|
Hah, guess I'm not the only one reading along that decided to pick one up. Got mine installed about an hour ago and am doing my best to fill it up with games. Coming from a 512gb it's so nice not to have to worry about running out of space.
|
# ? May 11, 2023 00:18 |
|
|
# ? May 24, 2024 05:38 |
|
BlankSystemDaemon posted:loving hell the Solidigm P44 Pro is fast. And DRAMless, apparently. I pumped the brakes on getting one, I forgot I don't have a second M.2 slot to work with. Am gonna try to get work to reimburse me for one for the work machine.
|
# ? May 11, 2023 00:37 |
|
SwissArmyDruid posted:And DRAMless, apparently. Is it? It's a straight hynix p41 clone, pretty sure it has dram.
|
# ? May 11, 2023 01:40 |
|
P44 Pro has dram. It’s an updated Platinum P41 https://www.tomshardware.com/reviews/solidigm-p44-pro-ssd-review E: not to be confused with the QLC Soldigim P41 Plus which is NOT an updated Platinum P41. I know. Cygni fucked around with this message at 02:14 on May 11, 2023 |
# ? May 11, 2023 01:49 |
|
Henrik Zetterberg posted:gently caress JIRA seriously Also gently caress Agile
|
# ? May 11, 2023 12:22 |
|
I thing Agile is fine it’s just when you get into dumb things like “Roadmaps” and “Deadlines” that everything falls apart.
|
# ? May 11, 2023 13:27 |
|
WhyteRyce posted:Also one time I had one of those software engineers tell me they weren’t going to deliver a fix that was blocking PRQ because they had some doxygen requirements that took priority lol I would've hit the roof if it's the same "Intel JIRA" that I used, I spent a week writing a CLI interface for simple updates for one of those Santa Clara DEC teams that never bothered to learn how to work a mouse. I call myself a dinosaur now but those folks were really long in the tooth
|
# ? May 11, 2023 17:00 |
I don't know why, but I swear installing the driver from Solidigm makes a difference in how snappy things feel, especially in Explorer.SwissArmyDruid posted:And DRAMless, apparently.
|
|
# ? May 11, 2023 18:29 |
|
JawnV6 posted:lol I would've hit the roof Probably the same reason we have the mozilla browser on the SUSE linux EC instances anno 2023.
|
# ? May 11, 2023 19:26 |
|
WhyteRyce posted:Solidigm is an standalone subsidiary under Hynix. Originally they were shipping their own products with their own controllers with their own NAND with their own firmware sold by their own teams. The joint venture separation from Micron was a huge mess. There were a couple hundred Intel people working in the Micron facilities in Idaho and Utah where there were no Intel sites. As part of the wind-down, one day Micron just said "naah, your badges don't work here anymore" and booted everyone off site. The official long term plan was to relo those teams to other sites, so they spent a fortune opening temporary offices locally so those people would have an office to come in to. This seemed like a dumb idea at the time, spending a pile of money on an office and prolonging the inevitable migration of people from the Intel/Micron joint venture to working directly for Micron. Shockingly, ~98% of the joint venture employees didn't want to move to a different state! They instead got immediately poached and hired by Micron and the offices stood up for 200 people ended up being used by fewer than 5 people. Also, I think Rob Crooke has the funniest name of anyone who ever worked there. Sounds very untrustworthy!
|
# ? May 11, 2023 21:33 |
|
I'm trying to remotely assist my stepson on a new Intel build w/DDR5. We can't get it to exit the bios and I'm wondering -- is this what the first-time "training" behavior of the RAM looks like? Kinda sits there and does nothing once you F10 to save/exit the bios? Do I just need to let it sit for five minutes or something?
|
# ? May 11, 2023 21:57 |
|
Training doesn't happen on the bios screen, it happens after you restart and the computer is booting up again, and your system won't output a video signal until it's done. Are you saying that the bios is staying on the screen when you tell it to save and exit? Because that's really not normal. Try updating the bios if possible.
|
# ? May 12, 2023 03:16 |
|
JawnV6 posted:lol I would've hit the roof I tried to, what I naively thought, end the discussion by saying the priority was getting the chip out the door because selling these chips is what we do. The guy looked at me smugly and said, "No. What we do is ship a healthy...product". It was that point I realized a non-zero number of internal tool developers had no idea about what they own and that no one outside of the lab will ever give two shits about their code. quote:if it's the same "Intel JIRA" that I used, I spent a week writing a CLI interface for simple updates for one of those Santa Clara DEC teams that never bothered to learn how to work a mouse. I call myself a dinosaur now but those folks were really long in the tooth Some graybeards in a largely defunct site conned my old group to let them develop our next generation automation framework. After the group agreed we found out it was 99% based off their rats nest of TCL code that they wrote over a decade ago and had no ability to scale and the installation process was basically "here take this .iso we made of our golden image...no we have no ability to package it or even point at a drat repo for you to clone because we just hacked up this Ubuntu image until things were working and we also too lazy to figure out how to setup a repository. If we need to fix something in the code we'll point to you a new .iso" canyoneer posted:
I was talking to some managers on the Intel side who were bitching because they absolutely wanted to retain some people in Colorado but Intel was so slow with coming out any kind of offer that they were losing key people left and right. quote:Also, I think Rob Crooke has the funniest name of anyone who ever worked there. Sounds very untrustworthy! We used to call him Gran Larsen WhyteRyce fucked around with this message at 03:56 on May 12, 2023 |
# ? May 12, 2023 03:51 |
|
did someone find an exciting new spectre/meltdown variant again https://twitter.com/tomshardware/status/1657496875044356096
|
# ? May 14, 2023 03:17 |
|
https://www.amazon.com/SolidigmTM-Internal-7000MB-6500MB-SSDPFKKW020X7X1/dp/B0BJGGL1SQ Those 2TB Solidigm P44 Pros came down in price, but only if you've got amazon prime, at which point it's $129. goons who bought one already might want to order and return for the better price.
|
# ? May 15, 2023 01:09 |
|
Is Solidigm going to be around to handle warranty requests in five years? I'm tempted to pick one up but always cautious about new companies even if they're spun off. I guess there's really no way to predict the future like that.
|
# ? May 15, 2023 02:56 |
|
Rexxed posted:Is Solidigm going to be around to handle warranty requests in five years? I'm tempted to pick one up but always cautious about new companies even if they're spun off. I guess there's really no way to predict the future like that. Solidigm is a subsidiary of SK Hynix, one of the big 3 memory manufacturers, and itself a part of a huge south korean conglomerate. So not a new company, just a new brand name. And not going anywhere.
|
# ? May 15, 2023 03:09 |
Solidigm appears to be SK Hynix + the firmware and driver folks from Intel - so basically the best of both worlds. Sure feels fast, though it annoys me that the price dropped for Amazon customers.
|
|
# ? May 15, 2023 09:42 |
|
Hynix has their own firmware people already. Fun fact, the head of the enterprise firmware team was the guy Rob Crooke canned from the same position at Intel
|
# ? May 15, 2023 15:17 |
|
So, I bought a 1 TB P44 Pro and put it in my new 10th gen Intel board with 10th gen CPU. I can't get Fast Lane to work. Clean install, just the their Synergy driver and toolkit installed. I don't have the error right now, but it's basically, "You can't enable this feature on this hardware" and Clear Cache is also unactionable. What's the deal, you think? Does this require a PCI 4.0 lane, which this board has but only when using an 11th gen proc. I'm forgoing the first M.2 slot and using the second. Happens on a clean install of 10 Pro as well as 11 Pro. Anybody else?
|
# ? May 15, 2023 19:49 |
|
Neither of those features are supported on the P44 Pro. Both of them are really more for QLC drives where caching is a huge deal. Fast Lane is a brand name of host-managed caching, for example. So don't worry about em, you are good, you bought the fast drive that doesn't need it.
|
# ? May 15, 2023 20:09 |
|
I'm gonna get in on this solidgm pro bandwagon. We'll all go down together.
|
# ? May 15, 2023 20:46 |
|
Rexxed posted:I'm gonna get in on this solidgm pro bandwagon. We'll all go down together. Mods thread title please Intel: We’ll all go down together
|
# ? May 15, 2023 20:59 |
|
Cygni posted:Neither of those features are supported on the P44 Pro. Both of them are really more for QLC drives where caching is a huge deal. Fast Lane is a brand name of host-managed caching, for example. So don't worry about em, you are good, you bought the fast drive that doesn't need it. Thank you, that makes perfect sense, and I appreciate the chart.
|
# ? May 15, 2023 21:00 |
|
wet_goods posted:Mods thread title please Fittingly most of the ex-Intel leadership team at Solidigm is on the outs according to Blind. Even Intel inertia means nothing to the new Korean bosses WhyteRyce fucked around with this message at 21:59 on May 15, 2023 |
# ? May 15, 2023 21:45 |
|
Rexxed posted:I'm gonna get in on this solidgm pro bandwagon. We'll all go down together. I mean, it's a drat sight more interesting that whatever the hell Intel currently has going on for it.
|
# ? May 16, 2023 04:17 |
|
SwissArmyDruid posted:I mean, it's a drat sight more interesting that whatever the hell Intel currently has going on for it. What's not nice is that they only allow those things to use a single DDR5 "channel".
|
# ? May 16, 2023 04:21 |
|
enjoy your ssd that switches off at the hard write limit. solidigm apparently inherited that poo poo from intel. team MP33, MP34, Cardea Zero Z440, Cardea Zero A440, HP EX950, FX900, and FX900 Pro are my current recommendations for 2TB-4TB TLC models right now lol. (yes, I know the HP drives are subvended... they've been subvended 100% competently for 5+ years. It's one of the SSD OEMs who does a ton of in-house work for other brands.) Paul MaudDib fucked around with this message at 04:42 on May 16, 2023 |
# ? May 16, 2023 04:37 |
|
are hard write limits even something that anyone short of a datacenter or hyperscaler will hit before the drive becomes obsolete? I thought this was proved YEARS ago with The Tech Report's long-term SSD endurance torture test, and that was on SATA edit: SATA, on quarter-terabyte class drives. Now we're on terabyte or multiple-terabyte drives, with even more sectors and spare flash to reallocate as bad ones are marked off. SwissArmyDruid fucked around with this message at 04:54 on May 16, 2023 |
# ? May 16, 2023 04:49 |
|
SwissArmyDruid posted:are hard write limits even something that anyone short of a datacenter or hyperscaler will hit before the drive becomes obsolete? I thought this was proved YEARS ago with The Tech Report's long-term SSD endurance torture test, and that was on SATA.
|
# ? May 16, 2023 04:52 |
|
SwissArmyDruid posted:are hard write limits even something that anyone short of a datacenter or hyperscaler will hit before the drive becomes obsolete? I thought this was proved YEARS ago with The Tech Report's long-term SSD endurance torture test, and that was on SATA. Solid state disks are both getting better and worse - there some truly dire modern QLC lifespans but also that 2TB nvme is rated for 1800TBW which should be more than fine for most people.
|
# ? May 16, 2023 04:54 |
|
SwissArmyDruid posted:edit: SATA, on quarter-terabyte class drives. Now we're on terabyte or multiple-terabyte drives, with even more sectors and spare flash to reallocate as bad ones are marked off. god, if it's sata SSD why even bother? and I'm only partially kidding. I'd love big QLC/QVO style 8tb+ SATA drives if it was sufficiently cheap, but outside of SATA being easier to scale than NVMe (tri-mode SAS cards etc) it's really just not worth a "high end" SATA ever now. Those drives I listed are reasonably cheap TLC drives, DRAMless TLC NVMe really isn't that much worse than SATA anything, and those drives go up to like 2TB at great value. The cost of disk is determined by the flash I think, NVMe vs SATA is no longer that big a price difference in the BOM of a high-end drive. Paul MaudDib fucked around with this message at 05:01 on May 16, 2023 |
# ? May 16, 2023 04:58 |
|
Paul MaudDib posted:enjoy your ssd that switches off at the hard write limit. solidigm apparently inherited that poo poo from intel. Wasn't this a myth? There was a lot of bluster about this a few years back, and I believe Intel clarified that it only entered read-only when the entire spare area was exhausted. It wasn't based on the writes hitting the specified endurance "limit".
|
# ? May 16, 2023 06:17 |
|
Anime Schoolgirl posted:video editing/3d modeling churns large enough amounts of write data for a P/E shutoff anti-feature to actually be a problem Paul MaudDib posted:god, if it's sata SSD why even bother? Right, and as I edited to add later: those SATA drives they tested, in some cases, wrote over 2 PB of continuous, nonstop writes over two years, and those were only 240-260 GB drives. With modern drives flexing four to eight times the raw volume? You could reserve an entire one of those quarter-terabyte hard drives as spare area, and not miss the capacity drop much. I think hitting a write limit within a reasonable product lifetime is *vastly* overblown, under anything short of DC use.
|
# ? May 16, 2023 06:26 |
|
SwissArmyDruid posted:Right, and as I edited to add later: those SATA drives they tested, in some cases, wrote over 2 PB of continuous, nonstop writes over two years, and those were only 240-260 GB drives. With modern drives flexing four to eight times the raw volume? You could reserve an entire one of those quarter-terabyte hard drives as spare area, and not miss the capacity drop much. I think hitting a write limit within a reasonable product lifetime is *vastly* overblown, under anything short of DC use. I think we're in a moment of unique flash cheapness. Being able to get near-max-density 2280 TLC configurations for $250-280/4TB PCIe 4.0 w/DRAM is super duper cheap and this level may not be sustained. Good SAS U.2 drives are also starting to hit the market in volume, I think there is the WD 7.92TB or something that's had some good prices. Like yup probably can skate $20 cheaper or whatever but the nicer drive that isn't going to SLC-throttle and will wear harder isn't that much more expensive. And it may not stay that way forever. It could stay low through say Q4 but it might not, and I think the long term trend in 1y+ might well be higher as flash/DRAM production corrects. Hynix, Samsung, and Micron have decided you're cut off lol. Paul MaudDib fucked around with this message at 06:49 on May 16, 2023 |
# ? May 16, 2023 06:43 |
|
i especially like how after you write more than the SLC cache on a qlc nand ssd, performance drops to below hdd levels
|
# ? May 16, 2023 06:52 |
|
SourKraut posted:Wasn't this a myth? There was a lot of bluster about this a few years back, and I believe Intel clarified that it only entered read-only when the entire spare area was exhausted. It wasn't based on the writes hitting the specified endurance "limit". As always: 1) Read the data sheet (or god willing some kind of technical sounding document) instead of press release 2) Behavior may differ from product to product Example: https://www.solidigm.com/support-page/product-doc-cert/ka-00079.html quote:Percentage Used Estimate (Value allowed to exceed 100%) A value of 100 indicates that the estimated endurance of the device has been consumed but may not indicate a device failure. The value is allowed to exceed 100. Percentages greater than 254 shall be represented as 255. This value shall be updated once per power-on hour (when the controller is not in a sleep state). If the value reaches or exceeds 105, drive will enter a write protect mode with write bandwidth reaching <10MB/sec Quite cheeky of them to copy/paste the NVMe spec definition and tack on their on implementation at the end of it. NVMe spec definition: quote:Percentage Used: Contains a vendor specific estimate of the percentage of NVM subsystem WhyteRyce fucked around with this message at 07:05 on May 16, 2023 |
# ? May 16, 2023 06:58 |
|
Can anyone with a Solidigm SSD running their fancy unique driver test if BypassIO is still working? I used to use the provided driver for my HP SSD, because I thought why not, but I found out that BypassIO was only supported by the inbuilt windows driver and I don't know if that has changed. my brain still reads BypassIO as rhyming with "Mario" and I love it
|
# ? May 16, 2023 07:41 |
|
Not sure how to properly check that. I have an Optane DC P4800 and a 6TB DC P4608. The Solidigm storage tool reports them from having an intel driver, but that might be the RST controller. Windows reports it uses the builtin windows drivers.code:
|
# ? May 16, 2023 08:57 |
|
|
# ? May 24, 2024 05:38 |
|
Paul MaudDib posted:enjoy your ssd that switches off at the hard write limit. solidigm apparently inherited that poo poo from intel.
|
# ? May 16, 2023 14:36 |