Hacker News Reader

Comment by Normal_gaussian | Hacker News Reader

Is there a good aggregation of cooler performance somewhere? I've recently completed a build for the first time in a decade and was pleasantly surprised by tools like PC Part Picker, the assembly was a breeze (many micro improvements by everyone), and everything 'just worked'.

Literally the only question I was out cold for was which cooler is good enough? With some coolers rivaling RAM costs and cases having fixings for radiators... it really leaves you thinking. Then again, I left it on stock and have hooked temps into prometheus and everything seems fine so far...

View on HN

Yeah, the ad-funded model and how that supports reviewers seems to have damaged well structured and comprehensive independent review sites. It's all in videos or really low-effort stuff nowadays, and that makes it hard to draw conclusions without significant effort - and even then, considerable risk of drawing the wrong conclusions.

I think this affects far more than just coolers; it's simply that for categories that have few options (say, CPUs) it's easier to slog through all the youtube videos, and there still exist a few text-based reviewers.

But cases, ram, motherboard AIB versions, GPU AIB versions... even power supplies often - reviews are suprisingly sparse and hard to compare and rarely comprehensive.

The Gamer Nexus reviews are independent and extremely methodical and honest. It's by far the best PC components reviewer on YouTube right now.

“It's all in videos or really low-effort stuff nowadays,”

GN doesn’t really update their website with written reviews.

https://techpowerup.com is still alive.

I can think of a few others, including an offshoot of [H] called thefpsreviews.com but I've seen a lot of crappy, low effort "reviews" just in time for the "video cards are finally cheap" manufactured craze last year, which was last straw for me.

They have degraded significantly over the last few years. Up to around 2015 it was about my #1 reference source. It's how I found out about the fact many GPUs used to triple or quadruple their power consumption just because of an extra monitor being connected, and that AMD's HBM-based GPUs were at that point a surprising exception (when otherwise AMDs GPUs are usually the worst at it). These days however I have to take everything they measure with a grain of salt, and at some point I even had a discussion with a member of staff on the merits of having idle power consumption measurements at all. Something has definitely changed on their side.

The one I really miss is techreport.com

Now, these guys used to deliver some quality stuff, including the big frame pacing thing that finally convinced the masses that no, fps number is not everything. One of the owners joined AMD and yeah, it's been dead for several years.

Community was quite nice, too.

YouTube is almost completely useless for reviews. You can't scroll the video, meaning you have to watch almost all of it to realize they're actually not evaluating the metrics you are interested in. One would think a video would make it harder for a reviewer to just parrot official PR, but alas...

The only time I ever find a YouTube review of interest is when e.g you are actually showing some physical characteristic of the product, which is almost never the case for PC components, but maybe the case in some portable devices (tablet PCs come to mind). Notice that on these types of physical aspect reviews the reviewer himself doesn't appear at all on the video. In most PC components "video reviews" you are simply staring at the face of the reviewer. They could literally do without the video at all (save for a couple of still images where tables are presented) and would still be as informative, if not more. Some could even do without the audio...

yd-dl / youtube-download helps.

Have you actually looked at the GN reviews or are you just assuming this... I agree that text would be nicer to have but man. Also there's chapters now at least, so if you don't care about the analysis you can just jump to the numbers.

> so if you don't care about the analysis you can just jump to the numbers.

I really disagree. See for example https://www.youtube.com/watch?v=VtVowYykviM . Does he measure idle power consumption ? It's impossible to tell from the chapter names. (Spoiler: he doesn't). Also whether he compares the performance of the 7900 sans X with the 7900X at similar power levels (Spoiler: he does, albeit only for 7950X and a minority of the benchmarks.) This is a video which is supposed to be about power efficiency, so it catches my attention; but most of it is about runtime benchmarks with only a couple power/efficiency comparisons early on. And after watching the full video, I cannot think of anything about the visuals that helped made his point clearer. A simple plain page with a couple of tables would have been at least as much helpful, or even more helpful, since it would have been searchable. If you disagree, can you elaborate what did you feel the video/audio provided over a static HTML page ?

No you're right that text would be better, especially when you're looking for specific things. I have a feeling though that we're in the minority with that unfortunately. What's sad, and what I had assumed was still the case, is that GN no longer uploads transcripts to their website.

I had the same gripe learning firebase with pretty great official tutorials, which would have taken half the time if I could just read them.

The automatic transcripts work fairly well which is how I cope with it.

If youtube is not your thing, they also post results on their site. Here's their most recent review:

https://www.gamersnexus.net/hwreviews/3635-fractal-meshify-2...

Look at that set of tests for a case. What more could you want?

The most recent review is from more than 2 years ago ?

Oh dear you're right, I didn't look at the date. I guess you're stuck with youtube then, sorry, forget everything I said.

notebookcheck.net has good written reviews (only the final percentage score is sometimes weird).They do occasionally review desktop hardware too

They are for me the only page remaining who still do decent, methodical reviews on metrics that I find of interest.

For example, idle power consumption.

Can you really blame reviewers? Doing a comprehensive thorough review is hard work that has to meet a very aggressive deadline to maximize utility. Meanwhile youtube channels like LTT are printing money on low quality low effort content. Obviously you'd rather do that.

Some of LTT’s money goes towards the not-yet-launched LTT Labs, which is meant to become a serious test facility.

> Meanwhile youtube channels like LTT are printing money on low quality low effort content.

I find this statement rather snarky and it is imho quite close to being against the HN Guidelines. Anyway I also think it is wrong. Sure not all videos from LTT are high-quality well researched and contain thorough testing but when they are really reviewing a product they put a lot of effort into it. They are even in the process of building a decked out lab to facilitate proper testing.

An example of this is their review of the recently released Ryzen 7000 non-X CPUs. They tested 11 different CPUs in 9 different games and 14 productivity benchmarks and further testing with regards to thermals and power consumption. The whole 18 minute video is basically Linus explaining result charts for these various tests. https://youtube.com/watch?v=CTiRNnSg0jA&t=9s&pp=2AEJkAIB

I encourage you to watch this or any of the other hardware review videos and think about reevaluating your opinion. I don’t want to come across as a fanboy but it think your under a quite outdated impression and I think it’s unfair with regards the investment and change LTT is undergoing as of late.

Despite all the effort, benchrmak results are still not consistent. The other day I wanted to check how Ryzen 7600 compares to i13600k in code compilation. Ratio in time between these 2 processors reported by different authors were wildly different:

LTT: 129% GN: 160% Techpowerup: 105%

With results fluctuating up to 2 generational differences it is hard to make sense, despite all the authors effort. This leaves that kind of content more of entertainment rather than informational and from that perspective low effort doesn't bother me.

Presumably "compilation" is such a broad basket of activities that they each use a different workload. Have you looked at phoronix? They tend to cover that niche a little better, and will potentially have numerous compilation benchmarks.

I find it ironical that Phoronix used to be considered a rather bad reviewer like a decade ago (just Google around!), while nowadays it is probably one of the best. I mean, Phoronix has definitely improved, but not that much.

e.g. they used to be considered clickbaity, but for today's standards you might as well call them "mild".

I'm not blaming anyone; it's just an observation that it's harder to find comprehensive reviews than it used to be.

I mean, stuff like https://www.anandtech.com/bench/CPU-2020/2758 still exists, but I get the impression they no longer have the resources they once did to really fill it the way they used to; and similar things goes for other sites. I think the best option I know about that remains is https://www.techpowerup.com/review/, but they're much less in depth than what used to be available. It's still great to have such a broad set of reviews, mind you, and they're doing a great job. I get the impression there's much less funding for stuff like this, making the last few survivors all the more impressive.

While not a replacement for independent reviews, Noctua provides compatibility lists with rough thermal headroom indicators[0].

They also have compatibility list for motherboards, cases and RAM to see if it will physically fit[1].

[0]https://ncc.noctua.at/cpus

[1]https://ncc.noctua.at/

The ice giant thermosiphon seems to be the state of the art right now from what I’ve seen. To the point where that’s being used by the world’s top overclocker instead of water cooling loops.

If I were building a system I’d just slap one of those in, buy some Noctuas for quieter fan operation and ignore it, and let the CPU turbo to its hearts content.

Edit: link https://www.icegiantcooling.com/

On my 32 core ThreadRipper I’ve been running an Ice Giant since it came out. It’s excellent. The only gotcha is to ensure your case can fit it in.

How big is this thing next to something like NH-D15?

I didn't know about them, so I've checked, IceGiant ProSiphon Elite vs Noctua NH-D15,

https://www.youtube.com/watch?v=a0OQxbMvo80

At min 26:30 it's shown test results, and the performance seems better with Noctua NH-D15. Less noise (less rpm) for the same temperature dissipation.

Noctua's performance with low noise is among the main reasons why I own one since several years ago ( And also why I choose air-cooling over water-cooling). It would be interesting to know if someone has beaten them (At same performance, less weight would be something for to take in account).

The IceGiant ProSiphon has fittings for 4 x 120mm pwm fans. Fitted on my ThreadRipper, I switched the stock IceGiant fans out for Be Quiet Silent Wings. A significant extra cost to be sure. I'm sure some Noctua fans would fit. Best of both worlds.

Any judgment on thermosiphon technology (where noise is largely not a factor) should not be conflated with a judgment on the quality of attached fans. Sold together as a product, though, I suppose it remains relevant.

Remember to compare any noise savings against the total noise. e.g. I have 8 other fans in the box. If memory serves when I built it, tech websites were reporting that the temperature under load with an IceGiant was lower.

World’s top overclockers use liquid nitrogen

The issue is that heatsinks for quiet operation are better off with larger gaps between fins (smaller number of fins per inch, FPI), and Thermal Siphon goes the other way for best operation with powerful, loud fans.

I did a lot of research for a theadripper, including AIO liquid cooling and all the reviews said noctua fans were as just as effective and as quiet for a fraction of the price and complexity.

The U14 or D15 cost about as much as a 360/420 AIO.

Whilst they can have similar performance you aren’t saving on costs especially when you buy a 2nd fan for the main cooler which is another $25 or so if it’s Noctua.

I switched from a Noctua to a 320mm AIO and was not happy. The cooling capacity with the AIO is slightly higher, but it makes much more sound at idle. The pump never stops and unlike a loafing fan it makes a high-Q tone that is clearly audible in a quiet room. That was very disappointing. All the reviews said the AIO is “silent” which would have been easily disproven by objective measurements, even with free apps on a smartphone.

I'm in similar boat; my current desktop has 360mm aio cooling, and it makes me wonder what is peoples baseline if it can in any way be considered quiet. It feels louder than any my past aircooled builds, I definitely expected more from it.

https://www.frostytech.com/ used to be good but I’ve not looked in many years so I’ve no idea if they’re still useful.

For just Noctua, they have a numerical rating scale for all their models, I think.

SilentPCreview?

Noctua always. For best performance on air cooling.

Reply on news.ycombinator.com