SSD-tråden: info om og diskusjon rundt SSD

[Syar-2003]

Det er feil at Windows kun kjører TRIM.
Om volsnap er enabled kjøres en intelligent defragmentering månedlig i tillegg til TRIM.

Hvis dere ikke har Volume Shadow Copying tjenesten enabled er det kun TRIM som kjøres og det er ikke noe problem.

https://www.hanselman.com/blog/TheRealAndCompleteStoryDoesWindowsDefragmentYourSSD.aspx

[Syar-2003]

G skrev (17 timer siden):

Må jeg slå av dette i hver og en VMware Windows 10 virtuell maskin også når Verts (Host) maskinen får det slått av?

Ja , det må justeres i hver enkelt. De virtuelle maskinene styres helt av sitt eget OS og ikke vertsmaskinens . Volsnap er også en del av vmware tools som du har installert i VMs.

Endret 20. september 2020 av syar2003

[Theo343]

Quote

Actually Scott and Vadim are both wrong. Storage Optimizer will defrag an SSD once a month if volume snapshots are enabled. This is by design and necessary due to slow volsnap copy on write performance on fragmented SSD volumes. It’s also somewhat of a misconception that fragmentation is not a problem on SSDs. If an SSD gets too fragmented you can hit maximum file fragmentation (when the metadata can’t represent any more file fragments) which will result in errors when you try to write/extend a file. Furthermore, more file fragments means more metadata to process while reading/writing a file, which can lead to slower performance.

NTFS relatert som G nevnte...

Og Trim som vi alle har vært klar over i mange år og dermed ikke stresset med det.

Quote

As far as Retrim is concerned, this command should run on the schedule specified in the dfrgui UI. Retrim is necessary because of the way TRIM is processed in the file systems. Due to the varying performance of hardware responding to TRIM, TRIM is processed asynchronously by the file system. When a file is deleted or space is otherwise freed, the file system queues the trim request to be processed. To limit the peek resource usage this queue may only grow to a maximum number of trim requests. If the queue is of max size, incoming TRIM requests may be dropped. This is okay because we will periodically come through and do a Retrim with Storage Optimizer. The Retrim is done at a granularity that should avoid hitting the maximum TRIM request queue size where TRIMs are dropped.

Og som Syar nevnte:

Quote

A bit of advanced info here - Defrag will only run on your SSD if volsnap is turned on, and volsnap is turned on by System Restore as one needs the other. You could turn off System Restore if you want, but that turns off a pretty important safety net for Windows.

Dette er hvordan Windows normalt sett fungerer med SSD men når det gjelder det G linket til så er det vel snakk om en bug som kjører full defrag mye oftere enn det som beskrives her. Må innrømme at jeg ikke leste linken til G og skal derfor gjøre det nå for å se hva det egentlig er snakk om.

Dette er jo faktisk også et poeng.

Quote

Yes, your SSD's file system sometimes needs a kind of defragmentation and that's handled by Windows, monthly by default, when appropriate. The intent is to maximize performance and a long life. If you disable defragmentation completely, you are taking a risk that your filesystem metadata could reach maximum fragmentation and get you potentially in trouble.

Si man har en Win10 installasjon man har kjørt i åresvis (guilty) og man begynner å få feil i filsystemet og må  kjøre sfc /now som man ser fikser ting. Da har man kanskje begynt å nå begrensninger i hvor mange filfragmenter NTFS kan håndtere.

Endret 20. september 2020 av Theo343

[Theo343]

1 hour ago, syar2003 said:

Ja , det må justeres i hver enkelt. De virtuelle maskinene styres helt av sitt eget OS og ikke vertsmaskinens . Volsnap er også en del av vmware tools som du har installert i VMs.

Og takk for denne. Jeg tok det for god fisk at Hyper-V faktisk var såpass intelligent og integrert i Windows at det forstod at det ble lagret på en fysisk SSD. 

Gjelder dette også Hyperkonvergert infrastruktur og eks. Storage Spaces Direct med SSD? I såfall kan man få en rimelig ekstrem mengde unødvendig IO om man har noen hundre VMer kjørende i et slikt cluster. Riktignok trenger man der "volsnap" ifb. med backup systemer.

Endret 20. september 2020 av Theo343

[Syar-2003]

Theo343 skrev (27 minutter siden):

Og takk for denne. Jeg tok det for god fisk at Hyper-V faktisk var såpass intelligent og integrert i Windows at det forstod at det ble lagret på en fysisk SSD. 

Gjelder dette også Hyperkonvergert infrastruktur og eks. Storage Spaces Direct med SSD? I såfall kan man få en rimelig ekstrem mengde unødvendig IO om man har noen hundre VMer kjørende i et slikt cluster.

G har vel wmware hypervisor på en Windows 10 vmware host (om jeg tyder han rett). Hyper-V har jeg ikke satt meg noe særlig inn i.
For virtuelle disker i wmare er det en god praksis å sette I/O intensive VMs om har mye writes med thick zero disk for å ha skriveoperasjoner begrenset til samme konsistente fil/storage blokker på host storage. Plass skal man ha nok av når men driver med VMs. Bruke partisjonering inne i VMs er også lurt sånn sett for å holde plassbehovet i sjakk.
I tillegg syns jeg bør man absolutt bør kjøre optimeringer på VMs , da er denne grei å bruke for å temme Windows 10 VMene litt (kun optimize funksjonen) https://techzone.vmware.com/resource/vmware-operating-system-optimization-tool-guide#_Toc47964441

 

Edit: Må ikke glemme at en VMs OS sin eventuelle derfamenteringer vil begrense seg til lese/skrive operasjoner på allokerert disk-file på host storage.

 

Endret 20. september 2020 av syar2003

[G]

5 hours ago, syar2003 said:

G har vel wmware hypervisor på en Windows 10 vmware host (om jeg tyder han rett). Hyper-V har jeg ikke satt meg noe særlig inn i.

Riktig, jeg er en litt GUI pek og klikk aktig bruker. Skyldig! Vet at det er litt mer kommandokode for å sette opp ESXi f.eks., det kan koste en del penger også. Har ikke sett på andre løsninger enda som gir "bare metal". Hørt om noe som kalles docker og noe sånn, men kan ikke noe om det. Ser at VMware Workstation oppgraderingsdialogen prater om noe støtte for noe slikt i versjon 16, men skal jeg gjette så blir det vel en del ikke-GUI / kommandoer for å styre med tingene.

Har jo vurdert om jeg burde kjøre VMware Workstation under Linux, og alt med Windows som gjeste-OS der. Men har aldri orket å flytte og være like dynamisk rundt OS som jeg var på 90-tallet. Den gang var det jo fett/rått å bare skvise nok OS inn i en bootloader, ja det var tider 

Jeg er absolutt ingen superbruker rundt IT, mangler utdannelser og greier rundt det også. Kunne nok valgt det løpet, men jeg fant rett og slett IT-verdenen som en del hodebry nokså tidlig, og har valgt det som en hobby. Og når en kun liker det litt, så har det begrensninger på hvor dypt man gidder å grave i materier innen IT også. Hilsen en pek-og-klikk bruker.

Det jeg kan om Linux er også nokså overfladisk, selv om jeg vet om en del rundt virkemåte, så er det verre med koding og det å orke å forstå alle slags typer kommandoer.

Endret 20. september 2020 av G

[G]

Herregud, ser jo ut som en slik kabelkanal 

ADATA Launches GAMMIX S70 PCIe Gen4 M.2 2280 Solid State Drive

Quote

For reliable operations and stability, the S70 is equipped with XPG’s proprietary CoolArmor heat spreader design. The aluminum heat spreader features a terraced structure, which resembles layers of armor, for increased surface area and hollow chambers underneath. This design provides for more effective heat dissipation that can reduce temperatures by up to thirty percent.

Perfekt for hybelkaninene å gjemme seg bort inni.. Artikkelen nevner heller ikke om det er TLC-basert, som jo er det mest sannsynlige, eller om det er QLC. Rart om det er QLC da (kanskje noe usannsynlig). Men har jo SLC-cache, så kan jo gjennomføres et QLC drittprodukt om de bare ønsker det sterkt nok.

https://www.guru3d.com/news-story/adata-launches-gammix-s70-pcie-gen4-m-2-2280-solid-state-drive.html

Endret 21. september 2020 av G

[Simen1]

G skrev (8 minutter siden):

Perfekt for hybelkaninene å gjemme seg bort inni..

Det er jo så populært med innsektshotell for tida..

[G]

Fant et bedre bilde av den: https://www.benchmark.rs/vesti/adata_predstavlja_xpg_gammix_s70_pcie_4.0_ssd-85353

@Nizzen Når du får anledning å se på data, hvor nære stiller denne seg mot Sage?

 

Sage her forresten:

Spoiler

Bør en fortsatt vente på Sage? Når kommer s70 til Norge etc.?

EDIT:

_______________________________

27. september. Endelig kom jeg over litt forvented forsendelsesinformasjon:

Quote

With support for NVMe 1.4 and PCIe 4.0×4 interface, the GAMMIX S70 is capable of reaching sequential read/write speeds up to 7400 MB/s and 6400 MB/s. Random 4K performance cannot be found in the product page. ADATA does not disclose any information in regards of the memory controller and NAND flash. We guess it may be equipped with the Innogrit Rainier IG5236 controller and Micron’s 96-Layer 3D TLC NAND flash.

The S70 is provided with a 5-year warranty, with two capacities (1 TB, 2 TB) to choose from. It is expected to ship in October, but no pricing is revealed yet.

https://www.unbxtech.com/2020/09/adata-xpg-gammix-s70-nvme-ssd-release.html

Endret 27. september 2020 av G

[Theo343]

30 minutes ago, Simen1 said:

Det er jo så populært med innsektshotell for tida..

Fint å bruke som forklaring når man ikke gidder skifte panel eller male huset egentlig.

[G]

En kjempefordel med å gå for den varianten kjøleribbe er at det er grisebillig å få ekstrudert en slik aluminiums/kobberprofil. Framfor å dille å få designet til å bli så lekkert som mulig, med buer, carbonfiber-utseende, RGB eller hva de måtte finne på.

Litt sånn tilbake til røttene på 90, tidlig 2000-tallet med kjøleribber.

 

En sin mening fra et forum:

Quote

Yeah, not really impressive, especially as there are no 4K Q1T1 numbers. I think Adata is going to have to do some more firmware work on this drive before they release it.

For the 4K Q64T8 and Q32T8 results, my "old" Patriot Viper VPN100 has better read performance and it even beats it in the Q16T8 test. Write speeds are also only slightly faster. It really shows that we need a wider controller/NAND bus to gain any useful performance advantage over current NVMe drives. Alternatively we need a new type of NAND.

https://www.techpowerup.com/forums/threads/adata-xpg-launches-gammix-s70-pcie-gen4-m-2-2280-solid-state-drive.272381/

Fra samme forumtråd, en annens input:

Quote

It's unfortunately more complicated than that. I'm lucky to now be on the forefront of flash analysis but it's moving rapidly to the point it will be very difficult for reviewers moving forward. For example, we have SLC caching algorithms that not only rely on behavioral (workload) profiles but can shuffle data into SLC for reads and utilize AI/ML for ECC per zones (e.g. static vs. dynamic SLC, with the latter shifting through native flash). We have adaptive read schemes (Samsung's ARC), tiles, SBL, etc. to improve tR for 4K requests. Samsung even uses 8KB subplanes on their new flash - see the 980 Pro. They also use a new TurboWrite. Gauging "user experience" is going to get very difficult with these new drives, relying on 4K random no longer cuts it as you have all sorts of tricks, as another example Kioxia has an adaptive scheme for programming BiCS5 that adjusts for detected tR needs. New QLC from Kioxia and Intel have parallel/independent plane reads which can mask QLC's poor tR (in addition to them "hiding" OS files in SLC cache for reads), etc.

Quote

They're not - it's just that E18s have to use older, 96L flash for now which requires more flash to saturate the channels. This is why the Samsung 980 Pro is using 256Gb dies at 250GB-1TB (likely 512Gb at 2TB) despite launching as 512Gb, SK hynix is using 512Gb dies in the Gold P31 despite launching that flash at 1Tb (likely to be used in the 2TB Platinum P31), etc. At 500GB you have only one 64GiB die per channel (technically 256Gb dies are possible as on lower capacity E16s, though) and 96L flash is too slow. Upcoming flash will be much faster per die, from 82 MB/s (TLC) with 6th Gen V-NAND to 132 MB/s with BiCS5, so we'll see 500GB SKUs again soon.

Quote

I don't consider 4K beyond around QD/T4 or maybe 8 to be relevant for consumers, but in any case - yeah, it's highly dependent on the flash. Some controllers do eek out more like the SMI designs (thanks to help from Intel) but even SK hynix didn't do bad with their new Cepheus controller. However, there's just a ton of tricks they're using on the flash now to improve tR, which I try to document over at Reddit for those that want the technical details. However we still only have SK hynix's "128L" flash on the market although I've seen 980 Pro results already...in any case, we need to wait for Micron's new TLC and BiCS5.

I haven't worked with ADATA but I have written a whitepaper on SSD ML recently (which is under NDA) - I largely agree it's fluff and primarily for marketing, however it would be a big mistake to disregard the fact that the vast majority of gains in flash and NAND performance have come from increasingly complex algorithms. AI/ML in particular benefits more on the enterprise end, especially with cTLC as a budget option, however it's already used and will be used on consumer drives (e.g. Samsung's flash) to improve tR (latency) but more importantly keep the drive performing better, longer. You have far less degradation of performance after writes and wear due to the improved read retry cycles, and it also enables you to do better programming (e.g. 8-16 vs. 4-8-16) including with interruptions (i.e. doing a read mid-program sequence). So while that seems minor to the consumer, it means you can get more out of the flash.

Quote

Phison supposedly has at least improved SLC algorithms (for their full-drive SLC caching), beyond that I can't comment although you do need new firmware for new flash. In any case, yes, it is a Micron issue, they've stay latched on to FG with Intel for example with the 96L QLC, they have to retool for something based on TCAT. I mean these time tables are almost always optimistic anyway and then you have the pandemic...

I'm all for better QLC performance, I more meant that it can be deceptively good for example in reviews or specification listings. One could argue that if the consumer doesn't ever "see" it then it's not relevant - which is a valid point, but underlines that NAND is a very murky business. See the recent Level1Techs stream with Malventano to see what I mean.

I can assure you this is something discussed at great length by reviewers since I talk to almost all of the prominent ones at this point. More generally the ability to benchmark has improved the last two years and information has been much better-shared. That being said, staying cutting-edge requires a lot of research and keeping up with the technical side (through conferences and such, livestreams more commonly now, but also reading academic journals and patents). Applying that to "real world" testing is more difficult as I find the engineers to be reticent to reveal proprietary information - for example, I had to track down patents for Crucial's P5 SLC caching scheme as they wouldn't tell me anything. And I had to study their flash geometry (tiles and circuitry) to determine the difference between B27A and B27B since they refused to say. I'm getting off track though - I suspect that, at least for now, 4K isn't really the bottleneck for most things (although software could and will improve), so a better measure is consistency, however most drives this generation will be focused on efficiency and sequentials.

 

Endret 21. september 2020 av G

[G]

Samsung 980 Pro review:

Quote

Fra konklusjon:

<snip> 

TLC Write holes

Technologies like TLC and QLC face some challenges writing more bits per cell of NAND, we, however, did not notice a dropoff in performance with mixed heavy workloads, so the dreaded TLC write hole did not kick in (and I do mean continuously sustained/linear writes minute after minute). See when then the SSD buffers are full and start to write directly to TLC NAND that could be an issue. This, in a nutshell, is what you need to be aware of with TLC and QLC SSDs. IOPS performance is good on this unit, really good. The overall workload traces also indicate this SSD to be extremely capable and fast. TBW values are average for a PRO model. 600 TBW for the 1 TB model. Albeit how companies calculate or test these values these days, is a bit of a mystery. A listed TBW value, however, is something you can pin them on warranty wise.

 

  

  

Concluding

The Samsung 980 PRO series is an impressive drive, however, it managed to impress me less than the release of the 970 PRO. The focus for Samsung has been clear, reach a peak read speed of ~7 GB/sec and ~5 GB/for writes. You'll only hit these numbers in very specific workloads, mostly sustained and linear writes. Also if you want to copy at 5 GB/sec to the Pro, surely your host/client setup should be the same. So you'll be hardpressed to reach these numbers in a real-world scenario.

</snip>

https://www.guru3d.com/articles-pages/samsung-980-pro-1tb-nvme-ssd-review,20.html

Endret 22. september 2020 av G

[G]

 

Spoiler

Quote

ADATA XPG Now offers SPECTRIX D50 Xtreme DDR4 RGB Memory Modules

by Hilbert Hagedoorn on: 09/22/2020 08:19 AM | source: | 1 comment(s)

XPG SPECTRIX D50 Xtreme DDR4 RGB memory modules deliver speeds of up to 5000Hz, the XPG SPECTRIX D50 Xtreme isn't just about performance but also form with its solidly constructed heat sink with gorgeous geometric lines, dazzling RGB lighting, and a stunning reflective surface.

Manufactured with only the highest quality chips and PCBs, the SPECTRIX D50 Xtreme offers the pinnacle in stability, reliability, as well as speeds of up to 5000 MHz. What's more, it supports the latest Intel and AMD platforms. The SPECTRIX D50 Xtreme is a looker too with the combination of its indented geometric lines, triangular RGB light bar, and a polished and electroplated surface with a mirror-like sheen. The SPECTRIX D50 Xtreme comes as a dual kit with a capacity of 8 GB per module and in two performance variants, 4800 MHz and 5000 MHz.

With the XPG RGB Sync app or a RGB software from a major motherboard brand, users can switch between three RGB modes - Static, Breathing, and Comet. In addition to the three modes, users can also set it to Music mode to sync with their favorite tunes.

The SPECTRIX D50 Xtreme has full XMP 2.0 compatibility to make overclocking effortless when installed on PCs that also support XMP 2.0. XMP 2.0 support means users have more ways to access memory overclocking, including directly from the operating system rather than via more complex BIOS settings.

 

 

https://www.guru3d.com/news-story/tsmc-would-use-gate-all-around-transistors-for-2nm-node.html

 

Å faen, det var jo RAM. RGB-glory tok vekk fokus i fra RAM-pinnene på slot  Sorry, litt feilpostet da. Kan pakke den inn i Spoiler..

Endret 22. september 2020 av G

[G]

Sekunder gammel nyhet hos denne nettblekka:

https://wccftech.com/samsung-finally-bringing-the-980-pro-pcie-gen-4-0-nvme-ssd-to-market/

Quote

Just under a month ago Samsung officially listed its next-generation PCIe 4.0 based 980 Pro NVMe drive on their website, now they've announced the launch window and some reviews have already gone live. To cut to the chase the Samsung 980 Pro SSD will be available worldwide 'this month', so assuming before the end of September 2020.

Så omtrent kr tusen for 250 Gig modellen..

Endret 22. september 2020 av G

[Nizzen]

Herreguuuuuud for spamming da!

[siDDis]

Like crap som den gamle 950 pro modellen. Gi meg heller en Intel 320, den er fremdeles raskere.

[Betroz]

4 minutes ago, siDDis said:

Like crap som den gamle 950 pro modellen. Gi meg heller en Intel 320, den er fremdeles raskere.

Raskere til hva da?

[mandela]

Til å få 8MB feilen ?

[siDDis]

Nå husker jeg ikke om det var 10x eller 100x raskere 4kb sync write.

Uansett det er lett å teste med fio, mens Optane greier 300MB/s på 4kb sync write, så sliter vell denne under 10MB/s

[Nizzen]

Ikkje så gale 4k random read @QD =1 på Samsung 980 pro da. Den slår tilogmed Adata sx8200pro  

(faen med på tide at noen slår den gamle ssden...)

 

Tilogmed kjørt på AMD

Endret 22. september 2020 av Nizzen