*Official* nVidia GTX 10xx Series notebook discussion thread

Not so sure about the false negatives, I haven't experienced any in the entire time I have used it.

Failures have always been debuggable and solveable.

What are the specifics you recall about the false negatives?

 

I'll say this, it will establish the upper limit for stable voltage on desktops, but I don't favor the use on laptops. It may leave performance on the table, though, if you never do tasks that push the CPU that hard. It should not be the sole test used ever. I've described passing p95 and crashing with heavy ram usage with hundreds of tabs open.

What specs are you working with (mine are in my signature)?

Sent from my SM-G900P using Tapatalk

 

You're saying that Prime95 won't always show up overclock instabilities? What do you recommend for testing?

Specs in my sig. No instability so far with gaming / browsing / prime95. (Haha, hundreds of tabs open you say - ok, I'll give it a try!)
EDIT: Haha, I got about 150 tabs open now in Chrome (all different web pages), and I'm sitting at 80% RAM usage out of 16GB, and about 5-30% CPU usage, why anyone would want hundreds of tabs open I don't know but I was curious for some stability testing! Although just now it did shut down Chrome just now after I ignored a message saying it was running out of memory (I don't have a page file). Not convinced about the insane tab method for stability testing!

 

occt large data set for 20 min is very failure sensitive also, 10 consecutive runs of CB15 turned out to be quite sensitive towards unstable Vcore voltage on skylake! who knew voltages stable after occt turned out to bsod or freeze on me after 1-7 consecutive CB15 runs

Sent from my Huawei Mate 8 NXT-AL10

 

You're supposed to let it cool off between runs, to the lowest idle temp from 100% fans...

 

I just did 11 consecutive runs of CB15 (just kept clicking the CPU test button immediately after previous run was finished - I've never used CB15 before), got a score of 981 and temperature max of 59 degC - so that was stable! I might try OCCT, I've never used that before either.

 

I don't recall so much, since it's been a couple years since I've used it because I started to consistently get a single thread failure when everything else I threw at it was 100% stable. I just couldn't rely on it for stress or stability testing because even a little adjustment outside of stock clocks and voltage would frequently result in a failure.

 

You had different turbo boosts for number of cores active? Say 4.5Ghz for 1 & 2 core, 4.0Ghz for 3 & 4 cores active? I've not had that problem yet that you're talking about, although my max turbo for single & all cores are the same: 4.5Ghz for everything.

EDIT: OCCT stable for 20mins, 55 degC average CPU temp, max CPU temp 66 degC, 88-102W CPU power. I'm confident my CPU is stable now, also what with running the 8hr Prime95 & also gaming for hours on it.

EDIT 2: to keep on topic of thread (kind of), I've finalised an overclock without additional voltage for my Zotac GTX 1070 Amp (albeit it's in a desktop of course): +200 core / +500 VRAM. I'll post up the Firestrike & 3DMark 11 scores after I run them again. How is this comparing against notebook GTX 1070 & 1080, because I know they're pretty impressive still in notebooks??
(pic of GPU overclock settings follow):


***Firestrike Benchmark: 17106 (20911, 14354, 8237) http://www.3dmark.com/3dm/16352201?

***3DMark11 Benchmark: 21552 (27532, 13032, 13075) http://www.3dmark.com/3dm11/11780393

(gaming stable boost clocks of 2037-2050Mhz at max 72 degC and 1.031V-1.062V)
Max power draw from the wall for my desktop has been around 280 odd Watts, with normal power draw being around 250W (not including monitor of course) - hell this could almost be a notebook??!!

 

Thread title?
Last ~10-ish pages made me question if I clicked on a right link...

 

Well, we are discussing stability testing tools - they could be used by owners OF GTX 10-series notebooks to figure out max temps or stability of undervolts.

Sent from my SM-G935F using Tapatalk

 

Fair point, I was reading this as "gtx 10xx "notebook", not as "gtx 10xx notebook", considering we are in the section called "software and graphics cards"

 

says who? its not about max temps here, but rather about voltage stability

Sent from my Huawei Mate 8 NXT-AL10

 

If you are over-temping, power throttling, thermal throttling, you are *not* exercising full voltage draw or testing voltage stability, you are testing power and thermal protection

You must maintain full performance to get full voltage stability, you have to be running pushing limits but not exceeding them.

I'm sure you can work out a method that works for you, but for someone just starting out, it's a good rule of thumb to let things cool down - and watch it happen, and then push hard again watching readings.

The process of doing that and watching the readings will bring about thoughts of optimizing for further intensity in testing.

It's a nice place to start, aware and with a purpose

 

(mine wasn't throttling, but luxury of a desktop, although I'm sure some laptops can also manage, so I ran back to back CB15 runs)

 

ah yes, power/thermal/current limits were all put aside as a base for testing voltage stability and once u reach the thermal ceiling ure done anyways (unless your hardware just gives out due to its frequency limit)

Sent from my Huawei Mate 8 NXT-AL10

 

981cb in Cinebench R15. But with What clock speed? 4.5 GHz? I can see your sig show [email protected]. If so the CB R15 score is a bit low. You should have been more like +1000cb. What was the highest vs. lowest score you got?

Spoiler: CB R15 with 4.5 GHz

 

I do everything, from running benches is loops to stress testing. If you use Intel burn test, p95 small fft, or OCCT, you can check what the biggest heat you are likely to incur (which is really nice if you use encoding software, especially if AVX2 is used). Using chrome and tabs for a browser is like lynpack testing or P95 random which uses high memory, or OCCT testing with high memory. It tests stability in a different way, which can show CPU or ram OC weaknesses. This can find issues that might show up in memory heavy tasks and programs. The key is to throw everything at it that will push the system further than your everyday uses. My system does p95 small fft @4.8@70C on the hottest core with a push pull on my radiator (something a laptop cannot do on its own well).

As for tabs, I do two screens with lots of research open, different tabs for each source. When you are verifying manuscripts with over 800 footnotes, you'd be surprised how well that works!

Also, why did you choose a 1070 over a 980 Ti?

Your scores are around where mine are at. I got 11.70 on cb 11.5 and 1065 on cb 15. I'm still futzing with Windows 7 install, but I'll get it soon enough... So we will have some fun!



Sent from my SM-G900P using Tapatalk

 

You have 1001 and I have 981, that's only 2% faster. You have Windows 8, I have Windows 10 - that could make a difference. Also, I can imagine you disabling every Windows service under the sun that you don't need & also probably making sure you're not running any programs unnecessarily in the background. I didn't do any of that 'tweaking', so I think both of those reasons would likely account for the 2% difference, I'm not gonna lose any sleep over it. (Oh yeah, I didn't answer your other question - range of scores were from 980-983, so stable scores over 11 runs).

That's why you're a tab whore then! I chose a GTX 1070 over a 980ti because its the newer tech & is slightly faster & uses less power - it's more future proof. Let me guess, this is where you're probably gonna prove me wrong with your OC'd 980ti! ;-) Either way, I'd still prefer the 1070 for the probably better compatibility with future games (DX12/Vulkan perhaps), and the lower power consumption, also has 8GB VRAM vs 6GB, all the better for furture proofing!

 

I plan on going to 1080 Ti or waiting on Volta. Pascal has some concerning built in throttle. I got an EVGA 980 Ti Classified with 14+3 passes for $315. I wasn't concerned about future proofing, just getting a good card that OCs like a beast, has a dual bios to play and learn with, all while at a decent price. I'm not going to bash your choice of a 1070 amp because our hardware is equivalent regardless of which card. Mine overclocks a bit better, but yours has the edge at stock. You are planning on longer ownership, so paying more makes sense. I just wanted to know why you chose it. Has it been conclusively down Pascal performs better than Maxwell at Vulkan, dx12, and async processing?

For power consumption, not a factor for me for a desktop! But understandable. As you can tell, not here to bash, just was wondering...

Sent from my SM-G900P using Tapatalk

 

Well yeah, I honestly didn't even consider for one moment getting a 980ti. I knew that at stock the GTX 1070 was faster - I knew that much. I didn't want to buy a second hand 980ti, but I didn't even consider it that far to be honest. If I think about it logically now, I think if I had gone 980ti then it would use more power, that heat would have to be dissipated into the case in a 'normal' setup, which increases case temperatures and cause other issues and potentially increase the cost of the rest of the build, it could change the whole build. (My whole tower uses only something like 250-260W when gaming normally when measured at the wall). I just straight up didn't consider 980ti for a moment. I don't regret it though, because of all those reasons I mentioned: power consumption (adds up costs over ownership life), case heat dissipation, having to buy second hand, 8GB vs 6GB VRAM, better compatibility with future games/drivers. I think I heard that Pascal performs with DX12 & Vulkan better than Maxwell, but that's just a vagary on my part from something I think I've read - haha!

EDIT: I'll probably go research the Pascal vs Maxwell DX12/Vulkan idea after I've cooked some food. Was it a guru3d article/chart I saw, hmm, don't know, I'll have to satisfy my curiosity later.

EDIT 2: And yeah, as you say, a 980ti like you've got is good to get second hand at $315, especially if you know you're not gonna keep it that long - it doesn't have to live & survive as long, and that's a bit more of an unknown with a second hand card. I'm happy to keep mine for a few years as it stands currently. (He says now!)

 

That's cool. I centered my build around future upgrading and eventually a custom water loop (which later can be changed to a chilled water loop). We had different build goals and came to different conclusions based on our goals. I found a college student that owned this card and never overclocked, so the card never had it's legs stretched (I'll be changing that soon). But there is always a risk with that approach, I agree.

As for cooling, I've got 6 Noctua nf-f12 iPPC 3000 rpm fans on my aio cooler, the original 100 CFM fans pulling air in, a 20cm fan in the front, a 12cm fan blowing out the back, and a ram cooler that pushes 26 cfm on my ram (Plus the two fans in the card itself). I have it set to be quiet for normal operations, but a beast when I benchmark (sounds about like a vacuum cleaner). So, once again, the beauty of different builds based on different goals and the beauty of customization...

Sent from my SM-G900P using Tapatalk

 

Well said! And sounds like an interesting build too.

 

While my food's cooking done a bit of quick research on Vulkan re Pascal vs Maxwell. My first bit of data comes from here:
http://www.gamersnexus.net/game-bench/2510-doom-vulkan-vs-opengl-benchmark-rx-480-gtx-1080
From that it shows GTX 1080 gaining 19% performance from OpenGL to Vulkan, whereas GTX 970 gains only 1% from moving to Vulkan (although maybe could be argued that GTX 1080 was initially underperforming in OpenGL vs the GTX 970, but not checked the ratio of difference yet). I'll still do some more research to see what I can find re more Vulkan & also DX12.

This one also on topic of Vulkan & Doom:
http://www.eteknix.com/doom-opengl-vs-vulkan-graphics-performance-analysis/5/
results at 1080p show the biggest disparity (the resolution I use for 144Hz gaming with my new build). Negative scaling for some Maxwell cards, slight positive scaling for other Maxwell cards (980ti offering biggest Maxwell scaling), somewhat larger scaling for Pascal cards.

Mmm, food time!....(I might see if I can find DX12 stuff later)...

EDIT: literally just 5 mins research re DX12, too tired to do more research right now (need to veg with a film or something!):
http://www.pcworld.com/article/3071...badass-graphics-card-ever-created.html?page=8
Pascal showing some gains with DX12 over DX11, Maxwell not showing any tangible gains going from DX12 to DX11. Need an AMD card really to love the DX12 & Vulkan, but those AMD cards start off at a lower performance baseline anyway, so swings & roundabouts!

 

Intel, Nvidia ready to unveil new platforms for CES 2017 ( January 5-8, 2017)

http://www.digitimes.com/news/a20161130PD203.html

 

Laptop version GTX 1050 to launch at CES
https://www.guru3d.com/news-story/laptop-version-gtx-1050-to-launch-at-ces.html

"Recently you have already been able to spot some benchmark leaks on it, but according to the latest rumors the GeForce GTX 1050 (and Ti) for mobile will be released during CES. Tests show that the Ti si as fast as a GTX 970M with the regular 1050 model roughly 10% slower.

Nvidia is also ready to announce its entry-level GeForce GTX 1050 GPU for notebooks, according to sources from the upstream supply chain reports digitimes. Nvidia GTX 1060 GPU-equipped notebooks are currently enjoying strong demand from China and Southeast Asia and the release of the GTX 1050 is expected to further increase sales."

Spoiler: 1050ti mobile score compared to 970m, 965m, 960m

Chip 3DMark Cloud Gate 3DMark Fire Strike Unigine Heaven 4
GTX 1050 Ti (Laptop) 49.976 7.757 1.836
GTX 970M (ASUS GL502) 45.541 7.271 1.691
GTX 965M (HP Omen 2016) 38.125 6.063 1.260
GTX 960M (Lenovo Y700) 31.097 4.451 9.89

 

OT: RX 470 performance similar to a 970m...at least according to the firestrike run done in this thread. This is only one sample though.
http://forum.notebookreview.com/thr...y-config-is-unavailable.798865/#post-10401227

 

The RX470 dominates the 1050ti in these tests, 22% faster on average

 

Sorry for the poor image, but I think this is an error @GenTechPC - I thought the GL502VM maxed out at 24GB RAM? At least, that is what XoticPC's and Pro-Star's customization page shows.

@hmscott - Is this right?

Sent from my SM-G935F using Tapatalk

 

According to the Asus GL502 Specifications page:
https://www.asus.com/us/ROG-Republic-Of-Gamers/ROG-GL502VM/specifications/

"Memory - 8 GB DDR4 2133MHz SDRAM Onboard Memory, 1 x SO-DIMM socket for expansion, up to 16 GB SDRAM"

If those slots support 2 x 16GB SO-DIMM's that would allow up to 32GB.

2 RAM Slots are visible in this internal shot:

From this review:
http://www.hardwarezone.com.sg/review-asus-rog-strix-gl502-gaming-notebook-preview-orange-new-black

 

Then all the resellers are wrong? That would be weird. Anyway, thanks for the useful info as always.

Sent from my SM-G935F using Tapatalk

 

Hmmm, seems to be a wording issue with the GL502 specification page, they must mean 1x 8GB + 1x 8GB = 16GB.

A little more reading, and an internal photo shows 2 x SO-DIMM slots, and if it supports 16GB SO-DIMM's, you would top out at 32GB.



I updated my previous post.

 

Interesting. I wonder why Xoticpc, Pro-Star and HIDevolution max out at 24GB?

Sent from my SM-G935F using Tapatalk

 

IDK, ask them?

 

Maybe the 8GB stick is soldered?

Sent from my SM-G935F using Tapatalk

 

The internal photo clearly shows 2 sockets for SO-DIMM's.

 

Let's just say everybody but asus got their specs wrong. Even Alienware 13 I was looking at messed up the dimensions making out it was as wide as the 15


Sent from my iPhone using Tapatalk

 

Thanks. This is very interesting.

Sent from my SM-G935F using Tapatalk

 

I just noticed a weird thing in this video :
the clock of the 1070 is in the 18xxMHz when load is low (50% more or less). It suddenly goes back down to 15xxMHz when fully loaded (with unlocked fps / no vsync)..
Is this some limitation of the vbios ? then MSI or Nvidia is making the card throttle at 59°C--63°C.. That's bad..
Ps : He got Trurbo Fans, a.k.a, hellicopter noise fans always ON while gaming, even if he lives in the Netherlands..

 

The more these Pascal notebook cards are loaded the lower the frequency they run at. It's something I've noted as well. However, mine doesn't get that low in frequency.. I'm always at or above boost clock speeds, and higher when overclocked even under 100% load.

 

http://www.notebookcheck.net/Razer-Blade-2016-Notebook-Review.186751.0.html

Dat throttling.

 

So the classic +135MHz OC must be done to get the proper frequencies back...

 

nvm; misread.

 

The review is updated with the info that an EC update solves the CPU throttling.

 

http://www.notebookcheck.net/Mobile...060-Laptop-Benchmarks-and-Specs.169547.0.html

The Blade has the lowest graphics score of any 1060 laptop tested.

 

This raises the question though - it is still above an 1050, no?

Seems that since they fixed their throttling, their problems are alleviated. Are they in line with other 1060 based laptops?

 

Well someone reported on the blade they had or seen 9400 as fire strike score as opposed to that 9100 score. I guess drivers play a part a lot more than you think. I read on the aorus forum earlier from a user with a 1070 getting 18000+ fire strike graphics score. Then update the driver and got 17000 but much lower temps of 69c. But personally I would rather my gpu running at full potential hitting 80c cos that's what I paid for...power and expect it to get hot. I assume notebook check don't update drivers. Im sure they run stock drivers. In the 1060s case. It's the other way around and the scores go higher instead of lower from reading results people have done


Sent from my iPhone using Tapatalk

 

i could imagine that at 800 mhz the cpu would bottleneck the gpu score immensely. i must admit, not bad for a bga book concerning build quality, (non)-throttling and physical size/weight. however, the onboard ram (wtf), weak heatsink system (bad for overclocking), slow cpu and only one m.2 slot for storage makes me shy away from such machines...

Sent from my Huawei Mate 8 NXT-AL10

 

May as well be a tablet or smartphone with such a static and anemic configuration. With only a single M.2 they may as well solder that in place and use a microSD card for storage to make it a fully disposable one-shot wonder. I can't see the point in being excited about build quality at the expense of literally every other aspect of the product. And, for $3200 USD it is severely overpriced (at least double what would be appropriate be for such a pathetic offering).

 

Pascal is just weird, it overclocks weirder than Maxwell did. It performs similarly on desktop but not to the same extent. In my Tornado F5 and in the GT62 I had it would hold spec'd boost clocks and usually above those. When overclocked it goes well beyond the boost clocks. I can overclock my 1070 card +250mhz, the +135mhz limit doesn't exist in this gen, however they've locked us out of voltage and power controls unlike the desktop cards.

 

So the clocks would be like 16xx + 250Mhz + boost ? like 1950 Impressive Pascal (when pushed to limits)..