Load Times

Community Support for the machines running the game
User avatar
thebs
Master
Master
Posts: 732
Joined: Sat Apr 16, 2016 6:49 pm
CMDR: thebs
CMDR_Platform: None Specified
Contact:

Re: Load Times

Postby thebs » Tue Nov 01, 2016 9:51 pm

RD-83 wrote:
thebs wrote:
RD-83 wrote:
Still want one but dont really need one. :D
For a home user? Probably not. You're usually running only 2-4 active programs/services at any time.

But on a server? Oh yeah, NVMe destroys AHCI. AHCI was designed for ATA's ST506 spindle legacy. NVMe is designed for random access. Although you can get some speed up with a lot of random reads on a single user system via NVMe, it really kicks butt on a multi-user system.

Just putting in a US$50 240GB NAND EEPROM (aka 'Flash') C: drive, even though it's using AHCI-SATA, where you load Windows and your most popular games suffices.

Of course, Windows 10 Anniversity Edition doesn't like it when you have a separate D: drive, and Microsoft tries to downplay it as a niche configuration. But that's what you get from a company that doesn't know how to design an operating system, and doesn't even bother maintaining its developers between releases (contracted/outsourced).
Image
CMDR TheBS - Yet Another Middle Aged American (YAMAA) and Extremely Casual Gamer often confused for a Total Noob

User avatar
JustSomeGuy
Master
Master
Posts: 573
Joined: Sun Oct 25, 2015 5:01 pm
CMDR: Murhapuro
CMDR_Platform: PC-MAC
Contact:

Re: Load Times

Postby JustSomeGuy » Tue Nov 01, 2016 11:24 pm

thebs wrote:Of course, Windows 10 Anniversity Edition doesn't like it when you have a separate D: drive, and Microsoft tries to downplay it as a niche configuration. But that's what you get from a company that doesn't know how to design an operating system, and doesn't even bother maintaining its developers between releases (contracted/outsourced).

I have drives up to F:. Oh and even still have a dvd-burner assigned to G:. And I could connect an external drive or two. I have considered this to be normal. Apparently I was wrong. :)

And no devs maintained between releases.. That just spells disaster. It is just a really stupid idea. I assume this is just so that the company (thinks it) saves $5 somewhere. Seen it before.. If an OK solution costs $10, and a really shitty born-dead solution costs $9,50.. the company will take the cheaper one, even if it costs a whole lot more indirectly and in the long run. But nobody thinks about that part, or even hears it even if they are told repeatedly before purchase decision.
I play on PC, in the Mobius PVE Eurasia group.
Image

User avatar
thebs
Master
Master
Posts: 732
Joined: Sat Apr 16, 2016 6:49 pm
CMDR: thebs
CMDR_Platform: None Specified
Contact:

Re: Load Times

Postby thebs » Wed Nov 02, 2016 12:46 pm

JustSomeGuy wrote:
thebs wrote:Of course, Windows 10 Anniversity Edition doesn't like it when you have a separate D: drive, and Microsoft tries to downplay it as a niche configuration. But that's what you get from a company that doesn't know how to design an operating system, and doesn't even bother maintaining its developers between releases (contracted/outsourced).

I have drives up to F:. Oh and even still have a dvd-burner assigned to G:. And I could connect an external drive or two. I have considered this to be normal. Apparently I was wrong. :)

Optical and removable drives shouldn't affect the issue. It's any general, fixed block device other than C: that seems to be the issue. It namely plagues those of us with a smaller C: NAND device to a bigger D: platter (or array).

Microsoft, as usual, doesn't care because most OEM's deploy their NAND+platter with Intel's SRT. But what people haven't been told is what happens when the NAND fails in SRT. It usually takes Windows with it, even though the platter is fine. I know, first-hand happened to my wife, and the result is always the same.

The fact that the NT kernel and its BOOTMGR cannot handle it is the problem in the first place -- both via SRT and separate NAND+platter.

JustSomeGuy wrote:And no devs maintained between releases.. That just spells disaster. It is just a really stupid idea. I assume this is just so that the company (thinks it) saves $5 somewhere. Seen it before.. If an OK solution costs $10, and a really shitty born-dead solution costs $9,50.. the company will take the cheaper one, even if it costs a whole lot more indirectly and in the long run. But nobody thinks about that part, or even hears it even if they are told repeatedly before purchase decision.

The main problem is that traditional, long-term, commercial software development and maintenance is dead. Everyone contracts and/or outsources and developers move on. This plagues MS Office as well, which is why several backers of Microsoft's ISO Office OpenXML (OOXML) standardization have filed documentation to null it as a standard, as Microsoft itself never adopted it (there are now 4 'transitional' versions that are not to spec).

Off Topic
In the Open Source world, there is no guarantee that someone will maintain things forever. But at least you have a 'heads up' when a key maintainer retires or moves on to another project. Fortunately many end up maintaining the software for decades, and you can find the culprits. Although the Linux kernel itself is pretty expansive at this point, which is why very few run the latest release. But at least the standards are followed, including with ISO OpenDocument Format (ODF) too.

In the latter case it helps that Boeing (the world's #1 documentation producer) sponsored the original OASIS standardizaton in 2001 (yes, 15 years ago) after 7 years of MS Office incompatibilities between versions (don't get me started, saw it from the inside), with Corel signing on quickly, and IBM following suit in the ISO standardization a few years later. That prompted Microsoft to rush to ISO with its own spec, even though it was 1/10th the size (and many old MS Office objects were better documented in the ODF filing), because of the UK's adoption of ODF, and then 'ballot stuff' using its corporate contacts. The UK has been more recently joined by Spain, Italy and, now it seems, Germany, on filings against the standard, and are moving back to ODF.

What the Mozilla (aka Firefox) guys are doing with the Rust language and the Redox OS has many in the Open Source world talking. The language, unlike Google's Go and others, is designed to be strongly typed and thread safe, without the traditional issues of such. We'll see if it catches on in phones, because it's clear Google is slowly losing mindshare with Android (which is just Linux-based, not GNU/Linux or pure Java for that matter), not just the forks like in Wearables, but how they are approaching the 'exclusivity.'

Huawei (the world's largest network ODM, and 3rd largest handset ODM for a long-time now) is continuing their break with them (long story), and if Huawei is looking elsewhere, a lot smaller Chinese, Taiwanese and Korean entities are as well. I don't see Korean Samsung's Tizen doing much more than some off-shoots like wearables, which many (including Huawei) are moving too. Something else will need to come in, and I'd love to see Redox OS with the Rust language be that solution.

It's a nice, clean break from the cruft of Linux, especially as more and more (stupidly) get away from GNU and Java standards in Linux implementations. Although the Linux kernel continues to be GPL and unfractured, the support is slowly not. Google knew this when it broke from GPL licensed GNU implementations, as well as the Sun-Oracle mandated Java Community Process (JCP). Which is why Android is so f'ing fractured and non-standardized.

But that's also my fear with Rust/Redox, it's not GPL, but MPL, and allows the same too. If Berkeley Software Distribution (BSD) taught us anything, from BSD UNIX to Ingress/Postgres code that makes up virtually all of the world's database origins (from Microsoft to Oracle), anything outside of GPL (or LGPL for libraries) means eventual fragmentation.
Image
CMDR TheBS - Yet Another Middle Aged American (YAMAA) and Extremely Casual Gamer often confused for a Total Noob

User avatar
thebs
Master
Master
Posts: 732
Joined: Sat Apr 16, 2016 6:49 pm
CMDR: thebs
CMDR_Platform: None Specified
Contact:

Re: Load Times

Postby thebs » Wed Nov 02, 2016 1:04 pm

*Al* wrote:I cannot speak Greek

ACHI is like having a lot of people in 1-2 lines with 2 TSA (or other airport authority) agents and their scanning machines, even though there are 100+ airplane gates.

NVMe is like having a train station with hundreds of gates, where each gate has multiple agents/scanners for people boarding those trains individually. Doesn't mean all trains are there, but the "TSA line" is not a freak'n bottleneck!

ACHI works fine for platter drives where the disk can only seek one set of sectors (the heads are usually synchronized to match across multiple platters in a drive). It absolutely sucks for NAND EEPROM (aka 'flash').

NVMe is far, far better for NAND EEPROM, especially reading. NAND EEPROM can be almost as fast as RAM at reads, even if writes are far, far slower (but can be buffered in on-board DRAM that most NAND devices include).

Off Topic
NVMe will also, finally result in centralized, NVRAM on a board.

Long story short, right now the biggest issue with 'data loss' (as well as power consumption) in an unplanned 'power off' event with a NAND drive is the fact that they usually have 1MiB DRAM per 1GB NAND. DRAM is power hungry and continually has to be refreshed, and doesn't last but a few dozen seconds with capacitors, and usually requires batteries to maintain state for hours.

So when everything is NVMe, you'll see even cheaper NAND devices because there will be little-to-no DRAM on-device. The DRAM will now be the system memory, and part of system memory will have a battery backup so it's NVRAM. This will not happen with Windows anytime soon, but is already available with non-standard PC systems running a standard GNU/Linux uEFI boot (and non-BIOS/non-uEFI firmware uBoot since the '90s before that).
Image
CMDR TheBS - Yet Another Middle Aged American (YAMAA) and Extremely Casual Gamer often confused for a Total Noob

User avatar
JustSomeGuy
Master
Master
Posts: 573
Joined: Sun Oct 25, 2015 5:01 pm
CMDR: Murhapuro
CMDR_Platform: PC-MAC
Contact:

Re: Load Times

Postby JustSomeGuy » Wed Nov 02, 2016 1:21 pm

thebs wrote:Optical and removable drives shouldn't affect the issue. It's any general, fixed block device other than C: that seems to be the issue. It namely plagues those of us with a smaller C: NAND device to a bigger D: platter (or array).

Microsoft, as usual, doesn't care because most OEM's deploy their NAND+platter with Intel's SRT. But what people haven't been told is what happens when the NAND fails in SRT. It usually takes Windows with it, even though the platter is fine. I know, first-hand happened to my wife, and the result is always the same.

The fact that the NT kernel and its BOOTMGR cannot handle it is the problem in the first place -- both via SRT and separate NAND+platter.

I don't think I have any Intel rapid storage things turned on in bios. Or uefi in this case. But now I do want to double check :)
I have Windows on an SSD and the rest of the drives are just more or less permanent storage, couple of HDD's and one "temp" SSD. Oh and one more SSD dedicated for Linux with very much manual dual boot, meaning "press F11 to choose which drive to boot from" - solution. Wanted to keep those two well separated. The important bits are backed up on externals, so in case of complete failure I'll be ok with the data. Of course it the whole house burns down with the ext drives.. But at that point the lost data is not on the top of the problem pile, and I do have the absolutely most critical files hidden in my phone too.
I play on PC, in the Mobius PVE Eurasia group.
Image

User avatar
thebs
Master
Master
Posts: 732
Joined: Sat Apr 16, 2016 6:49 pm
CMDR: thebs
CMDR_Platform: None Specified
Contact:

Re: Load Times

Postby thebs » Wed Nov 02, 2016 2:39 pm

JustSomeGuy wrote:I don't think I have any Intel rapid storage things turned on in bios. Or uefi in this case. But now I do want to double check :)

If you do, it actually 'prevents' the issue with the BOOTMGR loader and NT kernel, if your NAND+platter only appears as a single C: drive.
But if you have separate C: and D:, they have issues, like when RAID (RST/SRT) is turned off.

JustSomeGuy wrote:I have Windows on an SSD and the rest of the drives are just more or less permanent storage, couple of HDD's and one "temp" SSD.

That's the configuration that "confuses" BOOTMGR/NTkernel in NT6.4 (aka NT10).

JustSomeGuy wrote:Oh and one more SSD dedicated for Linux with very much manual dual boot, meaning "press F11 to choose which drive to boot from" - solution.

If you're using native uEFI (no CSM) to boot, then let me introduce you to my colleague Rod Smith's rEFInd Boot Loader. ;)
- http://www.rodsbooks.com/refind/

Rod Smith works at Canonical (Ubuntu) now, as much as I tried to get him hired at Red Hat (who moves way too slow on hires).

JustSomeGuy wrote:Wanted to keep those two well separated.
With native uEFI booting using the EFI System Partition (ESP), mutliple OSes coexist without issue. It's why I've gone 100% native uEFI since 2012 with Fedora/RHEL and Windows 7 x64, although Ubuntu and most other distros were also good by 2014.
Image
CMDR TheBS - Yet Another Middle Aged American (YAMAA) and Extremely Casual Gamer often confused for a Total Noob

User avatar
JustSomeGuy
Master
Master
Posts: 573
Joined: Sun Oct 25, 2015 5:01 pm
CMDR: Murhapuro
CMDR_Platform: PC-MAC
Contact:

Re: Load Times

Postby JustSomeGuy » Wed Nov 02, 2016 6:13 pm

thebs wrote:
JustSomeGuy wrote:I don't think I have any Intel rapid storage things turned on in bios. Or uefi in this case. But now I do want to double check :)

If you do, it actually 'prevents' the issue with the BOOTMGR loader and NT kernel, if your NAND+platter only appears as a single C: drive.
But if you have separate C: and D:, they have issues, like when RAID (RST/SRT) is turned off.

C: and D: are completely separate, all OS stuff is in C:, and D: is just a place to store stuff. Not even attached to the same controller.
Intel RST is disabled on my board, as well as "Intel Smart Connect Technology" which probably isn't related to this in any way.
I still have Win10 version 1511, and no idea when the anniversary update hits.

thebs wrote:
JustSomeGuy wrote:I have Windows on an SSD and the rest of the drives are just more or less permanent storage, couple of HDD's and one "temp" SSD.

That's the configuration that "confuses" BOOTMGR/NTkernel in NT6.4 (aka NT10).

I really should do a complete separation.. Dedicate one box for gaming on Windows only, and another box for everything else without Windows. I need to start planning that.

thebs wrote:
JustSomeGuy wrote:Oh and one more SSD dedicated for Linux with very much manual dual boot, meaning "press F11 to choose which drive to boot from" - solution.

If you're using native uEFI (no CSM) to boot, then let me introduce you to my colleague Rod Smith's rEFInd Boot Loader. ;)
- http://www.rodsbooks.com/refind/

Rod Smith works at Canonical (Ubuntu) now, as much as I tried to get him hired at Red Hat (who moves way too slow on hires).

CSM is enabled. It is my first and only uEFI board, ASrock Z77 Extreme4. CSM was enabled by default, and the text box next to it just said that 'do not disable unless running WHCK test'. And since I did not know what it is, and everything seemed to work fine, it was left enabled. Now I don't know what happens if I did try to disable it. Fearing the worst I have not tried.

I have seen that name rEFInd somewhere before, probably on the ArchWiki pages. Interesting, but sounds like I'm out of luck with that CSM.

thebs wrote:
JustSomeGuy wrote:Wanted to keep those two well separated.
With native uEFI booting using the EFI System Partition (ESP), mutliple OSes coexist without issue. It's why I've gone 100% native uEFI since 2012 with Fedora/RHEL and Windows 7 x64, although Ubuntu and most other distros were also good by 2014.

Anyway, if a disaster happens, it is a good reason to re-think and rebuild the whole setup the way I would like it to be today.
I play on PC, in the Mobius PVE Eurasia group.
Image

User avatar
thebs
Master
Master
Posts: 732
Joined: Sat Apr 16, 2016 6:49 pm
CMDR: thebs
CMDR_Platform: None Specified
Contact:

Re: Load Times

Postby thebs » Wed Nov 02, 2016 7:33 pm

JustSomeGuy wrote:
thebs wrote:
JustSomeGuy wrote:I don't think I have any Intel rapid storage things turned on in bios. Or uefi in this case. But now I do want to double check :)

If you do, it actually 'prevents' the issue with the BOOTMGR loader and NT kernel, if your NAND+platter only appears as a single C: drive.
But if you have separate C: and D:, they have issues, like when RAID (RST/SRT) is turned off.

C: and D: are completely separate, all OS stuff is in C:, and D: is just a place to store stuff. Not even attached to the same controller.

Correct! And that causes the Windows 10 Anniversary Edition issue! Microsoft requires your ...

\WINDOWS

to be on the same drive letter as all ...

\Program Files

\Program Files (x86)

If not, Windows 10 Anniversary Edition takes issue! Yes, it's beyond stupid.

I only mentioned RST/SRT because that is the only, Microsoft supported way you can have more than one device show up as a single C: drive where everything can be at. ;)

JustSomeGuy wrote:Intel RST is disabled on my board, as well as "Intel Smart Connect Technology" which probably isn't related to this in any way.
I still have Win10 version 1511, and no idea when the anniversary update hits.

Stay with it for now! ;)

JustSomeGuy wrote:
thebs wrote:
JustSomeGuy wrote:I have Windows on an SSD and the rest of the drives are just more or less permanent storage, couple of HDD's and one "temp" SSD.

That's the configuration that "confuses" BOOTMGR/NTkernel in NT6.4 (aka NT10).

I really should do a complete separation.. Dedicate one box for gaming on Windows only, and another box for everything else without Windows. I need to start planning that.

No, that's not the issue. It has nothing to do with Linux. It's 100% Windows side.

It has to do if you have more than 1 drive with installed "Program Files" in addition to the same drive where "Windows" is at. Yes, Microsoft now expects you to have only the C: drive for all programs. Only data can reside on D: and later, or whatever the drive letters are for not "Windows" and "Program Files" themselves.

Off Topic
JustSomeGuy wrote:
thebs wrote:
JustSomeGuy wrote:Oh and one more SSD dedicated for Linux with very much manual dual boot, meaning "press F11 to choose which drive to boot from" - solution.

If you're using native uEFI (no CSM) to boot, then let me introduce you to my colleague Rod Smith's rEFInd Boot Loader. ;)
- http://www.rodsbooks.com/refind/

Rod Smith works at Canonical (Ubuntu) now, as much as I tried to get him hired at Red Hat (who moves way too slow on hires).

CSM is enabled. It is my first and only uEFI board, ASrock Z77 Extreme4. CSM was enabled by default, and the text box next to it just said that 'do not disable unless running WHCK test'. And since I did not know what it is, and everything seemed to work fine, it was left enabled. Now I don't know what happens if I did try to disable it. Fearing the worst I have not tried.

You cannot change it once the OS is installed.

It has to do with disk label ...
- GUID Partition Table (GPT) instead of Master Boot Record (MBR), supporting unlimited number of partitions (although 128 seems to be the universally compatibile limit)
and layout ...
- Requires the EFI System Partition (ESP), Partition Type EF00h (yes, there are 4 hex digits to GPT, instead of 2 for MBR) that is formatted FAT*

*NOTE: the uEFI spec states FAT32 for non-removable, FAT12-16 for removable, but usually most do FAT12-32 without issue, although Windows installers sometimes take issue with FAT16, and create a 2nd ESP on a disk formatted FAT32.

It also requires a bootloader and OS that supports it. E.g., grub-efi (0.97 patched w/uEFI) or grub2-efi (1.99).

All Windows NT 6 releases in x64 version (Vista and later) support, it although I recommend NT 6.1+ (7+). PC OEMs are required by contract to ship Windows 8 x64 and later in native uEFI booting mode. This is because the Windows Licensing Key is stored in an uEFI area, so there is no label on the PC OEMs' box.

SIDE NOTE: Microsoft also requires a 128MiB Partiton Type 0C01h for "Reserved Data" where it can write uEFI/GPT specific information. Long, long story, Microsoft is quite specific on what it requires for GPT.

Red Hat backported uEFI boot support to GRUB 0.97, and many of the patches for 1.99 as well in Fedora and other distros. I was booting native uEFI on Fedora as early as 2010 (GRUB 1.9x) , and working with IBM in 2011 to fix early RHEL 6.0 GA support (backported GRUB 0.97), that made it into 6.1 and then 6.2.

JustSomeGuy wrote:I have seen that name rEFInd somewhere before, probably on the ArchWiki pages. Interesting, but sounds like I'm out of luck with that CSM.

rEFInd really addresses what most uEFI firmwares do not ...

A dynamic way of loading any OS bootloader/kernel in any EFI System Partitions (ESP) on any GPT disk in the system. Most uEFI firmwares require a specific 'label' in the uEFI firmare to target the exact disk, the exact ESP and the exact OS file.

If you've never done native uEFI boot, then you're probably not familiar with the details. In-a-nutshell, there no more MBR, just the ESP where files go. The ESP is mounted as
/boot/efi
in a Linux system. In fact, if you have an ESP, you shouldn't need a separate
/boot
any more. Unfortunately GRUB 1.99 is quite ignorant of many things, so a separate /boot is usually done by most installers.

JustSomeGuy wrote:
thebs wrote:
JustSomeGuy wrote:Wanted to keep those two well separated.
With native uEFI booting using the EFI System Partition (ESP), mutliple OSes coexist without issue. It's why I've gone 100% native uEFI since 2012 with Fedora/RHEL and Windows 7 x64, although Ubuntu and most other distros were also good by 2014.

Anyway, if a disaster happens, it is a good reason to re-think and rebuild the whole setup the way I would like it to be today.

As I mentioned, every system I've installed since 2012 has been native uEFI with GPT, including with Windows 7 x64. Despite what most people say, Windows 7 x64 is just fine installing native uEFI, as long as you have the required storage drivers.

The only kicker is that the stupid Windows USB Installation creator does not create an uEFI/GPT compatible USB device. So Rufus is what most people use, including myself.
Image
CMDR TheBS - Yet Another Middle Aged American (YAMAA) and Extremely Casual Gamer often confused for a Total Noob

User avatar
JustSomeGuy
Master
Master
Posts: 573
Joined: Sun Oct 25, 2015 5:01 pm
CMDR: Murhapuro
CMDR_Platform: PC-MAC
Contact:

Re: Load Times

Postby JustSomeGuy » Wed Nov 02, 2016 10:39 pm

thebs wrote:
JustSomeGuy wrote:I really should do a complete separation.. Dedicate one box for gaming on Windows only, and another box for everything else without Windows. I need to start planning that.

No, that's not the issue. It has nothing to do with Linux. It's 100% Windows side.

It has to do if you have more than 1 drive with installed "Program Files" in addition to the same drive where "Windows" is at. Yes, Microsoft now expects you to have only the C: drive for all programs. Only data can reside on D: and later, or whatever the drive letters are for not "Windows" and "Program Files" themselves.

That 2 machine setup is something I have been thinking about regardless, not because of this. Don't know if it's a good idea or not, but for now it sounds like a good one. :)

Anyway, currently I do have \windows and both \program files folders on C: but I do have most games installed elsewhere in custom folders. Like Elite for example, that is run from a traditional HDD and has no program files folder in it's path. Steam games are also on the same drive with Elite, but those are handled by Steam itself, and the steam app itself is in one of the program files folders, likely the x86 one. I hope these kind of installs are fine and won't make Windows stumble on its own feet.
I play on PC, in the Mobius PVE Eurasia group.
Image

User avatar
thebs
Master
Master
Posts: 732
Joined: Sat Apr 16, 2016 6:49 pm
CMDR: thebs
CMDR_Platform: None Specified
Contact:

Re: Load Times

Postby thebs » Wed Nov 02, 2016 10:53 pm

JustSomeGuy wrote:That 2 machine setup is something I have been thinking about regardless, not because of this. Don't know if it's a good idea or not, but for now it sounds like a good one. :)

I have a pair of $175 AMD quad-core, Zotac Zbox "nano" (5x5x2.5", 4x4" board) units with a 1TB SSHD in them for most of my Linux testing. They are now 2.5 years old.
- https://www.zotac.com/us/product/mini_pcs/overview

But I'm running Linux on this Dell Precision m4700 workstation notebook as I type this too.

JustSomeGuy wrote:Anyway, currently I do have \windows and both \program files folders on C: but I do have most games installed elsewhere in custom folders. Like Elite for example, that is run from a traditional HDD and has no program files folder in it's path. Steam games are also on the same drive with Elite, but those are handled by Steam itself, and the steam app itself is in one of the program files folders, likely the x86 one.

Which means you have a \program files folder on a drive other than the C: drive, correct? That's the bug with Anniversary Edition. ;)

JustSomeGuy wrote:I hope these kind of installs are fine and won't make Windows stumble on its own feet.

Again, I must be confusing you. The problem is having program files on anything but the same drive as Windows.
Image
CMDR TheBS - Yet Another Middle Aged American (YAMAA) and Extremely Casual Gamer often confused for a Total Noob


Return to “Hardware and Technical”

Who is online

Users browsing this forum: No registered users and 24 guests

i