Steamをインストール
ログイン
|
言語
简体中文(簡体字中国語)
繁體中文(繁体字中国語)
한국어 (韓国語)
ไทย (タイ語)
български (ブルガリア語)
Čeština(チェコ語)
Dansk (デンマーク語)
Deutsch (ドイツ語)
English (英語)
Español - España (スペイン語 - スペイン)
Español - Latinoamérica (スペイン語 - ラテンアメリカ)
Ελληνικά (ギリシャ語)
Français (フランス語)
Italiano (イタリア語)
Bahasa Indonesia(インドネシア語)
Magyar(ハンガリー語)
Nederlands (オランダ語)
Norsk (ノルウェー語)
Polski (ポーランド語)
Português(ポルトガル語-ポルトガル)
Português - Brasil (ポルトガル語 - ブラジル)
Română(ルーマニア語)
Русский (ロシア語)
Suomi (フィンランド語)
Svenska (スウェーデン語)
Türkçe (トルコ語)
Tiếng Việt (ベトナム語)
Українська (ウクライナ語)
翻訳の問題を報告
I highly doubt that. Your GPU doesn'ty have any interaction with the drives... textures and ops are read into system ram and then copied into VRAM. And again this is done in fixed regular intervals like when you enter a new map area. The only reason your Drive would affect your FPS is if its constantly paging data which basically means your system needs more RAM.
HAve't play skyline but I've yet to encounter a game that made such an oddly ass backwards use of Datastreaming.
[quote[hdd's are for storage, nuff said...[/quote]
And games require a ♥♥♥♥♥♥♥ of storage space :).
I wish you all well and know of no other way to hopefully help folks learn about what has been in Linux since the beginning and is a big factor in its speed and not needing optimizations.
In my days coding we had sequential and random access text files. Sequential had to load into RAM and there wasn't much in virtual memory on micro computers at that time, so if they file exceeded RAM, you couldn't load it.
Random Access text files required the coder to set the number of fields, the type of field, string, integer, floating point, etc., and also the length of the record itself. If you exceeded the record size, oh well you can't write the record to the database, and this also meant there were a lot of empty spaces left where information wasn't filled in.
Sequential files would have no index (unless the coder made one) and often due to the size of the database itself, you had little room to execute the search, so often the records took more time to bring up. If the file fit in memory, however, a shell sort would be almost instant.
With Random Access files finding a record was almost instantaneous no matter how big the database became because the record sizes were all uniform, and any index was specific to a field of fixed length etc. vs the code, essentially, "reading" the file looking for a particularly searched string.
Preallocation is the equivalent of a Random Access system with each file being a record so there is no fixed length, but setting specific allocation based on the DEVELOPERS view that these locations will be optimal.
Essentially the great uniformity that helped Random Access find records faster is what's being implemented, with an allocation system that assures no need to optimize the files.
Where it gets unwieldy is when the number of files is huge, essentially making the file index a significant sized database all its own, and that only exacerbates delay when coupled with the verification of each file's integrity prior to each preallocation, again this is pursuant to the Developer's specifications in how they structured the files for their product and the optimal access to those files.
This is the best I could do to try to explain what knowledge I am operating from.
Good luck to the OP and anyone else with is issue and I do hope this helps better appreciate and understand where to focus your complaints about it, to the Developer.
I suggested plenty of methods to start investigating upstream. This far zero reports on any of them even attempted. Others can be figured too.
The small amount of actual info we have suggests no disk activity in the "stuck" phase, that implies preallocation is already over. procmon timestamps could provide hard proof.
I just received an update for total warhammer II, a 6.5GB update, and experienced nothing out of the ordinary about it. That doesn't mean something is preallocated at the start of steam's launch when updates are queried based on library contents.
It could be that the update process itself is prepared ahead of the actual start of the download, though. But I have no idea, so I thought I'd offer my anecdotal experience to see if it matches others experiences.
yet modders and devs told me what i said, which is what matters as they know more than you do.
your comments are nonsense, stop trying to derail the thread.
where do you think textures ect.. come from, they come from the files in the game installed on your drive,which are then read when needed, so yes the gpu does interact with the drive, it has a huge factor in how quick the gpu gets its info, thats why you can easily see a difference in render times and fps on low end systems going from hdd to a ssd (heck even an msata).
tbh, unity engine is quite bad, for some reason the devs using unity seem to think forcing use of paging file is good, they also use this awful process called garbage collection which causes stuttering and fps drops and when the stutter is to long it will cause rubberbanding / desync, which is a huge problem in rust and everyone complains about it.
they may require a good deal of storage space but ssd's have gotten much bigger, no reason to be clinging to old hardware such as hdd's, unless ofc your using it strictly for storage (ie.. pictures, movies/vids, recordings, any sort of text files, or backups), gaming on them is just not standard now a days and rightfully so and the only people i see using them for gaming is budget builders or people still clinging to old outdated pc's, i guess i can say data servers as well, but we are talking gaming here.
After an initial pre-allocation, there's not much you can do to keep files from fragmenting if your update actually expands the size of the file on disk. You have to assume that your update comes days or weeks later, and that the user has added new files to disk in that period, so you can expand the file via immediate/unfragmented preallocation but short of running defrag on the whole disk first you can't guarantee that the new data isn't fragmented from the original.
On the upside, unless you're doing a major overhaul with added features and content (or you're just a really bad developer) most of your patches and updates should fit into the address space that's already there. You're fixing little things, changing a +1 to a +2 for example to fix a bug or make the "Easy" gameplay level even easier, so while the file changes its length does not.
lol
Having rebuilt the entire system from the ground up, there is a huge improvement in performance. However, while occurring much less, there is still some latency in the Pre-allocation process from the Steam platform. Other platforms do not experience such latency.
Start of comment:
I have finally finished my restore, whew! You really have no idea everything that has happened during all of this time, but I finally finished what we have set out to test. I will list what I did, along with my results.
Steps taken:
I began by deleting the RAID 5 array, separating the drives into 5 singular 10TB drives, and have spread my Steam library equally across 4 of them. But after having deleted the drive array, I began having issues with Windows itself after a large Microsoft update that occurred at the same time. So, instead of trying to figure out the Windows issue, or even uninstalling the update, it dawned on me that perhaps a fresh install of the OS might also aid in our testing, so at that point the C drive was wiped clean, and a fresh Windows install was performed, and in keeping with the test, I did not install any AV or any other programs for that matter. I then installed Steam. My C drive remains a RAID 0 across two M.2 SSDs, as I did not have issues with it, however, should the primary system drive respond poorly again, to another Windows update or in general, I will remove that array as well, and run the OS on a singular drive as well.
I began restoring my games library from the backups I had created. Although this takes up a large amount of space, and in my case I had to buy 3 USB 8TB, I highly recommend this for folks with large libraries. If I had downloaded my entire library, it would have taken me at least two weeks to complete, as it stands the restore took me less than 4 days.
On the very first batch of restores, I experienced high latency during the Pre-allocation process for SOME games. I also found that the issue appears to happen with those same games repeatedly, ie: the aforementioned Age of Fear.
HOWEVER, I see huge improvements across the board in several areas. For starters the Pre-allocation process, for the games that do still experience the problem, the latency is in fact improved. Such as during the first batch of restores, it included all of the As (restored as I backed up, alphabetically), which also included all of the Assassin's Creed games. The pre-allocation for some of those particular games was far more than some other games of like size, between 20-30 minutes. On severe offender is Remnant From The Ashes, no matter what is happening in that game the Pre-allocation process if vastly exaggerated as compared to other games of like size. For instance, a 340MB update for that game might take upwards of 30 minutes to complete.
Secondly, and this is a big one, I am no longer experiencing the broken up downloads, that is the stopping and starting up of retrieving data from the Steam servers. This issue alone had, at times, added hours of time to my downloading of games and updates. While it happened nearly every time something was downloaded, to what extent fluctuated greatly. Now, that is not to say that this issue shall not return, but I certainly expected to see it during the restore process, and I did not observe it happening one single time.
Third overall system wide performance appears to be greatly improved. Since Steam is literally the only thing installed, outside of Windows and the required drivers, it is the only thing I have to base my observations on. But, the disk activity to the drives that were once members of the array, from inside of the Steam app, has improved by no less than 4x. Whereas, under the array I would see Disk Activity measured at 54MB/s and below, Steam now lists this same activity at sustained between 180-220MB/s, with bursts to over 300MB/s.
My take away:
While the pre-allocation process latency is still occurring, it is much less exaggerated than previously experienced. This latency appears to be relegated to particular games now, where previously it affected those same particular games as well as countless others. However, I still am not experiencing a like issue from any of the other gaming download sights, this issue still appears to only affect Steam downloads. I know someone had asked if I had the same game to test across two separate platforms, and in fact I do, I have several, some through the Steam-GOG cross platform offerings, and others simply by mistake or intentionally purchased for other reasons, and I tested a few. None of these games have any issue when downloading/updating from other platforms, such as Singularity on GOG, but does experience some pre-allocation latency in Steam. Using the on-board RAID option appears to impact this issue greatly. Given that every previous system I experienced this issue on had an array of some sort, it appears that may well have been the source of many of my frustrations. It would seem that something Steam does inside of the Pre-allocation process differs from what other like platforms are doing, and is much more sensitive to the environment in which it operates.
Other areas of the system performance have vastly improved, and overall I am much happier with the performance of my system. To date I have had very few updates (previously I experienced near daily updates that exceeded 20 games a day), and I will report back over time as to if this remains the status quo. I will be happy to run any particular benchmark/analytics anyone might wish to see, just tell me what it is you would like to see specifically, and I can provide it.
So, there you have it, things are greatly improved across almost all areas, but the pre-allocation latency does still exist in some games, however to a much lesser extent than previously experienced. I do not experience a 30 minute wait before any game download begins inside of other platforms, but I do experience this, at times now, during the "Pre-allocation" process of Steam. However, even that 30 minutes is a vast improvement from the previous experience, where some games would take literal hours. I am much happier with my system's overall performance, as well as that of Steam, than I was when I started this thread!!
What I am reading is: "I had some things that didn't install or take properly when I first installed them, not sure what but in doing a full restore and making some significant changes I apparently fixed those things and maybe by some updates also went around some bottlenecks I didn't know I had before. Everything is better but I still see issues with Steam, but not all of Steam but particular games."
That's what I got out of it and pointing to this because it's very possible that the high number of files of one developer and the pre-allocation of those files to avoid optimization will take a lot longer than a game that has many less files. So this is the expected result to me and I am glad things are better, I am sorry they aren't perfect, and I strongly suggest taking up the issue with the Developer of each of the games with long pre-allocation as they need to know their method is creating a problem with their users and that it may even translate into lag and other issues during game play, ways they can do a better job, pick a better scheme or flowchart per se ("method" is a coding term and often gets confused if used in normal conversation).
So glad you got it restored and things are better and hope these Devs will hear you out on what they need to do to make their games pre-allocate better :)
And the RAID5 write penalty for much of the system as a general rule. It may be smoothed over with disk caching (meaning--3rd party software or something like IRST ssd caching) but won't solve problems caused by externalities--like strange tactics for the pre-allocation.
The fact that the system was rebuilt and many issues went away helps demonstrate the problem isn't entirely the fault of programs experiencing the worst performance hits.
It is an old tactic to blow away data on a disk array (such as a RAID) and restore from tape or other backup--it's then restored in a contiguous file sequence all at once, and is written to the disk in a mostly unfragmented state, unless efforts were made to start a bunch of different copies at once in an effort to use cpu threads or something to 'speed the restore up'.
In any event, it is good to hear the OP with similar hardware to mine is now performing a I'd have otherwise seen for myself--I don't have the RAID 5 (I made a RAID 10 out of 10TB drives instead; an upgrade from 1TB drives that had been holding a steam library for over ten years) and it only got faster as a result of the upgrade. Everything else in the OP's hardware list generally is similar. Here is hoping both my and the OP's RAID do not slow down over time! (And if it does, that the fix is easy...)
Also clarify what you mean by 'restore' in reference to those games.
Otherwise you are to be commended for your exhaustive efforts. If nothing else this thread may become the basis of a KB article.
You are correct Sir, the onboard RAID contributed to the issue, at least in part. My drives are Seagate Barracuda Pro 10TB 7200rpm 6g/bs ith 256MB of cache.
The restore I was referring to s the process of restoring games that were previously backed p sing the Steam utility.
Hopefully his may help someone, that might have a similar setup, and perhaps save someone a tremendously long time of frustration. I know I could have used this quite awhile ago.
Thank you for your help!