Re: [gentoo-user] Re: SDD, what features to look for and what to avoid.
On Saturday, 4 April 2020 10:51:42 BST Peter Humphrey wrote: > On Friday, 3 April 2020 16:14:33 BST Frank Steinmetzger wrote: > > Well, raw throughput is great ’n all, but in real-life you won’t notice > > much difference between a SATA and an NVME drive. > > Not so. The difference is dramatic. For example, when I dep-cleaned gentoo-sources-4.19.97 this morning, it took 11 seconds. That's real write performance. -- Regards, Peter.
Re: [gentoo-user] Re: SDD, what features to look for and what to avoid.
On Saturday, 4 April 2020 11:38:48 BST Alexandru N. Barloiu wrote: > On Sat, 2020-04-04 at 10:51 +0100, Peter Humphrey wrote: > Also, you don't need a fancy gui for nvme temperature. nvme smart-log > /dev/nvme0 will tell you the temperature from console. Thanks for the pointer. # nvme smart-log /dev/nvme0n1 Smart Log for NVME device:nvme0n1 namespace-id: critical_warning: 0 temperature : 49 C available_spare : 100% available_spare_threshold : 10% percentage_used : 31% [...] Interesting. -- Regards, Peter.
Re: [gentoo-user] Re: SDD, what features to look for and what to avoid.
On Sat, 2020-04-04 at 10:51 +0100, Peter Humphrey wrote: > On Friday, 3 April 2020 16:14:33 BST Frank Steinmetzger wrote: > > > Well, raw throughput is great ’n all, but in real-life you won’t > > notice much > > difference between a SATA and an NVME drive. > > Not so. The difference is dramatic. > > > The bottleneck quickly becomes > > the CPU again during boot or loading more complex applications > > (browser, > > office). The biggest improvement in those situation comes from the > > fast > > “seeking” and reading of many small files. HDDs are at a big > > disatvantage > > here due to their moving head and mechanical seeking. > > > > In fact I doubt you have many use cases for reading many gigabytes > > at a time > > over and over again every day without much CPU overhead, like video > > editing > > (loading previews in 4K or 8K), copying, archiving, checksumming > > and so on. > > > > Due to their immense speed, those NVMEs also tend to heat up quite > > a bit > > under load, eventually leading to throttling. So from a practical > > POV, and > > since you’re on a budget, I suggest cutting cost by staying with > > SATA. > > I can't say anything about temperature, because gkrellm can't see a > sensor of > it, but I certainly wouldn't go back to plain old SSDs. > sata was a major bottleneck. in most motherboards it was embeded in the southbridge WITH lan, sound, usb. When you saturate the sata controller, all others stop working, or lag terribly. That's why swap on sata systems only exacerbates the problem. nvme drives have been moved up in northbridge with cpu, memory and gpu. The bandwidth is superhuge now. Not only you get amazing speeds with one thread, but regardless of how much disk io you are doing, the computer is not lagging anymore. problem with heating is the way it is mounted. M2 drives sit too close to the motherboard, pciexpress drives are usually mounted on the bottom half of the case because of the way pciexpress ports are layed out. port one for gpu. last port from bottom, nvme. you could just put a fan on it. Also, you don't need a fancy gui for nvme temperature. nvme smart-log /dev/nvme0 will tell you the temperature from console. axl
Re: [gentoo-user] Re: SDD, what features to look for and what to avoid.
On Friday, 3 April 2020 16:14:33 BST Frank Steinmetzger wrote: > Well, raw throughput is great ’n all, but in real-life you won’t notice much > difference between a SATA and an NVME drive. Not so. The difference is dramatic. > The bottleneck quickly becomes > the CPU again during boot or loading more complex applications (browser, > office). The biggest improvement in those situation comes from the fast > “seeking” and reading of many small files. HDDs are at a big disatvantage > here due to their moving head and mechanical seeking. > > In fact I doubt you have many use cases for reading many gigabytes at a time > over and over again every day without much CPU overhead, like video editing > (loading previews in 4K or 8K), copying, archiving, checksumming and so on. > > Due to their immense speed, those NVMEs also tend to heat up quite a bit > under load, eventually leading to throttling. So from a practical POV, and > since you’re on a budget, I suggest cutting cost by staying with SATA. I can't say anything about temperature, because gkrellm can't see a sensor of it, but I certainly wouldn't go back to plain old SSDs. -- Regards, Peter.
Re: [gentoo-user] Re: SDD, what features to look for and what to avoid.
Frank Steinmetzger wrote: > On Thu, Apr 02, 2020 at 09:45:58AM -0500, Dale wrote: >> Grant Edwards wrote: >>> On 2020-04-02, Dale wrote: >>> Oooo. That nvme speed is fss. Do you happen to have the OS on that and if so, just how fast does it go from BIOS or Grub to bootup complete? I'm almost scared to ask. o_O >>> I've been wondering if older fairly generic motherboards (7-8 years >>> old) from the likes of Asrock would be able to boot from an NVMe card >>> using a PCIe adapter like this: >>> >>> https://www.amazon.com/QNINE-Adapter-Express-Controller-Expansion/dp/B075MDH28Y >>> >>> I suspect not... >>> >>> -- >>> Grant >> I have a Gigabyte 970 that is only a few years old and it doesn't >> support it. I wish it did. > Well, raw throughput is great ’n all, but in real-life you won’t notice much > difference between a SATA and an NVME drive. The bottleneck quickly becomes > the CPU again during boot or loading more complex applications (browser, > office). The biggest improvement in those situation comes from the fast > “seeking” and reading of many small files. HDDs are at a big disatvantage > here due to their moving head and mechanical seeking. > > In fact I doubt you have many use cases for reading many gigabytes at a time > over and over again every day without much CPU overhead, like video editing > (loading previews in 4K or 8K), copying, archiving, checksumming and so on. > > Due to their immense speed, those NVMEs also tend to heat up quite a bit > under load, eventually leading to throttling. So from a practical POV, and > since you’re on a budget, I suggest cutting cost by staying with SATA. > Yea, I don't see me getting a nvme anyway. They kind of pricey. While fast, I have more time than I do money. I do try to keep my rig in a good place as far as age tho. I may try to upgrade my mobo in a couple years but stick with my current CPU and memory. Mobos do go bad with age, I've read about caps blowing their top and stinking up a room quite well. While I don't edit videos, I do have several terabytes of video. I mostly just play them tho and they play fine with no skipping or anything so what I got is plenty fast for that. I mostly just want the OS itself on something faster but also newer since that HDD I have now has some age on it. I still find the nvmes interestingly fast. Wow!! Dale :-) :-) P. S. My top fan in my Cooler Master HAF-932 case got stuck the other day. It stopped spinning and gkrellm was kind enough to let me know that. I took it out, oiled it good and it works fine again. I also oiled the side fan and the front fan. I got some high dollar gun oil I use for those. It's super slick, handles a wide range of temps and lasts for ages. I've used it in several fans. Even my old CPU cooler fan still runs. I replaced it over a year ago. I use the old one to dry counter tops, dishes and cool batteries I'm charging.
Re: [gentoo-user] Re: SDD, what features to look for and what to avoid.
On Thu, Apr 02, 2020 at 09:45:58AM -0500, Dale wrote: > Grant Edwards wrote: > > On 2020-04-02, Dale wrote: > > > >> Oooo. That nvme speed is fss. > >> Do you happen to have the OS on that and if so, just how fast does it go > >> from BIOS or Grub to bootup complete? I'm almost scared to ask. o_O > > I've been wondering if older fairly generic motherboards (7-8 years > > old) from the likes of Asrock would be able to boot from an NVMe card > > using a PCIe adapter like this: > > > > https://www.amazon.com/QNINE-Adapter-Express-Controller-Expansion/dp/B075MDH28Y > > > > I suspect not... > > > > -- > > Grant > > I have a Gigabyte 970 that is only a few years old and it doesn't > support it. I wish it did. Well, raw throughput is great ’n all, but in real-life you won’t notice much difference between a SATA and an NVME drive. The bottleneck quickly becomes the CPU again during boot or loading more complex applications (browser, office). The biggest improvement in those situation comes from the fast “seeking” and reading of many small files. HDDs are at a big disatvantage here due to their moving head and mechanical seeking. In fact I doubt you have many use cases for reading many gigabytes at a time over and over again every day without much CPU overhead, like video editing (loading previews in 4K or 8K), copying, archiving, checksumming and so on. Due to their immense speed, those NVMEs also tend to heat up quite a bit under load, eventually leading to throttling. So from a practical POV, and since you’re on a budget, I suggest cutting cost by staying with SATA. -- Gruß | Greetings | Qapla’ Please do not share anything from, with or about me on any social network. “He doesn’t know how to use the three seashells!” signature.asc Description: PGP signature
Re: [gentoo-user] Re: SDD, what features to look for and what to avoid.
On Thu, Apr 2, 2020 at 10:28 AM Grant Edwards wrote: > > On 2020-04-02, Mark Knecht wrote: > > On Thu, Apr 2, 2020 at 7:29 AM Grant Edwards wrote: > >> On 2020-04-02, Dale wrote: > >> > >> I've been wondering if older fairly generic motherboards (7-8 years > >> old) from the likes of Asrock would be able to boot from an NVMe card > >> using a PCIe adapter like this: > >> > >> > > https://www.amazon.com/QNINE-Adapter-Express-Controller-Expansion/dp/B075MDH28Y > >> > >> I suspect not... > > > The questions section says it will only boot from the card if the BIOS > > supports the MB booting from NVMe drives > > Obvioiusly, boards can only boot from devices that are supported by > the BIOS. The question is when did NVMe support in BIOSes become > common? > > -- > Grant Common? Don't know. Possible? Since probably 2011 with the right UEFI. If your Asrock has UEFI/BIOS updates in the last couple of years I'd guess you have a good chance. Asrock gives you a place to ask questions: https://www.asrock.com/support/index.us.asp Give them a try. Mark
[gentoo-user] Re: SDD, what features to look for and what to avoid.
On 2020-04-02, Mark Knecht wrote: > On Thu, Apr 2, 2020 at 7:29 AM Grant Edwards > wrote: >> On 2020-04-02, Dale wrote: >> >> I've been wondering if older fairly generic motherboards (7-8 years >> old) from the likes of Asrock would be able to boot from an NVMe card >> using a PCIe adapter like this: >> >> > https://www.amazon.com/QNINE-Adapter-Express-Controller-Expansion/dp/B075MDH28Y >> >> I suspect not... > The questions section says it will only boot from the card if the BIOS > supports the MB booting from NVMe drives Obvioiusly, boards can only boot from devices that are supported by the BIOS. The question is when did NVMe support in BIOSes become common? -- Grant
Re: [gentoo-user] Re: SDD, what features to look for and what to avoid.
On Thu, Apr 2, 2020 at 7:29 AM Grant Edwards wrote: > > On 2020-04-02, Dale wrote: > > > Oooo. That nvme speed is fss. > > Do you happen to have the OS on that and if so, just how fast does it go > > from BIOS or Grub to bootup complete? I'm almost scared to ask. o_O > > I've been wondering if older fairly generic motherboards (7-8 years > old) from the likes of Asrock would be able to boot from an NVMe card > using a PCIe adapter like this: > > https://www.amazon.com/QNINE-Adapter-Express-Controller-Expansion/dp/B075MDH28Y > > I suspect not... > > -- > Grant > The questions section says it will only boot from the card if the BIOS supports the MB booting from NVMe drives
Re: [gentoo-user] Re: SDD, what features to look for and what to avoid.
On Thursday, 2 April 2020 03:57:00 BST William Kenworthy wrote: > mfsmaster ~ # hdparm -Tt /dev/nvme0n1 > > /dev/nvme0n1: > Timing cached reads: 8524 MB in 1.99 seconds = 4283.31 MB/sec > Timing buffered disk reads: 4252 MB in 3.00 seconds = 1416.93 MB/sec > mfsmaster ~ # > > > Samsung 970 plus NVME M.2 on an odroid H2 /dev/nvme0n1: Timing cached reads: 18158 MB in 1.99 seconds = 9124.28 MB/sec Timing buffered disk reads: 5262 MB in 3.00 seconds = 1753.46 MB/sec That's a "Samsung SM951 - NVMe Native Next Gen PCIe3 x4 M.2 256GB SSD", quoting from the invoice issued by Armari[1], the system builder. It's four years old, so it isn't even the latest whiz-bang toy. The motherboard is an Asus X99-A. The whole Gentoo system is on it, apart from /tmp, /var/tmp/portage and /home/ prh/boinc. Boinc runs 24/7/52 on 12 threads + Radeon GPU. 1. Armari build high-powered workstations for the London city financial markets, where fractions of a second count. Some serious iron. -- Regards, Peter.
Re: [gentoo-user] Re: SDD, what features to look for and what to avoid.
Grant Edwards wrote: > On 2020-04-02, Dale wrote: > >> Oooo. That nvme speed is fss. >> Do you happen to have the OS on that and if so, just how fast does it go >> from BIOS or Grub to bootup complete? I'm almost scared to ask. o_O > I've been wondering if older fairly generic motherboards (7-8 years > old) from the likes of Asrock would be able to boot from an NVMe card > using a PCIe adapter like this: > > https://www.amazon.com/QNINE-Adapter-Express-Controller-Expansion/dp/B075MDH28Y > > I suspect not... > > -- > Grant I have a Gigabyte 970 that is only a few years old and it doesn't support it. I wish it did. I'm not sure when mobos started supporting them tho and it may vary from one manufacturer to another. I googled my model number and nvme to find a site that says it isn't supported with mine. You may can do the same. Dale :-) :-)
[gentoo-user] Re: SDD, what features to look for and what to avoid.
On 2020-04-02, Dale wrote: > Oooo. That nvme speed is fss. > Do you happen to have the OS on that and if so, just how fast does it go > from BIOS or Grub to bootup complete? I'm almost scared to ask. o_O I've been wondering if older fairly generic motherboards (7-8 years old) from the likes of Asrock would be able to boot from an NVMe card using a PCIe adapter like this: https://www.amazon.com/QNINE-Adapter-Express-Controller-Expansion/dp/B075MDH28Y I suspect not... -- Grant
Re: [gentoo-user] Re: SDD, what features to look for and what to avoid.
On 02/04/2020 13:37, J. Roeleveld wrote: Same here, the colour cartridges have been saying they're "critically low" for the past couple of months. As they don't expire, I did order a new set when they got low. Those are still sealed in storage. Yup. I ordered a set when they hit critically low. And replaced the yellow when it started looking naff. The other colours still look fine. The printer has been registered for its "free" 3-yr warranty, but it's a requirement that you must use only genuine HP cartridges. So I'm reckoning that once this first set of replacement cartridges runs out, the printer will be over three years old. The second set of cartridges will be compatibiles, and if I pay £20 instead of £100 per cartridge, that'll save me enough to buy a new printer if the old one breaks. I think the set of four XL cartridges cost more than the printer did! Printers don't seem to last nowadays, anyway. I remember somebody asking about "old printers", and there were no "medium age" printers - they were all either less than five years old, or ancient LaserJet 4 era printers. Anything in-between had died (or, like one of my old printers, been sunk by production of consumables stopping). Cheers, Wol
Re: [gentoo-user] Re: SDD, what features to look for and what to avoid.
J. Roeleveld wrote: > On Thursday, April 2, 2020 12:49:11 AM CEST Dale wrote: >> antlists wrote: >>> On 01/04/2020 22:46, Dale wrote: I still haven't bought it yet. I ordered some toner cartridges a while back for my printer. The site said that the ones I ordered fits my printer. Well, it appears they found out that was a error because they removed that page and relisted it but did not include my printer model. So, I had to order a whole new set, at about $100.00 each for high yield. Needless to say, I'll be paying on that for a while. I'll try to sell the wrong ones later. I only opened one color so the others are still sealed. >>> I know you're not UK, but under European rules you'd be able to return >>> that opened cartridge at their expense - "not as described" and it's >>> their problem even if you had to open it to find out. >>> BTW, next time I'll find a printer that allows refilling and such too. I don't like that chip thing. It counts against my page count on color even if I print a black and white page. Still, printer does a awesome job. Beats those ink jet thingys by a country mile. >>> My HP laser is like that. Mind you, my colour cartridges "ran out" at >>> about 1900 pages. They're still going fine (well, the yellow got >>> replaced) at 4000 pages. But HP do say that a cartridge running empty >>> won't damage the printer - all the vulnerable parts are in the cartridge. >>> >>> Cheers, >>> Wol >> I did try to contact them but have not heard back, either way. They may >> be closed or something tho. No idea. Some businesses are closed due to >> the bug thingy. It may be that later on they allow me to return it. >> >> From my understanding, when it reaches zero pages, it stops that >> cartridge from working. In other words, even if there is toner in the >> cartridge, it won't feed it into the printer. It may be a Lexmark >> thing. HP may do it differently. I looked to see if I could find a >> replacement chip and some toner but not for my printer, yet anyway. >> Maybe in a year or so. > I use HP and don't have that issue. > Otherwise it would have stopped printing months ago :) > > -- > Joost > I'm surprised someone hasn't hacked the firmware in order to bypass all that page count crap. I wouldn't be surprised if it was some Linux type OS either. That should make it easier. It seems the printer is fairly new. Maybe by the time these new cartridges get low, something will have changed. Either a way to disable page counts, cheap replacement ships and toner refill or something else I'm not aware of yet. I may keep a eye out on a HP or other brand of printer. One thing tho, it will be duplex. I love printing on both sides. :-D Dale :-) :-)
Re: [gentoo-user] Re: SDD, what features to look for and what to avoid.
On Thursday, April 2, 2020 3:31:48 AM CEST Adam Carter wrote: > > A simple read test with hdparm -t reveals: > > 500 GB Crucial from 2016: Timing buffered disk reads: 1596 MB in 3.00 > > seconds = 531.46 MB/sec > > 128 GB Sandisk from 2014: Timing buffered disk reads: 1532 MB in 3.00 > > seconds = 510.60 MB/sec > > 120 GB Sandisk from 2017: Timing buffered disk reads: 968 MB in 3.00 > > seconds = 322.42 MB/sec > > FYI some other hdparm results in case anyone's interested. > > Host Disk Cached Reads Buffered Reads > FX-8350 nvme 512G Samsung 950 (951?) 4793 1435 this seems low, my 950pro (256GB) does: Timing cached reads: 21216 MB in 1.99 seconds = 10660.10 MB/sec Timing buffered disk reads: 5444 MB in 3.00 seconds = 1813.68 MB/sec (On an Asus Sabertooth X99 mainboard, X99 chipset with i7-5820K CPU) I though the 256GB was supposed to be slower than the 512GB version. Not the other way round. -- Joost
Re: [gentoo-user] Re: SDD, what features to look for and what to avoid.
On Thursday, April 2, 2020 12:49:11 AM CEST Dale wrote: > antlists wrote: > > On 01/04/2020 22:46, Dale wrote: > >> I still haven't bought it yet. I ordered some toner cartridges a while > >> back for my printer. The site said that the ones I ordered fits my > >> printer. Well, it appears they found out that was a error because they > >> removed that page and relisted it but did not include my printer model. > >> So, I had to order a whole new set, at about $100.00 each for high > >> yield. Needless to say, I'll be paying on that for a while. I'll try > >> to sell the wrong ones later. I only opened one color so the others are > >> still sealed. > > > > I know you're not UK, but under European rules you'd be able to return > > that opened cartridge at their expense - "not as described" and it's > > their problem even if you had to open it to find out. > > > >> BTW, next time I'll find a printer that allows refilling and such too. > >> I don't like that chip thing. It counts against my page count on color > >> even if I print a black and white page. Still, printer does a awesome > >> job. Beats those ink jet thingys by a country mile. > > > > My HP laser is like that. Mind you, my colour cartridges "ran out" at > > about 1900 pages. They're still going fine (well, the yellow got > > replaced) at 4000 pages. But HP do say that a cartridge running empty > > won't damage the printer - all the vulnerable parts are in the cartridge. > > > > Cheers, > > Wol > > I did try to contact them but have not heard back, either way. They may > be closed or something tho. No idea. Some businesses are closed due to > the bug thingy. It may be that later on they allow me to return it. > > From my understanding, when it reaches zero pages, it stops that > cartridge from working. In other words, even if there is toner in the > cartridge, it won't feed it into the printer. It may be a Lexmark > thing. HP may do it differently. I looked to see if I could find a > replacement chip and some toner but not for my printer, yet anyway. > Maybe in a year or so. I use HP and don't have that issue. Otherwise it would have stopped printing months ago :) -- Joost
Re: [gentoo-user] Re: SDD, what features to look for and what to avoid.
On Thursday, April 2, 2020 12:13:38 AM CEST antlists wrote: > On 01/04/2020 22:46, Dale wrote: > > BTW, next time I'll find a printer that allows refilling and such too. > > I don't like that chip thing. It counts against my page count on color > > even if I print a black and white page. Still, printer does a awesome > > job. Beats those ink jet thingys by a country mile. > > My HP laser is like that. Mind you, my colour cartridges "ran out" at > about 1900 pages. They're still going fine (well, the yellow got > replaced) at 4000 pages. But HP do say that a cartridge running empty > won't damage the printer - all the vulnerable parts are in the cartridge. Same here, the colour cartridges have been saying they're "critically low" for the past couple of months. As they don't expire, I did order a new set when they got low. Those are still sealed in storage. -- Joost
Re: [gentoo-user] Re: SDD, what features to look for and what to avoid.
> > 3900X@2133 nvme 512G Samsung 970 EVO PLUS 11904 2858 > 3900X@3200 nvme 512G Samsung 970 EVO PLUS 15213 3032 > > Oooo. That nvme speed is fss. Do > you happen to have the OS on that and if so, just how fast does it go from > BIOS or Grub to bootup complete? I'm almost scared to ask. > Yep, everything's on the nvme except /var. From grub to console login prompt is about 20 seconds, but it waits quite a while for the network service to come up, perhaps because its using a bonded interface. Booting is quick, but I really notice the speed when booting a win10 vm in virtualbox, which takes about 15 seconds. The system was delivered with the RAM running at 2133, so the auto speed detection hadnt worked. Hardcoding it to the spec speed of 3200 took the cache reads from 11.9Gig 15.2Gig...
Re: [gentoo-user] Re: SDD, what features to look for and what to avoid.
On 2/4/20 10:05 am, Dale wrote: Adam Carter wrote: A simple read test with hdparm -t reveals: 500 GB Crucial from 2016: Timing buffered disk reads: 1596 MB in 3.00 seconds = 531.46 MB/sec 128 GB Sandisk from 2014: Timing buffered disk reads: 1532 MB in 3.00 seconds = 510.60 MB/sec 120 GB Sandisk from 2017: Timing buffered disk reads: 968 MB in 3.00 seconds = 322.42 MB/sec FYI some other hdparm results in case anyone's interested. Host Disk Cached Reads Buffered Reads Phenom965 6T wd red WDC WD60EFRX-68L 3785 147 Phenom965 8T wd red WDC WD80EFAX-68K 3716 190 Phenom965 250G Samsung SSD 840 (sata) 3757 60 FX-8350 nvme 512G Samsung 950 (951?) 4793 1435 FX-8350 1T WDC WD10EARX-00N 3883 118 FX-8350 3T WDC WD30EFRX-68A 4422 111 3900X@2133 3T WDC WD30EFRX-68A 11623 143 3900X@3200 3T WDC WD30EFRX-68A 15225 139 3900X@2133 nvme 512G Samsung 970 EVO PLUS 11904 2858 3900X@3200 nvme 512G Samsung 970 EVO PLUS 15213 3032 i3-6100U INTEL SSDSCKJW12 128G M2 sata 3541 448 Oooo. That nvme speed is fss. Do you happen to have the OS on that and if so, just how fast does it go from BIOS or Grub to bootup complete? I'm almost scared to ask. o_O Dale :-) :-) mfsmaster ~ # hdparm -Tt /dev/nvme0n1 /dev/nvme0n1: Timing cached reads: 8524 MB in 1.99 seconds = 4283.31 MB/sec Timing buffered disk reads: 4252 MB in 3.00 seconds = 1416.93 MB/sec mfsmaster ~ # Samsung 970 plus NVME M.2 on an odroid H2 :) BillK
Re: [gentoo-user] Re: SDD, what features to look for and what to avoid.
Adam Carter wrote: > > A simple read test with hdparm -t reveals: > 500 GB Crucial from 2016: Timing buffered disk reads: 1596 MB in > 3.00 seconds = 531.46 MB/sec > 128 GB Sandisk from 2014: Timing buffered disk reads: 1532 MB in > 3.00 seconds = 510.60 MB/sec > 120 GB Sandisk from 2017: Timing buffered disk reads: 968 MB in > 3.00 seconds = 322.42 MB/sec > > > FYI some other hdparm results in case anyone's interested. > > Host Disk Cached Reads Buffered Reads > Phenom965 6T wd red WDC WD60EFRX-68L 3785 147 > Phenom965 8T wd red WDC WD80EFAX-68K 3716 190 > Phenom965 250G Samsung SSD 840 (sata) 3757 60 > FX-8350 nvme 512G Samsung 950 (951?) 4793 1435 > FX-8350 1T WDC WD10EARX-00N 3883 118 > FX-8350 3T WDC WD30EFRX-68A 4422 111 > 3900X@2133 3T WDC WD30EFRX-68A 11623 143 > 3900X@3200 3T WDC WD30EFRX-68A 15225 139 > 3900X@2133 nvme 512G Samsung 970 EVO PLUS 11904 2858 > 3900X@3200 nvme 512G Samsung 970 EVO PLUS 15213 3032 > i3-6100U INTEL SSDSCKJW12 128G M2 sata 3541 448 Oooo. That nvme speed is fss. Do you happen to have the OS on that and if so, just how fast does it go from BIOS or Grub to bootup complete? I'm almost scared to ask. o_O Dale :-) :-)
Re: [gentoo-user] Re: SDD, what features to look for and what to avoid.
> > A simple read test with hdparm -t reveals: > 500 GB Crucial from 2016: Timing buffered disk reads: 1596 MB in 3.00 > seconds = 531.46 MB/sec > 128 GB Sandisk from 2014: Timing buffered disk reads: 1532 MB in 3.00 > seconds = 510.60 MB/sec > 120 GB Sandisk from 2017: Timing buffered disk reads: 968 MB in 3.00 > seconds = 322.42 MB/sec > > FYI some other hdparm results in case anyone's interested. Host Disk Cached Reads Buffered Reads Phenom965 6T wd red WDC WD60EFRX-68L 3785 147 Phenom965 8T wd red WDC WD80EFAX-68K 3716 190 Phenom965 250G Samsung SSD 840 (sata) 3757 60 FX-8350 nvme 512G Samsung 950 (951?) 4793 1435 FX-8350 1T WDC WD10EARX-00N 3883 118 FX-8350 3T WDC WD30EFRX-68A 4422 111 3900X@2133 3T WDC WD30EFRX-68A 11623 143 3900X@3200 3T WDC WD30EFRX-68A 15225 139 3900X@2133 nvme 512G Samsung 970 EVO PLUS 11904 2858 3900X@3200 nvme 512G Samsung 970 EVO PLUS 15213 3032 i3-6100U INTEL SSDSCKJW12 128G M2 sata 3541 448
Re: [gentoo-user] Re: SDD, what features to look for and what to avoid.
antlists wrote: > On 01/04/2020 22:46, Dale wrote: >> I still haven't bought it yet. I ordered some toner cartridges a while >> back for my printer. The site said that the ones I ordered fits my >> printer. Well, it appears they found out that was a error because they >> removed that page and relisted it but did not include my printer model. >> So, I had to order a whole new set, at about $100.00 each for high >> yield. Needless to say, I'll be paying on that for a while. I'll try >> to sell the wrong ones later. I only opened one color so the others are >> still sealed. > > I know you're not UK, but under European rules you'd be able to return > that opened cartridge at their expense - "not as described" and it's > their problem even if you had to open it to find out. >> >> BTW, next time I'll find a printer that allows refilling and such too. >> I don't like that chip thing. It counts against my page count on color >> even if I print a black and white page. Still, printer does a awesome >> job. Beats those ink jet thingys by a country mile. > > My HP laser is like that. Mind you, my colour cartridges "ran out" at > about 1900 pages. They're still going fine (well, the yellow got > replaced) at 4000 pages. But HP do say that a cartridge running empty > won't damage the printer - all the vulnerable parts are in the cartridge. > > Cheers, > Wol > > I did try to contact them but have not heard back, either way. They may be closed or something tho. No idea. Some businesses are closed due to the bug thingy. It may be that later on they allow me to return it. >From my understanding, when it reaches zero pages, it stops that cartridge from working. In other words, even if there is toner in the cartridge, it won't feed it into the printer. It may be a Lexmark thing. HP may do it differently. I looked to see if I could find a replacement chip and some toner but not for my printer, yet anyway. Maybe in a year or so. Dale :-) :-)
Re: [gentoo-user] Re: SDD, what features to look for and what to avoid.
On 01/04/2020 22:46, Dale wrote: I still haven't bought it yet. I ordered some toner cartridges a while back for my printer. The site said that the ones I ordered fits my printer. Well, it appears they found out that was a error because they removed that page and relisted it but did not include my printer model. So, I had to order a whole new set, at about $100.00 each for high yield. Needless to say, I'll be paying on that for a while. I'll try to sell the wrong ones later. I only opened one color so the others are still sealed. I know you're not UK, but under European rules you'd be able to return that opened cartridge at their expense - "not as described" and it's their problem even if you had to open it to find out. BTW, next time I'll find a printer that allows refilling and such too. I don't like that chip thing. It counts against my page count on color even if I print a black and white page. Still, printer does a awesome job. Beats those ink jet thingys by a country mile. My HP laser is like that. Mind you, my colour cartridges "ran out" at about 1900 pages. They're still going fine (well, the yellow got replaced) at 4000 pages. But HP do say that a cartridge running empty won't damage the printer - all the vulnerable parts are in the cartridge. Cheers, Wol
Re: [gentoo-user] Re: SDD, what features to look for and what to avoid.
On Wednesday, 1 April 2020 22:25:08 BST Frank Steinmetzger wrote: > I’ve been using it for 5 years for a dual-boot system (2×64 G for Windows > and Gentoo). By the end of last year it became slower and slower when > writing. Especially eix-update became slw. My main suspicion is that it > was quite full and there probably is no overprovisioning for wear-leveling > built into the drive. So it was writing the same cells over and over when I > did my world updates. With SSDs it is important you overprovision your partition(s) in terms of individual partition size so they don't fill up. I tend to always leave a little empty space on an SSD device when I create partitions, which may be unnecessary these days, because the OEMs tend to do the same themselves: https://www.kingston.com/us/ssd/overprovisioning signature.asc Description: This is a digitally signed message part.
Re: [gentoo-user] Re: SDD, what features to look for and what to avoid.
Frank Steinmetzger wrote: > On Tue, Mar 17, 2020 at 02:22:36PM -0500, Dale wrote: > >> Thanks for the additional info. As I figured, they got most of the >> kinks worked out by now and we got some dependable SSDs to buy. >> >> I found a 240GB for a little over $42.00 USA. Not bad at all. For >> those curious: > Careful, you get what you pay for. > > When I built my first PC from scratch (had laptops before that), I bought a > 128 GB Sandisk SDSSDP128G. I didn’t have a lot of money back then, it was > relatively cheap, but not the very cheapest. It doesn’t even have real > branding to speak of. No series or model name, just a label with “Sandisk”. > > I’ve been using it for 5 years for a dual-boot system (2×64 G for Windows > and Gentoo). By the end of last year it became slower and slower when > writing. Especially eix-update became slw. My main suspicion is that it > was quite full and there probably is no overprovisioning for wear-leveling > built into the drive. So it was writing the same cells over and over when I > did my world updates. > >> Sandisk SDSSDA240GG26 >> >> Should last me a good long while. It's the /home that keeps growing. o_O > When I built my NAS 2 years ago, I wanted the cheapest (but still from a > notable brand) SSD for the OS. So I bought a Sandisk SDSSDA120G, so > apparently from the same series you mentioned. > > A simple read test with hdparm -t reveals: > 500 GB Crucial from 2016: Timing buffered disk reads: 1596 MB in 3.00 > seconds = 531.46 MB/sec > 128 GB Sandisk from 2014: Timing buffered disk reads: 1532 MB in 3.00 > seconds = 510.60 MB/sec > 120 GB Sandisk from 2017: Timing buffered disk reads: 968 MB in 3.00 seconds > = 322.42 MB/sec > > The theoretical maximum of SATA-III is around 550 MB/s. As you can see, even > a very simple read test already shows a considerable performance drop, even > though it is the newest in the bunch (by date of purchase). > A good SSD should always be able to saturate SATA-III when reading. Most do. > Sequential writing on the cheap Sandisk topped off at around 90 MB/s, IIRC. > This is slower than an HDD. For a NAS system drive this is enough, but not > for a desktop, methinks. > > So my message is: don’t by the cheapest. > I generally go by the model number. If a model number is the same between two different sellers but one has a better price, I'll go for the better price. If it's the same model, it shouldn't matter. One thing I did try to do, avoid the much older versions. I found some older models that were cheaper but I wanted a newer model since most of the kinks and quirks should be fixed in the newer ones, according to what folks posted on here anyway. I still haven't bought it yet. I ordered some toner cartridges a while back for my printer. The site said that the ones I ordered fits my printer. Well, it appears they found out that was a error because they removed that page and relisted it but did not include my printer model. So, I had to order a whole new set, at about $100.00 each for high yield. Needless to say, I'll be paying on that for a while. I'll try to sell the wrong ones later. I only opened one color so the others are still sealed. BTW, next time I'll find a printer that allows refilling and such too. I don't like that chip thing. It counts against my page count on color even if I print a black and white page. Still, printer does a awesome job. Beats those ink jet thingys by a country mile. Dale :-) :-)
Re: [gentoo-user] Re: SDD, what features to look for and what to avoid.
On Tue, Mar 17, 2020 at 02:22:36PM -0500, Dale wrote: > Thanks for the additional info. As I figured, they got most of the > kinks worked out by now and we got some dependable SSDs to buy. > > I found a 240GB for a little over $42.00 USA. Not bad at all. For > those curious: Careful, you get what you pay for. When I built my first PC from scratch (had laptops before that), I bought a 128 GB Sandisk SDSSDP128G. I didn’t have a lot of money back then, it was relatively cheap, but not the very cheapest. It doesn’t even have real branding to speak of. No series or model name, just a label with “Sandisk”. I’ve been using it for 5 years for a dual-boot system (2×64 G for Windows and Gentoo). By the end of last year it became slower and slower when writing. Especially eix-update became slw. My main suspicion is that it was quite full and there probably is no overprovisioning for wear-leveling built into the drive. So it was writing the same cells over and over when I did my world updates. > Sandisk SDSSDA240GG26 > > Should last me a good long while. It's the /home that keeps growing. o_O When I built my NAS 2 years ago, I wanted the cheapest (but still from a notable brand) SSD for the OS. So I bought a Sandisk SDSSDA120G, so apparently from the same series you mentioned. A simple read test with hdparm -t reveals: 500 GB Crucial from 2016: Timing buffered disk reads: 1596 MB in 3.00 seconds = 531.46 MB/sec 128 GB Sandisk from 2014: Timing buffered disk reads: 1532 MB in 3.00 seconds = 510.60 MB/sec 120 GB Sandisk from 2017: Timing buffered disk reads: 968 MB in 3.00 seconds = 322.42 MB/sec The theoretical maximum of SATA-III is around 550 MB/s. As you can see, even a very simple read test already shows a considerable performance drop, even though it is the newest in the bunch (by date of purchase). A good SSD should always be able to saturate SATA-III when reading. Most do. Sequential writing on the cheap Sandisk topped off at around 90 MB/s, IIRC. This is slower than an HDD. For a NAS system drive this is enough, but not for a desktop, methinks. So my message is: don’t by the cheapest. -- Gruß | Greetings | Qapla’ Please do not share anything from, with or about me on any social network. A good position is better than any job. signature.asc Description: PGP signature
Re: [gentoo-user] Re: SDD, what features to look for and what to avoid.
David Haller wrote: > Hello, an addendum without digging up the details ... > > On Tue, 17 Mar 2020, David Haller wrote: >> On Tue, 17 Mar 2020, Grant Edwards wrote: >>> I've put five Samsung SATA drives into various things in the past few >>> years with flawless results. Samsung is one of the big manufacturers >>> of flash chips, so I figure they should always end up with 1st choice >>> quality chips in their own drives... >> And they produce and use their own controllers, so they additionally >> know the ins and outs of those, i.e. they can easily optimize the >> whole SSD from Flash-Chip over controller up to the firmware... > [..] >> AFAIgathered, Samsung is the only one producing the whole product. > I guess Intel did (still does?) that too, but you'll have to check > that, ISTR that Intel now sells SSDs with non-Intel controllers and/or > non-Intel/"IM-Flash" flash-chips... Oh, wait, yes, Intel still does, > but those "pure Intel" SSDs come with a *very* hefty price (like 4 > times as much) and all the "normal" priced ones are those with either > and/or non Intel flash-chips and/or -controllers... But please go > check that yourselves though! > > The second thing I remembered: the german "c't"[2] magazine did a > torture test in late 2018 (IIRC), basically grabbing a few then > current SSDs and run their own testtool[1] on them until they died. Or > so was the plan. That was a "write till it dies" test. > > First of all: all SSD exceeded their specs, some IIRC just barely. The > bulk by a factor of 2 or more. ISTR some of those "just barely", but > wont name them without digging out the actual results, which I'll do > upon requests. > > The test had one problem though: a (IIRC) Samsung 850 Pro just refused > to die ;) They aborted the test after something like over 4 months > (all other drives had died inside of about a month) of _continous_ > writes (or write-verify cycles) to that one remaining SSD, which was > still happily chugging along... > > I do remember though, that even the Samsung EVO came out at the top of > the bunch > > (Note: c't does not award a "test-winner" or anything. Just data and > an conclusion aka "Fazit", the reader has to digest the data and make > up his own mind for _her/his_ own usecase). > > All IIRC, I can dig out and translate the details though! (and it's > month's later followup on what became of that Samsung ;) > > HTH, and please do PM (no need to clog the ML) if you want me to go > digging for the details, > -dnh > > [1] which name escapes me ATM, but tried and tested since 199[0-5] or > so ;) > > [2] https://en.wikipedia.org/wiki/C%27t (that page is sadly woefully > outdated) > I wouldn't mind seeing the info posted to the list myself. I'm interested and I suspect some others may be as well. After all, several people have SSDs now and buying good ones is something we all are interested in. Dale :-) :-) P. S. Sorry for the typo in the subject line. I just noticed I typed it in as SDD instead of SSD. I generally proof the body part but given my age, eyes getting bad, plus just being me, I may need to proof the subject line as well. o_O
Re: [gentoo-user] Re: SDD, what features to look for and what to avoid.
Hello, an addendum without digging up the details ... On Tue, 17 Mar 2020, David Haller wrote: >On Tue, 17 Mar 2020, Grant Edwards wrote: >>I've put five Samsung SATA drives into various things in the past few >>years with flawless results. Samsung is one of the big manufacturers >>of flash chips, so I figure they should always end up with 1st choice >>quality chips in their own drives... > >And they produce and use their own controllers, so they additionally >know the ins and outs of those, i.e. they can easily optimize the >whole SSD from Flash-Chip over controller up to the firmware... [..] >AFAIgathered, Samsung is the only one producing the whole product. I guess Intel did (still does?) that too, but you'll have to check that, ISTR that Intel now sells SSDs with non-Intel controllers and/or non-Intel/"IM-Flash" flash-chips... Oh, wait, yes, Intel still does, but those "pure Intel" SSDs come with a *very* hefty price (like 4 times as much) and all the "normal" priced ones are those with either and/or non Intel flash-chips and/or -controllers... But please go check that yourselves though! The second thing I remembered: the german "c't"[2] magazine did a torture test in late 2018 (IIRC), basically grabbing a few then current SSDs and run their own testtool[1] on them until they died. Or so was the plan. That was a "write till it dies" test. First of all: all SSD exceeded their specs, some IIRC just barely. The bulk by a factor of 2 or more. ISTR some of those "just barely", but wont name them without digging out the actual results, which I'll do upon requests. The test had one problem though: a (IIRC) Samsung 850 Pro just refused to die ;) They aborted the test after something like over 4 months (all other drives had died inside of about a month) of _continous_ writes (or write-verify cycles) to that one remaining SSD, which was still happily chugging along... I do remember though, that even the Samsung EVO came out at the top of the bunch (Note: c't does not award a "test-winner" or anything. Just data and an conclusion aka "Fazit", the reader has to digest the data and make up his own mind for _her/his_ own usecase). All IIRC, I can dig out and translate the details though! (and it's month's later followup on what became of that Samsung ;) HTH, and please do PM (no need to clog the ML) if you want me to go digging for the details, -dnh [1] which name escapes me ATM, but tried and tested since 199[0-5] or so ;) [2] https://en.wikipedia.org/wiki/C%27t (that page is sadly woefully outdated) -- "If you are using an Macintosh e-mail program that is not from Microsoft, we recommend checking with that particular company. But most likely other e-mail programs like Eudora are not designed to enable virus replication" -- http://www.microsoft.com/mac/products/office/2001/virus_alert.asp
[gentoo-user] Re: SDD, what features to look for and what to avoid.
On 2020-03-17, David Haller wrote: > And they produce and use their own controllers, so they additionally > know the ins and outs of those, i.e. they can easily optimize the > whole SSD from Flash-Chip over controller up to the firmware... Yep, that was definitely the gist of my (wishful) thinking when I decided on Samsung. > AFAIgathered, Samsung is the only one producing the whole product. > Albeit, I heard rumours, that the various divisions of Samsung could > as well be different manufacturers[1], I've worked with smaller corporations where various seemingly closely-related divisions might as well be on different planets. But at least there's chance that Samsung products can benefit from the vertical/horizontal integration. As it says on the coffee mug my neice gave me for christmas: Hold on, let me overthink this. -- Grant
Re: [gentoo-user] Re: SDD, what features to look for and what to avoid.
Hello, On Tue, 17 Mar 2020, Grant Edwards wrote: >I've put five Samsung SATA drives into various things in the past few >years with flawless results. Samsung is one of the big manufacturers >of flash chips, so I figure they should always end up with 1st choice >quality chips in their own drives... And they produce and use their own controllers, so they additionally know the ins and outs of those, i.e. they can easily optimize the whole SSD from Flash-Chip over controller up to the firmware... Dunno if the same is true for Micron+Intel (IM Flash) / Crucial. Seems no, as e.g. the Crucial MX500 uses a Silicon Motion controller, even though Micron does (or did) produce controllers. Nor does WD, it seems. AFAIgathered, Samsung is the only one producing the whole product. Albeit, I heard rumours, that the various divisions of Samsung could as well be different manufacturers[1], but how far apart the Flash/Controller/SSD parts I have not an inkling ;) HTH, -dnh [1] well, considering that includes washing-machines, ships and whatnot ... ;) -- May I ask a question?
Re: [gentoo-user] Re: SDD, what features to look for and what to avoid.
On Tue, Mar 17, 2020 at 12:33 PM Grant Edwards wrote: > > I've put five Samsung SATA drives into various things in the past few > years with flawless results. Samsung is one of the big manufacturers > of flash chips, so I figure they should always end up with 1st choice > quality chips in their own drives... > > -- > Grant Edwards grant.b.edwardsYow! Where's the Coke > at machine? Tell me a joke!! > gmail.com > I've had similar results with Crucial SSDs. Mark
[gentoo-user] Re: SDD, what features to look for and what to avoid.
I've put five Samsung SATA drives into various things in the past few years with flawless results. Samsung is one of the big manufacturers of flash chips, so I figure they should always end up with 1st choice quality chips in their own drives... -- Grant Edwards grant.b.edwardsYow! Where's the Coke at machine? Tell me a joke!! gmail.com
Re: [gentoo-user] Re: SDD, what features to look for and what to avoid.
Grant Edwards wrote: > On 2020-03-17, Andrea Conti wrote: > >> "NAND flash" (as opposed to "NOR flash") refers to the way memory cells >> are organized and connected. See for example >> https://www.embedded.com/flash-101-nand-flash-vs-nor-flash/ >> >> AFAIK all SSDs use some variant of NAND flash. > Correct. NOR flash density is far lower than NAND flash. NOR is only > used for small applications (e.g. BIOS, small embedded systems). > Anytime you see sizes quoted in 'Gb' rather than 'Mb', you're > certainly talking about NAND. > > -- > Grant Thanks for the additional info. As I figured, they got most of the kinks worked out by now and we got some dependable SSDs to buy. I found a 240GB for a little over $42.00 USA. Not bad at all. For those curious: Sandisk SDSSDA240GG26 Should last me a good long while. It's the /home that keeps growing. o_O Filesystem Size Used Avail Use% Mounted on /dev/mapper/Home2-Home2 8.1T 5.4T 2.7T 68% /home Dale :-) :-)
[gentoo-user] Re: SDD, what features to look for and what to avoid.
On 2020-03-17, Andrea Conti wrote: > "NAND flash" (as opposed to "NOR flash") refers to the way memory cells > are organized and connected. See for example > https://www.embedded.com/flash-101-nand-flash-vs-nor-flash/ > > AFAIK all SSDs use some variant of NAND flash. Correct. NOR flash density is far lower than NAND flash. NOR is only used for small applications (e.g. BIOS, small embedded systems). Anytime you see sizes quoted in 'Gb' rather than 'Mb', you're certainly talking about NAND. -- Grant
