If this is your first visit, be sure to check out the FAQ by clicking the link above. You may have to register before you can post: click the register link above to proceed. To start viewing messages, select the forum that you want to visit from the selection below. |
|
|
Thread Tools | Rate Thread | Display Modes |
#46
|
|||
|
|||
SSDs/HDDs, memory ... (was: Have hardware prices gone crazyduring Covid?)
On 29/06/2020 05.58, J. P. Gilliver (John) wrote:
On Sun, 28 Jun 2020 at 21:33:26, Mayayana wrote: "J. P. Gilliver (John)" wrote | No, programmers will not eliminate that advantage. On the contrary, | they will make use of it and expect that everybody has an SSD. Same as | | That _is_ eliminating the advantage. If, in a few years' time, softwares | run no faster even if you are using an SSD than they do now if you're | not, then the advantage has been eliminated. Â*I don't think it's really a big factor. The main bloat is wrappers. When you've got something like Java or .Net, or javascript posing as software, you're very far removed from the actual operations. The trouble As a _user_, I am not really that bothered whether it's real or "pretend" software. I know enough (just) about concepts, like assemblers, compilers, linkers, and so on, as well as general programming, to have some idea what people are talking about: but really, I just want it to do it. I know enough to appreciate efficient code like IrfanView. is that things are so fast now that people don't care about doing a good job. They care about easy. An Indeed. They're lazy (or, under pressure to produce results quickly that are good _enough_). Or neither. It is about the business case. The "boss" is not going to pay for reinventing the wheel when there exists libraries or call them whatever that already do everything they need. So they will use ready made frameworks even if huge or not optimized. Saves money, they are tested. -- Cheers, Carlos. |
Ads |
#47
|
|||
|
|||
SSDs/HDDs, memory ... (was: Have hardware prices gone crazyduring Covid?)
On 29/06/2020 03.33, Mayayana wrote:
"J. P. Gilliver (John)" wrote | No, programmers will not eliminate that advantage. On the contrary, | they will make use of it and expect that everybody has an SSD. Same as | | That _is_ eliminating the advantage. If, in a few years' time, softwares | run no faster even if you are using an SSD than they do now if you're | not, then the advantage has been eliminated. I don't think it's really a big factor. The main bloat is wrappers. When you've got something like Java or .Net, or javascript posing as software, you're very far removed from the actual operations. The trouble is that things are so fast now that people don't care about doing a good job. They care about easy. An SSD is really only going to be a factor with intensive disk operations, like moving a lot of files. Maybe not that intensive. Consider a database operation based on disk, where the tools need many thousands of seeks, but only retrieve minimal amounts of data. This type of operation runs hundreds or thousands of times faster on SSD than on rotating rust. Of course, a designer that develops the code on a normal disk detects the problem when testing himself, so he will design the algorithms appropriately to compensate. But a designer that has an SSD himself will not notice the problem, and when the software is distributed will play down the issue. Just buy an SSD, he will say. Yes, I have seen this. -- Cheers, Carlos. |
#48
|
|||
|
|||
SSDs/HDDs, memory ...
On 29/06/2020 10.39, Yousuf Khan wrote:
On 6/29/2020 1:13 AM, Paul wrote: There was the ReFS file system, but it was canceled (not available on the usual desktop SKUs). https://en.wikipedia.org/wiki/ReFS ExFAT is good for NAND flash. I don't know if it's good for the OS feature set. Â*Â*Â* Paul Microsoft should just adopt one of the Linux filesystems, they seem to be very happy promoting Linux nowadays. Well, the attribute/permission set is different, which affects both directions. Maybe more issues? Legal? -- Cheers, Carlos. |
#49
|
|||
|
|||
SSDs/HDDs, memory ... (was: Have hardware prices gone crazy during Covid?)
"J. P. Gilliver (John)" wrote
| I don't think it's really a big factor. The main bloat | is wrappers. When you've got something like Java or | .Net, or javascript posing as software, you're very | far removed from the actual operations. The trouble | | As a _user_, I am not really that bothered whether it's real or | "pretend" software. I know enough (just) about concepts, like | assemblers, compilers, linkers, and so on, as well as general | programming, to have some idea what people are talking about: but | really, I just want it to do it. I know enough to appreciate efficient | code like IrfanView. | I think it's worth understanding. And it's not always bad. Often it's just convenience. Sometimes it's for security, like with phone apps. But the farther one is from the CPU, the slower it will be. An SSD will mostly be noticeable at boot and when moving 2 GB from one partition or disk to another. Example: If you're in IView, or another graphics program, and you want to sharpen a 30 MB image, that will probably be relatively slow. But it has nothing to do with the disk. The image is in memory. The operation is millions of mathematical comparisons. In IView, compiled to native code, those operations are probably direct calls to the CPU itself. If you have a graphics library then you have calls to the library, which calls the CPU. If that library is not in the same process then each call will be slowed. If you have Java or .Net then you have semi-interpreted code, which calls the Win32 libraries, which calls the CPU. If you have javascript in a "web app", one of Microsoft's store apps, or something like that, then you have interpreted code, calling a library, calling Win32, calling the CPU. Every single time. Millions of times over. Interpreted code means it can't make a direct call. Another program reads the instruction and acts on it, like script or a BAT file. Imagine an old style bucket brigade where you have to pass the bucket to an extra, superfluous person. It gets slowed down. With interpreted code, you can't even pass the bucket yourself. You have to ask someone else to do it for you. With every bucket. I don't mean to say that's bad. It provides a lot of convenience and it has its place. The trouble is that the more efficient the hardware gets, the more people figure it's OK to use another wrapper and save themselves some work. And at the same time, the more people can write software without knowing what they're doing. So if you work as a graphic artist and edit photos all day, you want lots of RAM, top quality software (not Java or .Net wrappers. Definitely not Python or javascript.), and a fast CPU. The disk will be all but irrelevant. The operations are not happening on the disk unless you don't have enough RAM to hold the data. | | My concern is that if it happens to C: - which is where an SSD's speed | advantage would show best - it's like cutting the power; cloning such a | disc would I fear have at least some chance of not restoring a working | system. Most of the time, I imagine it'd just come up with a "Windows | did not shut down properly" message and fix itself, granted. | I wouldn't expect a problem, but I haven't faced that scenario so far. | So far I haven't lost one, and given the cost I probably | won't keep them running for their expected life. I'm not | a person who needs 2-6 TB. An $80 500 GB is ample | for my needs, and I mirror everything except the | actual OS on a second disk. | | I'm managing on a 1 TB (HD); I image the C: part of that, and copy the | D: part. I do that, too. I keep disk images of basic C with software installed. I back up to a second disk and to DVD. Once in awhile I also make a current disk image of C. But I try to have it set up in such a way that if it fails I just need to reinstall a basic disk image, copy over app data, do any needed program updates, and be ready to go again. Though it can take a couple of hours to do all of that. Just adjusting prefs and reinstalling extensions in Firefox is a bear of a job. (Though I back up copies of prefs.js and keep copies of all extensions.) |
#50
|
|||
|
|||
SSDs/HDDs, memory ... (was: Have hardware prices gone crazy during Covid?)
"Carlos E.R." wrote
| Consider a database operation based on disk, where the tools need many | thousands of seeks, but only retrieve minimal amounts of data. This type | of operation runs hundreds or thousands of times faster on SSD than on | rotating rust. | Maybe. If the database is too big to be loaded into memory. And if you really need to do that many seeks. That's a lot of maybes. Most people are not doing that. There will always be operations where newer and more expensive hardware can make a difference, but in general, for most uses, people won't see much difference except at boot. Most common desktop software will not be noticeably affected by an SSD, because writing to disk is not the bottleneck. (It still takes 4 seconds for Firefox to wake itself up on my system, with an SSD and fast CPU. Why? It's a mess. It's not because it takes 4 seconds to load the program. But it might be connected to the fact that Mozilla are using a giant C++ wrapper, the CRT "stub API", which seems to be a superfluous wrapper isolating the actual Win32 API. And there's probably also a lot of built-up crud. Then there are the features removed in Pale Moon/New Moon, making those much faster, but which can't be disabled in FF. (Things like parental controls that most people won't use.) In my experience, things that slow down code can be very much unexpected. But generally, any extra steps, external to the EXE, will slow things down. Once it's all loaded into memory an SSD won't help that. |
#51
|
|||
|
|||
SSDs/HDDs, memory ... (was: Have hardware prices gone crazy during Covid?)
In article , Mayayana
wrote: Most common desktop software will not be noticeably affected by an SSD, because writing to disk is not the bottleneck. not true. most software is i/o bound and will benefit from an ssd, as will virtual memory, giving an overall performance improvement to just about everything. the difference in performance with an ssd is *substantial*, especially with nvme, which removes the sata bottleneck. (It still takes 4 seconds for Firefox to wake itself up on my system, with an SSD and fast CPU. Why? It's a mess. It's not because it takes 4 seconds to load the program. something is very clearly a mess if it's taking 4 seconds. it should be *much* less than that. But it might be connected to the fact that Mozilla are using a giant C++ wrapper, the CRT "stub API", which seems to be a superfluous wrapper isolating the actual Win32 API. And there's probably also a lot of built-up crud. Then there are the features removed in Pale Moon/New Moon, making those much faster, but which can't be disabled in FF. (Things like parental controls that most people won't use.) nope. something else is causing the delay. |
#52
|
|||
|
|||
SSDs/HDDs, memory ...
nospam wrote:
it might be connected to the fact that Mozilla are using a giant C++ wrapper, the CRT "stub API", which seems to be a superfluous wrapper isolating the actual Win32 API. It's the sort of approach you need to make software run across multiple platforms ... |
#53
|
|||
|
|||
SSDs/HDDs, memory ...
"Andy Burns" wrote
| it might be connected to the fact that Mozilla are using | a giant C++ wrapper, the CRT "stub API", which seems to | be a superfluous wrapper isolating the actual Win32 API. | | It's the sort of approach you need to make software run across multiple | platforms ... That can sometimes be a factor. In this case, I don't know much about CRT, but it seems to be more of a lockout wrapper. It's Windows C++ runtime, not a cross-platform runtime. It wraps the Win32 API, not a cross-platform API. If they'd written it in an earlier version of Visual Studio, or with non-MS tools, it wouldn't have that limiting bloat. In the case of Java, Python, etc, it's true that one reason it's often used is to make cross-platform software more easily produced. But that's not really an excuse. Like webpage scripting, those things are not truly cross-platform. They just employ similar wrappers to wrap different APIs on different platforms. Python is becoming an epidemic problem in that respect. People want a shortcut to "cross-platform" and want they produce just ends up being bloated script. So it can be convenient, but it's not "needed". Just like you have different kinds of pans to cook different dishes, there are different APIs on each platform. You can cook anything with a microwave oven, but that's not a "cross recipe solution". It's just trading options and flavor in favor of simplicity and speed. |
#54
|
|||
|
|||
SSDs/HDDs, memory ...
Carlos E.R. wrote:
But you have made me curious to find out how SmartMonTools work in Windows. What about that new toolset in Windows 10 that allows to run bash? I don't think there is a /dev/sda1 layer in Bash shell on Windows 10. They provide /mnt/c/users/username/Downloads as an example of a namespace for your home directory. They pretend all the drive letters are mounted. But the device layer is not exposed, so calls to devices should not work. fdisk /dev/sda # should not work, and needs elevation anyway Maybe that will change with WSL2. As if WSL2 is enabled, it's supposed to have something to do with Hyper-V. Paul |
#55
|
|||
|
|||
SSDs/HDDs, memory ...
I need to buy a 3.5" HDD SATA for a ten year old desktop.
I've been meaning to go to Frys for weeks, but I have been sheltering in place more than ever. I rarely buy online, except when it's not in the brick and mortar shops, but maybe it's time to buy my first online HDD. If you know of an online source with a good price for a 1TB or 2TB (or so), let me know! -- Usenet is a polite public potluck of purposefully helpful advice. |
#56
|
|||
|
|||
SSDs/HDDs, memory ... (was: Have hardware prices gone crazyduring Covid?)
On 29/06/2020 14.54, Mayayana wrote:
"Carlos E.R." wrote | Consider a database operation based on disk, where the tools need many | thousands of seeks, but only retrieve minimal amounts of data. This type | of operation runs hundreds or thousands of times faster on SSD than on | rotating rust. | Maybe. If the database is too big to be loaded into memory. It can be. (In the case where I experienced the issue). And if you really need to do that many seeks. That's a lot of maybes. Most people are not doing that. Well, it happened to me, and to thousands of people using that particular software :-) There will always be operations where newer and more expensive hardware can make a difference, but in general, for most uses, people won't see much difference except at boot. Another example. Computer with 8 gigs of ram, using 10 gigs with virtual memory on hard disk. Slow as hell, specially Firefox. Replace disk with SSD, result was very tolerable handling even up to 16 gigs of virtualized memory. -- Cheers, Carlos. |
#57
|
|||
|
|||
SSDs/HDDs, memory ...
On 29/06/2020 18.07, Paul wrote:
Carlos E.R. wrote: But you have made me curious to find out how SmartMonTools work in Windows. What about that new toolset in Windows 10 that allows to run bash? I don't think there is a /dev/sda1 layer in Bash shell on Windows 10. They provide /mnt/c/users/username/Downloads as an example of a namespace for your home directory. They pretend all the drive letters are mounted. But the device layer is not exposed, so calls to devices should not work. Oh. Â*Â* fdisk /dev/sdaÂ*Â*Â*Â*Â* # should not work, and needs elevation anyway Maybe that will change with WSL2. As if WSL2 is enabled, it's supposed to have something to do with Hyper-V. Ah, yes, heard something of it. -- Cheers, Carlos. |
#58
|
|||
|
|||
SSDs/HDDs, memory ...
[Full context kept for reference:]
Carlos E.R. wrote: On 29/06/2020 00.50, Paul wrote: Carlos E.R. wrote: On 28/06/2020 22.16, J. P. Gilliver (John) wrote: On Sun, 28 Jun 2020 at 21:36:43, Carlos E.R. wrote: Maybe. Use SMART. Hmm. Anyone care to comment on that? Especially when I specify _reliably_ (including not false positives)? Sorry, I don't have advice about using SMART under Windows. I use it on Linux. I know you can, but I don't know exactly how. Reliably? Somewhat. Sometimes it fails to warn. Most times it does warn. False positives I have seen none (I suppose you mean warning about an impending failure that doesn't happen). SMART exists in SSD proprietary toolbox softwares. SMART exists in HDTune, but that's only for HDD and not SSD. SMART exists in SmartMonTools (Linux), but is it ported to Cygwin or not ? Devices are exposed in Cygwin as /dev/sda instead of using Windows own namespace for them, so *maybe* it works. When I use the Cygwin version of disktype.exe, the output is not exactly the same as a "native" version on another platform. It is a shame if there is no native Windows tool to at least show SMART data on Windows, but the preferred thing would be some type of timer job and alert. Cygwin no longer works on WinXP. Cygwin no longer receives funding. Volunteers keep it going now. All my (personal) machines that have Windows also have Linux, so I have little need to install Cygwin for my personal use. And on the jobs, if the machine is not administered by me, I can't install it. Result is I know little about Cygwin. But you have made me curious to find out how SmartMonTools work in Windows. What about that new toolset in Windows 10 that allows to run bash? You don't have to keep the entire Cygwin tree on your disk. You can grab the runtime DLLs, plus the executable package and that becomes a "portable Cygwin tool". You can carry those items to another machine and use the executable. For example, this is my portable (runs in Windows) version of disktype. disktype.exe cygwin1.dll cyggcc_s-1.dll And Cygwin keeps web pages about packages. https://cygwin.com/packages/summary/...tools-src.html ** Paul https://www.smartmontools.org/ I don't know the first thing about Smartmontools, but I just followed your URL and the page says: "About Smartmontools ... Sourcecode tarballs and precompiled packages for Darwin (macOS) and Windows are available at the project page at Sourceforge." The 'project page at Sourceforge' link points to https://sourceforge.net/projects/smartmontools/files/smartmontools That page lists all the smartmontools versions and if you click on the newest (7.1), it will bring you to https://sourceforge.net/projects/smartmontools/files/smartmontools/7.1 which lists smartmontools-7.1-1.win32-setup.exe So AFAICT, there *is* a 'native' Windows version of smartmontools and hence no need to go the Cygwin (nor Chocolatey) route. Hope this helps. N.B. Paul, Carlos, other(s), can you check if this is indeed usable/working under Windows? [Rest of context kept:] There seems to be something called "Windows Choco" download, and Cygwin download. https://chocolatey.org/packages/smartmontools To install smartmontools, run the following command from the command line or from PowerShell: choco install smartmontools This is completely new to me. https://cygwin.com/packages/summary/smartmontools.html Installs a Linux like file tree. |
#59
|
|||
|
|||
SSDs/HDDs, memory ...
Arlen Holder wrote:
I need to buy a 3.5" HDD SATA for a ten year old desktop. I've been meaning to go to Frys for weeks, but I have been sheltering in place more than ever. I rarely buy online, except when it's not in the brick and mortar shops, but maybe it's time to buy my first online HDD. If you know of an online source with a good price for a 1TB or 2TB (or so), let me know! At my computer sto 1) They make you wash your hands at the entrance to the store. 2) A sales rep accompanies you around the store. No browsing! You take a list with you, and select items. 3) Only CC at cash. Still, I can do a return on an item within 30 days. And that's more convenient, than doing it with an e-tailer. Part of the disk drive price is the warranty. A company seeking to offer "deals", buys gray market drives from another country, ones without warranty when checked via the web site, and that's how you save money. Anything cheaper than that, could be a refurb. During the flooding that caused production of hard drives to slow, Newegg offered a lot of refurbs for sale. At it's peak, maybe half their items were refurbs. https://arstechnica.com/gadgets/2020...r-hard-drives/ I would think a 1TB drive would be safe. https://www.tomshardware.com/news/wd...smr-techNOLOGY https://www.techspot.com/news/84973-...wing-user.html You would expect such a drive to be about $50. https://www.newegg.com/blue-wd10ezrz...82E16822235014 Paul |
#60
|
|||
|
|||
SSDs/HDDs, memory ...
Frank Slootweg wrote:
I don't know the first thing about Smartmontools, but I just followed your URL and the page says: "About Smartmontools ... Sourcecode tarballs and precompiled packages for Darwin (macOS) and Windows are available at the project page at Sourceforge." The 'project page at Sourceforge' link points to https://sourceforge.net/projects/smartmontools/files/smartmontools That page lists all the smartmontools versions and if you click on the newest (7.1), it will bring you to https://sourceforge.net/projects/smartmontools/files/smartmontools/7.1 which lists smartmontools-7.1-1.win32-setup.exe So AFAICT, there *is* a 'native' Windows version of smartmontools and hence no need to go the Cygwin (nor Chocolatey) route. Hope this helps. N.B. Paul, Carlos, other(s), can you check if this is indeed usable/working under Windows? [Rest of context kept:] Presumably command line ? There is one field in the smartmontools dump output, which "rates" the drive. I don't always agree with the rating, but if you have a fixation on the declarations of others, there's a place it can tell you the drive is "FAILED" or the like :-) What I worry about is when a drive gets a rating of "PASS" or "Cool and Groovy", when it's obvious in the table that the drive has become a bit smelly. It depends on your tolerance for trouble, as to when to pull such a drive out of service. I've pulled drives with perfectly good SMART out of service, when the drive develops a "bad spot", a swath of 50GB of compromised, error-riddled storage which benchmarks at 5MB/sec. This is why I like to read bench the drives, to see if there are any large areas with abnormally low read rates. Using that SmartMonTools rating field, avoids the user/viewer of the output from having to interpret tables of numbers. Where FAILED means "compromised in some way" or "in danger of passing away soon". As SMART is intended to help predict that something bad is about to happen, without knowing how many hours and seconds it might take. I have compromised drives here, that five years later still run, but that doesn't mean they've "earned my respect" over those five years either :-) Those drives stay in the sin bin, and only come out for short-term installs. If a short-term install fails, it's not the end of the world. A few of the other tools, just give you some table of numbers, and leave the interpretation to you. For example, HDTune (ten year old free version) has several yellow marks on brand new drives. These are some sort of bogus interpretation. Now, if you were a believer in the color scheme, you might be panicking at seeing yellow marks on your new drive. When everything gets those yellow marks and the drive is actually fine. That's where the "table interpretation" can kinda drive you nuts - when you first try your hand at it, you'll likely be "scared" by some field which isn't behaving the way it should. Paul |
Thread Tools | |
Display Modes | Rate This Thread |
|
|