of the tuning for the choices it makes about when and how aggressively to boost, though, too.


Ryzen Master doesnt seem to be available for linux so you end up with bunch of unnofficial hacks that may or may not work. I run sff setup myself, originally wanted to get 3600 but it was out of stock, and the next tdp friendly processor was 3700x.


That's an annoyance, but on Linux you have infinite more control of thermal throttling and you can get whatever thermal behavior you want. Thermald has been really good on Intel, and now that Google contributed RAPL support you can get the same benefits on AMD-- pick exactly your power envelope and thermal limits.


Yeah but can I get a metric ton of benchmarks at that 45w setpoint?

I don't really see the reason in paying for a 100w TDP premium if I'm just going to scale it down to 65w.


> Yeah but can I get a metric ton of benchmarks at that 45w setpoint?

Yup, they're out there.

> I don't really see the reason in paying for a 100w TDP premium if I'm just going to scale it down to 65w.

You might want the core count or peak performance for the very short term. When I was looking, running 65W parts in the 45W envelope was only about a 7% penalty, so you get a bunch more performance/watt.


I'm running a 2400G in a Mellori-ITX. Another issue is sizing the power supply.


Back when my daily driver was a Core 2 laptop, someone told me that capping the clock frequency would make it unusable.

As a petty "Take that", I dropped the max frequency from 2.0 GHz to 1.0 GHz. I ran a couple benchmarks to prove the cap was working, and then just kept it at 1.0 for a few months, to prove my point.

It made a bigger difference on my ARM SBC, where I tried capping the 1,000 MHz chip to 200 or 400 MHz. That chip was already CPU-bound for many tasks and could barely even run Firefox. Amdahl's Law kicked in - Halving the frequency made _everything_ twice as slow, because almost everything was waiting on the CPU.


The funny thing is, on modern processors-- throttling TDP only affects when running flat out all-core workloads. A subset of cores can still boost aggressively, and you can run all-core max-boost for short intervals.

And the relationship between power and performance isn't linear as processor voltages climb trying to squeeze out the last bit of performance.

So if you want to take a 105W CPU and ask it to operate in a 65W envelope, you're not giving up even 1/3rd of peak performance, and much less than that of typical performance.


You’re giving up 0 of peak single thread performance. A single core in turbo across Intel and AMD, mobile and desktop uses max 50W.


Here are some numbers on single core power consumption, ⅌ https://www.anandtech.com/show/16214/amd-zen-3-ryzen-deep-di...:

AMD Ryzen 9 5950X: 20.6W for a single core at 5050MHz, 49W for the whole package. (And it’s generally the package figure that you care about.)

AMD Ryzen 9 5900X: 17.9W/54W at 4875MHz.

AMD Ryzen 7 5800X: 17.3W/37W at 4825MHz.

AMD Ryzen 5 5600X: 11.8W/28W at 4650MHz (though the highest core reading is 13W, at three cores loaded).

You’re both correct: by simply restricting that power envelope by 40%, you shed a lot less multi-threaded performance than people realise, and no single-threaded performance.

Look at the 5950X figures, and you observe that at about 120W, it can run 6 cores at 4,650MHz (27,900 core–MHz), or 16 cores at 3,775MHz (60,400 core–MHz).

Expressed one way: by dropping the frequency by 20%, power per watt increased by around 2.7×.

Expressed another way: let’s skip a 65W envelope—put this particular 105W chip in a 40W envelope and you lose only 20% of your six-cores performance. Seriously. But I’m not sure what the curve would look like if you load all 16 cores at a 40W envelope, what speed they’d be going at.

(But do remember that “TDP” is a bit of a mess as a concept, and that we’re depending on non-core power consumption being generally fairly consistent regardless of load.)




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

Search: