Aleksey

Yes, encryption may slowdown read/write speed, but what I observed was something extraordinary. Both drives were encrypted. This probably slows down the read/write speed from 2200 to 2000 Mb/s, but the speed that I consistently observed in my case was like 10-20 times slower. All I can think of is that there is some bug in the Apple encryption software.

I have two external 4Tb SSDs. I was using it to store my memories, like photos, videos, etc. After the purchase both were initialized with Apple data encryption and filled with archive data, approximately 1.4 Tb of media files. One drive I kept in a secure place and the other I used for everyday work and kept adding files. After some time I run a test and realized that the read/write speed was incredibly slow. On a new drive and the one I kept separate, the read/write speed over the Thunderbolt connection was ~2000 Mb/s. But the one that I kept using from time to time, showed me ~200Mb write an ~100Mb read. I tested this using a different, USB 3.2 enclosure, but the speed was the same. The other SSD with the same data kept working as expected and the Thunderbolt enclosure was also working great with another SSD. So I decided the SSD was faulty. After filing for a return with Amazon, I erased the data and just in case, decided to test it one last time. To my surprise the read/write speed returned back to normal! In the 3.2 enclosure, it was less, than in the Thunderbolt, but still very respectable 900 Mb/s write and 1100 Mb/s read. OK, I thought, surely this is because the drive is nearly empty now. So I filled it with data, but nothing changed - the speed was still great. Now I do not understand - can Apple encryption be the cause of this issue? How can this be?

It means that it draws too much power. Just like Windows which does a lot of unnecessary work in the background like it runs on a wall powered PC, is not optimized for laptops. An Apple notebook will run macOS for hours on batteries, because they try to save energy. If you install Windows on the same laptop, it will drain the battery twice or three times as fast.

That's right. Chrome is not laptop optimized. Use Safari on Mac. Personally I like Razer Blade laptops. I was considering buying one, but then choose a MacBook. Razer Blades look pretty much like MacBooks, but are black. But they are more expensive than the Macs.

Yes, this seems to be a better option. I ditched P40 idea. It has no value. I decided to go with 3090. It has 24GB of RAM. Two 3060 may be a bit cheaper but it is slower than one 3090 and it may require some software tweaks.

OMG, that Gigaflops compared to Teraflops. That's bad...

I think you mixed it up with M40. M40 does not support half precision, P40 - does.

Actually if we compare P40 specification and 3060 specification the former is almost 15 times faster when using half precision - 183.7 GFLOPS vs 12.74 GFLOPS. Furthermore - 3060 does not give any performance advantage for half precision. This look too good to be true. I am not even sure if the data is correct.

Why? According to this data it supports half precision. According to this comparison chart, it is indeed slower, but there is a catch - 3060 only has 12GB of RAM, while P40 has 24GB, which is important for running modern AI models. Most LLM do not fit into 12GB.

There are many adapter models that can be 3D printed. E.g. - https://www.yeggi.com/q/nvidia+tesla+p100+p40+k80+by/ An alternative would be to pay Amazon over $1,000/month. I can build an entire computer from used parts for these money.

Thanks. I use Linux so that's OK. Specifically the Tesla P40 cards with 24Gb of RAM seems to be very interesting.

There are some used Nvidia Tesla cards on the market at relatively low price, like $200-$300, some of them seem to have come from China. Are they OK cards? Will they work for machine learning tasks?

It's actually incorrect to say that Instagram does not protect its users. Some time ago I logged in to Instagram, which I rarely use. Even though I logged in on the same device I use for years, it blocked all actions on the site. I could not follow anyone (the main reason I logged in this time), I could not change anything in my account for almost week. Only after approximately a week, I was able to use my account.

BTW, when I was replacing the cooler, I notice that thermal paste was distributed unevenly - there was more of it on the sides of the lid, while most of the central part was contacting the cooler directly. I went out to see if I can delid the CPU for better contact and found that I am not the only one who had this issue with slightly curved lid on an Intel CPU. Here is JayzTwoCents talking about the same problem:

I had replaced the cooler yesterday and decided to check the performance. Before this I did not run tests for a few years. I remember it went up to 100 deg Celsius with no issues. But I used a water cooler. Recently the pump started making noise and I replaced the cooler with "be quiet! Dark Rock 4", which is rated up to 200 W. My CPU TDP is 95 W, but according to the Intel Utility sometimes it goes up to 156W, which is still well within range of the cooler. The cooler itself looks pretty solid on the CPU. I actually checked Cinebench again and I do not see the performance drops. Could it be so that the thermal paste distributed more evenly after having been on for 24 hours?