A little something interesting... · 3:47pm May 24th, 2020
Hello everypony.
Did you know, when I announced the accelerated schedule for On the Implications of Parallel Worlds three days ago, I didn't tell anyone, but it was actually contingent on something happening?
The thing is, on May 20th, I finished Chap. 25, and one of my editors commented on the length of the backlog, suggested an accelerated schedule. I promised to go on a temporary accelerated schedule when I hit Chap. 30.
It was the 3rd chapter I finished that day. It was not the last, 26 was (albeit technically at 1AM the next day).
So on the 21st, with only Chap. 26 written, I made a prediction... and made the above-mentioned announcement. And not just here, but on Patreon as well.
Yesterday, on the 23rd, that prediction came true. I finished Chapter 30. The hidden condition for the accelerated schedule has been satisfied.
And today, I finished Chap. 31, and reached the end of a sub-arc. In order to continue writing, I'll have to introduce the next plot point.
... That was 9 chapters in just 5 days.
If the coming arc writes with any kind of comparable speed, the accelerated schedule will stick around until it hits a Tuesday on which I have less than 7 pending chapters after publishing.
And something tells me it will. Silversong didn't disappear, after all.
This reminds me. I placed 3 orders on Amazon recently... and they're shipping in 9 packages. Like, really? The first 3 or 4 arrive on Tuesday. Then one on Wednesday. Then the new motherboard for my computer hasn't shipped yet, but is "expected by" Thursday... I don't expect them to make it. All the other components will be arriving in the first wave- including the CPU, cooler, and special SSD my current motherboard doesn't support. At least the memory will be usable without the new motherboard... despite being specially timed for the new AMD CPU, rather than my current Intel CPU.
I really wish the Coronavirus rush didn't slow UPS down so far that their 'Next Day Air' takes 4-5 days to ship. Someone get them more planes/trucks/workers, I need those computer components- especially the motherboard- before my laptop finishes dying on me!
So, you basically bought an entire new computer, with the exception of the case itself. :P
At least, that's what I got out of that.
I really hope that this whole virus fiasco clears up soon. With the casualty count continually dropping and the actual infection numbers increasing as we find out that everyone apparently caught it and never noticed, I hope that the politicians finally see reason...
What was the hidden condition for the speedup, assuming it's not private or a plot point?
5268387
... The hidden condition for the speedup was, as stated in the post, that I reach (and write) Chap. 30. Which happened.
As for the computer, it's not far off. Basically an entire new computer, except the case... and the graphics card. The 1060 I have is already going to be an upgrade from what I had before, on my laptop... But 250GB of hard drive space, SSD or not? Not a chance, I used about 350 of my laptop's 1TB drive before it started giving warning signs. Celeron CPU? Also not a chance, that thing is literally the slowest processor Intel makes. 4 GB of memory? Not a chance- I can't play any games on it, and opening ~15 tabs of Google Docs (and absolutely nothing else) overloads the memory and very nearly causes the system to crash. (And that's with 2GB of page file on an SSD, so effectively 6GB of memory).
You'll notice the motherboard isn't on that list. That's because there's nothing really wrong with the one I have... except that it's designed for an Intel CPU, and the CPU I'm replacing it with is AMD, requiring a different socket type, and so a different motherboard. Oh, and the new motherboard has WiFi built-in, so I don't need to add a WiFi network card to the list.
BUT... The Ryzen 5 (roughly equivalent to the i7 of five years ago, or the latest i5 of today... Didn't need the Ryzen 9 that sweeps the board with Intel's latest/greatest) ships with a stock cooler that exactly matches it in TDP, or Thermal Design Power, the amount of heat it can expect to produce (or, for a cooler, dissipate) at maximum load. While technically enough, the room it's running in can hit 30C (86F) during the summer, despite normal room temperature being 20C (68F) or so. And I want this sucker to run just as fine in a hot room as in a cold room- my brother had his CPU die on him not too long ago, from overheat damage. Max processor temp 95C (203F) my foot. So...
Stock cooler that exactly matches the processor's 65W TDP? Not a chance, I'll upgrade it with a 180W TDP cooler! That CPU will stay below 60C or else! (No, I'm not overclocking it, I don't need THAT much power...)
5268487
I know the feeling with the Celeron and RAM--my old laptop, which died early this year, had a dual-core Pentium and 4GB RAM. It was slow, and that was after upgrading the drive a few years back to a modern SSHD (not quite an SSD, but still better than it had). The thing dated from around 2010 (a Satellite L655-S5078 IIRC), so I got a lot of use out of it, but if it wasn't swapping I was wondering what was wrong with it!
Fairly sure it was a motherboard issue; it would randomly lock up and power down (no, it wasn't a cooling issue; I checked), and finally it never woke up again. Even the charge/power LEDs no longer lit!
Built-in WiFi is one of the best features of a computer. It's great, and saves running Ethernet everywhere. Although there are times I'd like to have the speed.
I agree with you on the cooling issue. I have a stock Dell desktop now that runs rather warm... I'm fairly sure they put the cheapest PSU and heat-sinks in there that they possibly could. They also don't list specs on any of those, and the PSU's a weird shape that makes it hard to find replacements. I hope it holds up well...
5268565
The funny part is, the 7th gen Celeron desktop processor it's running right now is actually comparable to the 7th gen i5 mobile (laptop) processor I was using before.
Yeah, a hybrid drive... Haven't done any research into them, but it seems to me that they're kinda a bit of both worlds: The drawbacks of both the HDD and the SSD, in one package. (That is, HDD's vulnerability to vibration, and SSD's limited lifetime write). Better value, in my opinion, to get an SSD and an HDD and RAID 1 them together, so you get the SSD's speed and resistance to vibration paired with the HDD's unlimited read/write capability. One will fail before the other... replace it, let the RAID rebuild, and boom, you'll never lose your data!
And yeah, that sounds like either a PSU problem or a MB problem. Voltage spikes or drops from the power supply system could have caused the hardware to go into a safety shutdown, just like a bad motherboard chipset (or damaged motherboard PCB) could cause data issues big enough to trigger an 'emergency shutdown'. The charge LED never lighting again suggests (very strongly) that it was the power supply/charge controller that was going out; the battery charge LED is usually pretty direct from that, and motherboards tend to last a lot longer than the other components (by virtue of a much lower load factor). Most importantly, don't take my word for it, I'm not an IT professional.
And you just described the main reason I built my own computer (it's completely custom, I can send you links if you like, about $1.2K total, not including operating system; the PSU is 500W and no longer available, build is ~350W)... instead of buying a prefab for... Hangon. Looking it up, I could get a $900 prefab with the bare minimum- a 1050, i5 9th gen, 16GB, Win10, and a 1TB HDD... and probably no expansion capability. My build might be more expensive, but half the difference is the $150 M.2 drive I don't really need... and the rest is in the $300 top-of-the-line motherboard that'll handle literally anything I throw at it. Add Windows 10, and the PSU, and my build comes out as about $50-$100 more expensive on a per-power basis... Oh wait, the 1060 is $50 more expensive than the 1050 (and there are way newer, more capable cards for about the same price point), so that's $0-$50 difference. And I can upgrade it later with no issue, unlike the OEM stuff...
Building custom is the way to go, if you've got the knowhow.
5268578
Unless I'm missing something, that's not how RAID 1 works. It will only go as fast as the slowest drive (so SSD would match speed to the disk drive) and it will limit the usable size to the smallest drive, although you can get much larger capacity SSD's now, so that may not be a concern. I believe there is software that makes a hybrid driver of separate SSD's and disk drives, but I think that's different from RAID. Hybrid drives use the SSD has an extra large cache to store frequently used files, so it can serve those quicker and increase average speed. IMO, if you want the advantage of fast flash memory, it's better to just use flash memory drives like SSD and NVME and have a regular backup on longer lived disk drives if you're worried.
5268617
... Yes, I got the right RAID number.
And kinda, yes. RAID 1- complete duplication- does limit write speed to the slowest drive, and capacity to the smallest.
However, array read speed is (in theory) the sum of the drive speeds, as any given read request can be serviced by any of the drives on the array, unlike write requests that have to post to all of the drives. (I'm mildly curious if a hardware RAID between an M.2 and a spindle disk will recognize the speed difference between the drives, allow a write to run at the full M.2 speed, and start copying the new data from the M.2 to the spindle once the utilization falls below max... Probably depends on the controller)
Once my components arrive, if it turns out to be slow (as you say), I'll probably find some other way to set it up to be HDD/SSD-failure-resistant... I've had fifteen too many computers die to drive failure (including the laptop I used until now, though it's not fully dead yet- HDD's just giving me warning signs) to expose myself to that again.
Speaking of the laptop, once I get my data off of it, I plan on seeing if I can "restore" it to fully usable state (to be my new "backup"), recognizing that only 30% of its 1TB drive is/was in use, through the use of its recovery partitions...
And yeah, that's the idea: NVME-based SSD for the speed, then a RAID 1 (if it works) to have a "backup" on the HDD... Which, since the 4TB HDD is 4 times the size of the SSD, I'm going to see if the BIOS will let me set up a hardware RAID with the SSD and a 1TB partition of the HDD, freeing the rest for use as a "second" hard drive. I don't expect it to.
Edit after some quick research
Yeah... RAID 1 theoretical max read is the sum of the two, depending on I/O load. Considering it's a slow spindle drive I'd be pairing it with, I'll probably want something that I can be sure will let it lag behind. Either that, or a second NVME drive (of a different manufacturer or model, to get around the "similar drives tend to fail around the same time" issue) for the RAID... I mean, my new MB has 3 M.2 ports, soo...
Yeah, talking to someone I know that actually knows their computers (as different from me), you're right, it'd be immensely slow. New plan: Multiple NVMe drives!
5268637
I can think of ways to make what you were trying work, but niche and complicated solutions have their own problems. The easiest way to get the best of both worlds is just to have a stupidly large RAID array.
I suppose read speeds could be faster depending on the configuration, but I was just going off of the general rule that's been the same for a very long time. RAID 0 for fast and reckless, RAID 1 for slow and secure. Other types of RAID are mostly just some combination of those. I really don't know that much about it aside from basics.
I have 2 NVMe drives on RAID 0 because screw it.
Then again, I've never had an internal drive fail in the 20 or so years. 15 is some really bad luck dude! 
5268578
That might be interesting. I have no need for a new desktop PC (I have one already, and tend to use these things until they break, not seeing the point of too many upgrades) at the moment, but I would like to go for quality next time, and would like to learn how to actually put these things together. I'm not sure how to learn exactly what RAM to use (beyond the types, clock rates, voltage, and DDR version there's a bunch of stuff I don't understand) or how the motherboard/BIOS have to be picked out to work with all the rest of it. Or even that all the connectors for power and peripherals fit...
I'm a bit less interested in the Nvidia graphics, though. I run Ubuntu (may soon switch to something that doesn't have the migraine that is Gnome, though), so OS cost isn't an issue, but I have had a horrible time with the Nvidia drivers. Admittedly, that was on my replacement laptop, so it might also be the soft-muxing they use (the card is used as a coprocessor that pipes data to/from the embedded Intel graphics on most mobile systems) but I'm not sure. They don't release code, you have to log out to change anything, the firmware itself is signed and I believe is still not released for use with Nouveau (which was hamstrung as a result, if you want the open drivers), and I think they never got the thermal monitoring working (again with Nouveau, but the official ones caused an unholy string of errors...) for systems with independent fans (ie, non-mobile cards). Never mind having a CPU go out early, imagine having a graphics card not having the fans running at all!
My ideal system is a quality Intel CPU with a mid-to-top range AMD graphics card (they are fairly good quality, and their Linux drivers suck a LOT less now). That's what my desktop is, but I have still not seen a laptop with it. Admittedly, the card on my desktop is only 2GB, but it still runs most things I play, for when I do that.
5268667
To be fair, of those 15 machines, 14 were laptops with spindle drives. By the nature of a spindle drive, every time I lay my hands on the laptop anything other than carefully while it was doing something, it'd take a hit, and make a blemish on the platters. Too many blemishes, and good bye hard drive. That last one was a small SSD... I'm a programmer, so my programming work very quickly used up the thing's limited lifetime write capacity, and it went bad that way.
If you're not using the machine for drive-heavy tasks like programming and animating and have a sufficiently large SSD, it should last longer than the rest of the computer. SSDs might have a limit for how many times any given sector can be written, but that limit is generally so high you can ignore it (a very large number of multiples of the size of the drive, usually). For someone like me, who does do drive-heavy tasks, I need a spindle disk to take those tasks, because there's no limit to the number of times I can overwrite something on a spindle disk... and those tasks can very quickly eat up the limit of an SSD.
And yeah, I could pull off NVMe-level (sequential) performance by mounting a few dozen HDDs and doing a RAID 0... add a drive, make it a RAID 5 (min. 3 drives), and now it'll survive a drive failure. Add a second (additional) drive, make it RAID 6 (min. 4 drives), it'll survive two drive failures. Then of course, there's the RAID 10 for min. 4 drives, that is like a RAID 0 over the "drives" resulting from a few separate RAID 1 arrays. Of those, 0, 5, and 6 all experience (sequential) performance increases, though the latter two will tend to have more processing overhead involved from calculating XOR parity data on a write, or on a read after a drive failure (which would have a MASSIVE speed penalty, down to the performance of a single drive). 5 and especially 6 will suffer massive slowdowns on random writes, because of the need to read the entire block of data and the entire block of parity data, do processing on both, and write the new parity data. But yeah...
Say I wanted 64GB/s speeds out of a RAID of spindle disks, and number/speed of connecting ports on the motherboard was not an issue. Power supply, and space requirements, likewise non-issues. Considering each spindle disk will have a transfer speed of around 1-200MB/s, I'd need something like 300-600 of them, but I could do it... For a sequential read/write, at least. A random read/write, worst case scenario, would hit the performance of just a single drive.
... This is giving me flashbacks to my blog posts about Ollivander's stock counts and the length of the Hogwarts Express.
5268672
Well, you'd certainly want a board designed for the Intel CPU you wanted (which would be your first choice)... then memory (probably DDR4, possibly something newer if it's out by then) that's timed for an Intel CPU... after that, the power supply connectors are pretty standardized, so you decide what disks and/or expansion cards you want (such as graphics cards)... THEN get a power supply that has the right wattage and the right numbers of the right kinds of power leads. Many expansion cards will work in basically any machine with a big enough PCIe slot, though there are some that only work for some operating systems, among other things.
Yes, there are a LOT more details than that in an actual build, but it's not hard to build something that works. I'm open to helping pick some of the equipment, and helping with assembly (likely through video call or the like), when the time comes.
That said, I would definitely recommend the AMD CPUs. They have more bang for your buck anymore, and are generally more powerful as well... aside from often having greater numbers of cores which, as CPU single-threaded performance crawls closer and closer to the absolute limit of possibility and programs start being more and more multithreaded, will have a much bigger advantage in the years to come. Especially with a game I'm looking at writing, that will be so massively multithreaded we're thinking on using AMD's Threadripper CPUs (128 cores, to the beefiest PC-class processor's 12) for any kind of public game servers.
5268674
Laptops always seemed like a waste of money for my use cases, especially with it being hard to upgrade core components like I do with the desktop.
Yeah, aside from not running drive heavy tasks like you're talking about, I've stuck with desktops, so the drives aren't getting knocked around unless the bass gets heavy.
5268689
Yeah, laptops have always been one-off for me. And the 'bass gets heavy' often won't be enough to "knock around" a good HDD, since it doesn't experience the massive G-forces that cause the read/write head to touch the platters. It supposedly could be, but it often won't- the forces are too small and the frequency too low to cause any kind of damage. Unless you've got that sucker turned up so high it blows out your eardrums in a matter of milliseconds...
Back when I got that laptop, I was a college student, and so found the need to have my main computing platform be a mobile one. Now, though, I'm not a college student, so a static desktop is more than enough.
Honestly, for a $600 laptop (which I then added $200 in memory upgrades to), it lasted a good long time, at 5-6 years. "Average" lifespan of a computer is about four years.
5268387
I don't know where you get your information from.
In the USA, hospitalization and death rates are going up everywhere except the new york/new jersey area because of the massive measures those areas took once they got into serious trouble.
And infection rates going up you say? Yes. But hospitalization/death lags infection by about 2-3 weeks. We're seeing spikes in cases as we open up, and will be seeing massive increases in problems.
We also have the problem that we just don't know the real numbers. We are not testing people enough to know how many people have it (much higher than expected) or what it's spread is (probably high), or what the actual death rate is (many people have died from "pneumonia" when it's really Covid).
The best studies so far have looked at the "excessive" deaths -- how many more deaths have happened this year than normally would be expected. That's the combination of Covid, and things like "I don't want to go to the hospital and risk getting Covid from someone there".
The last time that I am aware of that a massive new infection went into a large population that did not have any immunity was when the European settlers came to America, and wiped out 95% of the native american population from disease alone. Do not take "new disease" lightly. We are actually very lucky that Covid seems to have a fatality rate of a little less than 1%.
And, 1% of 7 billion people world-wide is still 700 million. 1% of 350 million USA is still 35 million.
Even if we manage to get a passable behavior over the next few years, if the vaccines don't work -- and this is the same type of virus that gives things like colds (no vaccines), and has historically proven hard to make vaccines for -- we can still expect 10-15 million dead. Not this year (based on the numbers and projections I've seen, I'm personally expecting 2 million dead by November), but this virus is not going to go away, no more than any of the many viruses out there.
5273641
You do have a good point there. Unfortunately, where I live, there are a bunch of new hospitals that were built that are basically still empty (and now even laying people off since they have no money and no patients), plus regular doctoring and such got canceled entirely in some cases, so there are a bunch of deaths from unrelated crap (including cancer!), and then the hospitals all blamed it on COVID.
There was also a news report somewhere that stated it was really dangerous... because a bunch of people had it and didn't show symptoms. I know what they meant, and there's always the chance that it becomes more lethal while it's out and around, but it's still hard to swallow that it's dangerous if it's so minor that most people had it already and never knew.
5273641
35 million is 10% of 350 million, not 1%. You were looking for 3.5 million, which is 1%.
... Sorry, pet peeve.
Thank you for that correction.
And yes, 50% of the people that have it don't know (again, approximate, estimated. Unless and until we get MASSIVE testing, we don't know. There have been a few places where there has been universal testing, and that's where the 50% estimate comes from, but it's a few small (few hundred) test cases each.) The original estimates were 3% fatality, but that was based on a very low testing rate, and the assumption of "no symptoms = no sickness". We now know that more people have it, and it's less deadly than previously thought, but it's still around 10x as deadly as the normal flu, and has a spread rate several times higher than the normal flu.
Basically, people are concerned that:
1. The risk of catching it on any given outing, given physical distancing and masks, is low. But repeated exposure to low risks has a high failure rate.
2. Even if you are healthy, you can give it to someone that is not healthy. You can kill your parents this way.
3. There is no way to tell for certain if you are safe or not. There's not enough testing, and even if you test clean, you don't know about the people you interact with, and might still catch it.
Bottom line: It's a threat to society as we know it.