Whether you genuinely believe what you’re saying, or are just trolling, you’ve managed to nerd snipe me. Whether or not you read all of this doesn’t matter, I’ve channelled my inner no-life tech nerd just because I can. There’s a TL;DR at the bottom :)
Not really, or in the kilobytes at best.
Just for proof, I’ve booted my near-fresh Fedora 40 install just to write this comment. When I finish writing the comment, I’ll place the total write operations, amount written, and time period at the end of the comment as the final lines.
Shadowplay isn’t a thing on AMD nor Linux, so no.
AMD: has AMD Relive, which, last I checked is a team red solution with near feature parity to shadowplay
Linux: has a few options actually. I’m currently using a lovely flatpak package called GPUScreenRecorder. Is it perfect? No, but does it work when I need it to? Yes!
Every SSD has limited write cycles, no matter how good it is. I spent a good amount of money on my Samsung SSDs, that doesn’t mean I’m not going to conserve their lifespan with useless “features”.
First, I’ll expand on what I said last time.
Every SSD does have limited write cycles, here you are correct. I was never implying that they don’t. My point was, the write cycle limit is such a non issue now that this sort of fear mongering over what is generally a negligible amount of data is no longer necessary (I’ll come back to this). I’ve pulled some data from my own systems, as well as re-reviewing Backblaze’s AFR report.
My Anectodal Evidence:
/dev/sdb (Install Location for Fedora 40)
Model, Capacity, and “Warrantied Writes”: Samsung 850 EVO, 500GB, 150TBW
Power On Hours: 18517 (Approx just over 2 years)
TB Written: 51.89TB
Wear Levelling: 93% of the drive is still “healthy”
Other Raw SMART Values that would indicate failure: 0
/dev/sdb (Steam Install location since 2017, Drive shipped with faulty controller and was not RMAd as an experiment for my own curiosity)
Model, Capacity, and “Warrantied Writes”: WD Blue ‘3D NAND’, 1TB, 400TBW
Power On Hours: 8979 (Approx 1 year)
TB Written: 18.06TB
Wear Levelling: 100% of the drive is still “healthy”
Other Raw SMART Values that would indicate failure: 0
Backblaze’s Data: https://www.backblaze.com/blog/ssd-edition-2023-mid-year-drive-stats-review/
This data isn’t perfect as these SSDs aren’t necessarily primary write locations, however, as stated, they are the Boot drives for Backblaze’s servers, as well as the log and temp file storage for each server. To keep this long comment short, as of 06/30/2023, for (consumer) drives with 100+ models, with 10k+ days, they report:
2884 Drives
2,293,595 Total drive days
45 Drive failures
0.72% Annual failure rate
Seems pretty non-concerning to me.
Now to get back to that “negligible amount of data” I mentioned earlier. Obviously we don’t yet know how much steam’s solution will use… but using some of my old shadowplay clips as a reference…
5 minute clips @ 1080/60 = 1.8GB
Lets take that 1.8GB and assume you’re playing 5 hours of games per day, every day, and you’re only using that drive for your Shadowplay/AMD Relive/GPUScreenRecorder/Steam whatever cache…
Is that a lot of data? Yeah, it honestly is. But most people don’t play games 5 hours/day every day. And in this hypothetical situation where that’s the only use of this drive, it’d still take a little under 8 years to reach the TBW warranty rating that most (trusted) 500GB SSDs currently have.
TL;DR:
SSDs aren’t as fragile as you think they are
If the feature is useless to you, then disable it
If you want to use the feature, but you’re really that worried about your drive life, use ImDisk and set your caching locating as a Ramdisk
We’re finally at the end! Here’s the stats from atop (It’s no 1.8GB/5min, but it’s definitely more than you think):
Total Time: 1h36m19s
Total Write operations: 103318
Average MB Written per second: 0.5
Total MB Written (Rounded down to nearest 0.1 MB): 2584.2MB
Edit: Fixed the typos I noticed immediately after hitting post
Always nice when a nerd snipe target takes the bullet like a champ and returns to sender with interesting data to share. I can safely tell you at least one person enjoyed reading this :^)
Whether you genuinely believe what you’re saying, or are just trolling, you’ve managed to nerd snipe me.
Such an accusation isn’t a good start if you want to have an actual discussion. Besides, I don’t see anything in my comment that could be interpreted as trolling.
AMD: has AMD Relive, which, last I checked is a team red solution with near feature parity to shadowplay
Not on Linux. AMD drivers are part of the kernel and there’s no additional bloated driver software like you see it under Windows.
And in this hypothetical situation where that’s the only use of this drive, it’d still take a little under 8 years to reach the TBW warranty rating that most (trusted) 500GB SSDs currently have.
Which just proves my point. But I guess you all swap out your hardware every 2 years or so. That’s cool for you rich people I guess but as a poor person I typically try to preserve and keep my hardware for as long as I can. I think the major fallacy for you guys here is that you just look at this one dataset, as if that’s the only thing that would go on in your system. Especially under Windows you’d obviously see a lot more activity but the point works generally in that I’m not going to use unnecessary crap that wears down my very expensive hardware for gimmicky features that I maybe could use once every 5 years or so, assuming I wouldn’t just forget about it anyway.
Such an accusation isn’t a good start if you want to have an actual discussion. Besides, I don’t see anything in my comment that could be interpreted as trolling.
I worded this more strongly than I intended to. This was meant to come across more lightheartedly and joking than the opening to my original message. I apologize for this, sincerely.
Not on Linux. AMD drivers are part of the kernel and there’s no additional bloated driver software like you see it under Windows.
Again, like I said on the line underneath that, while not perfect, GPUScreenRecorder exists for Linux and works well, despite being less fully featured. And although I didnt include it in the previous message, it supports Nvidia, AMD, and Intel GPUs.
I’m not sure that it does given that I provided a pretty ridiculous hypothetical that still results in an 8 year period at minimum before reaching the warrantied write count. The drive could fail before then, or it could fail after, but the point was that it’ll still likely exceed the recommended replacement period for a spinning HDD (3-5 years).
But I guess you all swap out your hardware every 2 years or so. That’s cool for you rich people I guess but as a poor person I typically try to preserve and keep my hardware for as long as I can.
Please dont make this out like I’m rich and can afford to just throw away hardware. If one of my drives fails, I can’t afford to replace it right now. But also, I havent needed to replace an SSD that wasnt faulty due to a manufacturing issue, either in my own system or in any of the hundreds or possibly thousands of computers at this point that I have installed SSDs in for work. The shit SSDs will die within a year or two, but thats usually due to having faulty controllers and is unavoidable, and the faulty SSDs (like the first batch of Samsung 870s) are, well, faulty.
I think the major fallacy for you guys here is that you just look at this one dataset, as if that’s the only thing that would go on in your system.
I’m not sure what “one dataset” I provided but I’m more than happy to provide more :)
I’m not going to use unnecessary crap that wears down my very expensive hardware for gimmicky features that I maybe could use once every 5 years or so, assuming I wouldn’t just forget about it anyway.
Okay then, like I said at the end of the last post, just turn the feature offdont turn the feature onwhen it’s released. My point isn’t that you personally need to use this feature, but rather that your SSD lifespan fears are outdated. You’re more than welcome to use your hardware however you like, I was never intending to change that :)
Edit: for clarity and inclusion of a correction pointed out by @asexualchangeling@lemmy.ml
Whether you genuinely believe what you’re saying, or are just trolling, you’ve managed to nerd snipe me. Whether or not you read all of this doesn’t matter, I’ve channelled my inner no-life tech nerd just because I can. There’s a TL;DR at the bottom :)
Just for proof, I’ve booted my near-fresh Fedora 40 install just to write this comment. When I finish writing the comment, I’ll place the total write operations, amount written, and time period at the end of the comment as the final lines.
AMD: has AMD Relive, which, last I checked is a team red solution with near feature parity to shadowplay
Linux: has a few options actually. I’m currently using a lovely flatpak package called GPUScreenRecorder. Is it perfect? No, but does it work when I need it to? Yes!
First, I’ll expand on what I said last time.
Every SSD does have limited write cycles, here you are correct. I was never implying that they don’t. My point was, the write cycle limit is such a non issue now that this sort of fear mongering over what is generally a negligible amount of data is no longer necessary (I’ll come back to this). I’ve pulled some data from my own systems, as well as re-reviewing Backblaze’s AFR report.
My Anectodal Evidence: /dev/sdb (Install Location for Fedora 40)
/dev/sdb (Steam Install location since 2017, Drive shipped with faulty controller and was not RMAd as an experiment for my own curiosity)
Backblaze’s Data: https://www.backblaze.com/blog/ssd-edition-2023-mid-year-drive-stats-review/ This data isn’t perfect as these SSDs aren’t necessarily primary write locations, however, as stated, they are the Boot drives for Backblaze’s servers, as well as the log and temp file storage for each server. To keep this long comment short, as of 06/30/2023, for (consumer) drives with 100+ models, with 10k+ days, they report:
Seems pretty non-concerning to me.
Now to get back to that “negligible amount of data” I mentioned earlier. Obviously we don’t yet know how much steam’s solution will use… but using some of my old shadowplay clips as a reference…
5 minute clips @ 1080/60 = 1.8GB
Lets take that 1.8GB and assume you’re playing 5 hours of games per day, every day, and you’re only using that drive for your Shadowplay/AMD Relive/GPUScreenRecorder/Steam whatever cache…
1.8GB * 12 (for 1 hour usage) * 5 hours * 365 Days = 39.42TB/year
Is that a lot of data? Yeah, it honestly is. But most people don’t play games 5 hours/day every day. And in this hypothetical situation where that’s the only use of this drive, it’d still take a little under 8 years to reach the TBW warranty rating that most (trusted) 500GB SSDs currently have.
TL;DR:
We’re finally at the end! Here’s the stats from atop (It’s no 1.8GB/5min, but it’s definitely more than you think):
Edit: Fixed the typos I noticed immediately after hitting post
Always nice when a nerd snipe target takes the bullet like a champ and returns to sender with interesting data to share. I can safely tell you at least one person enjoyed reading this :^)
Yeah it kinda… Just kept expanding as I was writing it. I’m glad I was able to provide an interesting read to at least one person though!
Such an accusation isn’t a good start if you want to have an actual discussion. Besides, I don’t see anything in my comment that could be interpreted as trolling.
Not on Linux. AMD drivers are part of the kernel and there’s no additional bloated driver software like you see it under Windows.
Which just proves my point. But I guess you all swap out your hardware every 2 years or so. That’s cool for you rich people I guess but as a poor person I typically try to preserve and keep my hardware for as long as I can. I think the major fallacy for you guys here is that you just look at this one dataset, as if that’s the only thing that would go on in your system. Especially under Windows you’d obviously see a lot more activity but the point works generally in that I’m not going to use unnecessary crap that wears down my very expensive hardware for gimmicky features that I maybe could use once every 5 years or so, assuming I wouldn’t just forget about it anyway.
I worded this more strongly than I intended to. This was meant to come across more lightheartedly and joking than the opening to my original message. I apologize for this, sincerely.
Again, like I said on the line underneath that, while not perfect, GPUScreenRecorder exists for Linux and works well, despite being less fully featured. And although I didnt include it in the previous message, it supports Nvidia, AMD, and Intel GPUs.
Here’s the link to the project if you’d like to investigate it for yourself: https://git.dec05eba.com/gpu-screen-recorder/about/
I’m not sure that it does given that I provided a pretty ridiculous hypothetical that still results in an 8 year period at minimum before reaching the warrantied write count. The drive could fail before then, or it could fail after, but the point was that it’ll still likely exceed the recommended replacement period for a spinning HDD (3-5 years).
Please dont make this out like I’m rich and can afford to just throw away hardware. If one of my drives fails, I can’t afford to replace it right now. But also, I havent needed to replace an SSD that wasnt faulty due to a manufacturing issue, either in my own system or in any of the hundreds or possibly thousands of computers at this point that I have installed SSDs in for work. The shit SSDs will die within a year or two, but thats usually due to having faulty controllers and is unavoidable, and the faulty SSDs (like the first batch of Samsung 870s) are, well, faulty.
I’m not sure what “one dataset” I provided but I’m more than happy to provide more :)
Okay then, like I said at the end of the last post, just
turn the feature offdont turn the feature on when it’s released. My point isn’t that you personally need to use this feature, but rather that your SSD lifespan fears are outdated. You’re more than welcome to use your hardware however you like, I was never intending to change that :)Edit: for clarity and inclusion of a correction pointed out by @asexualchangeling@lemmy.ml
deleted by creator
Even better!