*** MSI 16L13 (Eurocom Tornado F5)/EVOC 16L-G-1080 15.6" Owner's Lounge ***

Here is a comparison with my GTX 980M in El Cazador (P750ZM)

http://www.3dmark.com/compare/fs/10122247/fs/11562781

But, it overclocks like there is no tomorrow...

 

@birdyhands I detailed my experience with installing Win7 on page 122 and dealt with the same issues. I ended up saving my own personal made W7 iso with the usb3.0 drivers loaded for future use.

I had a question for you since I never got a reply from anyone else and Im guessing almost no one else is running W7 on this laptop. Do you have the 'legacy' option in BIOS selected? Also, is your boot time much longer than it is/was with win10? My bootup times are noticeably longer with W7 and if I disable legacy it will not boot.

Also, I delidded my 6700k this weekend and applied CLLP (pro not ultra) to my die, IHS, and GPU and have had some good results. I will say that using the vice only (and maybe even vice and hammer)method with skylake is not a good idea. My substrate did not bend, but my vice has teeth that I covered with tape to flatten it and the teeth still bit into the substrate. Luckily, it still works fine but I was very anxious for a bit. This was my 5th delid and all others have gone fine, but they were all 47x0 CPUs with fat sturdy substrates. But this puny one did not cope well with my vice. The razorblade method worked within about 2 minutes which is odd since I had lots of trouble with haswell and devils canyon and the razor method. There was so much glue/epoxy.

Also, stripped one of the damn GPU screws while unscrewing!! I had to take a dremel to my F5 and cut a horizontal divot to fit a big flathead! Talk about uneasy....

Anyway, all is well! Just wanted to share some experiences!

 

Makes me miss my Batman.... I got a good offer on it from a friend so tried out an Alienware.... Was a bad mistake...Lol..

Nice scores for a 980m.

 

I do have legacy option still enabled because it doesn't seem to cause any problems with me. I haven't noticed much of a difference in startup times, in fact I think windows 7 boots up faster. Yea those gpu heat sink screws can be a bit troublesome that's why I ordered some stainless steel head ones which should be much more durable. I'll edit this post with the link for them once I find where I bought it from. I'll take some pictures too once they arrive. 15 for $2 is a pretty good deal I recommend them. Also wow what a coincidence that we both screwed up our cpus with the hand vice lol. The teeth dug into mine but it still works thankfully I'm about to delid my CPU once the tool I ordered arrives and repaste with clu. I haven't had much luck with the razor blade it's too fidgety for me.

Edit: also what temps are you getting with the clp? Also under load of games?


Sent from my iPhone using Tapatalk

 

did you have any driver issues ?

 

For me W7 hangs a bit on the splash screen, just before the 4 spiralling windows come onto the screen, and if I disable GUI boot it still takes just as long. I'll have to time it but it is longer than my 5 year old eurocom monster, which doesn't seem right.

As soon as I stripped that #4 screw on the gpu I ordered the stainless steel ones that you posted earlier in the thread, before I had even managed to remove it. I lurk this thread a lot haha! They are from laptopscrews.com or something like that.

Also, I used a bench vice, not a hand vice. I'm not sure I'd trust a hand vice, but even then, the teeth on it caused me issues. I'll check temps over the next day or so now that I can actually control my ambient, but I dont run benchmarks, I just fire up games to stress.

I have been having some issues, with Battlefield 4 I cannot enable vsync or limit the fps via the game time.variable command as when I do I get massive stuttering. If I let it run uncapped and hit the 200fps limit its fine but as soon as I limit to below about 180fps my latency increases (via the enable graph cmnd in console) and stutters are terrible.

Also just noticed the same behavior with GTA 5 today.

Obviously the solution is to let them run uncapped, but then temps skyrocket. Before my delid this weekend, letting BF4 run uncapped would make my temps hit 100c on CPU and 85 on my GPU. This is with auto fans, which from my reading means 'silent.' I have no custom fan profiles, I run silent or use the button for max.

Now with the delid and cllp I saw 85c on CPU and 76c GPU letting the fps hit 200 and auto fans, but this was in a bit cooler ambient than normal so I will have to retest this week.

Enabling vsync and everything is golden. Temps dont even hit 70s but again.....stuttering! Lots and lots! I'll make a video and post what I'm talking about later. But my desktop ran sli 980tis and I never have had stuttering like this.

EDIT: Should have added, all this is done with my CPU multiplier set at 42 for all 4 cores and -180mv undervolt!

 

All Windows traffic goes to the USA, as in the telemetry. This way MS doesn't need to go to court to have your identity and details about what your using windows for, or your surfing habits, etc...

 

Oh hey, got some news. I had to build two desktops for some friends today, and they had 7700K and MSI Z270 Carbonfiber motherboards. Anyways, at 5.0Ghz, with delids thier temps were like 65-70C during load, however, in benches, like XTU or firestrike , they basically got like 15-25% less performance out of the CPU then we get from the F5. I had to show them this, I couldn't believe my eyes. We updated we did everything we could, but for some reason, the results were so much lower. Is there some known issue where these laptops are FASTER then desktops ? Our 1070's got higher scores thier there MSI armor 1070's got too. I couldn;t find any reason for it, unless the scores are all unfair or thier bios for thier motherboards is completely gimping them. If anyone has insight into this, I'm curious to know more, like why is this laptop faster then a desktop with similar if not even greater hardware - or are the scoring systems used by these aware of being on a laptop and they give "bonus" points or something.

 

@woodzstack

I can't answer the CPU discrepancy, but the mobile versions of the 1070 have higher shader units (2048 vs 1920) compared to the desktop counterparts. This was done to make the laptop and desktop versions of the card 'the same' since the laptop version is clocked lower. However these laptop 1070s can still overclock quite well.

 

Got the donor parts from El Cazador (512GB X400 M.2 and 2TB HDD), dual-booting Windows 7 now.

Here's a 3DMark 11 with a moderate overclock.

http://www.3dmark.com/3dm11/11944150

 

Probably had the CPU voltage to low for it to perform correctly. 6700K is the same way. It may "look" stable because of no crashing or BSOD, but if the voltage is not high enough it will have lackluster performance. The power limits and ICCMax may also be lower than ideal. Temps will go up somewhat as performance goes up, but one might expect that.

 

I haven't had any so far, every driver installed correctly and the components are being recognized


Sent from my iPhone using Tapatalk

 

Something isn't right. My DT 7700K (5.3ghz) easily outperforms the 7700K in my DTR. I lay the blame on the mobo.

 

Do you happen to know what's happening at the silicon level to cause that? Intuitively, I would predict that a CPU with too low voltage would fail intermittently, but my intuition obviously doesn't hold water here. I'm curious about the electrical mechanism in play here to cause it to work, but work too slowly.

 

Hey guys so I took a picture of the stripped screw from the gpu heatsink so you can see how bad the material is. When I was trying to take it out I had to cut a little nook for a flathead too lol like space


Sent from my iPhone using Tapatalk

 

Based off your username, I was expecting you to have talons...

They probably used junk metal to make those screws.

 

I think Intel made some improvements to the voltage controls on Skylake and Kaby Lake to keep the CPU stable under a wide variety of conditions. Under normal circumstances, if the voltage is too low you will see 0x124 and 0x101 bugcheck codes. With Skylake, and even more so with Kaby Lake, I see those less frequently. Instead of crashing as easily, the performance sucks. Now, if you go too low, yes... you will still see those codes. Or, the machine will just reboot itself instead of throwing a BSOD.

Those GPU screws are just really soft junk. They (all OEMs) have been using crappy soft screws for MXM for years and some are so soft they are good for one use and easily strip the head with almost no effort. I did not see the link from @woodzstack to purchase the stainless screws, but I am guessing those are a lot more durable than the soft steel screws that are commonly used.

 

This weirdly has never happened to.me, neither on the desktops x99 and z270 or on the DM1.
Everytime, if the voltage is low for a certain type.of load it will crash. But it never underperformed. And if it was not getting enough power , then it would just throttle.
The voltage has either been too low (Crash/ reboot/ BSOD) / just right (scores remain the same even after hitting the throttle ceiling)/ or too high (throttle hell).

 

I have never seen it before either. It's really weird. Maybe is it something with Z170. It is more noticeable to me with 7700K, but even with 6700K you see people doing some pretty crazy undervolting and under normal workloads they are not crashing. But, those same people often have sucky physics scores in benchmarks. All other chipsets and CPUs I have had crash as do your X99 and Z270 setups.

 

Here's the link for the ones I bought if you wanna get some: http://www.laptopscrews.com/M2x4.htm


Sent from my iPhone using Tapatalk

 

That's true of undervolting the CPU cache, that's why the weirdness with Skylake / Kabylake, the two settings CPU Core / Cache are locked together... unless someone has figured out how to separate them?

Keep the CPU cache voltage at default, and use a CPU core undervolt only.

Otherwise run a benchmark at stock voltage, and then undervolt not just for stability but also for check for performance by running that benchmark after undervolting.

It's good to undervolt to drop temps, but you don't want to sacrifice too much performance while doing it.

Test each undervolt drop to make sure you aren't losing too much performance

For me on Broadwell / Haswell I don't need to make that choice, I can leave the CPU cache voltage offset at 0 (default), and lower the CPU core offset as much as I like.

On Skylake / Kabylake, I would imagine I'd OC to the max on CPU - and there's likely no undervolt available there to worry about - maybe even a slight overvolt. That would be my performance / benchmark profile.

Then for everyday use I'd undervolt for best temps, and double check against CPU performance drops - balancing out the lower temps for a slight performance loss - or none - stopping the undervolt just as performance starts to drop.

 

What is the refresh rate for this model?

 

So far it looks like every 6 months a new model comes out, give or take a couple of months.

It will depend on the Intel chipset / CPU and Nvidia CPU release schedules.

All the vendors refresh with a minor / major upgrade potential.

 

The 1070N has extra CUDA cores and if it keeped cool and nice, it will actually boost decently high. Its an absolute joke how the desktop 1070 is actually worse than the mobile 1070N at times.

As for the CPU, try testing both at stock conditions and see if the same performance difference occur.

 

Same ram speed as well? XTU is a wacky benchmark. Better ram but lower clock speed provides a better score in XTU vs. the oposite. Even the BGA junk [email protected] with ok ram can outperform a nice overclocked 4930mx

 

After 1 week , this is my BIOS in Tornado...
I hope that today @Eurocom Support send me a GPU replacement

http://imgur.com/a/WRLGu

 

Ah finally got that out, nice. Gratz!

Man did that crack me up hahhaa~ But @Talon is another user....

nah nothing special, laptop screws from www.laptopscrews.com or something of the rather.

Also, I have noted this - desktops do not have as high performance as these laptops do. Sound scrazy - but what the heck is going on ? Are laptops finally cheaper and more powerful then desktops ?~? I mean I know of us crazy people love LGA laptops and hate BGA garbage but how soon will desktops fall under the BGA category lol... this ALMOST sounds like NBR level conspiracy..

 

@unclewebb has. Use ThrottleStop to set them differently. The pre-Kaby @Prema BIOS also allowed it. At higher core ratios having core and cache voltage matched is more stable, but stock and more sedate clock speeds like 4.7GHz~ do not seem to be affected much.

 

So, is M2 x 4mm the correct size for the GPU spring arms? Do you order the stainless steel screws that have the larger head (pan head) on them? That seems like a better choice than the wafer head screws. They are longer than 4mm, but maybe that is just overall length (including the thicker head).

 

I ordered the the stainless steel ones which should be more durable than the wafer head and whatever weak material that was made of

Edit: hopefully the pan head will fit because it seems to be a lot longer than the others, if it doesn't then I'll get the flat head stainless steel ones (I didn't see them when I ordered the pan head ones, maybe they're new)

Sent from my iPhone using Tapatalk

 

Are you certain that was a Phillips head with bad material? That hole looks extremely square to me. Is it possible you mistook a square drive for a Phillips?

 

It's not square lol all of the others are Phillips


Sent from my iPhone using Tapatalk

 

@Mr. Fox. Once again you are a font of knowledge. Thanks much for the explanation; I didn't know Intel had made that type of design change.

I'd support that theory. I don't know overclocking as well as @Mr. Fox, but I do happen to do some metalwork as a hobby, and I worked in a steel-related industrial company for over a decade. Anyone who has ever worked sheet metal can attest that stainless is a lot harder than mild steel, and IMO @Mr. Fox is right that fastener makers have gotten really sloppy about their heat treating (usually a case-hardening process). You can also buy higher-grade fasteners that are load rated, but the easy path of going for stainless is probably just as good in this situation.

 

A lot depends on how amenable the workload is to massive parallelism. Taking a slight downclock to add a few hundred cores is a good trade, if the workload can exploit 100% of the available cores. If the workload already left some cores unused, the downclocking will reduce the performance of the active ones, so it's a bad trade. You can actually work the math algebraically if you know the workload at a deep enough level (not typically the case with games, but sometimes true if doing CUDA programming for neural networks or other general compute-node work).

 

LOL! It wasn't a square before, but it is now!

 

Except nvidia boost clock is dynamic in the sense that if only partial blocks are used, the clock will be boost higher. I think you can only address individual blocks of CUDA cores and not individual cores anyways. Its just one extra block.

 

60Hz or 120Hz if you use the new AUO panel that was just released last week. Very nice upgrade too, worth it in my opinion!

 

Hey got some news for you, little birdy whispered to me that you have a large package being shipped to you, expected to be there wednesday, and the nice people bringing it want your autograph, for souvenir's sake, I'm sure *wink wink*

 

I consulted the CUDA programming guide ("CUDA by Example") from Nvidia, and it seems there are different uses of the word "block". In the programming model, a "block" is dynamically defined, and there can be many thousands of them on a single GPU or multi-GPU SLI image, if you have enough total cores. My copy of the book is from 2011, so the limits may have changed, but in that edition the block limit is 64K per dimension for a 2 dimensional grid of blocks -- obviously no real GPU has that many cores, except in our dreams! I do not think you can allocate more CUDAs than what are physically present; I'm just referring to the method in which they are carved into "blocks" from the programmer's viewpoint.

I think -- and correct me if I'm wrong -- that you are referring to the "blocks" of CUDA cores at a macro architectural level, in which there are only a small number of blocks (depending on which GPU) and each block contains hundreds of cores.

As far as I can tell from the book, there is no fixed relationship between the macro blocks which correspond to physical silicon areas, and the API-defined blocks which represent assignment of CUDA groups to API kernel instances.

You are correct that I hadn't factored in the boost clock; I was assuming clock speed pinned at a single value. I stand corrected on that.

There's someone who's been posting on on of my other threads, who's doing extensive CUDA programming for machine learning. I'm going to see if that person can check me here, and see if I'm making sense or not.

EDIT: Now I can't find the thread I was remembering. It may have been on another board that I frequent, so I'll have to look there.

 

Wait a second guys I'm so forgetful, I actually did order the flat head stainless steel not the pan head ones.

They do look much more durable, I'm going to test them out soon.


Sent from my iPhone using Tapatalk

 

Be extremely careful tightening those. Countersunk screws could fracture the PCB if you overtighten them.

 

Before and after pics:



Sent from my iPhone using Tapatalk

 

Also the custom bios by svet makes things much easier with undervolting. I used to have to start up xtu every time my computer booted up and sometimes I would even have to reapply the undervolt because it was lost or something when I turned off the computer, now it's locked into the bios so much easier!!


Sent from my iPhone using Tapatalk

 

The panhead screws fit perfectly for me. No clearance issues at all.

 

ETA for 120hz w/GSYNC?

 

Here we go...


http://www.3dmark.com/3dm11/11946356

http://www.3dmark.com/fs/11572495

 

"antivirus software vendors are terrible; don't buy antivirus software, and uininstall it if you already have it (except, on Windows, for Microsoft's)"

http://robert.ocallahan.org/2017/01/disable-your-antivirus-software-except.html

 

Time to spank some more BGA turdbooks...

http://valid.x86.fr/igp6v8

 

Takes like 3-4 days for delivery.. is that what you want to know as an ETA ? If your oversea's shipping you a 120Hz panel can take up to 6-7 days.