Root Access

@NotDog Users looking for value too :P

Not Dog

@Bob Yeah

Anyway, must sleep. My life's a fuckup.

G'night.

'night

@NotDog :(

12:30 AM

Users looking for value would just use the onboard sound card

@DavidPostill Enya-like typing detected.

@ThatREDACTEDGuy that's irish gaelic.

I just realised how many directors of foo + 'special duties' we have 0_0

@JourneymanGeek Audiophiles looking for value? :P

maaaaybe

But then they'd just buy schitt or some chinese brand

12:58 AM

> Enya doesn’t write any lyrics. They’re written by Roma Ryan. The main languages used are English, Irish Gaelic, and several albums feature a song in Latin.

@JourneymanGeek Oh Google Translate just told me @DavidPostill speaks Scottish Gaelic.

Does that mean he's Scottish? I always though he was from the UK.

..... Scotland is part of the united kingdom.

I mean its the United Kingdom of England, Scotland, Wales and Northern Ireland right?

DAMN RECURSIVE COUNTRIES

How is it recursive? Then it would be the United Kingdom of the United Kingdom of the United....

I'm only used to countries being single countries.

That is confusing.

Well, this diagram helps a little.

I never knew the difference between Britain, England and UK.

Somehow my mind unconsciously tagged it all as being the same country.

Scotland was just "some country in Europe with all the kilts and bagpipes"

And Ireland was one entire country.

And what is a Wales?

Ben N

1:17 AM

Wait, only part of Ireland is in the UK?

<-- fulfilling the "Americans are bad at geography" stereotype

@BenN Dude. Didn't you hear? I thought England, Britain and UK were the same country.

Well, I kinda passively-thought, but still.

1:46 AM

@ThatREDACTEDGuy I thought Queen Elizabeth was Queen of New England, and that the UK was just a region of New England

@allquixotic: this is nearly finalized, but I'm wondering what suggestions you have here:

PCPartPicker part list: https://pcpartpicker.com/list/xFv2wV
Price breakdown by merchant: https://pcpartpicker.com/list/xFv2wV/by_merchant/

CPU: AMD - Ryzen 5 1600X 3.6GHz 6-Core Processor ($240.98 @ Newegg)
CPU Cooler: Corsair - H100i v2 70.7 CFM Liquid CPU Cooler ($109.99 @ Newegg)
Motherboard: Asus - CROSSHAIR VI HERO ATX AM4 Motherboard ($258.98 @ Newegg)
Memory: G.Skill - Ripjaws V Series 32GB (2 x 16GB) DDR4-3200 Memory ($249.99 @ Newegg)
Storage: Samsung - 960 PRO 512GB M.2-2280 Solid State Drive ($330.98 @ Newegg)

@BenN yeah. They had a very frank and very impolite discussion over that at one point. Folk still disagree but its gotten more civil, most of the time.

A variety of options are available, including faster processor, faster graphics, more storage (larger boot SSD and/or separate SATA storage SSD), Blu-ray drive, larger PSU (up to 1000W), and custom liquid cooling.

> Do not allow the laser light emitted from the Touch Unit to pass through or be reflected by any optical object, such as a magnifying glass or mirror.
Continued use of the Touch Unit may result in fire or accidents, and also cause injury

I'm kinda tempted to drop to 750W for the power supply, though.

1:56 AM

0_0

@bwDraco what are your priorities and budget?

Multipurpose, balanced gaming/workstation/productivity/everyday system.

Budget still not yet final but estimated at about $2000-2500.

Not Dog

Cats

@NotDog meow

@bwDraco with a budget of $2500, imo the 1080 is a no-brainer

with $2k, I'd still try and fit at least a 1070 into it

and unless super heavy I/O like 8k video editing with complex pipelines is a top priority, you can get more SSD space for much cheaper by sticking with SATA III ~550 MB/s drives like the 850 series

also, that build doesn't need more than a 550W PSU :P

2:12 AM

Yeah. The biggest problem is that it'll lock me into a graphics card vendor (especially because there's also a dynamic-refresh-rate monitor option), but I don't quite think Vega will have a big enough advantage over high-end GeForce (assuming it has one at its launch price) for this to be an issue.

The option that's most likely to be exercised is the graphics card upgrade.

I think AMD is sadly too little too late on Vega; they've delayed their competition to Pascal for so damn long that they've effectively forfeited their top-end and even mid-tier market, and in so doing, all but doomed FreeSync to failure unless Nvidia actively embraces FreeSync

G-Sync and Nvidia appear to be the way to go on desktops for the foreseeable future unless Vega does to the GPU market what Ryzen did to the CPU market -- but I'm very doubtful, because Volta is around the corner and Vega is still December or later

My mind is set on an AMD Ryzen processor, though.

I'm ok with Ryzen -- I think Ryzen is far more disruptive, far more competitive (especially on cost), and far more successful than Vega has hope of being

go for it

@JourneymanGeek Do not taunt the happy fun ball.

Despite the single-threaded performance deficits, one of the key philosophies behind this build is that it needs to be a platform for the future. The Socket AM4 platform provides room to scale for future hardware without the need to do too much surgery. This also explains the choice of premium system components throughout, from the big HAF X case to Seasonic's flagship PSU to a blisteringly-fast SSD.

2:17 AM

@bwDraco 400W would be enough.

(I only went with higher for silent operation at low load.)

If a second graphics card is to be added, this would create a need to replace the PSU, but that's a bit unlikely.

@bwDraco Even then 600-700W is probably enough, no?

Modern graphics cards aren't that power-hungry.

The difference between PSUs at this point is only about $30-40.

@allquixotic I keep seeing your surname come up on the dc.syd list :P

@bwDraco Eh, if your budget allows, go for it.

I'm wondering if the budget will be high enough to allow a GTX 1080 Ti, the highest of the graphics card options...

(that's +$400-$500)

Then again, if I had this much money to build the system, I might as well exercise the gaming monitor option...

Note that if no monitor is being purchased, I'll be using my existing Acer H226HQL, a low-cost 21.5" e-IPS display currently used as a second monitor for the Dragon. I'd much prefer to have two displays, though.

Feb 7 at 3:42, by bwDraco

I can always reuse my cheapie Acer 1080p IPS display, but I'd prefer a pair of 1440p IPS gaming monitors with G-SYNC. (sorry AMD, I'm probably not going to get a Radeon, though I am strongly considering Ryzen)

2:31 AM

@bwDraco ehhhhhhhh get a standard 24" :P

So, the renovations on this house are more extensive than initially anticipated. I don't think the system will be built until August, though exact dates are not known.

@Bob with a 17" display, the Bonobo WS workstation has headroom for peripherals (USB, etc.) with a 650W PSU brick for dual laptop-TDP'ed 1080s, don't know if that means desktop ones would fit comfortably in that too

or maybe they would as long as you don't OC/overvolt

@allquixotic desktop should fit, yea. IIRC something in the region of 200W per GPU

2:46 AM

180W.

With overclock, expect a peak load of 250W per card.

brb.

(the laptop GTX 1080 has a 165W TDP)

3:07 AM

@bwDraco 180W is nominal, 200W is what I would consider a fairly safe peak. iirc anyway

I usually round up when doing PSU calcs

What I do is take the nominal ratings and make sure the load does not exceed 80% of PSU capacity during a worst-case workload.

More room is typically allowed for future upgrades.

it depends on the PSU

some are rated very conservatively and can run at their theoretical peak or slightly higher for 10-15 years with good efficiency and no negative stability issues

others are rated "generously" and fall over if you run them near their peak for more than a minute or two

or run very hot / very inefficiently near their peak

Seasonic or the high-end Corsairs are pretty much going to work, and unless you're sticking more than 2 GPUs in there or run a lightning bolt through it, you're not going to be able to make it fail even if you try.

The rating on any name-brand PSU is a continuous workload rating. With 80 PLUS certification requiring that the supply hit certain efficiency targets, most supplies have at least 10% headroom beyond the nominal rating for short periods (something like no more than 15-30 seconds per 5-10 minutes).

I think, because PSUs are expensive and potentially hazardous to human health if they're badly made, good manufacturers of PSUs use the "elevator design philosophy": determine what the limits are, add in safeties for good measure, then add in safeties on top of safeties, then add in a little more just to be absolutely certain it won't fail

elevator with a placard that says 1500 lbs can handle quite a bit more on a regular basis no problem :P

> Please contact the website owners to inform them of this problem.

"Hey Bob, your site's down. Thanks, Bob"

3:14 AM

@Bob lol

This headroom is there partly because if the power supply was running at its true peak capacity, its efficiency would not pass the certification requirements.

Also, high-end PSUs are typically thermally limited based on the parts selected; stronger components raise the point at which efficiency and power quality fall off a cliff. As such, added cooling could increase maximum effective capacity. FSP, one of the world's biggest PSU OEMs, demonstrated a concept 1200W PSU with integrated waterblock for attachment to a custom liquid-cooling system.

> It is noteworthy that along the highlights of its product FSP also mentions something called “power overclocking technology”. Based on what the company told us, once LCS is applied, it “enables” the PSU to “handle a 1400 W output".

3:44 AM

So... an update on the Bifrons SSD write offloading experiment. With the profile stored on a portable SSD and using CrystalDiskInfo's logging function, I've determined that on this system, Firefox does about 2-4 GB of writes in a typical day of web browsing. Most of this consists of random writes of varying block sizes.

With the BarTab Heavy extension, memory usage is reduced by automatically unloading inactive tabs, which itself reduces writes to the fragile eMMC by an estimated 0.5-2 GB per day (depending on how many tabs I have open). All told, this diverts about 3-5 GB of writes per day away from the eMMC. At 64 GB of capacity, that's one drive write's worth every two or so weeks.

It's not a huge difference, but when you have cheap planar TLC NAND whose low endurance is compounded by low capacity, the more I/O you divert from the eMMC, the better. Your thoughts?

I've said it before. Stop worrying about "fragile" NAND. It survives fine for everyone else.

Take backups, stop messing around, and if it fails then get it replaced.

That's what I'd do anyway.

4:01 AM

as has evvvverryyyooonnneeee

@Bob though, in the unlikely event it fails, you can't 'just' replace it

@JourneymanGeek ?

@JourneymanGeek Yea, get the machine replaced (under warranty) :P

@Bob EMMCs

But it probably won't fail that quick.

nope

@JourneymanGeek ? "evvvv-"?

4:08 AM

Furthermore, it's almost certainly DRAMless. This will make the endurance problem even worse.

tomshardware.com/reviews/dramless-ssd-roundup,4833.html

> Unfortunately, DRAMless SSDs also have a sinister side. Updating the map directly on the flash requires small random writes, which takes a bite out of the SSD's endurance. This is a particularly vexing issue with low endurance planar 2D TLC NAND flash.

XKCD

Magnetohydrodyanmics combines the intuitive nature of Maxwell's equations with the easy solvability of the Navier-Stokes equations. It's so straightforward physicists add "relativistic" or "quantum" just to keep it from getting boring.

> [...] one SSD vendor told us about an OEM 2D TLC SSD that will burn through the rated endurance in a little over a year. The SSD has to last a year because of the notebook's one-year warranty, but anything beyond a year's worth of use is up to the user to fix. Tactics like that are the driving forces behind putting cheap DRAMless SSDs in $500 notebooks.

That's what makes endurance a very real concern.

Most likely, this is with 32 GB of storage, not 64 GB, but still, endurance is a concern.

Feb 12 at 17:56, by bwDraco

The trend towards cheaper SSDs has been a very disturbing one. First it's TLC NAND with poor sustained write performance, now it's DRAMless designs with poor random write performance and low endurance (due to the need to maintain the mapping tables in NAND).

@Bob or just replace the machine.

@JourneymanGeek That too.

Cause those things are disposably cheap

4:15 AM

I've still got that old tablet... the one actually known for NAND issues, and previously-refurbished. Vivotab Smart.

May 18 at 1:27, by bwDraco

Well, some machines have as little as 32 GB of NAND, which not only means half the raw NAND endurance, but more write amplification, meaning that real-world endurance is probably going to be a third or less of that of a 64 GB device. 64 GB is better than most in the sub-$300 laptop space.

And I really don't care for jumping through hoops to try to extend lifetime.

Yup

I guess endurance is a non-problem.

and Its worth remembering most of these systems do little but web browsing

4:16 AM

If it's that shoddy I probably don't want it anyway.

@bwDraco yup

And if you really care enough to make the tradeoff, turn off browser caching.

lol

Everything's gonna be hella slow and your data usage will go up, but... hey, you saved your disk! Yay!

and manual updates. and...

4:17 AM

browser.cache.disk.enable, have fun

An important factor in endurance is the fact that this system has more storage than most sub-$300 laptops. Again, double the raw NAND endurance, plus far less write amplification because the drive isn't constantly running close to full.

Best part? Even if you turn off the disk cache, the RAM cache will still work. Granted, it probably won't be very big.

Apparently the default for 2 GB RAM is a 24 MB memory cache.

lol

Configurable though.

@Bob oh and virtual memory....

4:21 AM

Interesting that apparently the default for 32 MB RAM is 2 MB cache. Like... they planned for 32 MB RAM?

The average consumer does not care about this and will probably load the eMMC up to the point where there's only a few GBs left. This means extremely high write amplification, and yet it still needs to work for at least a year.

I find your lack of faith disturbing.

"extremely high write amplification" => we're not in 2010 anymore...

With 64 GB of storage that's nowhere near full, I doubt endurance is a real issue.

I guess there's nothing to worry about.

Typical consumer use case: 3 GB/day on a near-full 32 GB eMMC, so we'll assume 5x WA (which is actually a bit conservative). That's 15 GB/day, which is about 0.5 DWPD; with TLC NAND, this should be good for about 2-3 years. Just how in Oblivion are they exhausting the SSD's endurance inside of an 18-month period?

Rahul2001

Good morning

If I reply to a really old message from the transcript of a user who isn't active or pingable, do they still get a notification?

@Rahul2001 No. Mods can use what's called a superping, but that's only used when the mod must be able to contact the user without doing something more drastic like a moderator private message.

Rahul2001

4:33 AM

@bwDraco ok thanks!

hmm. I've actually never tried with a direct reply

unfortunately floofy has been here too recently to test :P

Only if the user is recently active.

To wear out a TLC NAND SSD within an 18-month period would probably require at least 1 DWPD. I suppose that the random writes forced by the DRAMless nature of the SSD are compounding the endurance problem, especially when the write amplification is already quite high in a typical consumer use case with a 32 GB eMMC module...

@bwDraco that's for a normal ping. Don't think anyone's ever said anything about replies

I believe the same holds for replies to very old chat messages.

Hi folks. I just tried to boot my machine in UEFI mode, but got this message:

> the system cannot find the UEFI driver for the add-on network devices [0x10D38086]

I'm running Linux, but I'm not sure if that is relevant. I'm not sure if my system even got to the Linux part.

4:36 AM

@FaheemMitha Did you get a GRUB boot menu?

@bwDraco No, nothing.

I put that error message into Google, and precisely nothing came up.

@FaheemMitha Apparently, you have an old NIC that doesn't support UEFI.

This shouldn't happen with a recent machine, but it's not quite impossible.

@bwDraco Hmm. But how can the BIOS tell that before even entering the operating system?

@FaheemMitha It's the UEFI firmware that's throwing the error.

Are you using the built-in networking or an add-in card?

@bwDraco Ok. But where is it getting this from?

@bwDraco I think it's add-on. Let me check.

4:39 AM

10D3:8086 is an Intel Gigabit Ethernet card.

> 0a:00.0 Ethernet controller: Intel Corporation 82574L Gigabit Network Connection

@bwDraco Wow, you're well informed.

Looked up 10D38086, and it dawned upon me that this is a PCI ID.

@bwDraco Sorry, I should have said that was the output of lspci.

Should I be looking it up another way?

And is UEFI support at the level of firmware? Excuse my ignorance.

@FaheemMitha UEFI is the system firmware.

Because clearly the MB doesn't know or care about Linux drivers.

4:42 AM

It can run in a BIOS compatibility mode, though.

@bwDraco Right. I meant, how is it checking the card is not supported?

@bwDraco Hmm. I'm not sure what that means. I just turned off that compat module thing.

CSM?

@FaheemMitha Compatibility Support Module. That's the BIOS compatibility layer.

@bwDraco Right. It was enabled, but I disabled it because the Debian installer kept booting in legacy mode.

I'm trying to move to UEFI.

Jeez, what's with the animated paperclip?

2

Would a question about this be on topic on SU? I don't think it's on topic for U&L.

It looks your network card's option ROM does not support UEFI. An add-on card can have its own executable code for use in the preboot environment. On a network card, this is typically for PXE. UEFI requires redesigned option ROMs for this purpose but your NIC was designed for legacy BIOS systems and will not work on a UEFI system that doesn't use the CSM.

You will need to replace your NIC.

@FaheemMitha ^^^

@bwDraco Thanks, though that is a bummer. Shall I post this as a question, in case it helps someone else?

Is upgrading this "option ROM" an, er, option?

4:52 AM

So... looked at intel.com/content/www/us/en/support/network-and-i-o/…. Your network card is an Intel® Gigabit CT Desktop Adapter.

@FaheemMitha Try this: downloadcenter.intel.com/download/19186/…

Yes, I see it.

E46981-xxx Intel® Gigabit CT Desktop Adapter
EXPI9301CT
EXPI9301CTBLK

RJ45, 10/100/1000, PCIe*, 82574L
VendorID:8086, DevID: 10D3

Latest drivers

(via downloadcenter.intel.com/product/50395/…)

> The Intel® Ethernet Flash Firmware Utility (BootUtil) is a utility that can be used to program the PCI option ROM on the flash memory of supported PCI and PCIe*-based Intel® Network Adapters and to update configurations.

@bwDraco So is this one of those situations where you try to flash a firmware upgrade while closing ones eyes tight and praying?

There's your answer. Feel free to post a question on the main site anyway, and I'll post an answer summarizing this chat session. That way, we get reputation for this and more people will understand what's going on and how to fix it.

@bwDraco I'll do so. Unless it's a dupe.

And on a scale of 1 to 10, how risky is this flash upgrade?

4:56 AM

As long as it isn't interrupted, you should have no trouble. If it does fail, you're just out the network card. A GbE NIC is pretty cheap anyway.

@bwDraco superuser.com/questions/1219822/…

If you think that wording could be improved, let me know. I think I'm going to go back to sleep now. Thank you very much for your help.

Well, not go back to sleep immediately, but in a few minutes...

._.

MS distributes the office offline installer as... a .img file

whaaaaat?

@bwDraco Somehow I had the impression you were one of the mods. But apparently not - no diamond thingy.

@bwDraco IIRC the onboard NICs are also connected via PCI

@Bob Quite unlikely if it's a UEFI board.

4:59 AM

@bwDraco Is the linux tag relevant in this case? The error message clearly isn't coming from anything to do with Linux.

@bwDraco How so?

When I say PCI, I include both PCIe and Conventional PCI

@Bob There's no reason for a UEFI board to have an integrated network adapter whose option ROM doesn't support UEFI.

@bwDraco Yes, but... that's got little to do with my statement that onboard NICs are also connected via PCI...

Though I might've misinterpreted your original message.

*shrug*

@bwDraco Let me know if there is anything else relevant I should add to the question. Thank you.

And do I need Windows to do the flashing? It doesn't look like it, but I don't have Windows.

Never mind. I see I need to download Preboot.tar.gz.

5:17 AM

@bwDraco I take it back, the linux tag is relevant. Since it appears from the Intel docs that I can do the firmware upgrade from inside linux, if I understand this correctly.

I'd say so.

Hmm... I may be mistaken. UEFI allows drivers to be loaded from the ESP.

I'm not familiar with this process, but I suspect you can put the drivers supplied with the package into /boot/efi (which is where most Linux installations mount the ESP) and try again.

Running it from /usr/local/bin gives the message:

> Connection to QV driver failed - please reinstall it!

downloadmirror.intel.com/19186/eng/bootutil.txt

> This option ROM includes PXE, iSCSI, FCoE, and UEFI
drivers, and the image is programmed to the flash memory at once.

Presumably it's looking for something and not finding it.

So, flashing the firmware adds the necessary drivers to the card itself.

5:25 AM

If this flash fails, presumably there is a high priority it will brick the card. Is that a fair statement?

You need root to do low-level hardware stuff like this- are you sure you're running this with elevated privileges?

@bwDraco Yes, I ran it as root.

root@orwell:/home/faheem# bootutil64e -?
Connection to QV driver failed - please reinstall it!

Then a whole bunch of other stuff.

Breakfast. Back in a bit.

Try modprobe iqvlinux.

You may need to build the QV driver yourself to flash the firmware - sourceforge.net/projects/e1000/files/iqvlinux

> In order to run this tool on Linux*, the base driver must be installed on the system.

(from the release notes)

This should be part of your distro. Are you sure the e1000 or e1000e driver is loaded?

@bwDraco Yes.

root@orwell:/home/faheem# lsmod | grep e1000
e1000e                212128  0
ptp                    17692  1 e1000e

wiki.debian.org/Firmware#Devices.2FDrivers_Firmware

Hmm...

5:39 AM

@bwDraco Yes, I saw that. But I don't know what base driver it's talking about. Anyway, since this is supposed to be handled in Linux, this probably makes it on-topic for U&L.

The base driver is supposed to be e1000e.

@Bob, any ideas?

13 mins ago, by Faheem Mitha

root@orwell:/home/faheem# bootutil64e -?
Connection to QV driver failed - please reinstall it!

@bwDraco That sounds relevant. Can you add that to the answer, please?

@bwDraco That's definitely loaded.

@bwDraco get a new card?

@Bob lol

use the boot-time flash, not the from-OS flash

5:43 AM

Probably not a bad idea.

@Bob From inside the MB boot utility?

BTW, any recommendations for a new network card to get? I've found Intel very trouble-free in the past, so I'd probably prefer to go that way.

hmmm

trouble with intel nics?

This looks relevant - federicofr.wordpress.com/2011/07/01/…

@djsmiley2k Not trouble, exactly.

But my card doesn't support EUFI, it seems.

Oh right

might be that Debian ships with a old driver?

what kernel you running?

uname -a ?

uname -a
Linux orwell 3.16.0-4-amd64 #1 SMP Debian 3.16.43-2 (2017-04-30) x86_64 GNU/Linux

@djsmiley2k And it's Debian stable, so yes.

yah, hat's rather a old kernel

Tho, it seems @bwDraco has already figured ou tthe issue

5:52 AM

@djsmiley2k yes, it looks like it needs this qvlinux driver thing installed. Yay.

So, recommendations for network cards, anyone?

Like I said, I've always used Intel, so I tend to prefer them.

Why not just install the driver?