*OFFICIAL* Alienware 17 R2/R3 Owner's Lounge

Guttsy,

I have to disagree with that. Buffer flushing, as stated by Samsung and many others is totally unnecessary for a desktop or laptop.

My SSD's in RAID 0 have a throughput of over 3GB/s, the data is on the disk in microseconds, so to end up with any corruption, it doesn't even need a cache let alone flushing of it, I would have to be running the laptop as a Web Server for Amazon with hundreds of users connected, the disk usage would need to be far higher than the processor and memory could manage (it would overheat if run at that level continuously) before there was a need to be concerned.

As Samsung and all the other optimisation articles advise, turn it off on desktop/laptop to avoid the unnecessary extra disk requests.

 

The white paper you quoted on the previous page states the exact opposite, "Samsung strongly recommends that Write-Cache Buffer Flushing be enabled at all the times."

How caches are stored and committed to SSD's versus spinning disks may have a bearing on the likelihood of FS corruption in the event of power failure or crash, but given that Samsung themselves strongly advise to enable cache flushing to avoid data loss I would argue the risk is still quite real.

I mainly work on Linux systems but I understand Microsoft file systems can suffer from the same issues with cache commit failures. Enterprise disks have firmware that respects controller settings regarding cache flushing (since hardware controllers always have battery / capacitor backups) but consumer devices lack these mechanisms. The same goes for enterprise SSDs, many of which have capacitors allowing them to commit cache before power total power loss which I've not come across in consumer grade hardware.

Again, Samsung categorically don't advise to turn it off. In fact they strongly *strongly* advise the opposite: To enable it. They say you are welcome to disable it at the risk of data integrity, against their strong advice and warnings.

At the end of the day, it's your call. I really don't think fsync is expensive enough an operation to warrant the increased risk of FS corruption, especially given the already immense throughput of SSD's which aren't a performance bottleneck in consumer use cases. I'm with Samsung on this one.

 

Hi Guttsy,

"However, if you are more concered with performance than data integrity, you are welcome to disable this feature to eliminate extracommands being sent to the SSD to flush the data."

Later on, if you desire, I can post links to loads of SSD optimisation articles from renowned sites and they all say turn off Cache Buffer Flushing. They can't all be wrong!

 

Yeah, I think it's ok to turn off Write Cache Buffer Flushing, especially if you back up your data. I've been running with Write Cache Buffer Flushing disabled for years, even with Samsung RAPID installed, and haven't seen any corruption yet. I have heard of cases where data has been corrupted by this combination though.

 

Sorry, but you're misquoting here. The first quote is under "Write-Cache Buffer" and refers explicitly to enabling the Write-Cache Buffer, not to disabling Write-Cache Buffer flushing. The second quote is under "Write-Cache Buffer Flushing" and refers to Write-Cache Buffer Flushing. These are two separate things, as is the advice given.

To summarise:

• Write-Cache Buffer: Samsung recommends enabling, arguing it's an "acceptable risk for Client PC systems because they enjoy a lot of idle time and because SSDs write data to NAND very quickly."
• Write-Cache Buffer Flushing: Samsung strongly recommend enabling, unless "you are more concerned with performance than data integrity..."

Thanks, but please save your time. My argument is that the performance increase in almost all consumer scenarios is not worth the associated risk of data loss. FS corruption due to your non-ECC RAM failing and the write-cache buffer filing with garbage which is then written to disk is a rare and unfortunate occurrence. FS corruption due to uncommitted cache not being written in the event of power loss or software crash is a far more common occurrence in comparison. Again, it's up to the individual to make that call, but I'm with Samsung in my recommendation.

 

Hi Gutsy,

I like your SELECTIVE selecting to put my quoted text out of context and try to swing it in your direction, I can do exactly the same as you can see below, are you a politician by any chance??

• Write-Cache Buffer: Samsung recommends enabling, arguing it's an "acceptable risk for Client PC systems because they enjoy a lot of idle time and because SSDs write data to NAND very quickly."
• Write-Cache Buffer Flushing: Samsung strongly recommend enabling, unless "you are more concerned with performance than data integrity..."

In case you hadn't noticed, this is an AW17R3 thread, the AW17R3 is a gaming laptop, the people that buy it are in the vast majority interested in increased performance, NOT data integrity.

Please note the write performance increase by disabling Buffer Flushing in the below article. The numbers don't lie either.

"Take a look at these two Crystal Disk Mark results. The change occurred only after Turn Off Windows write-cache buffer flushing was checked. Notice how the all important 4k random writes are 5 times better."
http://www.thessdreview.com/our-reviews/raid0-the-ssd-optimization-and-realssdx4/

Maybe you live surrounded by Enterprise class Servers, so do I, my career is as a 3rd Level Infrastructure Analyst and SQL DBA. I know that at the Enterprise level of data throughput the buffer flushing is a necessity, however, on a desktop or laptop in general use (even in gaming) the level of data flow to an EVO 950 PRO is so small and therefore so fast, that the cache is pointless, based on that the Buffer Flushing is also pointless.

If my AW17R3 were to crash and get corrupted I would fix it, same as most everyone on this Forum. The chances of corruption on my AW17R3 through not having Buffer Flushing on, is about 1 in 1000000. And BTW I also am qualified in risk with M_o_R and I am quite happy to take the minuscule risk.

 

I'm not trying to swing anything. I merely pointed out that you stated the following:

Which is a factually incorrect statement that entirely mischaracterises Samsung's advice. I'm not trying to win an argument, I'm just pointing out an obvious error in either the way you've read and understood the white paper or the way you've chosen to interpret it.

I respect your opinion but disagree. If you want to undertake the increased risk of filesystem corruption in exchange for a bump in performance that's your choice. As I've already stated, my personal opinion is that for typical consumer workloads the increased performance is not worth the increased risk. We can agree to disagree on this and without either of us presenting data to back up our positions for particular hardware configurations in specific scenarios it's moot.

I'm only taking issue with the way Samsung's advice on the matter has been inaccurately framed because most people are far more likely to accept the advice of a manufacturer compared to a pair of random people arguing on a forum.

 

For those that aren't bored yet ....... as I keep saying ........ !!

-------------------------------------------
For your convenience, you can download the IRST driver we are using at the moment by clicking HERE. After installing IRST driver, there are two settings that we need to disable; the first is Windows buffer flushing.

This is an old setting that was relevant to HDDs, and is no longer relevant to SSDs. Here's why it is not relevant to SSDs: When write caching is enabled, data is cached and held in volatile memory until your slow spinning HDD has time to write it. SSDs are so speedy that data is basically immediately written to non-volatile memory, so there is nothing buffered to flush. Disabling Windows buffer flushing results in a nice bump in performance.

http://www.tweaktown.com/guides/680...sd-performance-installation-guide/index7.html

 

The author of this article is a "PC enthusiast" rather than a professional. Take their advice for what you will compared to a manufacturer like Samsung. He doesn't even bother to mention the risk involved in disabling write cache flushing, suggesting he either doesn't fully understand or doesn't care about the advice he's giving.

He's correct that SSDs flush cache faster than spinning disks but it doesn't logically follow that there won't be unsynced data in the cache in the event of a crash or unexpected power loss where uncommitted write-through requests forced into write-back cache will never get written in their intended order with no chance of replay due to lack of battery back up.

 

Hi Guttsy,

If you think an AW17R3 can't write to a 950 PRO fast enough, and because of that it will need disk caching, well you go right ahead and leave the Cache and Buffer Flushing on, I on the other hand know that the data is hitting the disk in microseconds, even under heavy load it's so fast that it doesn't need a cache, so I'll be leaving the buffer flushing off.

2 x 950 PRO in RAID 0

 

Those are great benchmark results. Must feel pretty nippy!

Do these benchmarks tag their write requests write-through or FUA by default, as this would produce artificially poor results with cache flushing enabled compared to real world usage? My understanding is that write-back is default for Windows write requests so standard workloads shouldn't be terribly affected by enabling cache flushing, rather it ensures that applications and OS processes which really do require their requests be tagged write-through to avoid FS corruption can do so.

 

Hi Robbo99999,

This is what's going on with the Windows Cache setting vs the IRST Cache setting. It's not for HDD Cache mode on the IRST, that's something else.

http://www.tenforums.com/drivers-hardware/13030-intel-rst-raid-write-cache-settings.html

 

Yes, you might be right, do you know if Intel Cache in IRST is just specific for RAID arrays or is it also there for single drives? I'd check on my own system, but I only have the bare IRST drivers installed without the software panel.

 

Hi Robbo99999,

Not sure about single drives but the Cache is for the Volume, which you'll have with and without RAID.

I am currently playing with settings now. Clicking Cache in IRST opens up a list of Cache options for the RAID Volume.

It says that you can't enable write-back Cache unless you first disable Cache Buffer Flushing in Windows.

The best settings by far have been with the settings I posted.

Turning off the Windows Cache completely leads to dire performance, as if there is some incompatibility issue between what Windows wants to do and what the IRST driver wants to do. 50MB/s sequential write, down from 2500, there must be an issue there!

Running Windows Cache and IRST Cache together wrecks the write times again.

As I said when I set this baby up, I would test all the options and stick with the ones that worked best, those are the settings that still work best, the one's I posted yesterday.

 

New 17R3 BIOS released:

Edit: Installed and seems to be working.

 

@JerseyBoy , have you started your laptop cooler project yet?

 

If relevant speed read and write speed is not that I would use RAID1. Intel RST had read simultaneously from both drives while being protected data, the data writing speed is like no raid.
my test...
http://rodi.sk/misc/intel-rapid-storage-technology-ssd-intel-535-240gb-raid-1-benchmark/

 

After talking all that junk; have you done this yet? Im interested, I want some ideas.
Make a tut while you are at it please!

 

Hi svkRoDi,

Not bad for SATA 6gbps in RAID 1. You'd get a lot more speed in RAID 0 and almost twice the disk space. Depends if you REALLY need RAID 1. Last time I had RAID 1 on anything other than a Server was about 20 years ago. With SSD's they last so long (Guarantee is 400TB written to a 512GB 950 EVO before you have to worry, in RAID 0 that becomes 800TB written to the Volume!) that RAID 1 is just a waste of disk space.

 

Hi Robbo99999 / DeeX,

I have started, but I've been held up by work, wedding (not mine but I was very involved), holidays, and some bits that have got lost in the post.

Don't fret, Rome wasn't built in a day, but it will appear!

 

On another front, I wanted to test the HyperX Impact 32GB I installed, to be sure it had no issues with OC (as it's not 'approved' memory yet). I'm running at Dell's preset L2 at the mo' and no issues so far. I'm running with the performance setting on the fans for this.

 

Another example and explanation of why Cache Buffer Flushing on NVMe SSD's results in reduced performance. The same effect happened last night when I changed the setting on my AW17R3. I can't install the Samsung driver with the disks in RAID 0, or rather I can but Windows will always use the RAID Driver and the Samsung 950 PRO driver is not (yet) a RAID Driver. Using the Intel Driver for the RAID, but the Driver listed in Devices is the Windows Driver from 2006, I think that is where the incompatibility issue must be coming from.

-------------------------------------------------------------------------

We're seeing that NVMe SSDs get radically different storage benchmarks when Windows Write Cache buffer flushing is enabled versus disabled:

(The following benchmarks were taken with Windows 10)

Write Cache Buffer flushing enabled (unchecked in Device Manager HDD Policies)
Storage Score: 4476
Bandwidth (MB/S): 107.35

Write Cache Buffer flushing turned-off (check-marked in Device Manager HDD Policies)
Storage Score: 5092
Bandwidth (MB/s): 623.7

https://community.futuremark.com/fo...s-when-Write-Cache-Buffer-Flushing-is-enabled

--------------------------------------------------------------------------------

Also this one

http://www.legitreviews.com/samsung-sm951-nvme-m-2-pcie-ssd-review_162219/9

 

For anyone that's installing Win7 anywhere.

http://www.engadget.com/2016/05/18/microsoft-windows-7-service-pack/

 

I've installed it and am also working OK with it.

 

Has anyone been able to install Fedora (or Ubuntu) on 17 R3 and successfully able to run it? If so, how did you go about dual booting your laptop?

I tried installing Fedora 23 but I'm not even able to enter the setup let alone install. When I boot from USB, I land up in emergency mode of Fedora, not able to do anything but restart the system. I've tried different pen drives, different USB ports, different software to create bootable pen-drive, all failing.

I could install Ubuntu Gnome but WiFi fails to work on it

Any help would be appreciated. I really need to install Fedora on my laptop for my college work

Thanks

 

Hi sagarbhathwar,

Try these:

https://www.reddit.com/r/Alienware/comments/3jldm3/alienware_13_install_linux_experience/

http://forum.notebookreview.com/threads/alienware-17-and-linux.781013/

 

Could you post a syslog / dmesg output from the Fedora installer? It's difficult to debug with so little information.

As for WiFi on Ubuntu, assuming you have the Killer 1535 AC I'm pretty sure it's Atheros based. Did you see whether the athk10 module was loaded? Perhaps it needed additional binary firmware?

 

Ok. Here is the screenshot. I don't know if there is way to capture the log files of Fedora installation.

Thanks

 

Did you check the md5 of the Fedora ISO? How did you prepare the USB drive? I know there used to be an issue with Unetbootin causing issues with Squashfs in certain situations, not sure if it was resolved.

The nouveau module may just need to be disabled until you've installed the OS and had a chance to install the binary nvidia driver. At the GRUB menu try a custom boot command:

linux rd.driver.blacklist=nouveau

 

Yes, Checked the md5. It's fine

I tried turning off secure boot, then launched fedora installer in troubleshoot mode. The graphics literally sucks here. Are there any drivers available? And I couldn't find wifi drivers either when I tried to. Any luck for anyone finding drivers for Fedora?

 

http://rpmfusion.org/Howto/nVidia

I assume Fedora has kmod-ath10k already installed, but if not grab it from ELRepo. You may need to provide firmware for the module manually if its not in the repos: http://www.killernetworking.com/sup.../20-killer-wireless-ac-in-linux-ubuntu-debian

 

Thanks for the reply. I will look into it.

By the way, some things are causing concerns for me

1. When I updated my BIOS, only main bios and boot block were written and it shut down in the middle. Laptop does work fine without any problems though. Is it fine for the bios update to shut down without any completed message?

2.Before, when I tried to install fedora, I used to get unverified signature warning from secure boot. Now, even though secure boot is turned on, fedora bypasses the sucure boot showing "Booting in insecure mode". This is really causing me concerns. Is something wrong with my BIOS?

Thank you

 

Can anyone who has the 17r3 with UHD/4k display comment on how it is for gaming? I'm wondering what the response time and ghosting are like on that panel compared to the FHD one.

Also, for those of you running linux, does the HDMI out work OOTB with no tearing issues on the r3? The fact that it is HDMI 2.0 makes me think it is wired to the dGPU, which could be problematic...

 

I've got 4knscreen but 970m. I game most at 1080p, but doom in 4k looks so great I live with **** ty frames. Can't see any ghosting or any display issues at all.

Sent from my SM-G930V using Tapatalk

 

You did the flash in Windows?

You installed with secure boot disabled and then re-enabled it in the BIOS afterwards?

 

The 4k panel is Excellent for gaming. No ghosting and it's nice and bright.

Powered by: Quad Core Exynos + 6820HK

 

Does anyone know if they use the same 4k panel I have in the precision 7710?
I absolutely love my 4k screen. Best screen I've seen on a laptop, and have just somehow talked my company into replacing my newer precision 4800 for a 7710 with 4k with high hopes it's the same panel.

Sent from my SM-G930V using Tapatalk

 

Sorry, to specify, the 4k panel that's on my Alienware r3, is that the same panel they use in the precision?

Sent from my SM-G930V using Tapatalk

 

No. I flashed it at reboot using USB. Previous versions I had flashed it through windows, but same result though

No. Secure boot is on always. Yet, while booting thru usb, Fedora media boots up saying booting in insecure mode. This was not the case some time ago.

I'm afraid my BIOS is irrecoverably corrupted. Some gut feeling.
Previously clonezilla used to boot normally after turning off secure boot, where the USB was made bootable by tuxboot. But now, I get media detection failure and it only works if I burn the pendrive using Rufus. I'm really tensed

Is there any way to clear the contents of the BIOS and flash it anew? Is there a way to check if BIOS is working as it should?

Thanks

 

You can load setup defaults in the BIOS, save and reboot. That's about it. You could also reflash with the current version and then repeat the load defaults, save and reboot procedure afterwards, then redo any BIOS configuration required, save and reboot one final time.

In Windows, what is the value of the Secure Boot State item in the System Summary pane of msinfo32? If it's "On" then Secure Boot it working as expected and your BIOS is probably fine.

Fedora's secure boot implementation will be similar to other distros insofar as a bootloader shim will be validated against the MS key in the bios's UEFI key db, which then loads GRUB2 as a second stage bootloader signed by the distro maintainers, which then boots a signed kernel, which then loads signed modules. Unless you're using an old shim or something, it's probably more likely that you've installed a different kernel or unsigned kernel modules (nvidia?) causing it to boot in insecure mode.

 

Hello I've a couple of questions..which are the main problems with AW17R3 4k 980m version?; can be upgraded to m.2 950 pro samsung, and w7 ultimate? Also the g sync is a must to not? I gonna use to video editing and anime or movies.. thanks

 

Yes to 950 pro, have 1 now. W7, not sure oj driver support. But Im running 2 win7 ultimate vms right now without issue. Gsync, um, well it's definitely nice and helps with occasional tearing , but I'm pretty sure theres not a 4k gync panel. Video editing is why I bought mine and what I use it for, and 4k or bust.

Sent from my SM-G930V using Tapatalk

 

Hello all. I left my computer on last night unplugged and came back today and the damn thing won't turn on. If I plug in the charger the alien logo lights up but it still refuses to turn on. I've tried holding down the power button for 30 seconds without the charger in and holding fn and power. I have also opened this ***** up and unplugged the battery. Nothing. Anyone have any ideas of what else to try ?

 

What BIOS version have u installed?

 

Unless there was a newer one released a month ago it's the new one

 

Sounds like a dead motherboard like what happened to mine.
Do you happen to know what BIOS your machine originally came with?

 

I've had the r3 for a little while now and upgraded to the newest bios a little over a month ago. Unless there was one recently released, it should be the new one. I do not know what the original bios was though

 

Ok. I've noticed a trend of people on this forum who have upgraded from 1.2.3 to 1.2.12 ending up with bricked machines. My R3 came with 1.2.3 and after upgrading to 1.2.12 it died after a few hours. Dell replaced the board which also came with 1.2.3 and it died again after upgrading to 1.2.12.

However, The second replacement came with 1.2.2 and I have upgraded to 1.2.12 and this board is fine.

 

And this is EXACTLY why I will never buy another Alienware again. I've never seen a company fu*k up what once used to be a solid brand (this bad) ever. Despite how great their warranty coverage is; I don't buy a high performance gaming laptop so I can keep on getting it repaired/exchanged over and over and over and over and over and over and over and over and over and over again.

 

I think the common denominator here with the 1.2.3 was that darn AMD crap....along with who knows what.

@Daniel1983, I feel you pain...but I wouldn't give up on them just yet. Any brand with a messed up BIOS update could do the same thing. That 1.2.3....darn....even my first 17R3 was on 1.2.3 when it blew a gasket. But now I have a brand new mobo with 1.2.2 and it's rock solid. For me, there is no need to upgrade to 1.2.12...same goes for the NVIDIA graphics drivers..no need to upgrade the latest. Even Prema doesn't recommend it.