264 karma · joined September 21, 2019
Sure they do. WD sold device managed SMR 3.5 disks without telling anyone. And then people noticed because of absolutely terrible raid re-silvering performance.
The consumer market is absolute trash for SMR devices, all the disadvantages with no benefit. But you can be sure that the enterprise sector gets host managed SMR drives at a better size/price ratio.
So consumer SSDs now depend on a limited dynamic part of flash being used as SLC to cache writes. Just that this dynamic part mostly gets smaller the fuller the device is.
And as soon as you do anything with large writes you will hit a brick wall of slowness, on full devices sooner then on empty ones.
Only device managed SMR is terrible. Host managed SMR is fine for data-hording if the savings in size/price are forwarded.
It hat ~350TB written, and according to Samsung is rated for 3 full device writes per day for 5 years = 17280 TB. So I have some wriggle room left.
They use lots of capacitors to hold written data in a dram and flush it out to flash on sudden power down. So they can optimize write operations for speed and low flash wear. Write speed and endurance is the part where consumer SSDs suck more and more ass.
Annoyingly many sellers do not put up the written amount of data. But I think that is more out of ignorance and less to try to dump worn down devices.
Then you have to use a raspberry PI or some other mini computer to get the same use out of it. And then why use a "smart" TV in the first place?
The only issues is the connectors. You need 8x PCI-E lanes (4x might work) or a SAS connector that nearly no consumer board has.
With Intels wave of TSX disabling patches going over all the previous CPU generations, AMD apparently got it right by never bringing their own version into production at all.
Traveling back in time to kill humans, indiscriminately from their answers to random questions he asks them beforehand!
An early version of this AI can be found in the Microsoft Windows privacy settings.
Why don't you complain about Apple not supporting Vulkan instead?
Some consumer SSDs do that, too. But with significant smaller amounts of dram.
This can intercept all the small writes that you e.g. do for metadata of the filesystem and via mapping can combine them into very few big writes. Or writes to the same position over and over again only need to touch the flash when the devices is powering down.
Are there any QLC enterprise SSDs? All the cheaper used enterprise SSDs I looked up where using TLC so far.
iServer with 64 cores and 256 GiB DDR4 for only $ 699,999.95
It uses C++20 features and macros. The macros limit it to a POD count of 8.
I can create you a shitty linked list that fits perfectly in L3, but still has terrible cold cache performance because each individual cacheline has to be pulled in one by one.
Nothing a few fuses can't solve.
> Makes sense for massive quantities of crappy cheap flash, not so much for this.
Coming from the x86 world, they use plenty of fuses for everything. I can't imagine a reason why they would not absorb production variation with this for esp32 dies.
A sudden implosion of demand combined with a flooding of the used GPU marked. Better not plan on bringing out any new GPU architectures when your warehouses are still full of the previous generation.