Was Rocket Lake Power Efficient? | Hacker News Reader

Was Rocket Lake Power Efficient?

chipsandcheese.com

44 pointsrbanffy36 comments

I feel like a bit of a dope, but I have to go look up these "Lake" chips from Intel by name whenever somebody writes about them.

Often these are delayed so you've been hearing about "such-and-such Lake" for three years and by the time it arrives you are so sick of hearing it that it sounds like it is something old and obsolete.

It looks like Rocket Lake is an advanced chip on the 14nm node which is not so bad as it sounds because the process of things getting better when you shrink has been derailed as of late. Before 2005 you had explosive progress in hardware because the clock speed and power consumption both go down. Until recently the cost per transistor went down, but now it is heading up which is one of the reasons why the 4090 is as crazy expensive as it is. I've seen 22nm parts still running strong after 10 years, I think 14nm parts will hold up, but it might be 7nm parts won't have that kind of lifespan.

I have an i7-2600 (32nm) still in service and it's shocking how well it keeps up with modern tasks. The cheapest current Ryzen you can buy would destroy it benchmarks but you'd never notice day-to-day.

Same here, mine has been running overclocked to 5ghz for about 10 years (?) now and I refused to upgrade it by quite a few "lakes". Sure now it's getting pummeled by hugely multicores newer processors, but for a long time, it was top of the range.

The #1 reason to upgrade is Microsoft is demanding it for Windows 11. Otherwise, older CPUs with plenty of RAM and an SSD still do great for non-intense workloads.

This is so incredibly lame of Windows. So much e-waste created so that secretaries can use excel and look at their Facebook on Windows 11.

It's not an unreasonable decision for an OS vendor to occasionally say that a certain level of processor is necessary for future versions of the OS. Apple does it with MacOS. I have a 2013 Mac Mini that is going strong except that it doesn't run the latest version of MacOS. The only real problem I had with it is that Safari for a long time didn't support WebP on that version of MacOS which kept me using JPG images for a while.

The trouble is that Intel's rollout of the future has been shambolic at best. Potentially revolutionary extensions such as TSX and SGX failed. It is still too early to require AVX-512 which could speed up and reduce power consumption of your web browser since many of Intel's "latest and greatest" client CPUs don't support it. (Don't buy the false claim that AVX-512 is a power pig: sure it will consume a lot of power running at full clock because it is doing a lot of work... AVX-512 helps "atom class" processors quite a bit because it avoids instruction decoding which Intel is bad at.)

It's not clear that Microsoft's new line in the sand is going to buy them very much extra functionality for Windows.

The problem is they haven't added anything significantly new to Windows 11. I don't see any features that requires new processor features.

> The #1 reason to upgrade is Microsoft is demanding it for Windows 11.

Isn't that also the #1 reason not to upgrade? In other words, is Windows 11 compelling yet? As of now, mostly I just hear complaints, and ocassionally tidbits about better scheduling for newer chips with multiple core types (which is compelling, but only if you upgraded)

No, Windows 11 is not compelling. MS is trying to be a worse apple, and doing a great job at it.

WSL 2 is significantly better on Windows 11 than Windows 10. If you do remote development on WSL, it's worth upgrading to Windows 11 purely for the ability to get the wsl2 improvements.

Those have all (including WSLg) recently been completely backported to and made available on Windows 10 (https://devblogs.microsoft.com/commandline/the-windows-subsy...).

Same here, I relatively recently* upgraded from my OC'd first gen i7, that thing was great for a very long time.

*Good lord, has it been 3 years already?

Interesting, because when I upgraded from a Ryzen 3000 series to a Ryzen 7000 series, I experienced significant speedups in various tasks: - Compiling code - Running multiple services in docker containers for local development that are a mix of cpu and io bound

Whenever I use an older laptop I have that is on intel i7, 8th gen, these tasks are much much slower. I think it really depends on your workload.

For sure. I moved from setup with an i7-6700k and a gtx 1070 to one with a 5900x and an rx 6900xt, and in games the upgrade wasn't really as strong as you might expect.

Code compilation, hashing, rendering were all markedly faster though. That old PC became a perfectly serviceable hand me down after 5 years of service, and I hope to get another 5 out of this rig. Pretty cool that we're seeing longevity and some progress at the same time.

Oh, absolutely for what you describe newer chips are going to make a huge difference. The system I'm describing is used by my wife for Steam games, remote desktop, media consumption etc.

My work laptop is an M1 Max and personal laptop an ancient i5 Macbook. I had it upstairs recently and thought I'd do a little coding but gave up after a single `brew install whatever` was still going after 20 minutes and I lost my motivation...

The i7-2600 is a desktop part though, they stand up pretty well.

I'm rocking an i5-4670k with a mild overclock, and I'm just getting to the point where it's starting to bottleneck my framerate in games (the new MS Flight Sim was the straw that broke the camel's back)

It's probably a silly thing to do, but if you can pickup a 4770K or Xeon 1281v3 for the price of a pizza I bet you could get a a bit more life out of your machine. More cache and hyperthreading always helps.

I had a 1271v3 in my son's PC until the motherboard died earlier this year and it was still quite beefy.

Rocking an i5-2500 myself. We timed that purchase perfectly. Prior to our Sandy Bridge each new generation improved performance by ~25%. Subsequently, each new Intel generation improved by ~10%.

There have been a lot of 10%'s in the meantime, though...

Off-topic, but why do we still use nm to refer to things? I assume everyone on here is already aware that it's purely a marketing term, but we still use it.

People have been using nm for 20 years so it's going to take a while for people to adjust to using node names, if ever. People still don't use kibis and gibis.

I don’t know if this will help you but it is at least a nice list in a plain text file — intel-family.h from the Linux source tree:

https://github.com/torvalds/linux/blob/master/arch/x86/inclu...

Intel solutions always seem to have better idle consumption. Maybe this has changed, I'm still playing with older equipment.

From the stuff I have access too physically (xeons(well gold cpus) and thread rippers) the intel stuff is much cooler until you really rag it, where as the threadrippers are just plain warm.

This isn't scientific, so I might be comparing big 64 thread monsters to smaller intel jobbies.

My Zen3 still pulls >25W at idle which is an absolute embarrassment.

Is that a from-the-wall number or just the CPU?

My 5950x desktop pulls like 60W from the wall at idle in linux.

Damn. I'll have to test again, but mine idles around 120 W IIRC.

My UPS reads 150W when my monitors are both powered on.

It doesn't help that the x570 chipset uses like twice the idle power of the x470 it replaced.

I have a b550 chipset. It's got Gigabyte's high end VRM circuit, though.

One major drawback of the chiplet arch.

I found that having steam open caused my ryzen 5 3600 to go from 3 watts idle to multiple tens of watts. Chromium under the hood is so stupid.

If anything Intel has become better with each newer generation. My 12700K idled at < 3W package power. My 13700K pulls even less. They have a lot of good stuff going on like not enabling the L3 cache if the CPU in only briefly awake, or dynamically resizing L3 if they can get away with it.

Fellow 12700k owner, I am very impressed with the idle power consumption. Mine reports idle at ~7W, which is pretty close to my laptop's R7 4750U.

Came from an R5 2600, which I'm almost positive idled higher. While the full-core power consumption is quite excessive, I think Intel overall does a good job of efficiency under idle/light load.

If you object to the K-series CPU's ability to draw current, you have the option to use the linux tool `powercap-set` to limit it to 65W, 95W, or whatever arbitrary limit you think is best. Intel RAPL accepts its parameter down to the microwatt.

Reply on news.ycombinator.com