*OFFICIAL* Alienware 17 R5 Owner's Lounge

@doofus99 - I did more research on the pads. So it seems I want Gelid Extreme, those are the spongy ones, right? Like this:
https://www.amazon.com/Gelid-GP-Extreme-TP-GP01-B-1-0-Thermal/dp/B01AAHM1WW

Should I get ones thicker than 1.0 mm, or just 0.5 mm and 1.0 mm will be sufficient? It seems I could stack them also if needed although not completely ideal. Actually Amazon is only in stock on the 1.0 so I guess I will get those and hope they compress enough if there is not enough space. Just wouldn't want to have to close it up and open it up twice.

Also, while other pads say they are conductive, the Gelid Extreme says they are not - so basically, as long as I don't block airflow, I can go to town and put them wherever I want, right? Did you do anything to block short circuiting from happening?

And I assume I should get the cleaner and do a complete scrape and clean as if I was re-pasting.

Thanks.

 

Also, it seems Gelid Extreme are hard to find in general. Shipping is expensive. On the other hand, Arctic pads (rating of 6) and Thermal Grizzly (rating of 8) are easy to find on Amazon. But neither are as highly rated as the Gelid Extreme, which is a rating of 12. Also doesn't seem confidence-inspiring that on Thermal Grizzly's own site, they rate the pads much lower than their compound/paste solutions:

http://thermal-grizzly.com/en/products/13-minus-pad-8-en

Whereas other reviews I have read seem to say pads are just about as good as paste! And I know from watching some tests on YouTube, the Conductonaut by Grizzly didn't do much differently from other manufacturers' compounds, so perhaps their internal rating system is hyped/wishful thinking.

Anyway I guess I am not in a rush, I'll wait to see what anyone here has to recommend. I could order from another shop in the UK I suppose for the Gelid, but with Amazon I can just order all thicknesses and return what I don't use, easily.

For some reason in the pictures, the Arctic one looks the most spongy. Usually spongy is worse for heat transfer, but I guess in this case, we need the sponginess to ensure there's actually contact everywhere, right?

 

For anyone who is interested - I found this comparison of Arctic and Thermal Grizzly. This is the final page with results. Basically it seems like, fi you're going to compress the thicker Arctic pad, it works well, otherwise the Thermal Grizzly works better due to its density (but from the picture, it still appears to have plenty of ability to mold around the surface it's placed on).

I think I'll just go with Thermal Grizzly then most likely. Costs three times as much as Arctic though, uggh. Wish they'd tested Gelid Extreme also.

https://www.hwcooling.net/en/test-of-chewing-gums-3-arctic-and-thermal-grizzly-minus-pads/3/

 

Good choice!!!! Go with Gelids!!! They are really like a sheet of thermal paste so after assembly heat-sink will only go direction dies, and never 'up and down' with low mounting pressure heat-sinks combined with silicone pads. The Wm/k for the price is also good.

I do highly recommend. I have tried Fujjipoly 17Wm/k and stock DELL and both failed because of their material properties that did not suit me.

EDIT re. you new post:

Three people (confirmed) tried Alienware re-pad with Gelids. Me, @propeldragon and @rinneh all have awesome results.

Remember, with Gelids you will progressively only get better die vs. heat-sink contact by time.

 

Thanks! I ended up getting both Gelid and Thermal Grizzly. Grizzly also appears to be basically paste cut into a strip. I got the Gelid for the 3 mm thickness (no other brand comes that thick), and price was reasonable on Amazon with shipping. Had to go Grizzly for the other three sizes (0.5, 1, and 1.5). Also got the Arctic two-solution cleaner and the Gelid Extreme gel just to get the tool it comes with. (and if I decide to go with compound at least I'll have it on hand)

I guess I can try it and if it doesn't look as good as the Gelid, since I'll have both, I can return and order more Gelid, or I guess I can just compress the Gelid 3mm down (either manually beforehand, like with a rolling pin, or just by tightening down the heat sink).

I am planning on putting those thick ones on my SSD also, I've noticed anytime I do a large file copy the fans rev up high.

Thanks again! Hopefully I will have good news to report. Too bad it will take a week to arrive.

 

I have placed Gelids 1mm on locations where, based on Dell pad thickness, were 0.5mm and they adjusted automatically without compressing them before-hands.
I would not recommend going too thick on memory chips even with Gelids, as they do not get hot enough and the area to flatten is too large. VRM's are no problem with +50 % Gelids thickness guaranteed.

 

I think you need to open it up and have a look for yourself so you understand it better.

Thermal paste and "repasting" is what goes on the CPU, the GPU and in other places (other ICs) which have a dedicated heatsink (copper plate) over them. Typically, paste being paste, it will squeeze out / leak sideways but those CPUs and GPUs have enough space around them so as not to make a complete mess of the motherboard.

Thermal pads are what goes in between the heatsink and other ICs / components (eg chokes) and the heatsink metal does run over them but there is no dedicated copper plate, it's just the body of the heatsink itself, therefore much, much less cooling but at the same time there is much less heat generated and is less critical. In other words, you could remove all thermal pads and the machine would still run, but you would not be able to remove the thermal paste and expect it to last more than a few seconds in Windows.

On laptops however and specifically on this laptop, due to space confines, there is very bad ventilation and lack of sufficient cooling of peripheral components, such as the GTX 1080 VRAMs, a lot of chokes, and a lot of VRMs. And then there is also the PCH but my personal theory is that the PCH has been misreported to the tune of +20C-25C by HWinfo and scares people unnecessarily.

As a result we have found out that even when your CPU and the GPU are cooled sufficiently. those peripheral ICs, especially those chokes and VRs below the GPU get very hot, over 100C, this then gets detected, and the CPU drops to 800MHz until those VRs reach around 95C and then the CPU goes full speed again, for a while, and the cycle repeats.

When you remove the heatsink, any heatsink, you will need to re-paste, because thermal paste is creamy and it's all a mess, you need to clean it up, reapply, and position only once, not pulling up to see what happened. If you take the heatsink off for any reason, you'd need to clean up and repaste. Only takes a smidgeon of thermal grease anyway.

The thermal pads, being pads, are much tougher, and you can remove and reset the heatsink many times, they do not leak or degrade.

However as has been described dozens of times, the distances between those ICs and the body of the heatsink are semi-random and the heatsink has been designed to make good contact with the CPU/GPU but not with the various other components. As a result, due to the large number and different heights of such components, the heatsink has a very complex 3D shape and it is impossible to have good contact everywhere. You fix one area and you break something else. One solution would be to use jelly-like thermal pads, easily compressible, and DELL do use them in 2 places out of say 10. They should use them everywhere in my opinion. I have not found such jelly like thermal pads anywhere except one obscure ebay listing.

When it comes to cooling the CPU/GPU, the best way is to use LM, but this is precarious, or to use graphite pads which are almost as good.

Finally you have another battle on your hands. The heatsink bolts onto the motherboard with 7 screws on sockets supported by a strengthening steel band incorporated in the motherboard itself. However this steel band is far from man enough to compete with the heatsink which is solid thick metal and as a result when you tension the screws the motherboard flexes upwards. This combined motherboard/heatsink unit is then deposited on top of a very thin and precarious, flimsy base running over the equally flimsy keyboard. There is no level surface there to bolt anything onto anything, it's like a sandwich where all layers are weak and flex - all except the heatsink being the sturdiest. As a result depending on those flexes when you try to screw down the motherboard/heatsink you will notice there is like 2-3 mm of flex from the back to the front that need to be brute forced in... That will cause the not-so-good contact areas to become even worse and then during EVERY open and refit your results will be almost random.

I hope the above helps.

 

This is probably THE most comprehensive explanation of the finer points of repasting and how to make the best of a poor design.

 

@doofus99 That certainly is a complete explanation, thanks. More complete than the last time you took the time to explain it in the thread!

I haven't used pads before - I was under the impression it was better to use pads instead of paste (on this machine) due to the heat sink not being close enough to make good contact with the entire area of the paste (from what I thought I picked up from this and other forums). I take it this is incorrect though, and it's still best to use the compound on the CPU/GPU chips, and just use pads to shore up any other heat-generating chips to try to get them to shunt some more heat to the heat sink pipes, even though they're not supposed to have to do that.

Anyway, sounds like I might have to disassemble and re-assemble a few times.... uggh. Almost feel like cancelling my $150 order of heat pads and just living with it.

(To add insult to injury, the internet connection stopped working again, and then when I went to debug, not even running anything, I got a blue screen, and this is only running Alienware's own overclocking... although it had just played some games for a few hours straight before that with no issues, weird. I almost wonder if the internet dropping is due to internal heat also on the Killer card somehow.)

 

I went back further in this thread to the page 170 range/July - man you guys have been going at it for a while on the heat issues and other general issues! Honestly makes me want to see if I can get them to take it back. Tearing it down and targeting the different areas and then watching out for that flex and so forth doesn't seem as appealing as it did several hours ago.

I did see all the posts with photos though which helps. Now I am realizing I should take better stock of what temperature sensors beyond the cpu/gpu might be reporting as too hot before taking it apart, and try to target those areas. (but keeping in mind HWInfo might not report correctly such as on the PCH temp)

Trying one more thing too on my connection dropout - I had already set the WiFi device in Device Manager to not be turned off for power savings, but I did the same setting for the Ethernet controller and also all USB devices that have the setting, just to see if it makes a difference. Also did a network reset in Windows.

The thing that scares me is that there were all those basic issues with the command center and BIOS updates just a couple of months ago. I can't keep dealing with issues like this for the lifetime of the computer. It's somewhat fun I guess to try to optimize things at the start, like I would do with a computer I build myself, but I don't want to have to continuously face unwanted changes due to BIOS or Dell software stuff. (and don't want to have to reinstall Windows from scratch and not have any of their tools either)

I mean, even only getting in the 1100s for Cinebench is more than I need 99.9% of the time. Plus that's for constant load; if something just needs max speed for a second, it does get into the mid 4 range. Maybe I should just forget about doing the teardown. Well, I have a week to think about it before my full 30 days is up and before the pads arrive.

Does HIDevolution also work on customer's laptops or only ones they sell? Maybe it's better and not more expensive to just send it to them, I don't know. I thought it would just take 1-2 hours to do the re-pad/paste, but seeing the photos now, and not having done this before on any laptop, I can imagine it taking twice as long as that. I really don't want to end up tearing it down a few times, using up all the pads, and possibly end up in a worse condition. (and then being beyond the 30 days completely - I think with the WiFi dropping, the touchpad not scrolling all the time, and the inability to reach advertised CPU clock speeds, I could return beyond the 14 days... otherwise what is the 30 day period for, right.. they can ask me all the "questions" they want, I have plenty of reasons as to why I might not want to keep it beyond "change of mind").

 

For the software issues you need to install a clean Windows from the installation media, aka SD card. There is a "media creation tool" at Microsoft, look it up. Do AWAY with Alienware command centre and any other kind of junk, not until you have a stable system. I have not installed Tobii and my trackpad uses the standard Windows drivers so I cannot switch it off or do fancy things with it, and I cannot set the lights but it's a price I am willing to pay so as not to contaminate my laptop unnecessarily. After you install a very clean Windows you will need very few drivers, actually I have: GTX 1080 from Nvidia (or windows), Intel drivers (chipset + thunderbolt), Killer drivers (wireless, bluetooth, ethernet) and the touchpad which I am not sure what I did with it. That is all.

 

What's the best windows 10 build that has no issues?

Sent from my Phone using Tapatalk

 

I clean installed latest 1809 and all is good.

 

The "Gelid GP Extreme Thermal Pad" is not spongy at all. It is not at all suitable. You'd need to do a search and then tell the rest of us which ones you found to be good.

Examples (but no direct experience) :

https://www.ebay.com/itm/COOL-GAPFI...xible-Interface-Thermal-Gap-Pad-/150803017472

http://thermal-grizzly.com/en/products/13-minus-pad-8-en​

 

From my experience Gelid thermal pads did not make a big difference on my first system. In fact a 0.5 thickness difference (1.0 instead of 1.5) at the pads at the edges has made a difference. This does not support the spongy feature at all. I personally did not see the spongy feature anyways on the material itself (unlike K5pro which is something between thermal paste and thermal pad). In addition the Gelid thermal pads are easily destroyed meaning that if you ever open again for repaste they will probably be destroyed (especially the 0.5mm ones unlike the Dell default pads). This means you gonna have to replace the destroyed pads as well. Finally if they are really spongy why are there so many different sizes (check K5pro which covers a big range of sizes because it is highly compressable)

I do not know if anyone ever tried the cool gapfill above. Would be nice to know.

I think the spongy/thermal conductivity ratio is the number to be looking at but I am yet to see a thermal pad advertising its spongy features

 

@anyone with the Per-key lighted keyboard....what is the part number that shows up under your configuration on Dell's website? I've got an order in for a replacement and the keyboard is 580-AHCR.

 

Sure. Read my posts http://forum.notebookreview.com/threads/windows-10-redstone-5.823238/page-8#post-10804417

 

Looks like I am your guy on this one too - I made sure to get it (and verified I can light one key separately from the others), and my line item was as follows:

580-AHDC

Per key, multi-color RGB, keyboard with AlienFX lighting control

$23.99

So if yours is also a different price it would seem you're getting the segment and not per-key lighting. They had it wrong on mine also until two days after the order was placed. I complained, not sure if that fixed it but I doubt it, just took some time for the system to show the options I actually chose for some reason. Perhaps because I placed the order via phone.

 

@Aristotelhs2060 - in an older post of yours, you said you got the 1500 CB score and x46 JUST doing a re-paste of the CPU and GPU though right? Without touching any pads. Just wanted to confirm that since you are mentioning pads now; I think you had tried the pads on an EARLIER machine you had though right, which got replaced. Trying to get the history right.

And even if you just re-paste, you still had to remove the motherboard completely from the chassis, correct?

 

To no one in particular - just need to vent (again!)… my network is still dropping. I am not too thrilled about the re-paste/padding now that I see the motherboard has to come out completely.

I would estimate my chances of keeping this machine are now down to 40%. Might have to go to the dark side (or is it the light side, lol).

The thing that really irritates me, is I was leaning towards building my own mini-ATX since I have GPUs sitting around. But since I only have a mobo from the last generation chipset, and I didn't want to put the time I knew it would take into tuning a custom-built system, and really preferred a mobile machine, I went for this AW. But now I have put FAR more time into debugging this product than I would ever had to into anything I built myself.

If the darn network would just work, I'd still be highly likely to keep it. But I have no idea on how to fix it at this point, I'm out of ideas. It's just not acceptable to require a reboot once a day! I wouldn't have accepted that on my own custom computer even back in 2000. (maybe once a week back then; I once ran my MacBook for 6+ months straight without rebooting and under heavy usage for work)

Considering the Tobii is only giving me the auto-dimming and Hello, I could give it up. To be honest, the constant red lights are distracting after a while. The few times I've turned it off, it's ben relieving to NOT have those red lights aimed at me. I could always buy an IR webcam for cheap anyway and prop it on top of the Clevo if I really miss Hello that much. I am sure they make ones that can stick on for a somewhat more permanent solution. (Heck I could even just buy the separate Tobii and hang it on top of the screen, I'm sure.)

I just remembered the other thing I liked about this was the 9 macro keys; but honestly, I had a Logitech keyboard with them once and I never ended up using the macros. Can always use an external numpad or other solution; plenty of external device controllers for audio and video work especially to give hands on/shortcut keys.

I also remember now the thing I dislike about the AW, the lack of USB 3 ports; that is a nice plus that the Clevo has 4 or 5 of them, plus SD card slot. Not to mention room for two 2.5" drives, an extra m.2 slot (well, three full length ones, I think, but all are PCIe NVMe), and possibility to go up to 64 GB of ram. Doubtful I'd ever need that, but hey, maybe I'll need to do some VMware consulting again and would need it, who knows.

Thanks everyone for listening. This thread is like my Alienware shrink session. (seriously - if I didn't have a place to not only ask questions but just complain, I'd go nuts)

 

how did you buy it just like that? thought they couldnt sell separate parts

 

No, this was just a line item from my build summary. I specified it as an upgrade when I ordered (it was in the web order form but I ordered via phone - from all the reviews that complained about the lack of per-key lighting, I am guessing it is a recent option).

 

Damn I would pay more than 30 if i have to for one, since installation is easy anyways.

 

If it helps anyone, this is the most comprehensive video I have found so far on the teardown required to re-paste. Unfortunately, I was misled because on YouTube there is a "17 R5" re-paste video that is only 7 minutes long, but that is four years old and is a previous generation (when you really did just remove the bottom panel and then remove your heat sink).

I mean, I am handy, I can solder stuff, I can change my car's oil if need be, etc., I don't mind getting dirty. If I knew exactly what result just a re-paste would get me, I might be inclined to try it. But to guess where to put or replace pads, etc., and then put it all back together "just to see", not seeing it happen. (and like others have said, kind of ruins the replacement warranty - yeah they'll replace it, and you will probably have to do all that again)

Around 22:30 is where he finally gets to getting the motherboard separated from everything else.



It does make the bottom look cleaner if you're just taking the cover off to upgrade drives or RAM, but yeah, I can fully see now how there's no rigidity except for the heat sink itself.

I would say this laptop is at the point of thinness where, yeah, you CAN take it apart, but would you want to - i.e., this is a machine where it really should never have to be taken apart unless there is some situation for Dell to save on service by actually replacing a part rather than the whole machine. Kind of like where Macbooks were before they started soldering everything to the motherboard. [IN other words - they should really have done an A+ paste and pad setup from the factory... especially considering this is the Alienware line.]

When I bought into this I was keeping it in my back pocket that I could just upgrade the cooling, but like I said, I thougth it was just removing the heatsink after taking the bottom off. It wasn't until re-reading and getting the latest replies from everyone here who has done it that I realized it's a total teardown effort.

Anyway, just thought the video would help make it clear for anyone who hasn't seen it done. Even just seeing pictures of the motherboard doesn't really make you realize exactly how much you have to tear off/apart to get it out...

 

Like he did in this video you do NOT have to replace the pads, just the thermal grease under the CPU and the GPU. If you are lucky you will do a good job and achieve tighter core temp grouping and lower temps overall.

One thing to note is you need to take your time, start in the morning. Use guitar picks to pull the metal bottom out by inserting and running the pick along the edges. Use a guitar pick to pull out the plastic cover too, it is held with plastic tabs/clips, do not yank it open go really slowly with the pick, it will open without any force, if it needs force you are stressing the clips. Run the pick along the edge and you will hear the clips coming loose.

 

You have a software issue most likely. My laptop has not been switched off in days, however it does sleep every night, and sometimes it crashes during sleep and then I reboot it remembering always to press F2 and redo the "performance / OC1" BIOS settings that are lost each time Windows crashes during sleeps.

I strongly advise you to go to Microsoft, run the "media creation tool" for Windows 10, do a full clean install, remember to zap the disk partitions and redo them, leave no trace of the previous install behind. If you have a spin drive for data like most of us, you do not have to zap that too. I have already listed the ONLY drivers you will need to install so that there is no unrecognised hardware on the Device Manager. After a CLEAN Windows install you will need the KILLER device drivers, three of them for Ethernet, WiFi and Bluetooth, and the Intel Chipset and Intel Thunderbolt drivers. And the GTX 1080 drivers too of course. And that's it.

 

I have to be honest - crashing during sleep is not acceptable to me either. But I always set my computers to never sleep on power anyway. About the BIOS setting - this is in the OLDER BIOS version, right? I am not able to change the BIOS setting for performance setting in mine (I have the update that was two weeks ago). I checked again two nights ago just to be sure it wasn't something simple there holding me back, but there's nothing I can change there.

I did see your post about re-installing with the minimum drivers. But if I recall, you also have no ability to set the backlighting colors at all of the trackpad, keyboard, or other lighting areas, is that right? To me, that is also unacceptable. If I can't customize any of that, what's the point of having such a machine! If I got this for a deep discount from eBay or something, yeah, I would do whatever to make it work, but to have to do a clean install just to try and make the network work, and then not install anything from Dell in fear of issues creeping back in, and lose the customization -- that is not ideal. If I am mistaken about any loss of lighting control, please correct me. Thanks as always.

 

I think then you could install the Command Centre, set the lights how you want them, and then de-install it. The lights are stored in NVRAMs somewhere and stay forever as you have set it.

 

Ok I gotcha. Yeah I'd read in the posts (yours probably) about the settings being stored in NVRAM. I just thought you could NEVER install the command center because it would install other stuff and cause issues. I didn't realize you meant you can install it temporarily and uninstall it, without causing the issues that the clean install was trying to prevent. (I thought the best case would be to boot a separate install just to run the command center and set that NVRAM, then reboot to the main, clean install.)

 

Oh yes do it now, before you install clean Windows, and it will also stick without needing Command Centre at all in the new install afterwards, of course!

 

Everything you said is right. From my experience changing pads did not change anything, in fact you may also get it wrong. I would recommend to just re-paste and start from there.

 

For me I fresh installed windows october update in a new ssd but cannot change keyboard colors in command center, most likely bugged because of the new update

 

For anyone interested in adding the per-key lighting to your 17R5, I had a couple of chat's with Dell parts & sales folks. Apparently the part number 580-AHDC that shows up on a build sheet is unavailable to order as a spare part. They couldn't find anything else but in their "internal tool" they found this part that's available for $119.99 USD: Keyboard,103,United States,Eng land/English,M17IXF-BW. I have no idea if this is the actual keyboard with per-key lighting but after searching for a few minutes the Parts agent said this is the item that corresponds to 580-AHDC.

YMMV. $120 seems like a reasonable price for a replacement keyboard on a laptop but I'm not in the mood to be the guinea pig for this - at least for the time being. I might see if the Exchange team will cancel my order and re-order with the per-key lighted keyboard but I'm not holding my breath.

 

Uggh, sounds expensive actually... maybe the lighting costs more, but it's not mechanical. Regular backlit replacements for ThinkPads are only $35, I believe.

Anyway I am sorry to hear they messed up your order and that you aren't getting the per key lighting. Funny thing is, if they didn't have per key lighting, I very well may have not even bought the Alienware to begin with!

 

Can someone point me to the laptop fans that fit at the back of the laptop at the air exits? They work with suction mechanism enforcing the exit of air. I know someone posted them before in this thread but has been ages ago. They have like a BMW front appearance but I cannot find them.

 

amazon.com, search for laptop exhaust fan. Simple.

 

Not so simple. I cannot find the one I am talking about. Otherwise I would not ask. The one I was referring to was white and it was not usb powered. I think it was with battery

 

Battery powered? My god, why???

If it's not showing up on Amazon's search it's probably not available any longer. Bu let's be serious for a second, if you know the specific item you want was posted earlier in this thread why not go find the exact item in the thread that you're looking for? Seems pretty straight forward to me. Just go through past posts *done*

 

​

I found the post - took just a few seconds to search. They are not battery powered - each has a USB cable going to it.

 

Well I cannot find it. Maybe I do not know how to use advanced search. What is the post number?

 

@equalizer2000

Hey guys, I found a solution to my dropped Wi-Fi (and ethernet) issue. Just to recap the issue, my internet was going down after a few days of being up. I could ping sites, but not actually get any content. A reboot was the only fix I found. I tried the latest drivers, uninstalling all other software, etc. Nothing worked for me. It took a day or two test each thing because I had to wait and see if it went down.

Well tonight it went down and I devoted some time in testing things and I tried this site on my phone:
https://deploymentresearch.com/research/post/532/fix-for-windows-10-exhausted-pool-of-tcp-ip-ports

I added the registry items and BAM, my internet was back without rebooting. I had originally saw a message in my event log saying I was out of TCP ports which allowed me to stumble across that site during a google search.

I just got it working tonight, so I don't know if this just "prolongs" the situation or is in actual fix. Perhaps the ports will eventually max out again (or something).

For now, I'm happy I didn't have to reboot and I'll see how long this lasts. If the problem happens again, I'll post about it to disregard the fix above.

Hope this helps others.

Thanks!

FYI, Use at your own risk. If you don't want to click on the site above, here's the relevant details:

Add these registry values:

HKLM\SYSTEM\CurrentControlSet\Services\Tcpip\Parameters

TcpTimedWaitDelay
REG_DWORD: 0000001e (hex)

MaxUserPort
REG_DWORD: 0000fffe (hex)

TcpNumConnections
REG_DWORD: 00fffffe (hex)

TcpMaxDataRetransmissions
REG_DWORD: 00000005 (hex)

 

Wow, you are a hero! I haven't tried yet but yep, I have those same entries in my event log. From the message in the event log, I wonder what it is that is creating so many connections. Message is pasted below. I wonder if Support Assist, in those 345+ registry changes it made, tweaked one of these, and caused it (or somehow the OEM image has it set oddly). I know PLENTY of people have this issue on the Microsoft forums (laughable how I have NEVER seen a single thread on those have a real resolution, always template responses and not even understanding what people's issues are), so I don't even think we can blame this one on Dell?

This is quite a relief. I actually ordered a 9260 card, just arrived today. Guess I won't need to install it quite yet to test. (and it wouldn't have worked anyway, this is definitely an OS/registry issue!)

I started debugging more the last few times it happened to me, but I got lazy - after 5 minutes of poking around, it's so much easier to just restart and get back to what I was doing. Kudos to you sir for completing the mission!

"TCP/IP failed to establish an outgoing connection because the selected local endpoint was recently used to connect to the same remote endpoint. This error typically occurs when outgoing connections are opened and closed at a high rate, causing all available local ports to be used and forcing TCP/IP to reuse a local port for an outgoing connection. To minimize the risk of data corruption, the TCP/IP standard requires a minimum time period to elapse between successive connections from a given local endpoint to a given remote endpoint."

 

I added these values to a .reg file if it makes it easier for anyone else to import. {Disclaimer: USE AT YOUR OWN DISCRETION AND RISK.} I can't upload a file here but the text is below, just add it to a .reg and it will add those four keys. I figure if I need to re-add them after Support Assist does something or if I re-install I won't forget now.

I wonder if it's just one of those making it work, like the max port number to use. I also wonder if this will keep it from ever running out as they get properly recycled, or if it will still run out eventually if not rebooted.

If anyone who hasn't has any network issues could check their registry to see if those four (or any of the four) keys exist, that would be really helpful.

.reg file content:
Windows Registry Editor Version 5.00
[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\Tcpip\Parameters]
"TcpTimedWaitDelay"=dword:0000001e
"MaxUserPort"=dword:0000fffe
"TcpNumConnections"=dword:00fffffe
"TcpMaxDataRetransmissions"=dword:00000005

 

I found this post - some comments seem to blame it on DNSCache eating up all the available ports, and disabling that and not caching locally; the other person seems to have found a health awareness utility to be buggy and doing the same:
https://superuser.com/questions/590686/tcpip-error-request-failed-due-to-all-such-ports-being-in-use

I ran "netstat -anob" on mine (have to run as Administrator for command prompt), and while I do see dozens of entries for things like Dropbox and Edge (yeah I sometimes use Edge, lol, especially on new computers before I sync up Chrome stuff), I don't see thousands like that person does. I was wondering if OneDrive and Dropbox were using up the ports, if they use a new one for every little file they sync. But the connection went down even after they'd caught up on stuff. Well, perhaps the total number of ports was limited or something. I did see in other threads that past windows updates have cause the issue to start as well.

Weird how this has happened apparently on Windows 8 and various server versions, yet I've never encountered it. Truly a gremlin type of issue!

 

@equalizer2000 , Yea, those logs were in mine too!

Good call on that .reg file. The guy on that site said these values may get reset when you do a Windows Update, so I'll have to keep an eye out for it. I still wonder if it will eventually require a reboot, I'll report back after a while.

HA! I did the exact same thing as you, spent a few min trying to figure it out and then decided to reboot and be back working in 10-15 seconds. This time I was really fed up though

I also have seen lots of reports of this issue across various machines, cards, etc. I'm surprised nobody has actually fixed it, seems systemic enough to look into. My last alienware didn't seem to have this issue (AW17R3), also had Killer cards. That machine was Win 10 Home (vs Pro I have now), not sure if that's a potential problem. Granted the card was a different model from Killer, but I've seen similar reports on AW17R3 as well.

I'm happy for now though!

Thanks,
Jeff.

 

Interesting!

I ran that "netstat -anob", since my machine had the "problem" and while there was a bunch of connections, I wouldn't say there were thousands at all. Hard to say how many though, maybe 150-ish?.

 

Looking at what those flags mean on Wikipedia, it seems those results are only active connections. I like you have what appears to be 100+ but no more than 150-200 total - but if those are only active ones, hmm.

Do you use OneDrive, DropBox, or Edge, by chance? Only other thing I leave running is Outlook.

I am thinking if that list is only active stuff, it could be still a lot of old connections that used up the ports. So many ways to think where the issue is coming from - is the recycle time set too high in Windows somehow, and thus the old ports don't get reused like they should? Is the number of ports available to bind to artificially low? Or is there some rogue process in these machines or in Windows in general that isn't binding properly and using them up.

I'll laugh if it's Edge somehow causing it... should've known better if that's the case, hehe. (in my last debugging, I had turned off some developer option in Edge when you press Ctrl-Shift-D on the about:flags interface - someone had an issue where they could still ping and not connect to websites, but when they disabled Enable TCP Fast Open in Edge it fixed it - didn't do anything for me, but I wonder if Edge is somehow causing it)

I'd even disabled anything possible with Cortana because I thought maybe it was keeping some connection open and causing the issue (but it didn't change the behavior either).

 

In preparation for my thermal treatment (re-pasting, padding, perhaps padding memory and SSDs to bridge them with the rear cover, and doing the iUnlock suggestion of a heat sink on the PCH), I did some benchmarks to get a baseline of temperatures.

For the SSD, what I did is reset HWInfo right before starting CrystalDiskMark, then I ended it right after, so that the average temperature is accurate. (since the max temps can spike higher than average, and I won't remember exactly how each temperature changed during each test of each drive, then the GPU, and then the CPU cores, I figured the average would be a good single value to compare in addition to max)

I assumed the 2 TB I had installed would be hotter from being more dense, but actually the factory SM961 512 runs hotter!

I also read that Dell only includes the copper heat plate on their OEM drive if one select the 1 TB option. That explains why I didn't get one in mine.

I assume these temperatures are normal for a high performance SSD in a laptop. I can't seem to find much info about whether that will harm them over time or not. Is anything under 100 pretty much OK for SSDs, considering it's not doing maximum speed writes all the time? (the max temp was only during write - read tests were at least 10 degrees cooler if not more, off the maximum)

I've had SSDs in laptops for the longest time, never really cared about the temps before. Only reason I noticed in this one is being more aware of temperature in this unit overall, and also, when doing the large writes, the fans turn up.

 

I see Dell still prefer throw out boiling ssd's in their notebooks This isn't normal temperatures.



Could you run ATTO Disk Benchmark 4 times in a row (Just click start again right after each finish). Then post pict of results and temps for both ssd's.

 

Here's the temp of my ssd

Sent from my Phone using Tapatalk