*** Official Clevo P37xSM-A/Sager NP9377 owner's lounge ***

MX4 was terrible on the GPUs for me. I applied it five different times and the result was terrible. Not enough mounting pressure for it.

My converter box showed up today but it's pretty pointless to use it now so it will sit there unused for awhile. Just wanted to update that laptop Parts Now sent it and I received it.

Sent from my HTC One_M8 using Tapatalk

 

I believe the 1248 is the max stock clocks, which is what I am seeing almost the entire time. The average was barely below that:thumbsup: It was probably confusing the way I worded some of that. I was mixing clocks and counts from the logs in the same sentence and it doesn't make it very easy to read correctly.

 

That's why you get Generic Log Viewer and put your Hwinfo logs in it and let it sort them put

Unfortunately it doesn't record the second GPU properly so logging those has to be done with two instances of GPU-Z with one set on each GPU but the same log viewer can make the charts.

Did you happen to check out the OSD? I've put my laptop to rest for now, waiting to see what happens with nVidia investigation before I do anything else. Using my desktop for now.

Sent from my HTC One_M8 using Tapatalk

 

I did google RTSS and it looks like what I used previously and did not like.

I guess I just don't like having multiple programs running for something so simple when HWinfo64 does it so well and no configuration at all. I also don't need to see the info in game, and would be distracted by having a bunch of stuff on the screen anyway. Fraps is bad enough and it's just one number! It's the same reason I cannot play windowed games with a 2nd display showing browser/desktop. Full screen is the only way I can enjoy a game and feel engrossed in it.

If it was super simple and one click to run I would be more willing to try it I guess.

 

Paste still looked fresh after 3 months of heavy use, but a nice big pile had gathered on the side with the least pressure. I know this isn't due to excessive paste alone as there was literally a small "hill" of it to one side of the die and that side only. So it seems like it pumped to the side with the least contact pressure.

I think there's definitely something wrong with your slave heatsink, either a severely warped plate, or retention tabs gone bad.

 

It should be the opposite, MX4 is specifically good for low mounting pressures because it's a soft paste, which will squash into all the micro crevices of the heatsink & core while at low pressure, whereas the thicker pastes need a higher pressure to achieve the same effect, see here for a Toms Hardware link explaining the effect (see "How Thermal Paste Spreads under Pressure" section):
The Differences Between AMD And Intel Heat Spreaders - Thermal Paste Comparison, Part One: Applying Grease And More

And this article showing mx4 good for low mounting pressure:
Results: CPU Air Cooler; Low Mounting Pressure - Thermal Paste Comparison, Part Two: 39 Products Get Tested

Both of those articles are really good reading to understand thermal pastes.

 

Its very well possible that I could have a bad tube. It's not unheard of.

Sent from my HTC One_M8 using Tapatalk

 

Yeah, maybe. I edited my post after you replied, adding another link in there too. Both those links are really good reading to learn about thermal pastes, how to choose one, also how to apply it.

 

Lol the link showed MX-4 not being a better choice than the stock paste on their test GPU with GC Extreme beating both it and IC Diamond.

GC Extreme also beat MX-4 in high and low mounting pressure air testing.

I guess I made a good choice with GC Extreme for future applications.

I thought IC Diamond 7 and 24 were the same but different sizes yet it shows a big difference between 7 and 24?

Sent from my HTC One_M8 using Tapatalk

 

Yes, Gelid Extreme did really well in all tests, I think I might buy some to use next time I repaste my CPU & GPU. If nothing else just to see if it's any better than the Arctic Silver 5 I've been using for 2 years on this laptop.

 

Btw for those curious, Far Cry 3 gives almost +10C to both the master and slave GPU temps, both peak and average, compared to Watch Dogs. Also this time round Gelid or MX-4 made no difference on the slave, as it maxed out at 81-82C, and constantly fluctuated between 78-81C. Master GPU hit 77C and averaged 73C. Settings were Ultra quality + 2x MSAA + HBAO. Even Crysis 3 doesn't push my GPUs this hard, it's absurd.

Though these temps were recorded after a 4 hour gaming session, so I guess I really can't complain. Max fans kicked in after a certain period of time, maybe 3 hours into the session?

 

Watch dogs is so poorly coded I'm not surprised. It just can't use the resources properly.

 

^I've always wondered whether that was the case. The visiual fidelity isn't as good with Watch Dogs, yet it soaks up more vRAM. For no apparent reason I never get above 80 frames, but the cards don't seem to be working all that hard given the low temperatures I'm seeing (relative to other graphically intense titles). And the frame variance, oh my god just shoot me. Even when I'm just staring a static object in the game, I've seen frames jump from 62 to 70, then to 58, and back to 65. dafuq

 

Its built for the unified memory in consoles, simple as that. The game has zero optimization for running on desktop hardware with split memory.

I guess the reason it heats up my cards so much is my use of MSAA

Sent from my HTC One_M8 using Tapatalk

 

I use TXAA 2 because for me it is the best balance of performance and image quality. TXAA 4 is too blurry and also really taxing on the frames.

 

Maybe I'm blind but I honestly couldn't tell any difference between having no AA and MSAA 4x. In the end I went with SMAA just to have something, and especially because there is 0 performance penalty with SMAA.

 

Txaa is too blurry... I much prefer MSAA. The reason I stopped using smaa was because of the graphics glitches that the game had with it where things would get stuck on the screen. Always freaked me out and made me think it was the GPU until i searched watch Dogs artifacts on Google and found out it is common with smaa

Sent from my HTC One_M8 using Tapatalk

 

Yes, for me TXAA is too blurry (also the same for FXAA). x4 MSAA or x8CSAA are my favourites for performance & quality.

 

So back to temps. I have a stack load of IC Diamond so I decided to do some testing. Really sorry but did not do any pictures. I got myself a second tube of Liquid Ultra just to compare because as of yesterday I was still sceptical of CLU. Granted this was testing on an Alienware with very different heatsink configurations but my findings should apply to most.

Basically CLU is excellent when the application is perfect ie a smooth uniform spread across the die with no balls or excess droplets. Due to the less viscous nature of the compound it only works well when there is ample contact pressure and when there is not too much of it. One CPU application gave me bad results and I removed to the heatsink to find it had pooled simply because there was too much of it and it actually affected temps more than I thought. My custom CPU rendering script used to reach 89C each time with ICD, now with a good application of CLU it only tops out at 84C. The bad application of CLU went over 90C.

CLU really didn't work well on my GPU though and here's why. IC Diamond is really viscous and really hard to spread. I know it scratches and stains things but I've used it for many years and I swear by it for one single reason. That reason is that it performs very well under low pressure environments. My GPU heatsink is not lapped and therefore when using CLU, the middle section of made very little contact compared the upper and lower edges of the die. With IC Diamond, because it is very tick, the paste equalised in the middle nicely when using the line method.

This next step is extremely controversial but it has benefited me immensely. When I apply IC Diamond, because the heatsink pressure does the spreading, and due to the fact that it is also extremely thick, I never know the exact amount that will cover the die. After a brief burn in after the initial application, I remove the heatsink and check the spread. Usually there is both overspill as well as areas which the paste has not spread far enough such as corners. I actually scoop up the overspilled paste and I redistribute them to the unpasted corners and I put the heatsink back on. I run a burn in for the second time (this can include games, rendering, anything to heat it up to over 65C). Then for the final time, I remove the heatsink and check the spread and make sure the corners also have paste, I then scoop up any more overspill (if there is any) and I discard it leaving a perfectly covered die and I reinstall the heatsink.

I know this method may introduce air bubbles and other variables but in my experience, ensuring paste efficiency in the spread has yielded me better temps overall. It may be unnecessary but it's piece of mind for me and there is a noticeable temp difference from the initial application. I would never use any form of pressure spreading application with CLU simply because I am scared of where the overspilled metal would roll off to. Better to always use the brush for that one.

Just my 2c and of course, YMMV.

 

I think your'e spot on with the needing good contact pressure for CLU to work properly. Mr. Fox's AW18 has excellent CPU-die contact, and mine is not shabby either. Temps drop about 7-10C under load using CLU compared to MX-4.

Interesting, CLU seems to work well with the GPUs in Fox's AW18 even though the contact isn't as good (two sides only, completely lacking in the middle).

EDIT: Pics

My CPU heatsink:


Mr. Fox's CPU heatsink:


Mr. Fox's GPU heatsink:


I've lost my pressure map for the GPU heatsink, but suffice to say it looks quite similar to Mr. Fox's.

 

That's interesting. Wouldn't it be better to do the spread method directly on the GPU die, considering that you end up spreading it manually anyway (after your initial line method application), you'd probably get less air bubbles if you did the spread method right from the get go. Have you tried that approach before with the IC Diamond?

 

I have considered this and yes it would be much easier but IC Diamond is so thick it is really difficult to actually spread with something like a credit card. You actually need the heat from the components to melt and soften it. I have tried.

 

Ah, I see, how about warming up the tube of paste in really hot water first, or does it cool too quickly once you squeeze it onto the die?

 

It can be done but is a lot of trouble when the usual method can work with care. Cheat and put tiny bits in the corner if you like.

 

Well my paste job appears to be holding up well. I'm @ 77C at 3.5GHz 8 thread 1024M, 2 thread runs 3.85GHz @ 69C

That's with auto fans for both which is so much more bearable than max fans. I really like the GC Extreme and its so much easier to apply than IC Diamond to boot.

The slave GPU is still holding up with the liquid ultra, running at the same temperature that the master card runs at with IC Diamond.

I'm happy with the machine now. I'm not absolutely blown away by it but its much better than it was stock.

 

After a week of heavy gaming (30+ hours), MX-4 is already showing signs of degradation, and temps are 3C higher already. Whereas the Gelid Extreme paste that has gone through maybe another 15 hours of gaming (45 total hours) is still holding on like a champ, and temps are still as good as day 1. When I pulled apart the heatsink for inspection, the Gelid paste showed very little pumping out, but MX-4 had already started to pool at the sides of the die.

So yeah, MX-4 is complete garbage for longevity. Switching to Gelid for now.

 

That was the same issue I noticed with MX-4 is that the temps went up quickly.

Sent from my HTC One_M8 using Tapatalk

 

mx4 does seem to struggle with certain heatsinks.

 

Well that slave card? It kicked the bucket today.

I had fired up 3DMark 11, testing to see if K-Boost would stop the clock drop and it froze a few seconds in and the computer shut off and restarted itself, came back up and froze again at the desktop, rebooted again and my screen resolution was 640x480 so I figured I had to remove Precision X and reinstall the drivers. Uninstalled Precision X and used DDU to remove the drivers, rebooted and reinstalled them, everything looked fine until I went to the nVidia control panel and the 2x880M is now replaced with 880M and HWiNFO64 shows the second card but only shows that 217MB of memory are allocated and the link speed, no core info, nothing.

GPU-Z looks like this:



It still shows the frequencies but it shows it being revision FF and just says GDDR5 instead of reporting it was Samsung and its missing the numbers in parenthesis after the BIOS version.

Even with no drivers, the master shows its core temp. *sigh*

EDIT: Wow... I don't even know what to think right now, I was browsing the web and got a pop up to enable SLI, enabled it and now it works again? ?!

Here's the new GPU-Z from that card



Everything went back to normal. I don't even remotely know what to think at this point. I'm definitely going to RMA it (Xotic was fantastic - I submitted my request and not even 10 minutes later got an email that it was submitted to Sager, very impressed), an intermittently working card is a card on its way to failing but I've never seen something like that before... just wow.

Thinking that maybe it was my CLU application, I opened up the machine and checked the paste and it was perfect, no shimmering anywhere (which I would have been able to see because I had IC Diamond all over my VRMs still from the stock paste job, took a flash light and a magnifying glass to the whole thing looking for any trace of it and didn't find a single bead) so I pulled the video card out of the chassis looking for a short and couldn't find one either. I repasted it with GC Extreme and fired up Bioshock Infinite and it worked just fine, let it sit there for 15 minutes and the behavior was the same as ever, master card was running at 953.7MHz when I came back while the slave dropped to 901.5MHz. Master was at 79C, slave at 82C.

I'm positively baffled... In all my years working on computers I've never seen an intermittently-working video card.

 

It's possible either a solder joint is not quite making contact and warming allows it to make contact or the SLI connector cable has issues.

 

I thought about the cable but I have disconnected that cable before and the card still reports its temperature and other stats. The first time I repasted I had knocked the cable loose and couldn't enable SLI but everything else was normal then I opened it up and realized I hadn't snapped the cable down.

It baffles me. Hopefully they replace the card, this is now three exchanges with Xotic since i got the machine highlighting issues with that card so there is a history I can forward to Sager if they try to argue that the card is fine.

 

An intermittent cable could make the card go odd though.

 

Which might be the reason for the 3 exchanged cards in the past too - (depending on what they were replaced for).

 

What exchanged cards? These are the original cards that came in my machine.

Sent from my HTC One_M8 using Tapatalk

 

Oh, ok, forget that then.

 

I think you are thinking of Arotished who ordered some cards off eBay and had issues with them in his Alienware but he now has them in a new P570WM and they are working fine last I heard.

Sent from my HTC One_M8 using Tapatalk

 

Ethrem,

I believe I am working with you and Sager on this issue and will follow your case We want to make sure your issues are resolved and you have the machine working to your expectations.

 

Yep, that's me you're working with. I have no doubt that it will get resolved, this 880M has been a headache but you guys are awesome

 

Txaa x2 is my go to for every game basically

Sent from my Nexus 5 using Tapatalk

 

I get motion sickness from TXAA, the blur literally makes me dizzy, especially if there are fast changing scenes like driving a vehicle.

Sent from my HTC One_M8 using Tapatalk

 

It definitely does have quite a lot of motion blur I can definitely agree with you there, it was just after extensively testing all the different types of AA in black flag that I arrived at the conclusion that despite that I still thought it was the best looking of the bunch.

Sent from my Nexus 5 using Tapatalk

 

Really? TXAA 4x is inferior to 8x MSAA although the performance hit from MSAA is severe in comparison. 4xTXAA definitely looks better than 4xMSAA and would work in a game that doesn't have the blur problem but the only game I play that sucks so bad to need such high AA is Watch Dogs. I'm okay driving a car but get on a bike and I'm ready to vomit with TXAA and MSAA drops the frames into the 20s and 30s in places.

Sent from my HTC One_M8 using Tapatalk

 

Or maybe I should say best looking with smallest performance hit lol

Sent from my Nexus 5 using Tapatalk

 

Now you're talking lol!

Sent from my HTC One_M8 using Tapatalk

 

Ok, so I got this monster of a notebook in yesterday and immediately tore into it and installed my 240Gb m500 and a 1tb sshd. I spent all night downloading all my crap and let it download overnight. Played around with it quite a bit and this thing is a beast even with the 870's in sli. The only thing I don't like so far is the touchpad but maybe it just takes time to get used to. Now i'm wondering if anyone is running this thing as a hackintosh and if they are can they point me in the right direction to get started? I haven't done one in a very long time.

 

What display do you have? If you have the 120hz screen, it won't work at all and you will have to disable Apple's nVidia drivers and download the nVidia optional drivers. You also need to use the xpcm-free kernel because the BIOS has locked MSRs which will cause an instant reboot.

There are a lot of issues with it.

Sound has to be done with voodoo hda and it will pop every time the machine starts, you might or might not be able to get your 870M working (I have the 120hz screen so i was unable to get a display with the nVidia alternate drivers because it won't initialize an eDP display), and there are other glitches as well. Best bet is Niresh but you won't be able to update the system unless you make sure to replace the kernel with a prepatched xpcm-free kernel every time.

In the end I decided it was too much trouble and went back to Windows 8.1 on my laptop and just set up Mavericks on my desktop with my 780 Ti.

Sent from my HTC One_M8 using Tapatalk

 

I ended up just going with the stock screen so no worries there. I just read about the locked MSR issue which is exactly what happened when I just tried to install with unibeast. I have windows up and running with an ssd and sshd so i have a free 1tb drive to play with. I've only had a hackintosh one other time and that was back when you could get an amd machine running (just barely though). It seems like you leave the hackintosh scene for a day and your left in the dust.

 

That can happen a lot in any tech field

 

Fortunately for me, my previous experience working with the original developer image and figuring out how to get the DVD booting and installing on x86 machines has left a lot of knowledge about how the whole thing works that doesn't seem to go away so a lot of the old tricks still work (like using single user mode to rm -R NV* in the S/L/E folder when nVidia drivers cause a kernel panic or black screen) but the new stuff like dsdt drives me nuts.

Oh yeah use Clover EFI as your bootloader, it builds your dsdt for you and applies the necessary patches plus you will likely need to pass EFI strings to the nVidia driver to get it working properly later.

I'm not gonna lie though, it is a ton of work getting everything up and running and if I were you, I would not use automatic updates and instead get the single updates from Apples site to get up to 10.9.2 and then get the nVidia alternative driver for 10.9.2 and stop there the latter updates break.

You'll also need MacDrive so you can copy and replace the Mach_kernel with an xpcm-free version

Sent from my HTC One_M8 using Tapatalk

 

For me, 0xAA seems to work the best, and no performance hit either!