|
repiv posted:Doing it at the drive level isn't ideal though, since you'd need to shift uncompressed data over PCIe in that case. To match what the consoles are doing we'd need decompression built into the CPU so it can stream compressed data in over PCIe and dump it straight into RAM. I think it is more a for encoding/transcoding video so the raw data is streamed and it encodes and stores on the fly. Additionally it could transcode for streaming out lower quality if streaming to a handheld device vs a full 4k. Probably a bunch of different use cases.
|
#
?
May 14, 2020 00:17
|
|
|
|
| # ? Apr 16, 2024 11:22 |
|
|
I assume that it really kicks rear end for huge sequential transfers like maybe level loading and getting an equivalent hardware decompression system in the PC ecosystem will be a near impossible task
WhyteRyce fucked around with this message at 01:32 on May 14, 2020 |
|
#
?
May 14, 2020 00:38
|
|
|
It's not impossible but standardizing and shipping a compression codec and interface across all NVMe controllers (if drive-side) or Intel and AMD (if cpu-side) is going to take years. Don't hold your breath.
|
|
#
?
May 14, 2020 00:52
|
|
|
Remember the UE3 Samaritan demo, the UE3 DX11 demo, the UE4 demo, and every other previous UE demo has ended up being "this is what real-time cutscenes will look like in the next 1-2 years, and what actual gameplay will look like 3-6 years from now." Epic is in the business of selling engines, and you better believe money changed hands so that this demo was on a PS5 and not an XboxXXXXX. So take both the demo and the whole PS5 SSD is godly with a grain of salt. Be excited that next-gen games will be built for SSD type storage and that PCs will undoubtedly see improvements as well. Don't run out to buy a PCIe g4 NVMe drive now, because games that fully utilize that bandwidth are not gonna be here on day 1. The transition is probably gonna be slow and games will have a lot of cross-gen overlap ala the 360/PS3 for a year or so, for the same reasons that gen had.
|
|
#
?
May 14, 2020 00:59
|
|
|
Remember when games were 32bit and single threaded and designed to load from a spinning optical disk.
|
|
#
?
May 14, 2020 01:08
|
|
|
repiv posted:It's not impossible but standardizing and shipping a compression codec and interface across all NVMe controllers (if drive-side) or Intel and AMD (if cpu-side) is going to take years. Don't hold your breath. Oh I know theoretically it wouldn't be impossible but getting the entire ecosystem to agree is for all intents impossible
|
|
#
?
May 14, 2020 01:22
|
|
|
repiv posted:It's not impossible but standardizing and shipping a compression codec and interface across all NVMe controllers (if drive-side) or Intel and AMD (if cpu-side) is going to take years. Don't hold your breath. I can't see how it would be that hard. Most useable place would probably be to have hardware or GPU assisted decompression in GPUs, since most games will probably have a lot of data to be fed to the GPU and this data is quite similar across large amount of games, so a single compression algorithm works for all of them. If you have advanced enough GPU the game can send the compressed data straight to it. If the GPU can't handle it but drivers are new enough, then the GPU driver uses CPU to decompress the data. If also the drivers are too old, them the game will have to do the decompression on CPU.
|
|
#
?
May 14, 2020 09:33
|
|
|
DrDork posted:It's about how Sony has baked a decompression engine into the SoC. Games are apparently going to be delivered as pre-compressed blobs. The SSD (which is fast on its own) just spits out compressed data onto the PCIe interface, then it gets decompressed going into RAM. The result of all that is a theoretical effective maximum read bandwidth of up to 9GB/s, which is indeed considerably faster than what you can get out of a normal consumer PC SSD. Thanks, that’s really interesting. I know that SSD controllers already perform transparent compression, but this takes it to the next level. It will have some interesting implications on the whole consumer PC market if next gen titles use high bandwidth to stream very large, detailed assets from storage in real time. Technically current PCs could end up in a situation where not CPU or GPU but the storage subsystem (no matter if M.2 or SATA) ends up being a bottleneck for graphic detail. Obviously we are years from seeing that happen but the possibility wasn’t even on my radar until recently.
|
|
#
?
May 14, 2020 12:17
|
|
|
taqueso posted:Ok, so apocalypse happened. we really need to watch superman though. how do we decode this lost video content. Use a few of the billions of bluray drives and billions of 90% destroyed bluray discs and run a bunch of error correction passes to make sure everything agrees; or do we try to find one of the one known readers of the quartz cube That's like literally the plot of the Mechwarroir franchise. Though replace 'perfectly ok-ish movie about a guy' with 'priceless military and industrial secrets'. Also that DRM is poo poo, and good luck fixing poo poo when all the parts are tamper proofed black boxes with 80 shitzillion bit keys. eames posted:Thanks, that’s really interesting. I know that SSD controllers already perform transparent compression, but this takes it to the next level. It'll be real interesting seeing a game system that performs BETTER than an equivalent PC for more than a year or so. Even with PCIe4 SSDs, that kind of technological chicanery will take time to overcome. Now I wanna see a side-by-side on Call of Shooty: Unreal 5 Tech Demo Edition, where even on a top-end PC, the textures load in like the bad parts of Halo 2. Methylethylaldehyde fucked around with this message at 12:22 on May 14, 2020 |
|
#
?
May 14, 2020 12:18
|
|
|
eames posted:It will have some interesting implications on the whole consumer PC market if next gen titles use high bandwidth to stream very large, detailed assets from storage in real time. Technically current PCs could end up in a situation where not CPU or GPU but the storage subsystem (no matter if M.2 or SATA) ends up being a bottleneck for graphic detail. Obviously we are years from seeing that happen but the possibility wasn’t even on my radar until recently. Well, consoles have often performed favorably compared to affordable PCs for a little bit after they're launched before PCs catch up. Though this time they do seem to be making a more aggressive performance jump, and finally fixing some of their bottlenecks that have really held things back. I think we'll get much better games because of it, even if more people than usual find they need to upgrade. Before people get to "the sky is falling" over he PS5 SSD, I think it's worth remembering that a big part of why SSD read times matter at all on a console is because they have historically been specced out with anemic amounts of RAM, and simply couldn't afford to actually cache all the assets needed into memory. PCs have generally had a lot more freedom on that front (A 1080 has almost as much VRAM by itself as a PS4 has total RAM), and while the next-gen consoles are bumping up to 16GB, it won't be long before PCs are way ahead on that front again--a lot of gaming builds already suggest 16GB min, and even the 1060 has 6GB VRAM, and those numbers are only going to go up. So there's a lot you can do with intelligent asset caching if you have the space for it that can get around a "slow" drive subsystem. The other part, as I already noted, is that XBox isn't doing anything nearly as fancy, so expect the vast majority of games aren't really going to take full advantage of it in the first place. And while they don't really exist quite yet, there's nothing really stopping SSDs from getting faster once PCIe 4.0 actually takes hold. A PCIe 3.0 x4 slot (which is normally what's used for NVMe drives) maxes at ~4GB/s, which is why that's more or less where the top-end drives have stopped for a bit now. PCIe 4.0 x4 goes up to ~8GB/s, so expect to see drives that take advantage of that in a year or two when it's in more systems.
|
|
#
?
May 14, 2020 14:31
|
|
|
Lol 149 sold. https://www.ebay.com/itm/Samsung-86...deuKTx&LH_BIN=1
|
|
#
?
May 14, 2020 20:01
|
|
|
Hmmm, selling for 1/3 the price of major retailers from an account with no sales for years and then suddenly has 200+ SSDs? And posted in German and Euros despite the account being listed as a US-based account? Seems totally legit. In for 3!
|
|
#
?
May 14, 2020 20:09
|
|
|
That reminds me of the usb drive from china someone that supposedly had a hdd in it but just had a flash drive spoofing the storage space with a large metal nut in it to provide the feeling of weight. Will try to dig up a picture. E: wasn’t this pic but similar  GIS for “fake usb hdd” is hilarious 
priznat fucked around with this message at 20:14 on May 14, 2020 |
|
#
?
May 14, 2020 20:11
|
|
|
Someone had to make that. Hey boss, why are we gluing in bolts?
|
|
#
?
May 14, 2020 20:24
|
|
|
they're more consistent than rocks
|
|
#
?
May 14, 2020 20:29
|
|
|
priznat posted:GIS for “fake usb hdd” is hilarious that one looks like it would be suspiciously heavy also gising fake ssd is extra funny because some of the results are people who opened up a legit SSD, and the 1/4th size PCB makes them think they have a fake
|
|
#
?
May 14, 2020 20:32
|
|
|
Have you guys never seen fake capacitors
|
|
#
?
May 14, 2020 21:14
|
|
|
there's a smaller one inside that, right?
|
|
#
?
May 14, 2020 21:18
|
|
|
Thats odd. The value is all wrong and the smaller cap is a Rubycon which is very high quality.
|
|
#
?
May 14, 2020 22:02
|
|
|
bold of you to assume it's not a fake rubycon
|
|
#
?
May 14, 2020 22:06
|
|
|
I like to pretend it's a nesting doll of fakes
|
|
#
?
May 14, 2020 22:08
|
|
|
repiv posted:bold of you to assume it's not a fake rubycon I know but what the gently caress is the point of hiding a fake one in a fake cap. The mind boggles.
|
|
#
?
May 14, 2020 22:53
|
|
|
redeyes posted:I know but what the gently caress is the point of hiding a fake one in a fake cap. The mind boggles. If you’re making the external fake shell you just use the cheapest poo poo that’ll work well enough to get away with the scam for a while before the victim figures out their device is behaving oddly and failing at higher rates because the caps aren’t up to spec. Fake Rubycon fits the bill!
|
|
#
?
May 14, 2020 23:33
|
|
|
Yeah but im thinking a used Rubycon.
|
|
#
?
May 14, 2020 23:36
|
|
|
Hey, those eWaste recyclers have to sell their poo poo to someone, right?
|
|
#
?
May 14, 2020 23:40
|
|
|
Bought a somewhat cheap 850 Evo (500G/$60). Lets see if its fake. Magician can tell right? Assuming everything else checks out.
|
|
#
?
May 14, 2020 23:45
|
|
|
Shaocaholica posted:Bought a somewhat cheap 850 Evo (500G/$60). Lets see if its fake. Magician can tell right? Assuming everything else checks out. That might be a cheap used 860 EVO
|
|
#
?
May 15, 2020 00:00
|
|
|
BobHoward posted:If you’re making the external fake shell you just use the cheapest poo poo that’ll work well enough to get away with the scam for a while before the victim figures out their device is behaving oddly and failing at higher rates because the caps aren’t up to spec. Fake Rubycon fits the bill! You know those Saved By the Bell episodes where Zach spends so much effort trying to cheat on the exam that you think he could have just studied instead for the same result. That's what I think about scammers sometimes
|
|
#
?
May 15, 2020 00:34
|
|
|
redeyes posted:I know but what the gently caress is the point of hiding a fake one in a fake cap. The mind boggles. If you're making fake caps, you probably have other fake caps on hand?
|
|
#
?
May 15, 2020 04:18
|
|
|
Might have a real one to hand, and just don't gaf because you're working in a sweatshop.
|
|
#
?
May 15, 2020 04:43
|
|
|
Is it a stupid idea to sell my Samsung 860 EVO sata drive to buy an equivalent capacity Inland Premium NMVE drive? There's not that much price difference and the Inland drive has like 6-7x the r/w speed.
|
|
#
?
May 16, 2020 00:49
|
|
|
You probably won't notice the speed difference in practice, IMO just keep the 860 until you're ready to move to a larger capacity.
|
|
#
?
May 16, 2020 00:55
|
|
|
What differences would I find between: - Western Digital SN750 2 TB M.2-2280 NVME Solid State Drive - HP EX950 2 TB M.2-2280 NVME Solid State Drive - ADATA XPG SX8200 Pro 2 TB M.2-2280 NVME Solid State Drive other than a big swing in prices? My applications are test VMs with occasionally high R/W I/O load, GaMiNg, Fusion 360, software compilation. (I don't actually know how those ended up being my short list, but I assume it's from either this thread or the PC Building one.)
|
|
#
?
May 20, 2020 15:33
|
|
|
Subjunctive posted:What differences would I find between: None whatsoever, all 3 are comparable and at the 1TB size are usually priced about the same. WD charging an extra $60 at the large size is big brands doing big brand things. (The EX950 and the Adata 8200 Pro are even the same controller and are identical in performance.
|
|
#
?
May 20, 2020 16:08
|
|
|
Klyith posted:None whatsoever, all 3 are comparable and at the 1TB size are usually priced about the same. WD charging an extra $60 at the large size is big brands doing big brand things. Thank you!
|
|
#
?
May 20, 2020 16:43
|
|
|
Klyith posted:None whatsoever, all 3 are comparable and at the 1TB size are usually priced about the same. WD charging an extra $60 at the large size is big brands doing big brand things. I recall reading (from NewMaxx, the SSD guy on reddit) that SN750 will still maintain decent performance when nearly full, which isn't the case for the 8200 Pro (as it relies on a dynamic SLC cache to achieve it's performance, but if the drive is close to full, most of that cache space is instead being used for storage). Check out the empty vs full performance here. Though even if that's the case, that might not matter much for a lot of workloads. $60 premium for the 2TB SN750 though is kind of crazy considering at 1TB these models are all around the same price.
|
|
#
?
May 20, 2020 19:31
|
|
|
Splinter posted:I recall reading (from NewMaxx, the SSD guy on reddit) that SN750 will still maintain decent performance when nearly full, which isn't the case for the 8200 Pro (as it relies on a dynamic SLC cache to achieve it's performance, but if the drive is close to full, most of that cache space is instead being used for storage). Check out the empty vs full performance here. Though even if that's the case, that might not matter much for a lot of workloads. $60 premium for the 2TB SN750 though is kind of crazy considering at 1TB these models are all around the same price. On the Canadian pcpartpicker right now, it's showing the SX8200 Pro at $360 and the SN750 at $498--so a pretty big swing! I don't think it'll get that full; if it does I'll fish out an older SSD for another TB of space or something.
|
|
#
?
May 21, 2020 20:23
|
|
|
I probably should have asked here before spending money... Group opinion on the Western Digital Blue 1TB? My laptop has a 256GB m.2 drive with room for a 2.5in. Went to download GTA5. 90+GB laughed in my face. It's an older machine and absolute speed wasn't a priority, I think the m.2 drive is still SATA anyways. I know it's not bleeding edge, but it was cheap enough on Amazon. I remember buying my first SSD, a 240GB something, for like $250 dollars, so a 1TB for around 100 from a brand I know seemed like a reasonable impulse purchase.
|
|
#
?
May 23, 2020 00:10
|
|
|
It's good. The only "uh oh" would be if you bought the older, non-3d NAND version, as the performance isn't as good.
|
|
#
?
May 23, 2020 00:32
|
|
|
|
| # ? Apr 16, 2024 11:22 |
|
|
Actuarial Fables posted:It's good. The only "uh oh" would be if you bought the older, non-3d NAND version, as the performance isn't as good. Says it's "3D Nand". Well that's good to know. I'm so behind on my ssd knowledge. Like "Have you heard of that fancy MLC flash storage" behind. Are there any current generation drives that fail catastrophically and early these days?
|
|
#
?
May 23, 2020 00:58
|
|